The Johns Hopkins ACG System. Version 11.0 Technical Reference Guide November 2014

Transcription

1 The Johns Hopkins ACG System Version 11.0 Technical Reference Guide November 2014

2 Important Warranty Limitation and Copyright Notices Copyright 2014, The Johns Hopkins University. All rights reserved. This document is produced by the Department of Health Policy and Management at The Johns Hopkins University, Bloomberg School of Public Health. The terms The Johns Hopkins ACG System, ACG System, ACG, ADG, Adjusted Clinical Groups, Ambulatory Care Groups, Aggregated Diagnostic Groups, Ambulatory Diagnostic Groups, Johns Hopkins Expanded Diagnosis Clusters, EDCs, ACG Predictive Model, Rx-Defined Morbidity Groups, Rx-MGs, ACG Rx Gaps, ACG Coordination Markers, ACG -PM, Dx-PM, Rx-PM and DxRx-PM are trademarks of The Johns Hopkins University. All materials in this document are copyrighted by The Johns Hopkins University. It is an infringement of copyright law to develop any derivative product based on the grouping algorithm or other information presented in this document. This document is provided as an information resource on measuring population morbidity for those with expertise in risk-adjustment models. The documentation should be used for informational purposes only. Information contained herein does not constitute recommendation for or advice about medical treatment or business practices. No permission is granted to redistribute this documentation. No permission is granted to modify or otherwise create derivative works of this documentation. Copies may be made only by the individual who requested the documentation initially from Johns Hopkins or their agents and only for that person's use and those of his/her co-workers at the same place of employment. All such copies must include the copyright notice above, this grant of permission and the disclaimer below must appear in all copies made; and so long as the name of The Johns Hopkins University is not used in any advertising or publicity pertaining to the use or distribution of this software without specific, written prior authorization. Disclaimer: This documentation is provided AS IS, without representation as to its fitness for any purpose, and without warranty of any kind, either express or implied, including without limitation the implied warranties of merchantability and fitness for a particular purpose. The Johns Hopkins University and the Johns Hopkins Health System shall not be liable for any damages, including special, indirect, incidental, or consequential damages, with respect to any claim arising out of or in connection with the use of the documentation, even if it has been or is hereafter advised of the possibility of such damages.

3 Contents Contents Chapter 1: Getting Started...6 Introduction to The Johns Hopkins ACG System...6 Customer Commitment and Contact Information...9 Chapter 2: Diagnosis-based Markers...10 Morbidity Types Aggregated Diagnosis Groups (ADGs)...10 Patterns of Morbidity Adjusted Clinical Groups (ACGs)...17 Clinically Oriented Examples of ACGs...29 Resource Utilization Bands (RUBs)...35 Disease Markers - Expanded Diagnosis Clusters (EDCs)...41 Special Population Markers...49 Chronic Condition Count...49 Hospital Dominant Morbidity Types...55 Frailty Conditions...56 Compassionate Care Allowances (CAL-SSA)...58 Pregnant...58 Delivered...58 Pregnancy without Delivery...59 Low Birth Weight (Less than 2500 Grams)...59 Chapter 3: Pharmacy-based Markers...60 Rx-Defined Morbidity Groups (Rx-MGs)...60 Active Ingredient Count...66 Chapter 4: Diagnosis+Pharmacy-based Markers...67 Condition Markers...67 Pharmacy Adherence...73 Chapter 5: Utilization and Resource Use Markers...88 Business Rules for Utilization Markers...88 All Cause Inpatient Hospitalization Count...88 Inpatient Hospitalization Count...88 Inpatient Hospitalization Days...89 Unplanned Inpatient Hospitalization Count...89 Readmission 30 Day Count...89 Unplanned Readmission 30 Day Count...89 Emergency Visit Count

4 Contents Outpatient Visit Count...90 Dialysis Service Marker...90 Nursing Service Marker...91 Major Procedure Marker...91 Cancer Treatment Marker...91 Mechanical Ventilation Marker...92 Psychotherapy Service Marker...92 Resource Bands...93 Chapter 6: Coordination...94 Assessing Care Coordination...94 Management Visit Count...96 Majority Source of Care (MSOC)...96 Unique Provider Count...98 Specialty Count...99 Generalist Seen Risk of Poor Coordination Care Density Conclusion Chapter 7: Risk Modeling Elements of a Risk Model: Five Key Dimensions Concurrent Cost Models Prospective Cost Models Chapter 8: Predictive Models for Hospitalization Typology of Acute Care Hospitalizations How the ACG System Predicts Hospitalization Predictive Models for Hospitalization include Utilization Measures Likelihood of Hospitalization Empiric Validation of the Likelihood of Hospitalization Model Likelihood of Unplanned Readmission within 30 Days Empiric Validation of the Readmission Model Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models Introduction Individual ACG Categories Included in the Risk Models ACGs Included in Three Resource Groups Pregnancy Without Delivery Delivered EDCs Included in the Risk Models

5 Contents Rx-Defined Morbidity Groups (Rx-MGs) Included in the Risk Models Special Population Markers Demographic Markers Optional Prior Cost Markers Utilization Markers Prior Hospitalization Markers Appendix B: Acknowledgements Documentation Production Staff Dedication Support Third-Party Library Acknowledgements Index

6 Chapter 1: Getting Started Chapter 1: Getting Started Introduction to The Johns Hopkins ACG System The ACG System characterized by excellence in both research and practice is based at the Johns Hopkins Bloomberg School of Public Health and has been used in performing risk measurement and casemix categorization for more than 25 years. The ACG System Team s mission is to support healthcare systems use of their available health information to measure the health care needs of their population, provide equitable distribution and remuneration of services, and identify ways to improve the efficiency of the delivery of care, with the primary objective being the improved health status of the population. The ACG System Team consists of dedicated Johns Hopkins University faculty and staff with broad experience in US and global health care systems. Their expertise includes extensive research on primary health care delivery and financing, the ability to interpret morbidity patterns from a clinical as well as managerial perspective, and the capacity to adapt models to fit the unique aspects of health care systems financial and organizational structures. Moreover, the Team possesses cutting-edge knowledge of electronic health records and health information technology, advanced analytics including model development, and a comprehensive hands-on understanding of international coding systems (both diagnostic and pharmaceutical). The ACG System has been used to support basic and complex applications in finance, administration, care delivery, and evaluative research for over two decades. These applications have been both concurrent (retrospective) and forward-looking (prospective). The applications may involve simple spreadsheet calculations or complex multi-variate statistical models. No other risk adjustment methodology has been used for so many purposes in so many places, while at the same time showing such high levels of quantitative and qualitative success. The flexibility offered by the ACG System demonstrates the recognition that one size does not fit all. The Risk Adjustment Pyramid figure illustrates the variety of ways in which risk adjustment is most commonly applied within healthcare organizations today. Some implementations, such as needs assessments or payment/financial applications apply to the entire population base. Other implementations, such as care-management or disease-management interventions, focus only on targeted sub-populations. 6

7 Chapter 1: Getting Started Figure 1. Risk Adjustment Pyramid The Overview of The Johns Hopkins ACG System figure illustrates the inputs to, and outputs from, the ACG-based analytics. In very simple terms, a patient file (identifying eligible individuals) is merged with diagnoses and pharmacy codes, to produce a series of risk factors and risk scores. The Installation and Usage Guide provides basic instructions on how to prepare input data files, select model options with which to process the data, and describes how to validate the output produced by the ACG Software. The full listing of risk factors assigned within the software is in Appendix A of the Installation and Usage Guide. The theoretical and conceptual basis of each of these markers as well as how they are combined and evaluated for modeling purposes is explicated in the Technical Reference Guide. 7

8 Chapter 1: Getting Started Figure 2. Overview of The Johns Hopkins ACG System Finally, the Common Applications figure describes how the ACG System is typically applied. Each potential application is described in more detail in the Applications Guide. Figure 3. Common Applications of the ACG System The ACG System has been used for a wide variety of applications in many contexts. The full ACG System bibliography is available in the Resource Center of the ACG website: /resource-center-83/acg-bibliography 8

9 Chapter 1: Getting Started Customer Commitment and Contact Information As part of our ongoing commitment to furthering the state-of-the-art of risk-adjustment methodology and supporting users of the ACG System, we will continue to perform evaluation, research, and development. We will look forward to sharing the results of this work with our user-base through white papers, our web site, peer-reviewed articles, and in-person presentations. After you have carefully reviewed the documentation supplied with this software release, we welcome your inquiries on any topic of relevance to your use of the ACG System within your organization. Technical support is available during standard business hours by contacting your designated account representative directly. If you do not know how to contact your account representative, please We thank you for using the ACG System and for helping us to work toward meeting the ACG Team's ultimate goal of improving the quality, efficiency, and equity of health care across the United States and around the globe. 9

10 Chapter 2: Diagnosis-based Markers Chapter 2: Diagnosis-based Markers The Johns Hopkins ACG System is a statistically valid, case-mix methodology that allows healthcare providers, healthcare organizations, and public-sector agencies to describe or predict a population s past or future healthcare utilization and costs. The ACG System is also widely used by researchers and analysts to compare various patient populations prior health resource use, while taking into account morbidity or illness burden. The ACG System provides a number of markers derived from a patient's diagnosis code history from all encounters during a 12-month period. This chapter provides definition for the ACG System markers derived from diagnosis information. Morbidity Types Aggregated Diagnosis Groups (ADGs) There are thousands of International Classification of Disease (ICD) diagnosis codes that clinicians can use to describe patients health conditions. The first step of the ACG grouping logic is to assign each diagnosis code to one or more of 32 diagnosis groups referred to as Aggregated Diagnosis Groups, or ADGs. The diagnosis-to-adg mapping embedded in the ACG Software includes an ADG assignment for all 1 ICD codes. Where a single diagnosis code indicates more than one underlying morbidity type, more than one ADG may be assigned. For example, in ICD-10 the code E11.31 (Type 2 diabetes mellitus with unspecified diabetic retinopathy) would trigger both ADG 18 (Chronic Specialty: Unstable-Eye) and ADG 11 (Chronic Medical: Unstable). Diagnosis codes within the same ADG are similar in terms of both clinical criteria and expected need for healthcare resources. Just as individuals may have multiple diagnosis codes, they may have multiple ADGs (up to 32). The following table lists the 32 ADGs and exemplary diagnosis codes. ADGs and Common Diagnosis Codes Assigned to Them ADGs ICD9-CM ICD-10 Diagnosis 1. Time Limited: Minor K52.9 Noninfectious Gastroentritis L22 Diaper or Napkin Rash 2. Time Limited: Minor-Primary B09 Unspecified Viral Infection Infections J05.0 Croup 3. Time Limited: Major I80.3 Phlebitis of Lower Extremities K56.7 Impaction of Intestine 4. Time Limited: Major-Primary K75.9 Hepatitis, Unspecified Infections M00.9 Pyogenic Arthritis 1 Because they indicate the cause of injury rather than an underlying morbidity, ICD-9 codes beginning with E and ICD-10 codes beginning V through Y have generally been excluded from the Diagnosis-to-ADG mapping. The source of codes is the Center for Medicare and Medicaid Services (CMS) list of ICD-9 and ICD-10-CM codes (available for download at ICD-10 codes are sourced from the Official ICD-10 Updates published by the World Health Organization (WHO). 10

11 Chapter 2: Diagnosis-based Markers ADGs ICD9-CM ICD-10 Diagnosis 5. Allergies J30.0 Allergic Rhinitis, Cause L50.9 Unspecified Unspecified Urticaria 6. Asthma J45.0 Extrinsic Asthma J45.1 Intrinsic Asthma 7. Likely to Recur: Discrete M10.9 Gout, Unspecified M54.9 Backache, Unspecified 8. Likely to Recur: Discrete J35.1 Chronic Tonsillitis Infection N39.0 Urinary Tract Infection 9. Likely to Recur: Progressive E11.1 Adult Onset Type II Diabetes I66.9 w/ketoacidosis Cerebral Thrombosis 10. Chronic Medical: Stable E10.9 Adult-Onset Type 1 Diabetes I10 Essential Hypertension 11. Chronic Medical: Unstable D57.1 Sickle-Cell Anemia E84.0 Cystic Fibrosis 12. Chronic Specialty: Stable M48.9 Cervical Spondylosis Without Orthopedic M24.9 Myelopathy Other Joint Derangement 13. Chronic Specialty: Stable H90.5 Central Hearing Loss Ear, Nose, Throat H71 Cholesteatoma 14. Chronic Specialty: Stable H52.1 Myopia Eye H11.9 Unspecified Disorder of Conjunctiva 16. Chronic Specialty: Unstable M48.0 Spinal Stenosis of Lumbar Orthopedic M92.8 Region Osteochondritis Dissecans 17. Chronic Specialty: Unstable H81.0 Meniere's Disease Ear, Nose, Throat H70.1 Chronic Mastoiditis 18. Chronic Specialty: Unstable H40.9 Unspecified Glaucoma Eye H15.0 Scleritis/Episcleritis 20. Dermatologic A63.0 Viral Warts I78.1 Nevus, Non-Neoplastic 21. Injuries/Adverse Effects: S13.4 Neck Sprain Minor T09.0 Injury to Trunk 22. Injuries/Adverse Effects: S06 Intracranial Injury Major T46.0 Poisoning by Cardiotonic Glycosides and Similar Drugs 11

12 Chapter 2: Diagnosis-based Markers ADGs ICD9-CM ICD-10 Diagnosis 23. Psychosocial: Time Limited, F12.1 Cannabis Abuse, Unspecified Minor F32.0 Brief Depressive Reaction 24. Psychosocial: Recurrent or F41.0 Panic Disorder Persistent, Stable F50.3 Bulimia 25. Psychosocial: Recurrent or F20.2 Catatonic Schizophrenia Persistent, Unstable F10.3 Alcohol Withdrawal Delirium Tremens 26. Signs/Symptoms: Minor G44.1 Headache M79.6 Pain in Limb 27. Signs/Symptoms: Uncertain M25.4 Effusion of Lower Leg Joint R53 Malaise and Fatigue 28. Signs/Symptoms: Major I51.7 Cardiomegaly R55 Syncope and Collapse 29. Discretionary K40 Inguinal Hernia (NOS) L72.1 Sebaceous Cyst 30. See and Reassure N62 Hypertrophy of Breast E65 Localized Adiposity 31. Prevention/Administrative V20.2 Z00.1 Routine Infant or Child Health V72.3 Z01.4 Check Gynecological Examination 32. Malignancy C50 Malignant Neoplasm of Breast C81.9 (NOS) Hodgkin's Disease, Unspecified Type 33. Pregnancy V22.2 Z33 Pregnant State Delivery in a Completely Normal Case 34. Dental K02 Dental Caries K05.1 Chronic Gingivitis Note: Only 32 of the 34 markers are currently in use. When the lenient diagnostic certainty option is applied, any single diagnosis qualifying for an ADG marker will turn the marker on. However, the stringent diagnostic certainty option can also be applied. For a subset of chronic diagnoses, there must be more than one diagnosis qualifying for the marker in order for the ADG to be assigned. This was designed to provide greater confidence in the ADGs assigned to a patient. For more information, refer to Chapter 4 in the Installation and Usage Guide. ADGs are distinguished by several clinical characteristics (time limited or not, requiring primary care or specialty care, or addressing physical health or psycho-social needs) and the degree of refinement of the problem (diagnosis or symptom/sign). ADGs are not categorized by organ system or disease. Instead, they are based on clinical dimensions that help explain or predict the need for healthcare resources over time. 12

13 Chapter 2: Diagnosis-based Markers The need for healthcare resources is primarily determined by the likelihood of persistence of problems and their level of severity. Example A patient with both Obstructive Chronic Bronchitis (ICD-9-CM code 491.2) and Congestive Heart Failure (ICD-9-CM code 428.0) will fall into only one ADG, Chronic Medical: Unstable (ADG-11), while a patient with Candidiasis of Unspecified Site (ICD-9-CM code 112.9) and Acute Upper Respiratory Infections of Unspecified Site (ICD-9-CM code 465.9) will have two ADGs, Likely to Recur: Discrete Infections (ADG-8), and Time Limited: Minor-Primary Infections (ADG-2), respectively. The criteria for ADG assignment depends on those features of a condition that are most helpful in understanding and predicting the duration and intensity of healthcare resources. Five clinical criteria guide the assignment of each diagnosis code into an ADG: duration, severity, diagnostic certainty, type of etiology, and expected need for specialty care. The Duration, Severity, Etiology, and Certainty of the ADGs table illustrates how each of these five clinical criteria is applied to the 32 ADGs. Duration What is the expected length of time the health condition will last? Acute conditions are time limited and expected to resolve completely. Recurrent conditions occur episodically with intermediate disease-free intervals. Chronic conditions persist and are expected to require long-term management generally longer than one year. Severity What is the expected prognosis? How likely is the condition to worsen or lead to impairment, death, or an altered physiologic state? The ADG-taxonomy divides acute conditions into minor and major categories corresponding to low and high severity, respectively. The system divides chronic conditions into stable and unstable based on the expected severity over time. Unstable conditions are more likely to have complications (related co-morbidities) than stable conditions and are expected to require more resources on an ongoing basis (i.e., more likely to need specialty care). Diagnostic Certainty Will a diagnostic evaluation be needed or will treatment be the primary focus? Some diagnosis codes are given for signs/symptoms and are associated with diagnostic uncertainty. As such, they may require watchful waiting only or substantial work-up. The three ADGs for signs/symptoms are arranged by expected intensity of diagnostic work-up, from low to intermediate to high. Etiology What is the cause of the health condition? Specific causes suggest the likelihood of different treatments. Infectious diseases usually require anti-microbial therapy; injuries may need emergency medical services, surgical management, or rehabilitation; anatomic problems may require surgical intervention; neoplastic diseases could involve surgical care, radiotherapy, chemotherapy; psychosocial problems require mental health services; pregnancy involves obstetric services; and, medical problems may require pharmacologic, rehabilitative, or supportive management. 13

14 Chapter 2: Diagnosis-based Markers Expected Need for Specialty Care Would the majority of patients with this condition be expected to require specialty care management from a non-primary care provider? The routine course of care for some ADG categories implies that specialty care is more likely. Duration, Severity, Etiology, and Certainty of the ADGs Note: ADGs 15 and 19 are no longer used. Expected Need Diagnostic for Specialty ADG Duration Severity Etiology Certainty Care 1. Time Limited: Acute Low Medical, non- High Unlikely Minor infectious 2. Time Limited: Acute Low Medical, High Unlikely Minor-Primary infectious Infections 3. Time Limited: Acute High Medical, non- High Likely Major infectious 4. Time Limited: Acute High Medical, High Likely Major-Primary infectious Infections 5. Allergies Recurrent Low Allergy High Possibly 6. Asthma Recurrent or Low Mixed High Possibly Chronic 7. Likely to Recurrent Low Medical, non- High Unlikely Recur: Discrete infectious 8. Likely to Recurrent Low Medical, High Unlikely Recur: Discrete- infectious Infections 9. Likely to Recurrent High Medical, non- High Likely Recur: infectious Progressive 10. Chronic Chronic Low Medical, non- High Unlikely Medical: Stable infectious 11. Chronic Chronic High Medical, non- High Likely Medical: infectious Unstable 12. Chronic Chronic Low Anatomic/Muscu High Likely: Specialty: Stable- loskeletal orthopedics Orthopedic 14

15 Chapter 2: Diagnosis-based Markers Expected Need Diagnostic for Specialty ADG Duration Severity Etiology Certainty Care 13. Chronic Chronic Low Anatomic/Ears, High Likely: ENT Specialty: Stable- Nose, Throat Ear, Nose, Throat 14. Chronic Chronic Low Anatomic/Eye High Likely: Specialty: Stable- ophthalmology Ophthalmology 16. Chronic Chronic High Anatomic/Muscu High Likely: Specialty: loskeletal orthopedics Unstable- Orthopedics 17. Chronic Chronic High Anatomic/Ears, High Likely: ENT Specialty: Nose, Throat Unstable-Ear, Nose, Throat 18. Chronic Chronic High Anatomic/Eye High Likely: Specialty: ophthalmology Unstable- Ophthalmology 20. Dermatologic Acute, Recurrent Low to High Mixed High Likely: dermatology 21. Acute Low Injury High Unlikely Injuries/Adverse Effects: Minor 22. Acute High Injury High Likely Injuries/Adverse Effects: Major 23. Psychosocial: Acute Low Psychosocial High Unlikely Time Limited, Minor 24. Psychosocial: Recurrent or Low Psychosocial High Likely: mental Recurrent or Chronic health Chronic, Stable 25. Psychosocial: Recurrent or High Psychosocial High Likely: mental Recurrent or Chronic health Persistent, Unstable 26. Uncertain Low Mixed High Unlikely Signs/Symptoms: Minor 15

16 Chapter 2: Diagnosis-based Markers Expected Need Diagnostic for Specialty ADG Duration Severity Etiology Certainty Care 27. Uncertain Uncertain Mixed High Uncertain Signs/Symptoms: Uncertain 28. Uncertain High Mixed Low Likely Signs/Symptoms: Major 29. Discretionary Acute Low to High Anatomic High Likely: surgical specialties 30. See and Acute Low Anatomic High Unlikely Reassure 31. N/A N/A N/A N/A Unlikely Prevention/Admi nistrative 32. Malignancy Chronic High Neoplastic High Likely: oncology 33. Pregnancy Acute Low Pregnancy High Likely: obstetrics 34. Dental Acute, Recurrent, Low to High Mixed High Likely: dental Chronic Major ADGs Some ADGs have very high expected resource use and are labeled as Major ADGs. The following table presents major ADGs for adult and pediatric populations. Major ADGs for Adult and Pediatric Populations Pediatric Major ADGs (ages 0-17 years) Adult Major ADGs (ages 18 and up) 3 Time Limited: Major 3 Time Limited: Major 9 Likely to Recur: Progressive 4 Time Limited: Major-Primary Infections 11 Chronic Medical: Unstable 9 Likely to Recur: Progressive 12 Chronic Specialty: Stable-Orthopedic 11 Chronic Medical: Unstable 13 Chronic Specialty: Stable-Ear, Nose, Throat 16 Chronic Specialty: Unstable-Orthopedic 18 Chronic Specialty: Unstable-Eye 22 Injuries/Adverse Effects: Major 25 Psychosocial: Recurrent or Persistent, Unstable 25 Psychosocial: Recurrent or Persistent, Unstable 32 Malignancy 32 Malignancy While the primary use of ADGs is as a means for collapsing all diagnosis codes into clinically meaningful morbidity types as a first step in the ACG assignment process, ADGs are useful as a risk assessment tool in their own right. There are many examples in the literature of using ADG markers as generic case-mix control variables. The most common application is the introduction of individual ADG-markers as binary 16

17 Chapter 2: Diagnosis-based Markers flags in a regression model, but something as simple as a count of ADGs or Major ADGs can be a very powerful indicator of need as well. Relationship Between Number and Major Morbidities in Year 1 and Likelihood of Subsequent High Cost Positive Predictive Value Number of Year 1 Major Percent High Cost in Percent High Cost in Morbidities Percent of Members Year 2 Year 3 0 Major ADGs 77.1% 9.6% 11.0% 1 Major ADG 17.3% 20.9% 21.5% 2 Major ADGs 4.2% 34.7% 34.1% 3 Major ADGs 1.1% 43.6% 45.6% 4+ Major ADGs 0.4% 72.4% 70.1% Patterns of Morbidity Adjusted Clinical Groups (ACGs) Adjusted Clinical Group actuarial cells, or ACGs, are the building blocks of The Johns Hopkins ACG System methodology. ACGs are a series of mutually exclusive, health status categories defined by morbidity, age, and sex. They are based on the premise that the level of resources necessary for delivering appropriate healthcare to a population is correlated with the illness burden of that population. ACGs are used to determine the morbidity profile of patient populations to more fairly assess provider performance, to reimburse providers based on the health needs of their patients, and to allow for more equitable comparisons of utilization or outcomes across two or more patient or enrollee aggregations. ACGs are a person-focused method of categorizing patients illnesses. Over time, each person develops numerous conditions. Based on the pattern of these morbidities, the ACG approach assigns each individual to a single ACG category. The concept of ACGs grew out of research by Dr. Barbara Starfield and her colleagues in the late 1970s when they examined the relationship between morbidity or illness burden and healthcare services utilization among children in managed care settings. The research team theorized that the children using the most healthcare resources were not those with a single chronic illness, but rather were those with multiple, seemingly unrelated conditions. To test their original hypothesis, illnesses found within pediatric health maintenance organization (HMO) populations were grouped into five discrete categories: 1. Minor illnesses that are self-limited if treated appropriately, e.g., the flu or chicken pox. 2. Illnesses that are more severe but also time-limited if treated appropriately, e.g., a broken leg or pneumonia. 3. Medical illnesses that are generally chronic and which remain incurable even with medical therapy, e.g., diabetes or cystic fibrosis. 4. Illnesses resulting from structural problems that are generally not curable even with adequate and appropriate intervention, e.g., cerebral palsy or scoliosis. 5. Psychosocial conditions, e.g., behavior problems or depression. The Johns Hopkins team s findings supported the hypothesis that clustering of morbidity is a better predictor of health services resource use than the presence of specific diseases. This finding forms the 17

18 Chapter 2: Diagnosis-based Markers basis of the current ACG methodology and remains the fundamental concept that differentiates ACGs from other case-mix adjustment methodologies. There are four steps in the ACG assignment process: 1. Mapping Diagnosis Codes to a Parsimonious Set of Aggregated ADGs 2. Creating a Manageable Number of ADG Subgroups (CADGs) 3. Frequently Occurring Combinations of CADGs (MACs) 4. Forming the Terminal Groups (ACGs) The first step is described in the preceding section while the remainder are summarized in the following tables and figures depicting the ACG-decision-tree logic. Creating a Manageable Number of ADG Subgroups (CADGs) The ultimate goal of the ACG algorithm is to assign each person to a single morbidity group (i.e., an ACG). There are 4.3 billion possible combinations of ADGs, so to create a more manageable number of unique combinations of morbidity groupings, the 32 ADGs are collapsed into 12 CADGs or Collapsed ADGs (presented in the following table). Like ADGs, CADGs are not mutually exclusive in that an individual can be assigned to as few as none or as many as 12. Although numerous analytic techniques could be used to form CADGs from ADGs, the ACG System has placed the emphasis on clinical cogency. The following three clinical criteria are used: The similarity of likelihood of persistence or recurrence of diagnoses within the ADG, i.e., time-limited, likely to recur, or chronic groupings The severity of the condition, i.e., minor versus major and stable versus unstable The types of healthcare services required for patient management--medical versus specialty, eye/dental, psychosocial, prevention/administrative, and pregnancy ADGs and CADGs can be used for various analytic and research applications that do not require mutually exclusive categories such as multivariate predictive or explanatory models. Collapsed ADG Clusters and the ADGs that Comprise Them Collapsed ADG (CADG) ADGs in Each 1. Acute Minor 1 Time Limited: Minor 2 Time Limited: Minor-Primary Infections 21 Injuries/Adverse Events: Minor 26 Signs/Symptoms: Minor 2. Acute Major 3 Time Limited: Major 4 Time Limited: Major-Primary Infections 22 Injuries/Adverse Events: Major 27 Signs/Symptoms: Uncertain 28 Signs/Symptoms: Major 18

19 Chapter 2: Diagnosis-based Markers Collapsed ADG (CADG) ADGs in Each 3. Likely to Recur 5 Allergies 7 Likely to Recur: Discrete 8 Likely to Recur: Discrete-Infections 20 Dermatologic 29 Discretionary 4. Asthma 6 Asthma 5. Chronic Medical: Unstable 9 Likely to Recur: Progressive 11 Chronic Medical: Unstable 32 Malignancy 6. Chronic Medical: Stable 10 Chronic Medical: Stable 30 See and Reassure 7. Chronic Specialty: Stable 12 Chronic Specialty: Stable-Orthopedic 13 Chronic Specialty: Stable-Ear, Nose, Throat 8. Eye/Dental 14 Chronic Specialty: Stable-Eye 34 Dental 9. Chronic Specialty: Unstable 16 Chronic Specialty: Unstable-Orthopedic 17 Chronic Specialty: Unstable-Ear, Nose, Throat 8 Chronic Specialty: Unstable-Eye 10. Psychosocial 23 Psycho-social: Time Limited, Minor 24 Psycho-social: Recurrent or Persistent, Stable 25 Psycho-social: Recurrent or Persistent, Unstable 11. Preventive/Administrative 31 Prevention/Administrative 12. Pregnancy 33 Pregnancy Frequently Occurring Combinations of CADGs (MACs) The third step in the ACG categorization methodology assigns individuals into a single, mutually exclusive category called a MAC. This grouping algorithm is based primarily on the pattern of CADGs. The MACs and the Collapsed ADGs Assigned to Them table shows the MACs and the Collapsed ADGs which comprise them. There are 23 commonly occurring combinations of CADGs which form MACs 1 through 23: The first 11 MACs correspond to presence of a single CADG. MAC-12 includes all pregnant women, regardless of their pattern of CADGs. MACs 13 through 23 are commonly occurring combinations of CADGs. MAC-24 includes all other combinations of CADGs. MAC-25 is used for enrollees with no service use or invalid diagnosis input data. MAC-26 includes all infants (age <12 months), regardless of the pattern of CADGs. 19

20 Chapter 2: Diagnosis-based Markers MACs and the Collapsed ADGs Assigned to Them MACs 1. Acute: Minor 1 2. Acute: Major 2 3. Likely to Recur 3 4. Asthma 4 5. Chronic Medical: Unstable 5 6. Chronic Medical: Stable 6 7. Chronic Specialty: Stable 7 8. Eye/Dental 8 9. Chronic Specialty: Unstable Psychosocial Prevention/Administrative 11 CADGs 12. Pregnancy All CADG combinations that include CADG Acute: Minor and Acute: Major 1 and Acute: Minor and Likely to Recur 1 and Acute: Minor and Chronic Medical: Stable 1 and Acute: Minor and Eye/Dental 1 and Acute: Minor and Psychosocial 1 and Acute: Major and Likely to Recur 2 and Acute: Minor and Acute: Major and Likely to Recur 1, 2 and Acute: Minor and Likely to Recur and Eye and 1, 3 and 8 Dental 21. Acute: Minor and Likely to Recur and Psychosocial 1, 3, and Acute: Minor and Major and Likely to Recur and 1, 2, 3, and 6 Chronic Medical: Stable 23. Acute: Minor and Major and Likely to Recur and 1, 2, 3, and 10 Psychosocial 24. All Other Combinations Not Listed Above All Other Combinations 25. No Diagnosis or Only Unclassified Diagnosis No CADGs 26. Infants (age less than one year) Any CADGs combinations and less than one year old 20

21 Chapter 2: Diagnosis-based Markers Forming the Terminal Groups (ACGs) MACs form the major branches of the ACG decision tree. The final step in the grouping algorithm divides the MAC branches into terminal groups, the actuarial cells known as ACGs. The logic used to split MACs into ACGs includes a combination of statistical considerations and clinical insight. During the ACG development process, the overarching goal for ACG assignment was to identify groups of individuals with similar needs for healthcare resources who also share similar clinical characteristics. Yale University s AUTOGRP Software (which performs recursive partitioning) was used to identify subdivisions of patients within a MAC who had similar needs for healthcare resources based on their overall expenditures. The variables taken into consideration included: age, sex, presence of specific ADGs, number of major ADGs, and total number of ADGs. Note: Because prevention/administrative needs do not reflect morbidity, ADG 31 is not included in the count of total ADGs 2. See the Final ACG Categories, Reference ACG Concurrent Risks, and RUBs table on page 36 for a complete listing and description of all ACGs. ACG Decision Tree The ACG Decision Tree figure illustrates the main branches of the ACG decision tree. Some MACs are not subdivided by the characteristics listed above because doing so did not increase the explanatory power of the ACG model. Some include only a single CADG: for instance, MAC-02 is composed of individuals with only acute major conditions. Others, such as MAC-01, acute conditions only, are subdivided into three age groups: ACG 0100 (Age = one year), ACG 0200 (Age = two to five years), and ACG 0300 (six or more years) because resource use differs by age for individuals with this pattern of morbidity. MAC-10, including individuals with psychosocial morbidity only and MAC-17, including individuals with psychosocial and acute minor conditions, are further split by the presence of ADG-24 (recurrent or chronic stable psychosocial conditions) and ADG-25 (recurrent or chronic unstable psychosocial conditions). 2 Refer to Weiner (91) and Starfield (91) for more detail on the historical origins of the current system including the original Version 1.0 development process. 21

22 Chapter 2: Diagnosis-based Markers Figure 4. ACG Decision Tree 22

23 Chapter 2: Diagnosis-based Markers Decision Tree for MAC-12 Pregnant Women The Decision Tree for MAC-12 Pregnant Women illustrates the grouping logic for pregnant women. All women with at least one diagnosis code indicating pregnancy are assigned to MAC-12. The ACGs for pregnant women are formed with subdivisions first on total number of ADGs (0-1, 2-3, 4-5, 6+) and second, for individuals with two or more ADGs, a split on none versus one or more major ADGs. These two splits yield seven ACGs for pregnant women. The standard seven ACGs for pregnant women can optionally be further subdivided according to whether delivery has occurred during the time period of interest, yielding a total of 14 ACGs for women with a diagnosis of pregnancy. Either diagnosis codes for delivery or a user-supplied delivery flag can be used to separate pregnant women according to delivery status. Because of the marked differences in resource consumption for women with and without delivery and generally adequate validity of diagnoses associated with delivery, most organizations will find this option desirable to use. By default, the software will use diagnosis codes to subdivide based on delivery status. Refer to Pregnant on page 58 and Delivered on page 58 for a more detailed discussion of appropriate means of identifying pregnancy and delivery status using user-defined flags. 23

24 Chapter 2: Diagnosis-based Markers Figure 5. Decision Tree for MAC-12 Pregnant Women 24

25 Chapter 2: Diagnosis-based Markers Decision Tree for MAC-26 Infants The Decision Tree for MAC-26 Infants illustrates the branching algorithm for MAC-26, which includes all infants, regardless of their pattern of CADGs. The first bifurcation is made on the total number of ADGs. Each group is further subdivided by presence of the number of major ADGs. These two splits yield four ACG groups. For the infant ACGs, there is an optional additional split on birth weight. If there is accurate birth weight information that can be linked with claims and enrollment files, the four standard infant ACGs can be further split into low birth weight (<2,500 grams) and normal birth weight (>2,500 grams). Our developmental work suggests that this additional split improves the explanatory power of the ACG System. However, two caveats are important to consider before using this ACG option. First, our research indicates poor validity for existing ICD-9-CM birth weight codes in some administrative data sets. Second, some populations may have such low rates of low birth weight infants that the number of infants grouped into an ACG may be too small for accurate estimates. In general, we recommend that at least 30 individuals per ACG are needed to obtain stable estimates of average resource use for that ACG. By default, the ACG System will divide infants based upon the presence or absence of a diagnosis code indicating low birth weight. Refer to Low Birth Weight (Less than 2500 Grams) on page 59 for a more detailed discussion of appropriate means of identifying low birth weight status using user-defined flags. 25

26 Chapter 2: Diagnosis-based Markers Figure 6. Decision Tree for MAC-26 Infants 26

27 Chapter 2: Diagnosis-based Markers Decision Tree for MAC-24 Multiple ADG Categories The Decision Tree for MAC-24 Multiple ADG Categories illustrates the last branch of the ACG tree, MAC- 24, which includes less frequently occurring combinations of CADGs. There are 33 ACGs within MAC-24. With MAC-24, the first two splits are total number of ADGs (2-3, 4-5, 6-9, and 10+) and then, within each of these four groups, by age. The age splits separate children (1-17 years) from adults (18+), and in some cases further subdivides within these groups. Additional divisions are made on sex and number of major ADGs. 27

28 Chapter 2: Diagnosis-based Markers Figure 7. Decision Tree for MAC-24 Multiple ADG Categories 28

29 Chapter 2: Diagnosis-based Markers Clinically Oriented Examples of ACGs Patients are categorized into an ACG based on the pattern of ADGs experienced over a predetermined interval and, in some cases, their age and sex. This approach focuses on the totality of diseases experienced by a person rather than any specific disease. Because this method diverges from the traditional biomedical, categorical method of examining morbidity, we show how ACGs classify patients with specific types of diseases. In the examples that follow, we categorize patients by choosing a specific clinical feature that they have, such as a disease, pregnancy, or by their age. These examples show how the presence of other diseases or total number of ADGs changes ACG assignment. Chronic Illnesses In the following examples, Example 1: Hypertension presents three patients with hypertension and Example 2: Diabetes Mellitus presents three patients with diabetes. These individuals were actual patients selected from a private healthcare organization database. The input data used by the ACG grouping software, the output produced by the software, and the associated resource consumption variables are presented. As these patients demonstrate, there is substantial variability in patterns of morbidity and need for healthcare for different patients classified by a specific condition such as hypertension or diabetes. Thus, knowing only that a patient has a particular medical problem, even if it is a chronic condition, provides little information about the need for medical services. In general, as the number of different types of morbidities increases, the total number of ambulatory visits increases as does total expenditures. However, the total burden of morbidity as represented by the ACG that is, the constellation of ADGs and presence of major ADGs is the most important determinant of resource consumption. In Example 1: Hypertension, during the assessment period Patient 1 had diagnosis codes given for only hypertension and a routine medical exam and is therefore classified into the ACG for patients with stable, chronic medical conditions (ACG-0900). In contrast, Patient 3 with hypertension is in an ACG that branches from MAC-24 (combinations of ADGs not otherwise classified). This occurs because the combinations of ADGs occur too infrequently to merit a separate ACG. Patients in MAC-24 have both high levels of morbidity and high levels of health need. There is a strong link between the total number of ADGs/major ADGs and resource consumption. There are two additional ACGs that describe commonly occurring combinations of morbidity for individuals with stable, chronic medical conditions. ACG-2300 (Chronic Medical--Stable and Acute Minor) is assigned to patients with uncomplicated diabetes, hypertension, or other stable chronic conditions and a minor illness, injury, or symptom. As shown in Patient 2 with Hypertension, individuals in ACG-3600 have four types of morbidities: stable chronic medical conditions (which include the diagnosis of hypertension), acute minor conditions, conditions of low severity likely to reoccur, and acute major conditions. Example 1: Hypertension The following patient types demonstrate the levels of hypertension, ADGs, and associated costs. 29

30 Chapter 2: Diagnosis-based Markers Patient 1: Low Cost Patient with Hypertension Input Data/Patient Characteristics ACG Output Resource Consumption Variables Age/Sex: 51/Male ACG-0900: Chronic Medical, Total Cost: $128 Stable Ambulatory visits: 1 Emergency visits: 0 Conditions: Hypertension, General ADGs: 10 and 31. Medical Exam Chronic Medical: Stable, Prevention/Administrative Patient 2: High Cost Patient with Hypertension Hospitalizations: 0 Input Data/Patient Characteristics ACG Output Resource Consumption Variables Age/Sex: 54/Male ACG-3600: Acute Minor/Acute Total Cost: $3,268 Major/Likely Recur/Eye & Dental Ambulatory visits: 1 Conditions: Hypertension, ADGS: 07, 10, 26, and 27 Emergency visits: 1 Disorders of Lipid Metabolism, Likely to Recur: Discrete Chronic Hospitalizations: 0 Low Back Pain, Cervical Pain Medical: Stable Syndromes, Musculoskeletal Signs Signs/Symptoms: Minor and Symptoms Signs/Symptoms: Uncertain Patient 3: Very High Cost Patient with Hypertension Input Data/Patient Characteristics ACG Output Resource Consumption Variables Age/Sex: 52/Male ACG : 10+Other ADG Total Cost: $45,937 Combinations, Age >17, 4+ Major Ambulatory visits: 17 ADGs Emergency visits: 0 Conditions: Hypertension, General ADGs: 02, 03*, 06, 07, 09*, 10, Hospitalizations: 1 medical exam, Cardiogenic Shock, 11*, 16*, 24, 27, 28, and 31. Asthma, Low back pain, Time Limited: Minor-Primary Peripheral Neuropathy, Anxiety, Infections Depression, COPD, Acute Upper Time Limited: Major, Asthma Respiratory Infection, Likely to Recur: Discrete Gastroesophageal Reflux, Iron Likely to Recur: Progressive Deficiency, Cervical Pain Syndromes, Sleep Problems, Chronic Medical: Stable Chronic Obesity, Sinusitis, Joint Pain Medical: Unstable Chronic Specialty: Unstable-Orthopedic Psychosocial: Recurrent or Persistent Stable Signs/Symptoms: Uncertain Signs/Symptoms: Major, and Prevention/Administrative *Major ADG 30

31 Chapter 2: Diagnosis-based Markers Example 2: Diabetes Mellitus The following patient types demonstrate the levels of diabetes mellitus, ADGs, and associated costs. Patient 1: Low Cost Patient with Diabetes Input Data/Patient Characteristics ACG Output Resource Consumption Variables Age/Sex: 49/Female ACG-0900: Chronic Medical, Total Cost: $296 Stable Ambulatory visits: 1 Conditions: Diabetes mellitus ADGs: 1 0 Emergency visits: 0 Chronic Medical: Stable Hospitalizations: 0 Patient 2: High Cost Patient with Diabetes Input Data/Patient Characteristics ACG Output Resource Consumption Variables Age/Sex: 49/Female ACG-3600: Acute Minor/Acute Total Cost: $1,698 Major/Likely Recur/Eye & Dental Ambulatory visits: 6 Conditions: Diabetes mellitus, ADGS: 01, 07, 08, 10, 22*, 26, and Emergency visits: 1 Disorders of Lipid Metabolism, 27 Hospitalizations: 0 Peripheral Neuropathy, Otitis Time Limited: Minor Media, Gastroesophageal Reflux, Likely to Recur: Discrete Acute sprain, Joint disorder, Likely to Recur: Discrete- Bursitis, Arthropathy Infections Chronic Medical: Stable Injuries/Adverse Effects: Major Signs/Symptoms: Minor Signs/Symptoms: Uncertain *Major ADG 31

32 Chapter 2: Diagnosis-based Markers Patient 3: Very High Cost Patient with Diabetes Input Data/Patient Characteristics ACG Output Resource Consumption Variables Age/Sex: 51/Female ACG : 10+Other ADG Total Cost: $33,073 Combinations, Age >17, 4+ Major Ambulatory visits: 23 ADGs Emergency visits: 2 Conditions: Diabetes mellitus, ADGs: 01, 02, 03*, 04*, 05, 07, Hospitalizations: 1 General medical exam, Ischemic 08, 09*, 10, 11*, 12, 16*, 17, 21, Heart Disease, Hypertension, 22*, 23, 26, 27, 28, 29, 30, 31 Disorders of Lipid Metabolism, and 34. Low Back Pain, Peripheral Time Limited: Minor Neuropathy, Cerebrovascular Time Limited: Minor- Primary Disease, COPD, Acute Lower Infections Respiratory Tract Infection, Time Limited: Major Allergic Rhinitis, Gingivitis, Otitis Time Limited: Major-Primary Media, Hearing Loss, Chest Pain, Infections, Allergies Syncope, Chronic Cystic Disease of the Breast, Tobacco Use, Likely to Recur: Discrete Abdominal Pain, Sinusitis, Sleep Likely to Recur: Discrete- Apnea, Contusions and Abrasions, Infections Headache, Cough, Fatigue Likely to Recur: Progressive Chronic Medical: Stable Chronic Medical: Unstable Chronic Specialty: Stable-Orthopedic Chronic Specialty: Unstable- Orthopedic Chronic Specialty: Unstable-Ear, Nose, Throat Injuries/Adverse Effects: Minor Injuries/Adverse Effects: Major Psychosocial: Time Limited, Minor Signs/Symptoms: Minor Signs/Symptoms: Uncertain Major, Discretionary, See/Reassure, and Prevention/Administrative *Major ADG Pregnancy Using diagnosis codes for pregnancy, the ACG System identifies all women who were pregnant during the assessment period and places them into the pregnancy MAC. ACGs are formed based on (1) total number of ADGs, (2) presence of complications (i.e., presence of a major ADG), and (3) whether the woman delivered (the default level of assignment can be overridden). Example 3: Pregnancy/Delivery with Complications shows how the ACG System groups women with a complicated pregnancy/delivery. Both women in the example had ICD-9-CM codes that map to ADG-03 (an acute major morbidity). The salient difference between the two that explains the difference in resource consumption is that Patient 2 had a greater number of ADGs and more major ADGs and thus fits 32

33 Chapter 2: Diagnosis-based Markers into a more resource intensive ACG. That is, Patient 2 had a higher level of morbidity than Patient 1, even though both women experienced a complicated pregnancy/delivery. Example 3: Pregnancy/Delivery with Complications The following patient types demonstrate the levels of pregnancy and delivery with complications, ADGs, and associated costs. Patient 1: Pregnancy/Delivery with Complications, Low Morbidity Input Data/Patient Characteristics ACG Output Resource Consumption Variables Age/Sex: 32/Female ACG-1731: 2-3 ADGs, 1+ Major Total Cost: $8,406 ADGs, Delivered Ambulatory visits: 3 Conditions: General medical ADGs: 01, 03*, 31, and 33. Emergency visits: 0 exam, Pregnancy and delivery - Time Limited: Minor Hospitalizations: 1 uncomplicated and Pregnancy and Time Limited: Major delivery - with complications. Prevention/Administrative, and Pregnancy *Major ADG Patient 2: Pregnancy/Delivery with Complications, High Morbidity Input Data/Patient Characteristics ACG Output Resource Consumption Variables Age/Sex: 36/Female ACG-1771: 6+ ADGs, 1+ Major Total Cost: $19,714 ADGs, Delivered Ambulatory visits: 13 Conditions: General medical exam ADGs: 03*, 07, 08, 10, 11*, 21, Emergency visits: 2 Hypertension, Low Back Pain, 22*, 28, 31, and 33. Hospitalizations: 1 Urinary tract infection, Renal Time Limited: Major Calculi, Cervical Pain Syndromes, Likely to Recur: Discrete Joint disorder, Pregnancy and Likely to Recur: Discretedelivery-with complications. infections Chronic Medical: Stable Chronic Medical: Unstable Injuries/Adverse Effects: Minor Injuries/Adverse Effects: Major Signs/Symptoms: Major Prevention/Administrative, and Pregnancy *Major ADG The Clinical Classification of Pregnancy/Delivery ACGs table presents an alternative clinical categorization of the pregnancy/delivery ACGs. Three dimensions are used to classify the ACGs number of ADGs, presence of major ADGs, and whether the women delivered during the assessment period. Resource consumption increases along each of the three axes: presence of delivery, presence of a major ADG, and number of ADGs. Using various combinations of these ACGs, a clinician, or manager can determine the proportion of women with complicated pregnancies and deliveries overall, and with different levels of 33

34 Chapter 2: Diagnosis-based Markers morbidity. The need for specialty services will be greatest for those women with higher levels of morbidity and complications as defined by presence of a major ADG. Clinical Classification of Pregnancy/Delivery ACGs ACG Levels Pregnancy Only Delivered Uncomplicated (No Complicated (1+ Uncomplicated (No Complicated (1+ Morbidity Level Major ADGs) Major ADGs ) Major ADGs) Major ADGs) Low (1-3 ADGs) 1712, , Mid (4-5 ADGs) High (6+ ADGs) Infants The ACG System places all infants into an infant MAC. By definition, all had at least one hospitalization (at time of delivery). ACG groups are formed based on total number of ADGs and the presence of a complication or major ADG. Example 4: Infants compares an infant in the low morbidity/no complications ACG (5310, the most frequently assigned infant ACG) to an infant in the higher morbidity/with complications ACG (5340, the most resource intensive infant ACG). Infant 1 had a typical course: hospitalization at birth, routine check-ups, and illness management for upper respiratory tract infection and otitis media. Infant 2 presents a completely different picture in terms of pattern of morbidity and resource consumption, both of which are substantially greater in comparison with Infant 1. Example 4: Infants The following patient types demonstrate the levels of infants with complications, ADGs, and associated costs. Patient 1: Infant with Low Morbidity, Normal Birthweight Input Data/Patient Characteristics ACG Output Resource Consumption Variables Age/Sex: 0/Female ACG 5312: 0-5 ADGs, No Major Total Cost: $3,208 ADGs, Normal Birthweight, Ambulatory visits: 17 Conditions: General medical exam ADGs: 02, 08, 26, and 31 Emergency visits: 0 Otitis media, Acute upper Time Limited: Minor Hospitalizations: 1 respiratory tract infection, Fungal Likely to Recur: Discreteinfection and Gastroesophageal Infections Reflux Signs/Symptoms: Minor, and Prevention/Administration 34

35 Chapter 2: Diagnosis-based Markers Patient 2: Infant with High Morbidity, Low Birthweight Input Data/Patient Characteristics ACG Output Resource Consumption Variables Age/Sex: 0/Male ACG 5341: 6+ ADGs, 1+ Major Total Cost: $165,142 ADGs, Low Birthweight Ambulatory visits: 19 Conditions: General medical ADGs: 03*, 04, 07, 10, 11*, 22, Emergency visits: 0 exam, Respiratory symptoms 26, 27, 28, and 31 Hospitalizations: 1 Congenital Heart Disease, Cardiac Time Limited: Major Arrhythmia, Septicemia, Nausea, Time Limited: Major-Primary vomiting, Gastroesophageal Infections Reflux, Neonatal Jaundice, Renal Likely to Recur: Discrete Disorders, Endocrine disorders, Chronic Medical: Stable, Chronic Vesicouretal reflux Medical: Unstable Injuries/Adverse Effects: Major Signs/Symptoms: Minor Signs/Symptoms: Uncertain Signs/Symptoms: Major, Discretionary, and Prevention/Administrative *Major ADG The Clinical Classification of Infant ACGs table provides a clinical classification of the infant ACGs. Clinical Classification of Infant ACGs Low Birthweight Normal Birthweight No Complications Complication (1+ No Complications Complication (1+ Morbidity Level (no Major ADGs) Major ADGs) (no Major ADGs) Major ADGs) Low (0-5 ADGs) Mid (6+ ADGs) Resource Utilization Bands (RUBs) ACGs were designed to represent clinically logical categories for persons expected to require similar levels of healthcare resources (i.e., resource groups). However, enrollees with similar overall utilization may be assigned different ACGs because they have different epidemiological patterns of morbidity. For example, a pregnant woman with significant morbidity, an individual with a serious psychological condition, or someone with two chronic medical conditions may all be expected to use approximately the same level of resources even though they each fall into different ACG categories. In many instances it may be useful to collapse the full set of ACGs into fewer categories, particularly where resource use similarity, and not clinical cogency, is a desired objective. ACGs are collapsed according to concurrent relative resource use in the creation of Resource Utilization Bands (RUBs). The software automatically assigns six RUB classes: 0 - No or Only Invalid Dx 1 - Healthy Users 35

36 Chapter 2: Diagnosis-based Markers 2 - Low 3 - Moderate 4 - High 5 - Very High The relationship between ACG categories and RUBs is defined in the following table Final ACG Categories, Reference ACG Concurrent Risks, and RUBs Non-Elderly (0 to Elderly (65 Years ACG Description 64 Years) and Older) RUB 0100 Acute Minor, Age N/A Acute Minor, Age 2 to N/A Acute Minor, Age > Acute Major Likely to Recur, w/o Allergies Likely to Recur, with Allergies Asthma Chronic Medical, Unstable Chronic Medical, Stable Chronic Specialty, Stable Eye/Dental Chronic Specialty, Unstable Psychosocial, w/o Psych Unstable Psychosocial, with Psych Unstable, w/o Psych Stable 1500 Psychosocial, with Psych Unstable, w/ Psych Stable 1600 Preventive/Administrative * Pregnancy: 0-1 ADGs N/A Pregnancy: 0-1 ADGs, delivered N/A Pregnancy: 0-1 ADGs, not delivered N/A * Pregnancy: 2-3 ADGs, no Major ADGs N/A Pregnancy: 2-3 ADGs, no Major ADGs, delivered N/A Pregnancy: 2-3 ADGs, no Major ADGs, not N/A 3 delivered 1730* Pregnancy: 2-3 ADGs, 1+ Major ADGs N/A 4 36

37 Chapter 2: Diagnosis-based Markers Non-Elderly (0 to Elderly (65 Years ACG Description 64 Years) and Older) RUB 1731 Pregnancy: 2-3 ADGs, 1+ Major ADGs, delivered N/A Pregnancy: 2-3 ADGs, 1+ Major ADGs, not N/A 3 delivered 1740* Pregnancy: 4-5 ADGs, no Major ADGs N/A Pregnancy: 4-5 ADGs, no Major ADGs, delivered N/A Pregnancy: 4-5 ADGs, no Major ADGs, not N/A 3 delivered 1750* Pregnancy: 4-5 ADGs, 1+ Major ADGs N/A Pregnancy: 4-5 ADGs, 1+ Major ADGs, delivered N/A Pregnancy: 4-5 ADGs, 1+ Major ADGs, not N/A 3 delivered 760* Pregnancy: 6+ ADGs, no Major ADGs N/A Pregnancy: 6+ ADGs, no Major ADGs, delivered N/A Pregnancy: 6+ ADGs, no Major ADGs, not N/A 3 delivered 1770* Pregnancy: 6+ ADGs, 1+ Major ADGs N/A Pregnancy: 6+ ADGs, 1+ Major ADGs, delivered N/A Pregnancy: 6+ ADGs, 1+ Major ADGs, not N/A 4 delivered 1800 Acute Minor and Acute Major Acute Minor and Likely to Recur, Age N/A Acute Minor and Likely to Recur, Age 2 to N/A Acute Minor and Likely to Recur, Age > 5, w/o Allergy 2200 Acute Minor and Likely to Recur, Age > 5, with Allergy 2300 Acute Minor and Chronic Medical: Stable Acute Minor and Eye/Dental Acute Minor and Psychosocial, w/o Psych Unstable 2600 Acute Minor and Psychosocial, with Psych Unstable, w/o Psych Stable 2700 Acute Minor and Psychosocial, with Psych Unstable and Psych Stable 37

38 Chapter 2: Diagnosis-based Markers Non-Elderly (0 to Elderly (65 Years ACG Description 64 Years) and Older) RUB 2800 Acute Minor and Likely to Recur Acute Minor/Acute Major/Likely to Recur, Age N/A Acute Minor/Acute Major/Likely to Recur, Age N/A 3 to Acute Minor/Acute Major/Likely to Recur, Age N/A 3 to Acute Minor/Acute Major/Likely to Recur, Age > , w/o Allergy 3300 Acute Minor/Acute Major/Likely to Recur, Age > , with Allergy 3400 Acute Minor/Likely to Recur/Eye & Dental Acute Minor/Likely to Recur/Psychosocial Acute Minor/Acute Major/Likely Recur/Eye & Dental 3700 Acute Minor/Acute Major/Likely Recur/Psychosocial Other ADG Combinations, Age < N/A Other ADG Combinations, Males Age 18 to N/A Other ADG Combinations, Females Age 18 to N/A Other ADG Combinations, Age > Other ADG Combinations, Age < 18, no N/A 3 Major ADGs Other ADG Combinations, Age < 18, N/A 3 Major ADGs Other ADG Combinations, Age 18 to 44, no N/A 3 Major ADGs Other ADG Combinations, Age 18 to 44, N/A 3 Major ADGs Other ADG Combinations, Age 18 to 44, N/A 4 Major ADGs Other ADG Combinations, Age > 44, no Major ADGs Other ADG Combinations, Age > 44, Major ADGs 38

39 Chapter 2: Diagnosis-based Markers Non-Elderly (0 to Elderly (65 Years ACG Description 64 Years) and Older) RUB Other ADG Combinations, Age > 44, Major ADGs Other ADG Combinations, Age < 6, no Major N/A 3 ADGs Other ADG Combinations, Age < 6, 1+ Major N/A 4 ADGs Other ADG Combinations, Age 6 to 17, no N/A 3 Major ADGs Other ADG Combinations, Age 6 to 17, N/A 4 Major ADGs Other ADG Combinations, Males, Age 18 to N/A 3 34, no Major ADGs Other ADG Combinations, Males, Age 18 to N/A 3 34, 1+ Major ADGs Other ADG Combinations, Males, Age 18 to N/A 4 34, 2+ Major ADGs Other ADG Combinations, Females, Age N/A 3 to 34, no Major ADGs Other ADG Combinations, Females, Age N/A 3 to 34, 1+ Major ADGs Other ADG Combinations, Females, Age N/A 4 to 34, 2+ Major ADGs Other ADG Combinations, Age > 34, Major ADGs Other ADG Combinations, Age > 34, 2 Major ADGs Other ADG Combinations, Age > 34, 3 Major ADGs Other ADG Combinations, Age > 34, Major ADGs Other ADG Combinations, Age 1 to 17, no N/A 3 Major ADGs Other ADG Combinations, Age 1 to 17, N/A 4 Major ADGs Other ADG Combinations, Age 1 to 17, N/A 5 Major ADGs Other ADG Combinations, Age > 17, Major ADGs 39

40 Chapter 2: Diagnosis-based Markers Non-Elderly (0 to Elderly (65 Years ACG Description 64 Years) and Older) RUB Other ADG Combinations, Age > 17, Major ADGs Other ADG Combinations, Age > 17, Major ADGs Other ADG Combinations, Age > 17, Major ADGs 5110 No Diagnosis or Only Unclassified Diagnosis ( input files) 5200 Non-Users (2 input files) * Infants: 0-5 ADGs, no Major ADGs N/A Infants: 0-5 ADGs, no Major ADGs, low birth N/A 4 weight 5312 Infants: 0-5 ADGs, no Major ADGs, normal birth N/A 3 weight 5320* Infants: 0-5 ADGs, 1+ Major ADGs N/A Infants: 0-5 ADGs, 1+ Major ADGs, low birth N/A 5 weight 5322 Infants: 0-5 ADGs, 1+ Major ADGs, normal birth N/A 4 weight 5330* Infants: 6+ ADGs, no Major ADGs N/A Infants: 6+ ADGs, no Major ADGs, low birth N/A 4 weight 5332 Infants: 6+ ADGs, no Major ADGs, normal birth N/A 3 weight 5340* Infants: 6+ ADGs, 1+ Major ADGs N/A Infants: 6+ ADGs, 1+ Major ADGs, low birth N/A 5 weight 5342 Infants: 6+ ADGs, 1+ Major ADGs, normal birth N/A 4 weight 9900 Invalid Age or Date of Birth Source: PharMetrics, Inc., a unit of IMS, Watertown, MA; national cross-section of managed care plans; population of 3,310,540 commercially insured lives (less than 65 years old) and population of 501,987 Medicare beneficiaries (65 years and older), Note: The default is to subdivide these groups* on delivery or low birth weight status. Grouping the ACGs without these divisions is optional and must be turned on in order to be used. 40

41 Chapter 2: Diagnosis-based Markers Disease Markers - Expanded Diagnosis Clusters (EDCs) The Johns Hopkins Expanded Diagnosis Clusters (EDCs) complement the unique person-oriented approach that underpins the ACG System. EDCs are a tool for easily identifying people with specific diseases or symptoms, without having to create your own algorithms. The EDC methodology assigns diagnosis codes found in claims or encounter data to one of 282 EDCs, which are further organized into 27 categories called Major Expanded Diagnosis Clusters (MEDCs ). As broad groupings of diagnosis codes, EDCs help to remove differences in coding behavior between practitioners. As a stand-alone tool, EDCs can be used to select patients with a specific condition or combination of conditions and can also be used to compare the distribution of conditions in one population with another. When combined with ACGs, the result is a powerful combination tool for demonstrating variability of cost within disease categories This is useful for many profiling applications and can help to target individuals for case-management purposes. The Development of EDCs EDCs build on the methodology developed by Ronald Schneeweiss and colleagues in In Schneeweiss original work, diagnosis codes were classified according to clinical criteria. Schneeweiss original 92 categories, called Diagnosis Clusters, underwent considerable updating and extensive expansions based on significant additional development at Johns Hopkins University, led by Dr. Christopher Forrest. In 2000, Dr. Forrest s team of generalist and specialist physicians used their clinical judgment to assign the most commonly used ICD-9 codes to EDCs. In all, the team assigned approximately 9,400 diagnosis codes to 190 EDCs. The development team reviewed the Clinical Classification for Health Policy Research produced by the Agency for Health Care Policy and Research 4, another disease-oriented grouping methodology, to identify additional conditions that were excluded from Schneeweiss original taxonomy. Other modifications made to the original Schneeweiss clusters included deletion of some clusters, expansion of the content of other clusters by updating the range of ICD codes, improving the clinical homogeneity of certain clusters, and splitting existing clusters into new categories. In addition, each of the EDCs was classified into one of 27 broad clinical categories, termed a Major EDC (MEDC). For example, three EDCs allergic reactions, asthma, and allergic rhinitis all fall within the Allergy MEDC classification. The original assignments of EDCs to MEDCs were made based on the assessment of our clinician panel as to the physician specialty most likely to provide care for the class of conditions, when other than primary care services were received. For example, Prostatic hypertrophy (GUR04) was mapped to the genito-urinary MEDC and gout (RHU02) to the rheumatologic MEDC. Further review of data and actual specialist use within each EDC category led to changes in the final assignment of EDCs to their respective MEDC. It should be noted that for many EDCs, the most common service provider is the primary care clinician, and not a specialist. 3 Schneeweiss R, Rosenblatt RA, Cherkin DC, Kirkwood R, Hart G. Diagnosis Clusters: Tool for Analyzing the Content of Ambulatory Medical Care. Med Care 1983; 21: Elixhauser A, Andrews RM, Fox S. (1993) Clinical classifications for health policy research: Discharge statistics by principal diagnosis and procedure (AHCPR Publication No ). 41

42 Chapter 2: Diagnosis-based Markers Since the initial development of the EDC typology in 2000, the EDC categories have undergone several revisions and expansions so that now all ICD (-9-CM, 10- and -10-CM version) codes are assigned. There are 282 EDCs in the current version of the ACG System. Understanding How EDCs Work Each diagnosis code maps to one or more EDCs. Diagnosis codes within an EDC share similar clinical characteristics and are expected to evoke similar types of diagnostic and therapeutic responses. Where a single diagnosis code indicates more than one underlying condition, more than one EDC may be assigned. For example, in ICD-10 the code E11.21 (Type 2 diabetes mellitus with diabetic nephropathy) would trigger both END07 (Type 2 diabetes, w/ complication) and REN04 (Nephritis, nephrosis). There are 65 ICD-9-CM and 41 ICD-10 codes (shown in the following tables) that practitioners can record as a diagnosis for otitis media. The EDC for otitis media combines these codes into a single rubric, which lessens the impact of physicians coding styles on your analyses related to this condition. ICD-9-CM Codes Assigned to Otitis Media EDC (EAR01) ICD-9-CM Codes Description Postmeasles otitis media 381 Nonsuppurative otitis media and Eustachian tube disorders Acute nonsuppurative otitis media Unspecified acute nonsuppurative otitis media Acute serous otitis media Acute mucoid otitis media Acute sanguinous otitis media Acute allergic serous otitis media Acute allergic mucoid otitis media Acute allergic sanguinous otitis media Chronic serous otitis media Simple or unspecified chronic serous otitis media Other chronic serous otitis media Chronic mucoid otitis media Simple or unspecified chronic mucoid otitis media Other chronic mucoid otitis media Other and unspecified chronic nonsuppurative otitis media Nonsuppurative otitis media, not specified as acute or chronic Eustachian salpingitis Unspecified Eustachian salpingitis 42

43 Chapter 2: Diagnosis-based Markers ICD-9-CM Codes Description Acute Eustachian salpingitis Chronic Eustachian salpingitis Obstruction of Eustachian tube Unspecified obstruction of Eustachian tube Osseous obstruction of Eustachian tube Intrinsic cartilagenous obstruction of Eustachian tube Extrinsic cartilagenous obstruction of Eustachian tube Patulous Eustachian tube Other disorders of Eustachian tube Dysfunction of Eustachian tube Other disorders of Eustachian tube Unspecified Eustachian tube disorder 382 Suppurative and unspecified otitis media Acute suppurative otitis media Acute suppurative otitis media without spontaneous rupture of eardrum Acute suppurative otitis media with spontaneous rupture of eardrum Acute suppurative otitis media in diseases classified elsewhere Chronic tubotympanic suppurative otitis media Chronic atticoantral suppurative otitis media Unspecified chronic suppurative otitis media Unspecified suppurative otitis media Unspecified otitis media 384 Other disorders of tympanic membrane Acute myringitis without mention of otitis media Unspecified acute myringitis Bullous myringitis Other acute myringitis without mention of otitis media Chronic myringitis without mention of otitis media Perforation of tympanic membrane Unspecified perforation of tympanic membrane Central perforation of tympanic membrane 43

44 Chapter 2: Diagnosis-based Markers ICD-9-CM Codes Description Attic perforation of tympanic membrane Other marginal perforation of tympanic membrane Multiple perforations of tympanic membrane Total perforation of tympanic membrane Other specified disorders of tympanic membrane Atrophic flaccid tympanic membrane Atrophic nonflaccid tympanic membrane Unspecified disorder of tympanic membrane Otorrhea Unspecified otorrhea Other otorrhea Otalgia Unspecified otalgia Otogenic pain Referred otogenic pain Unspecified disorder of ear ICD-10 Codes Assigned to Otitis Media EDC (EAR01) ICD-10 Codes Description B05.3 Postmeasles otitis media H65 Nonsuppurative otitis media and Eustachian tube disorders H65.0 Acute serous otitis media H65.1 Other acute nonsuppurative otitis media H65.2 Chronic serous otitis media H65.3 Chronic mucoid otitis media H65.4 Other and unspecified chronic nonsuppurative otitis media H65.9 Nonsuppurative otitis media, not specified as acute or chronic H66 Suppurative and unspecified otitis media H66.0 Acute suppurative otitis media H66.1 Chronic tubotympanic suppurative otitis media H66.2 Chronic atticoantral suppurative otitis media H66.3 Other chronic suppurative otitis media 44

45 Chapter 2: Diagnosis-based Markers ICD-10 Codes Description H66.4 Suppurative otitis media, unspecified H66.9 Otitis media, unspecified H67 Otitis media in diseases classified elsewhere H67.0 Otitis media in bacterial diseases classified elsewhere H67.1 Otitis media in viral diseases classified elsewhere H67.8 Otitis media in other diseases classified elsewhere H68 Eustachian salpingitis and obstruction H68.0 Eustachian salpingitis H68.1 Obstruction of Eustachian tube H69 Other disorders of Eustachian tube H69.0 Patulous Eustachian tube H69.8 Other disorders of Eustachian tube H69.9 Unspecified Eustachian tube disorder H72 Perforation of tympanic membrane H72.0 Central perforation of tympanic membrane H72.1 Attic perforation of tympanic membrane H72.2 Other marginal perforation of tympanic membrane H72.8 Total or Multiple perforations of tympanic membrane H72.9 Unspecified perforation of tympanic membrane H73 H73.0 Acute myringitis H73.1 Chronic myringitis Other disorders of tympanic membrane H73.8 Other specified disorders of tympanic membrane H73.9 Unspecified disorder of tympanic membrane H92 H92.0 Otalgia H92.1 Otorrhea H92.2 Otorrhagia Otalgia and effusion of ear The EDCs related to Diabetes receive special treatment in the EDC assignment process. Each Diabetes diagnosis is associated with either EDC END06 Type 2 Diabetes without complication or EDC END08 Type 1 Diabetes without complication. The EDC assignment process then looks for the presence of a potential complication diagnosis code. If a complication is found (see the Examples of Diabetes Complications table), 45

46 Chapter 2: Diagnosis-based Markers a patient with END06 will be assigned to END07 while a patient with END08 will be assigned to END09. The implication of this method is that a patient cannot be assigned to a diabetes condition with and without complications simultaneously. Examples of Diabetes Complications Patients will not be assigned to a Type 1 Diabetes EDC and a Type 2 Diabetes EDC concurrently. If the patient has diagnoses indicating both Type 1 Diabetes and Type 2 Diabetes, the EDC assignment will reflect Type 1 Diabetes only. ICD-9-CM Description Diabetes with ketoacidosis Diabetes with hyperosmolar coma Diabetes with coma NEC Diabetes with renal manifestation 410 Acute myocardial infarction 411 Other acute ischemic heart disease 412 Old myocardial infarction 413 Angina pectoris 414 Other chronic ischemic heart disease 581 Nephrotic syndrome 582 Chronic nephritis 583 Nephritis NOS 584 Acute renal failure 585 Chronic renal failure 586 Renal failure NOS V56 ICD-10 E10(.0-.8) E11(.0-.8) E12(.0-.8) E13(.0-.8) E14(.0-.8) I20 I21 I22 Dialysis encounter Description Insulin-dependent diabetes mellitus with complications Non-insulin-dependent diabetes mellitus with complications Malnutrition-related diabetes mellitus with complications Other specified diabetes mellitus with complications Unspecified diabetes mellitus with complications Angina Pectoris Acute Myocardial Infarction Subsequent Myocardial Infarction 46

47 Chapter 2: Diagnosis-based Markers ICD-10 I23 I24 I25 I51 N00 N01 N02 N03 N04 N05 N17 N18 N19 N28 Z49 Description Complications following acute myocardial infarction Other acute ischaemic heart diseases Chronic ischaemic heart disease Complications of heart disease Acute nephritic syndrome Rapidly progressive nephritic syndrome Recurrent and persistent haematuria Chronic nephritic syndrome Nephrotic syndrome Unspecified nephritic syndrome Acute renal failure Chronic renal failure Unspecified renal failure Other disorders of kidney Care involving dialysis Diagnostic Certainty The recording of provisional diagnosis codes can improperly attribute chronic conditions to a patient subsequently misidentifying individuals for care management programs and elevating cost predictions. The goal of stringent diagnostic certainty is to provide greater certainty of a given diagnosis for subset of chronic conditions. When the user applies the lenient diagnostic certainty option, any single diagnosis qualifying for an EDC marker will turn the marker on. However, the user may also apply a stringent diagnostic certainty option. For a subset of chronic diagnoses, there must be more than one diagnosis qualifying for the marker in order for the EDC to be assigned. This was designed to provide greater confidence in the EDC conditions assigned to a patient. For several conditions, namely Hypertension, Asthma, Diabetes, and Pregnancy, there are multiple EDCs to describe the condition. When applying stringent diagnostic certainty, if multiple diagnoses are required, they must qualify for the same EDC. For example, a single diagnosis for Type 1 Diabetes and a single diagnosis for Type 2 Diabetes will not turn on any diabetes-related EDC, but 2 diagnoses for Type 2 Diabetes will turn on EDC END06. For more information about the use of the diagnostic certainty options, refer to Chapter 4 of the Installation and Usage Guide. MEDC Types To assist analysts, the 27 Major EDCs may be further aggregated into five MEDC Types. Specifically, the categorization is presented in the following table. MEDC Type MEDCs 1. Administrative Administrative 47

48 Chapter 2: Diagnosis-based Markers MEDC Type MEDCs 2. Medical Allergy, Cardiovascular, Endocrine, Gastrointestinal/Hepatic, General Signs and Symptoms, Genetic, Hematologic, Infections, Malignancies, Neonatal, Neurologic, Nutrition, Renal, Respiratory, Rheumatologic, Skin, Toxic Effects 3. Surgical Dental, ENT, Eye, General Surgery, Genito-urinary, Musculoskeletal, Reconstructive 4. Obstetric/Gynecologic Female reproductive 5. Psychosocial Psychosocial In summary, EDCs can be examined at three levels. The broadest level is the MEDC -Type and includes five categories. MEDC -Types are based on 27 clinically oriented groupings of MEDCs. The full EDC taxonomy is composed of 286 groups. High, Moderate, and Low Impact EDCs Within the detailed EDC output from the system as well as in the Comprehensive Patient Clinical Profile Report, the EDCs associated with an individual are categorized as High, Moderate, or Low impact. The definition for these categories is based on the individual contribution of the condition to the predictive modeling score. The specific criteria are as follows: High Impact EDC: Coefficient >1 in the diagnosis derived predictive model without prior cost predicting total cost Moderate Impact EDC: Coefficient between 0.1 and 1.0 in the diagnosis derived predictive model without prior cost predicting total cost Low Impact EDC: Coefficient less than 0.1 or not included in the diagnosis derived predictive model without prior cost predicting total cost Important Differences between EDCs and ADGs Both EDCs and Aggregated Diagnosis Groups (ADGs) are aggregations of diagnosis codes. However, there is a significant difference in the methodology underlying the grouping of diagnosis codes: ADGs are groups of diagnoses with similar expected healthcare need, while EDCs are clinically similar clusters. The main criterion used for the diagnosis-to-edc assignment is diagnostic similarity, whereas the diagnosis-to-adg assignments are based on a unique set of clinical criteria that captures various dimensions of the range and severity of an individual s co-morbidity. This subject is discussed in more detail in Patterns of Morbidity Adjusted Clinical Groups (ACGs) on page 17. In brief, the key dimensions of the diagnosis to ADG assignment process are: Expected duration of illness (e.g., acute, chronic, or recurrent) Severity (e.g., expected prognosis with respect to disability or longevity) Diagnostic certainty (i.e., signs/symptoms versus diagnoses) Etiology (e.g., infectious, neoplastic, psychosocial conditions) Expected need for specialist care or hospitalization In the ADG assignment process, diagnosis codes representing multiple diseases and conditions may be assigned to the same ADG. This occurs when the diseases are expected to have a similar impact on the need for healthcare resources. In contrast, the EDC assignment does not account for differences in disease severity, chronicity, or the expected need for resources. 48

49 Chapter 2: Diagnosis-based Markers Because of the marked differences between the conceptual underpinnings of EDCs and the ADG framework, each ADG will usually contain diagnosis codes that map into more than one EDC. Likewise, diagnosis codes contained within a single EDC may map into multiple ADGs. Applications of EDCs EDCs have many applications, particularly in areas of profiling and disease/case management. EDCs can be used to: 1. Describe the prevalence of specific diseases within a single population; 2. Compare disease distributions across two or more populations; and, 3. Aid disease management/case management processes by identifying individual patients by condition and displaying a patient condition profile. Each of these applications is highlighted within the Applications Guide chapters for Population Health Monitoring and Clinical Screening. Special Population Markers The ACG System includes a number of special population markers to identify patient populations requiring specialized care. These markers enhance the clinical screening process by providing meaningful filtering criteria and clinical context. Several of the markers are also independent variables in the ACG risk models. Chronic Condition Count The ACG System includes a chronic condition count as an aggregate marker of case complexity. A chronic condition is an alteration in the structures or functions of the body that is likely to last longer than twelve months and is likely to have a negative impact on health or functional status. The ACG System defines a limited set of Expanded Diagnosis Clusters (EDCs) that represent high impact and chronic conditions likely to last more than 12 months with or without medical treatment (see the following table). From this list of EDCs, individual diagnosis codes were tested against the criteria for chronic conditions stated above. The diagnoses identified by these EDCs were compared against a chronic condition list provided by the Center for Child and Adolescent Health Policy, Mass General Hospital for Children, in Boston, Massachusetts. Differences between the lists were identified as psychological or medical/surgical conditions. The psychological codes were reviewed by a practicing psychologist and health services researcher for congruence with the operational definition. Several anxiety and substance abuse codes were deleted from the chronic condition list as a result of this review. Several acute conditions of the retina were deleted from the chronic condition list after a similar review of medical/surgical conditions. The Center for Child and Adolescent Health Policy list and the ACG System differ in definitions related to infectious diseases such as tuberculosis, peptic ulcer disease, congenital heart disease (which is generally resolved through surgical interventions at birth), gastrointestinal obstructions and perforations (likely to be acute and treatable conditions), osteomyelitis, and prematurity. These conditions are not considered chronic conditions in the ACG System chronic condition marker. The chronic diagnosis codes that were identified were further classified and aggregated into EDC categories. Not all diagnosis codes that map to these EDC categories are considered chronic. The chronic diagnosis codes trigger a chronic EDC flag (see the following table). The number of unique EDC flags found for a patient represents the final chronic condition count. 49

50 Chapter 2: Diagnosis-based Markers EDCs considered in the Chronic Condition Count Marker EDC Description Acute hepatitis Acute leukemia Acute lower respiratory tract infection Acute myocardial infarction Acute renal failure Acute sprains and strains Adjustment disorder Administrative concerns and non-specific laboratory abnormalities Adverse events from medical/surgical procedures Age-related macular degeneration Anxiety, neuroses Aplastic anemia Arthropathy Asthma, w/o status asthmaticus Asthma, with status asthmaticus Attention deficit disorder Autism Spectrum Disorder Autoimmune and connective tissue diseases Benign and unspecified neoplasm Bipolar disorder Blindness Cardiac arrhythmia Cardiac valve disorders Cardiomyopathy Cardiovascular disorders, other Cardiovascular signs and symptoms Cataract, aphakia Central nervous system infections Cerebral palsy Cerebrovascular disease EDC Flag GAS04 MAL16 RES02 CAR12 REN03 MUS02 PSY13 ADM05 TOX03 EYE15 PSY01 HEM05 RHU03 ASTH ASTH PSY05 NUR26 RHU01 GSU03 PSY12 EYE02 CAR09 CAR06 CAR07 CAR16 CAR01 EYE06 NUR20 NUR18 NUR05 50

51 Chapter 2: Diagnosis-based Markers EDC Description Chromosomal anomalies Chronic cystic disease of the breast Chronic liver disease Chronic pancreatitis Chronic renal failure Chronic respiratory failure Chronic ulcer of the skin Cleft lip and palate Congenital anomalies of limbs, hands, and feet Congenital heart disease Congestive heart failure Cystic fibrosis Deafness, hearing loss Deep vein thrombosis Degenerative joint disease Dementia Delirium Depression Developmental disorder Diabetic retinopathy Disorders of lipid metabolism Disorders of Newborn Period Disorders of the immune system Eating disorder Emphysema, chronic bronchitis, COPD Endometriosis ESRD Eye, other disorders Failure to thrive Fluid/electrolyte disturbances Gastrointestinal signs and symptoms EDC Flag GTC01 GSU06 GAS05 GAS12 REN01 RES13 REC03 REC01 MUS11 CAR04 CAR05 RES03 EAR08 HEM06 MUS03 NUR24 NUR25 PSY09 NUR19 EYE13 CAR11 NEW05 ALL06 PSY15 RES04 FRE03 REN06 EYE14 NUT01 REN02 GAS01 51

52 Chapter 2: Diagnosis-based Markers EDC Description Gastrointestinal/Hepatic disorders, other Generalized atherosclerosis Genito-urinary disorders, other Glaucoma Gout Hematologic disorders, other Hemophilia, coagulation disorder High impact malignant neoplasms HIV, AIDS Hypertension, w/o major complications Hypertension, with major complications Hypothyroidism Impulse control Inflammatory bowel disease Inherited metabolic disorders Irritable bowel syndrome Ischemic heart disease (excluding acute myocardial infarction) Kyphoscoliosis Lactose intolerance Low back pain Low impact malignant neoplasms Malignant neoplasms of the skin Malignant neoplasms, bladder Malignant neoplasms, breast Malignant neoplasms, cervix, uterus Malignant neoplasms, colorectal Malignant neoplasms, esophagus Malignant neoplasms, kidney Malignant neoplasms, liver and biliary tract Malignant neoplasms, lung Malignant neoplasms, lymphomas EDC Flag GAS14 CAR10 GUR12 EYE08 RHU02 HEM08 HEM07 MAL03 INF04 HYPT HYPT END04 PSY16 GAS02 GTC02 GAS09 CAR03 MUS06 GAS13 MUS14 MAL02 MAL01 MAL18 MAL04 MAL05 MAL12 MAL07 MAL08 MAL09 MAL10 MAL11 52

53 Chapter 2: Diagnosis-based Markers EDC Description Malignant neoplasms, ovary Malignant neoplasms, pancreas Malignant neoplasms, prostate Malignant neoplasms, stomach Migraines Multiple sclerosis Muscular dystrophy Musculoskeletal disorders, other Nephritis, nephrosis Neurologic disorders, other Neurologic signs and symptoms Newborn Status, Complicated Obesity Organic brain syndrome Osteoporosis Other endocrine disorders Other hemolytic anemias Other skin disorders Paralytic syndromes, other Parkinson's disease Peripheral neuropathy, neuritis Peripheral vascular disease Personality disorders Prostatic hypertrophy Psychological disorders of childhood Psychosexual Psych-physiologic and somatoform disorders Pulmonary embolism Quadriplegia and paraplegia Renal disorders, other Respiratory disorders, other EDC Flag MAL06 MAL13 MAL14 MAL15 NUR22 NUR08 NUR09 MUS17 REN04 NUR21 NUR01 NEW02 NUT03 NUR23 END02 END05 HEM01 SKN17 NUR17 NUR06 NUR03 GSU11 PSY08 GUR04 PSY14 PSY18 PSY17 RES08 NUR12 REN05 RES11 53

54 Chapter 2: Diagnosis-based Markers EDC Description Retinal disorders (excluding diabetic retinopathy) Rheumatoid arthritis Schizophrenia and affective psychosis Seizure disorder Short stature Sleep apnea Sickle cell disease Skin keratoses Spinal cord injury/disorders Strabismus, amblyopia Substance use Thrombophlebitis Tracheostomy Transplant status Type 1 diabetes Type 2 diabetes Vesicoureteral reflux EDC Flag EYE03 RHU05 PSY07 NUR07 END03 RES06 HEM09 SKN10 NUR16 EYE11 PSY02 HEM03 RES09 ADM03 DIAB DIAB GUR01 Distribution of the Chronic Condition Count Marker The chronic condition count represents the number of unique chronic EDC Flags present in the individual s diagnosis history. When the stringent diagnostic certainty option is selected, more than one diagnosis from the chronic condition list may be required for the marker to be turned on. The following table provides a representative distribution of the chronic condition count for elderly and non-elderly populations. Chronic Condition Count Non-Elderly (%) Elderly (%)

55 Chapter 2: Diagnosis-based Markers Chronic Condition Count Non-Elderly (%) Elderly (%) Source: PharMetrics, Inc., a unit of IMS, Watertown, MA; national cross-section of managed care plans; population of 3,310,540 commercially insured lives (less than 65 years old) and population of 501,987 Medicare beneficiaries (65 years and older), Hospital Dominant Morbidity Types Hospital dominant morbidity types are based on diagnoses that, when present, are associated with a markedly greater probability of hospitalization among affected patients in the next year. All these diagnoses are setting-neutral, i.e., they can be given in any inpatient or outpatient face-to-face encounter with a health professional. The variable is a count of the number of morbidity types (i.e., ADGs) with at least one hospital dominant diagnosis. When the stringent diagnostic certainty option is selected, more than one diagnosis from the hospital dominant condition list may be required for the marker to be turned on. The following table provides examples of diagnostic codes that have been identified as hospital dominant. Examples of Diagnostic Codes Included in the Hospital Dominant Morbidity Types ICD-9-CM Code ICD-10 Code Description C34.1 Malignant Neoplasm, Upper Lobe, Bronchus or Lung 261 E41 Nutritional Marasmus 2638 E43 Other Protein Calorie Malnutrition Nec D61.9 Other specified Aplastic Anemia 2894 D73.1 Hypersplenism F10.3 Alcohol Withdrawal F31.6 Bipolar affective disorder, mixed episode 0380 A40 Streptococcal Septicemia 4150 I26.0 Acute cor pulmonale 4821 J15.1 Pseudomonal Pneumonia J44.1 Obstructive Chronic Bronchitis with acute exacerbation J96 Acute Respiratory Failure 5722 K72.9 Hepatic coma N17.9 Acute Renal Failure, unspecified A48.0 Gangrene 7895 R18 Ascites 55

56 Chapter 2: Diagnosis-based Markers The risk of hospitalization, total healthcare costs, and medication costs rise dramatically in association with the number of hospital dominant morbidity types (see the following table). The count of hospital dominant morbidity types is a powerful predictor of greater resource use next year. Commercially insured individuals who have a single hospital dominant morbidity type have a five-fold greater chance of having at least one hospitalization next year, compared with individuals without such a condition. There is a 10- fold increase for individuals in the group with two or more hospital dominant morbidity types. Effects of Hospital Dominant Morbidity Types During a Baseline Year on Next Year s Hospitalization Risk, Total Healthcare Costs, and Pharmacy Costs HOSPITAL DOMINANT MORBIDITY TYPES Next Year s Outcomes Mean total healthcare Baseline Year Risk Factors % with 1+ hospitalization costs Mean pharmacy costs Commercially Insured Individuals (<65 years-old) Number of hospital dominant morbidity types None 2.5 $3,401 $ $17,634 $4, $40,457 $5, $79,483 $6,459 Medicare Beneficiaries (65 years and older) Number of hospital dominant morbidity types None 11.4 $10,152 $1, $25,901 $3, $42,964 $4, $67,102 $5,186 Source: PharMetrics, Inc., a unit of IMS, Watertown, MA; national cross-section of managed care plans; population of 3,310,540 commercially insured lives (less than 65 years old) and population of 501,987 Medicare beneficiaries (65 years and older), Frailty Conditions The Frailty Flag is a dichotomous (on/off) variable that indicates whether an enrollee over the age of 18 has a diagnosis falling within any one of 10 clusters that represent medical problems associated with frailty. Further, assignments associated with each frailty concept are also available in the output. When the stringent diagnostic certainty option is selected, more than one diagnosis from the frailty condition list may be required for the marker to be turned on. Examples of these problems are shown in the following table. Presence of any one of these diagnoses turns on the FRAILTY indicator (yes/no) variable. Among commercially insured populations, less than one percent of individuals have at least one of these frailty diagnoses, whereas among Medicare beneficiaries the proportion is greater than seven percent. 56

57 Chapter 2: Diagnosis-based Markers Medically Frail Condition Marker Frailty Concepts and Diagnoses Frailty Concept Malnutrition and/or Catabolic Illness (MAL) Dementia (DEM) Severe Vision Impairment (VIS) Decubitus Ulcer (DEC) Major Problems of Urine Retention or Control (URC) Loss of Weight (WEI) Absence of Fecal Control (AFC) Social Support Needs (SSN) Difficulty in Walking (WLK) Fall (FAL) Diagnoses (Examples) Nutritional Marasmus Other severe protein-calorie malnutrition Senile dementia with delusional or depressive features Senile dementia with delirium Profound impairment, both eyes Moderate or severe impairment, better eye/lesser eye: profound Decubitus Ulcer Incontinence without sensory awareness Continuous leakage Abnormal loss of weight and underweight Feeding difficulties and mismanagement Incontinence of feces Lack Of Housing Inadequate Housing Inadequate material resources Difficulty in walking Abnormality of gait Fall On Stairs Or Steps Fall From Wheelchair The risk of hospitalization, total healthcare costs, and medication costs rise in association with a frailty condition (see the following table). Effects of Frailty Conditions During a Baseline Year on Next Year s Hospitalization Risk, Total Healthcare Costs, and Pharmacy Costs FRAILTY CONDITIONS Next Year s Outcomes Mean total healthcare Baseline Year Risk Factors % with 1+ hospitalization costs Mean pharmacy costs Medicare Beneficiaries (65 years and older) Number of frailty conditions: None 11.9 $11,054 $1, $21,260 $3,009 Commercially insured lives (less than 65 years old) 57

58 Chapter 2: Diagnosis-based Markers FRAILTY CONDITIONS Next Year s Outcomes Mean total healthcare Baseline Year Risk Factors % with 1+ hospitalization costs Mean pharmacy costs Number of frailty conditions: None 2.7 $3,724 $ $16,200 $2,812 Source: PharMetrics, Inc., a unit of IMS, Watertown, MA; national cross-section of managed care plans; population of 3,310,540 commercially insured lives (less than 65 years old) and population of 501,987 Medicare beneficiaries (65 years and older), Compassionate Care Allowances (CAL-SSA) The intent of the Compassionate Care Allowances is to identify persons who have an increased risk of hospitalization. This marker complements other ACG System markers that identify vulnerable groups such as Frailty and Hospital Dominant Morbidity Types. The Compassionate Care Allowances marker was created by the U.S. Social Security Administration to identify types of impairment that invariably meet disability standards 5 and are, thereby, sufficient to allow an individual to become immediately eligible for benefits. The conditions that qualify for Compassionate Allowances include many cancers, amyotrophic lateral sclerosis (ALS), some types of muscular dystrophy and muscular atrophy, early-onset Alzheimer's disease, and a few other illnesses. The presence of an eligible diagnostic code triggers this dichotomous (on/off) marker. Pregnant Pregnancy status is a key differentiator for the ACG decision tree. The ACG Software will use the diagnoses provided in the medical services file to identify pregnancy as follows: ICD-9-CM: 640.xx-677.xx, V22.xx, V23.xx, V24.xx, V27.xx, V28.xx, V91.xx ICD-10: O00.x-O99.x, Z33.x-Z37.x, Z39.x It is possible for analysts to provide the software with a flag indicating that a woman is pregnant. The rationale for including this option is that, in some plans, it is not uncommon for the charges associated with a woman s pregnancy and subsequent delivery to be reimbursed as a global or fixed payment at the time of delivery. In this reimbursement scenario, a woman s claims history may not include a pregnancy diagnosis until she actually delivers. However, given the importance of this information, the plan often does know that the woman is pregnant, despite this lack of related diagnosis codes during the prenatal care period. In cases where the plan wishes to supplement the standard claims data (i.e., if a pregnancy registry is believed to be more accurate than standard claims data), the user may submit a special pregnancy flag that can supplement the standard diagnosis stream. Refer to Chapter 3 of the Installation and Usage Guide for a discussion on implementing this approach. Delivered Delivery status is an important branch for pregnancy in the ACG Decision tree

59 Chapter 2: Diagnosis-based Markers Each ACG from 1710 through 1770 is split into two categories (1711, 1712 through 1771, 1772) based on whether or not the women within these categories have delivered during the period of analysis. After extensive testing, the ACG System development team at Johns Hopkins is confident the standard diagnosis codes used by the software for identifying deliveries are effective with positive predictive accuracy (that is, the women did actually deliver) averaging greater than 96% among all plans tested. However, for a variety of reasons diagnosis codes for delivery may not appear in a woman s claim history even though she did in fact deliver. For example, the delivery may have occurred in an outpatient birthing center or other non-traditional venue, and claims were never submitted containing any delivery codes. Also, if an analyst is using only ambulatory data (not generally recommended) where the delivery diagnosis codes are not available, it is suggested that the user provide a delivered flag in the input data stream. Refer to Chapter 3 of the Installation and Usage Guide for a discussion on implementing this approach. Pregnancy without Delivery Pregnancy without delivery is an important condition for predicting future resource use and is a marker used by the ACG Predictive Models. If date of service is not available, this marker will be set to Y when the ACG assignment does not indicate delivery: 1712, 1722, 1732, 1742, 1752, 1762 or If date of service is available, this marker will consider the order of events to capture the latest pregnancy status during the observation period in the event that a pregnancy does not continue to delivery or the patient experiences multiple pregnancies during the year. If the patient delivered during the year, but there are additional pregnancy diagnoses more than six weeks after the latest delivery diagnosis, the pregnancy without delivery marker will be set to Y. If the patient was pregnant during the year, but the pregnancy terminated, the pregnancy without delivery marker will be set to N. If a subsequent pregnancy is identified more than six weeks after the latest termination diagnosis, the pregnancy without delivery marker will be set to Y. Low Birth Weight (Less than 2500 Grams) Because low birth weight is a significant predictor of resource use, diagnosis can be used to identify low birth weight newborns when available. The ACG Software uses the following list of codes to identify low birthweight newborns: ICD-9-CM: 764.0*; 764.1*; 764.2*; 764.9*; 765.0*; 765.1, V21.3* (where * = 1-8) ICD-10: P05.0, P05.1, P07.0, P07.1 In a manner similar to the way pregnant women are subdivided by delivery status, infants can be subdivided into subcategories based on their birth weight. Diagnosis codes allow for identification of low or normal birth weights among neonates. Historically, validation analysis across a variety of health care organizations indicated that within most plans 2% to 5% of infants were identified as low birth weight. Based on vital records and other sources, the actual percentage should be somewhere between 6% and 9%. If diagnoses do not seem to be a reliable source of the recording of birth weight, analysts may wish to take advantage of this feature to appropriately categorize low birth weight infants. Analysts can flag such infants before passing the data to the ACG Software. Refer to Chapter 3 of the Installation and Usage Guide for a discussion on implementing this approach. 59

60 Chapter 3: Pharmacy-based Markers Chapter 3: Pharmacy-based Markers This chapter is intended to describe the pharmacy based morbidity markers, Rx-MGs, and the clinical criteria used to assign medications to morbidity groups. The Rx-MGs provide a medication-based approach for describing the morbidity profile of an individual or population. Rx-MGs also form the basis of the pharmacy based predictive models. Specifics of the predictive models are discussed in greater detail in later chapters discussing Risk Modeling and Predictive Models for Hospitalization. Rx-MGs are groupings of medication codes. The system is compatible with NDC or ATC codes and may be further augmented with HCPCS 6 codes. Please contact the account representative to understand other localizations available. Rx-Defined Morbidity Groups (Rx-MGs) Rx-Defined Morbidity Groups (Rx-MGs) are a medication classification system that was developed according to the following principles: 1. Rx-MGs should be both comprehensive (includes all medications) and parsimonious (groups all medications into a manageable set of categories). 2. Rx-MGs should be useful for clinical and population management applications. 3. Rx-MGs should be adaptable for domestic and international uses, and should accommodate domestic and international medication codes (NDC and ATC, respectively). 4. Rx-MG assignment methodology should use a medication s mechanism of action, clinical indications, and route of administration information to make classification decisions. These principles were used to develop 67 Rx-MG categories. Each unique active ingredient/route of administration combination is assigned to a single Rx-MG. By extracting the key morbidity-related information from each active ingredient/route of administration combination, drugs within conventional therapeutic classes were sometimes logically assigned to different Rx-MGs. This observation reinforced the need to make assignments at the individual drug level, rather than at the level of a therapeutic class. Clinical Basis of Rx-MGs Medications can be classified according to their intended therapeutic use (i.e., their clinical indications) or their mechanisms of action. The mechanism of action of a medication refers to the molecular interactions that lead to its effect on health. Molecular interactions may include such actions as inhibition of enzyme activity, binding to a receptor site, or direct effects on the body. Rx-MGs use both the intended therapeutic uses, including off-label usage, and mechanism of action to categorize medications along four clinical dimensions of medication use: Primary organ system affected by the medication. Morbidity differentiation, which refers to whether the medication is used to relieve symptoms or treat a specific disease. Expected duration of targeted diseases as time-limited or chronic. 6 Healthcare Common Procedure Coding System, maintained by the Centers for Medicare and Medicaid Services. 60

61 Chapter 3: Pharmacy-based Markers Severity of illness, which refers to the long-term stability of the targeted diseases. These four clinical dimensions of medication use are also relevant for predictive modeling. For example, higher levels of differentiation, chronicity, and greater severity would all be expected to require higher levels of resource use. Each of these criteria is discussed in the following sections. Primary Organ System Affected By combining information on a medication s clinical indications and mechanism of action, we can identify the primary organ system that it influences. Medications generally are most useful for treatment of conditions related to a particular organ system. However, this is not a universal attribute of medications, as some can be used for treatment of conditions that affect one or more organ systems. In this case, we determined the most common, or primary, organ system that is targeted by the medication. Using this approach, we derived 19 categories, which categorized all unique medication/route of administration combinations: 16 represent organ systems, and the other three signs and symptoms, toxic effects/adverse events group, and non-specific categories. These groupings are called Major Rx-MGs. Medications assigned to the non-specific category are not reflective of any underlying morbidity type; examples of these medication types are diagnostic agents and multi-vitamins. Morbidity Differentiation Another dimension of medication usage is the differentiation of the health conditions that the drug is intended to treat. Three general categories of health condition differentiation are possible: 1. symptoms, which are poorly differentiated because a given symptom like pain can occur as a result of many health conditions; 2. a general class of health conditions within an organ system, an intermediate level of differentiation; or 3. a highly differentiated indication with a single health condition target. The Rx-MGs therefore include categories for symptoms (assigned to an acute minor category within an organ system grouping or to the general signs and symptoms category), fully developed diseases that have a 1-to-1 correspondence with medication, and more general morbidity-types. Expected Duration This dimension refers to the anticipated period of time that a medication will be used to manage a health condition(s): time-limited for acute conditions expected to last a few days to a few weeks. repeated episodes of treatment for recurrent conditions. long-term treatment for chronic conditions, which typically last several months to years. Severity This dimension refers to the expected impact of the health condition(s) on the physiological stability of the patient or the patient s functional status without the use of appropriate medication therapy. Acute conditions can be separated into low and high impact, the latter has significant effects on the ability of an individual to perform daily activities. For chronic conditions, morbidity severity is related to the stability of the problem over long periods of time. Without adequate treatment, unstable chronic conditions are expected to progressively worsen, whereas stable conditions are expected to change less rapidly. 61

62 Chapter 3: Pharmacy-based Markers Clinical Characteristics of Medications Assigned to Each Rx-MG The final set of 67 Rx-MG categories, along with their distinguishing clinical characteristics, is shown in the following table. Severity Chronic and Rx-Defined Acute Conditions: Recurrent Morbidity Group Differentiation Duration Impact Conditions: Stability Allergy/Immunology Acute Minor Symptoms/General Acute Low Impact Chronic General Chronic Unstable Inflammatory Immune Disorders General Chronic Unstable Transplant General Chronic Unstable Cardiovascular Chronic Medical General Chronic Stable Congestive Heart Disease Chronic Unstable Failure High Blood Pressure Disease Chronic Stable Disorders of Lipid Disease Chronic Stable Metabolism Vascular Disorders General Chronic Stable Ears-Nose-Throat Acute Minor Symptoms/General Acute Low Impact Endocrine Bone Disorders General Chronic Stable Chronic Medical General Chronic Stable Diabetes With Insulin Disease Chronic Unstable Diabetes Without Disease Chronic Stable Insulin Thyroid Disorders General Chronic Stable Growth Problems General Chronic Stable Weight Control General Chronic Stable Eye 62

63 Chapter 3: Pharmacy-based Markers Severity Chronic and Rx-Defined Acute Conditions: Recurrent Morbidity Group Differentiation Duration Impact Conditions: Stability Acute Minor: General Acute Low Impact Curative Acute Minor: Symptoms Acute Low Impact Palliative Glaucoma Disease Chronic Stable Female Reproductive Hormone Regulation General Chronic Stable Infertility General Acute High Impact Pregnancy and General Acute High Impact Delivery Gastrointestinal/Hepatic Acute Minor Symptoms/General Acute Low Impact Chronic Liver Disease General Chronic Unstable Chronic Stable General Chronic Stable Inflammatory Bowel Disease Chronic Stable Disease Pancreatic Disorder General Chronic Unstable Peptic Disease Disease Recurrent Stable General Signs and Symptoms Nausea and Symptoms Acute Low Impact Vomiting Pain Symptoms Acute Low Impact Pain and Symptoms Acute Low Impact Inflammation Severe Pain Symptoms Acute High Impact Genito-Urinary Acute Minor Symptoms/General Acute Low Impact Chronic Renal Failure Disease Chronic Unstable Hematologic Coagulation General Chronic Unstable Disorders Infections 63

64 Chapter 3: Pharmacy-based Markers Severity Chronic and Rx-Defined Acute Conditions: Recurrent Morbidity Group Differentiation Duration Impact Conditions: Stability Acute Major General Acute High Impact Acute Minor General Acute Low Impact HIV/AIDS Disease Chronic Unstable Tuberculosis Disease Acute Low Impact Severe Acute Major General Acute High Impact Malignancies Malignancies General Chronic Unstable Musculoskeletal Gout Disease Recurrent Stable Inflammatory General Chronic Unstable Conditions Neurologic Alzheimers Disease Disease Chronic Unstable Chronic Medical General Chronic Stable Migraine Headache Disease Recurrent Stable Parkinsons Disease Disease Chronic Unstable Seizure Disorder General Chronic Stable Psychosocial Attention Deficit Disease Chronic Stable Disorder Addiction General Chronic Unstable Anxiety Symptoms/General Recurrent Stable Depression Disease Chronic Stable Acute Minor Symptoms/General Acute Low Impact Chronic Unstable General Chronic Unstable Sleep disorders Symptoms/General Acute and Recurrent Tobacco Cessation Symptoms/General Acute Bipolar disorder Disease Chronic Unstable Respiratory Acute Minor Symptoms/General Acute Low Impact 64

65 Chapter 3: Pharmacy-based Markers Severity Chronic and Rx-Defined Acute Conditions: Recurrent Morbidity Group Differentiation Duration Impact Conditions: Stability Airway Hyper- Disease Chronic Stable reactivity Chronic Medical General Chronic Stable Cystic Fibrosis Disease Chronic Unstable Skin Acne Disease Chronic Stable Acute and Recurrent Symptoms/General Acute and Recurrent Low Impact Chronic Medical General Chronic Stable Toxic Effects/Adverse Effects Acute Major General Acute High Impact Other and Non- N/A N/A N/A N/A Specific Medications Rx-MG Medication Assignment Methodology The Rx-MG assignment methodology evaluates each unique active ingredient / route of administration combination along with the four clinical dimensions of medication use described above. This review encompasses U.S. Food and Drug Administration (FDA)-approved and common off-label uses of a drug. Each unique active ingredient/route of administration combination was assigned to a single Rx-MG according to this clinical review. The ACG team recognized that medications may be used differently, particularly for off label indications, in international settings. The biomedical literature was extensively reviewed for evidence of alternate and prevalent international use of medications. A team of Johns Hopkins physicians and pharmacist health professionals made the initial Rx-MG assignments. Disagreements were reconciled during consensus meetings. The face validity of the assignments was tested by having a second team of physicians and pharmacist health professionals examine the preliminary assignments. This second review led to some modifications, but generally confirmed the validity of the initial assignments. Assignment with NDC codes involved collapsing the 100,000 codes to 2,700 unique active ingredient/route of administration combinations which then were assigned to an Rx-MG. Assignment of ATC codes involved assigning close to 5,100 ATC codes (900 4th Level and nearly 4,200 5th Level ATC codes) to an Rx-MG. Most NDC codes representing drugs are assigned to a single Rx-MG, and just 9.8% are mapped to the Non-Specific category. NDC codes representing medical supplies, exclusively over-the counter medications, and other non-pharmaceutical items are not assigned to a Rx-MG. A single NDC may be mapped to two Rx-MGs if the drug is a combination therapy. For example, amlodipine besylate/atorvastatin calcium is a combined calcium channel blocker and statin therapy to treat patients with hypertension and disorders of lipid metabolism. This single oral tablet will be assigned to both CARx030 (high blood pressure) and 65

66 Chapter 3: Pharmacy-based Markers CARx040 (disorders of lipid metabolism). Drugs that combine over the counter medications with prescription medications will only be assigned to the Rx-MG for the prescription component (e.g., codeine/pseudoephedrine). Combination therapies that are treating a single underlying condition are assigned to only one Rx-MG (e.g., HAART made up of emtricitabine/nelfinavir/tenofovir for treating HIV). If HCPCS codes are available in the Medical Services input file, the Rx-MG assignments will also consider medications administered (typically as injections) in the physician's office. The ACG team updates the assignment of new NDC codes to Rx-MG regularly. High, Moderate, and Low Impact Rx-MGs Within the detailed Rx-MG output from the system as well as in the Comprehensive Patient Clinical Profile Report, the Rx-MGs associated with an individual are categorized as High, Moderate, or Low impact. The definition for these categories is based on the individual contribution of the condition to the predictive modeling score. The specific criteria are as follows: High Impact Rx-MG: Coefficient >1 in the pharmacy derived predictive model without prior cost predicting total cost Moderate Impact Rx-MG: Coefficient between 0.1 and 1.0 in the pharmacy derived predictive model without prior cost predicting total cost Low Impact Rx-MG: Coefficient less than 0.1 in the pharmacy derived predictive model without prior cost predicting total cost Active Ingredient Count An active ingredient count is calculated as the count of unique active ingredient/route of administration combinations encountered in the patient s drug claims. If HCPCS codes are available in the Medical Services input file, the active ingredient count will consider medications administered (typically as injections) in the physician's office in addition to drugs identified in the pharmacy file. This marker is a proxy for identifying poly-pharmacy members and contributes independently and significantly to the prediction of cost. This marker is incorporated into all pharmacy-based predictive models. Members with an active ingredient count of 14 or greater get additional weight in the predictive models. 66

67 Chapter 4: Diagnosis+Pharmacy-based Markers Chapter 4: Diagnosis+Pharmacy-based Markers The ACG System outputs a number of markers derived from a combination of the patient's diagnosis code history and medication fills from all encounters during a 12-month observation period. This chapter provides definitions for the ACG System markers derived from diagnosis and pharmacy information. Condition Markers The ACG System uses condition markers to highlight specific conditions that are high prevalence chronic conditions, commonly selected for disease management or warranting ongoing medication therapy. Conditions that are used to measure pharmacy adherence are discussed in more detail in this chapter. The specific criteria used to define each condition marker are listed in the following table. The values within these condition markers indicate the evidence used to identify members with the condition. The following values are possible: NP The condition is not present; there is no evidence of the condition from any data source. TRT The condition was identified according to specific treatment criteria (listed in the following table) which include a minimum of two prescriptions within 120 days of each other (but on different dates of service) from an appropriate chronic medication drug class. Regardless of the diagnostic certainty option selected, minimum visit counts will apply as listed in the following table. It is possible to select the stringent diagnostic certainty option requiring two diagnoses for a related EDC, yet assign a condition marker to the value of TRT based on a single inpatient or emergency department diagnosis. ICD The condition was identified only from diagnosis information. Each condition is identified by one or more EDCs. When the stringent diagnostic certainty option is selected, more than one diagnosis may be required for the condition to be identified. Rx The condition was identified only from pharmacy information and minimum treatment criteria were not met. BTH The condition was identified by both diagnosis and pharmacy criteria but minimum treatment criteria were not met. For 17 of the conditions, a separate column indicates if the condition is untreated. The specific criteria for an untreated condition are listed in the following table. Regardless of the specific visit criteria specified, the patient must also have the EDC assigned under the diagnostic criteria column to be identified as untreated. It is possible to meet the untreated criteria based on a single inpatient or emergency department diagnosis. If the stringent diagnostic certainty option is applied requiring two diagnoses for a related EDC, the patient will not be designated as untreated. The untreated column will have the following values: Y The patient has no prescriptions in an appropriate chronic medication drug class. P - The patient is classified as potentially untreated based upon one of the following criteria: 1. The patient has only one prescription in an appropriate chronic medication drug class. 2. The patient has more than one prescription in an appropriate chronic medication drug class but all prescriptions occurred on a single fill date. 3. The patient has more than one prescription from multiple chronic medication drug classes but does not have two or more from a single chronic medication drug class. 67

68 Chapter 4: Diagnosis+Pharmacy-based Markers 4. The patient has more than one prescription in an appropriate chronic medication drug class but there are not two prescriptions within 120 days of each other. N The patient meets treatment criteria and has two or more prescriptions in an appropriate chronic medication drug class. D The patient meets treatment criteria, but it appears all medications within an appropriate chronic medication drug class have been discontinued, and the patient is now likely untreated based on the fact that greater than 120 days have elapsed between the latest supply end date and the observation period end date. Blank There is insufficient evidence to confirm the presence of the condition. Definitions of Condition Markers Condition Diagnostic Criteria Pharmacy Criteria Treatment Criteria Untreated Criteria Bipolar disorder PSY12 N/A 1 inpatient or ER 1 inpatient or ER diagnosis or 2 diagnosis or 2 outpatient diagnoses outpatient diagnoses plus 2 prescription with less than 2 fills (within 120 days prescription fills of each other) in at (within 120 days of least one of the drug each other) in a classes: single drug class: Anti-convulsants, Anti-convulsants, Anti-psychotics Anti-psychotics Congestive heart CAR05 CARx020 1 inpatient or ER 1 inpatient or ER failure diagnosis or 2 diagnosis or 2 outpatient diagnoses outpatient diagnoses plus 2 prescription with less than 2 fills (within 120 days prescription fills of each other) in at (within 120 days of least one of the drug each other) in a classes: single drug class: ACEI/ARB ACEI/ARB Aldosterone receptor Aldosterone receptor blockers blockers Beta-blockers Beta-blockers Diuretics Diuretics Inotropic agents Inotropic agents Vasodilators Vasodilators Depression PSY09, PSY20 PSYx040 1 inpatient or ER 1 inpatient or ER diagnosis or 2 diagnosis or 2 outpatient diagnoses outpatient diagnoses plus 2 prescription with less than 2 fills (within 120 days prescription fills of each other) in the (within 120 days of drug class: each other) in the Anti-depressants drug class: Anti-Depressants 68

69 Chapter 4: Diagnosis+Pharmacy-based Markers Condition Diagnostic Criteria Pharmacy Criteria Treatment Criteria Untreated Criteria Diabetes END06, END07, Meglitinides 2 prescription fills 1 inpatient or ER END08, END09 Non-sulfonylureas (within 120 days of diagnosis or 2 Sulfonylureas each other) in at outpatient diagnoses Thiazolidinediones least one of the drug with less than 2 classes: prescription fills Other Meglitinides (within 120 days of antihyperglycemic Non-sulfonylureas each other) in a agents Sulfonylureas single drug class: Long and short- Thiazolidinediones Meglitinides acting Insulins Other Non-sulfonylureas antihyperglycemic Sulfonylureas agents Thiazolidinediones Long and shortantihyperglycemic Other acting Insulins agents Long and shortacting Insulins Disorders of Lipid CAR11 Bile acid 2 prescription fills 1 inpatient or ER Metabolism sequestrants (within 120 days of diagnosis or 2 Cholesterol each other) in at outpatient diagnoses absorption inhibitors least one of the drug with less than 2 Fibric acid classes: prescription fills derivatives Bile acid (within 120 days of HMG-CoA reductase sequestrants each other) in a inhibitors Cholesterol single drug class: Miscellaneous absorption inhibitors Bile acid antihyperlipidemic Fibric acid sequestrants agents derivatives Cholesterol HMG-CoA reductase absorption inhibitors inhibitors Fibric acid Miscellaneous derivatives antihyperlipidemic HMG-CoA reductase agents inhibitors Miscellaneous antihyperlipidemic agents Glaucoma EYE08 EYEx030 1 inpatient or ER 1 inpatient or ER diagnosis or 2 diagnosis or 2 outpatient diagnoses outpatient diagnoses plus 2 prescription with less than 2 fills (within 120 days prescription fills of each other) in the (within 120 days of drug class: each other) in the Ophthalmic drug class: glaucoma agents Ophthalmic glaucoma agents 69

70 Chapter 4: Diagnosis+Pharmacy-based Markers Condition Diagnostic Criteria Pharmacy Criteria Treatment Criteria Untreated Criteria Human INF04 HAART* 2 prescription fills 1 inpatient or ER Immunodeficiency (within 120 days of diagnosis or 2 Virus each other) in the outpatient diagnoses drug class: with less than 2 HAART* prescription fills (within 120 days of each other) in the drug class: HAART* Hypertension CAR14, CAR15 CARx030 1 inpatient or ER 1 inpatient or ER diagnosis or 2 diagnosis or 2 outpatient diagnoses outpatient diagnoses plus 2 prescription with less than 2 fills (within 120 days prescription fills of each other) in at (within 120 days of least one of the drug each other) in a classes: single drug class ACEI/ARB ACEI/ARB Aldosterone receptor Aldosterone receptor blockers blockers Anti-adrenergic Anti-adrenergic agents agents Beta-blockers Beta-blockers Calcium channel Calcium channel blockers blockers Diuretics Diuretics Vasodilators Vasodilators Hypothyroidism END04 Thyroid replacement 2 prescription fills 1 inpatient or ER drugs (within 120 days of diagnosis or 2 each other) in the outpatient diagnoses drug class: with less than 2 Thyroid drugs prescription fills (within 120 days of each other) in the drug class: Thyroid drugs Immunosuppression ADM03 ALLx050 1 inpatient or ER 1 inpatient or ER /Transplant diagnosis or 2 diagnosis or 2 outpatient diagnoses outpatient diagnoses plus 2 prescription with less than 2 fills (within 120 days prescription fills of each other) in the (within 120 days of drug class: each other) in the Immunologic agents drug class: Immunologic agents 70

71 Chapter 4: Diagnosis+Pharmacy-based Markers Condition Diagnostic Criteria Pharmacy Criteria Treatment Criteria Untreated Criteria Ischemic heart CAR03 N/A 1 inpatient or ER 1 inpatient or ER disease diagnosis or 2 diagnosis or 2 outpatient diagnoses outpatient diagnoses plus 2 prescription with less than 2 fills (within 120 days prescription fills of each other) in at (within 120 days of least one of the drug each other) in a classes: single drug class: Antianginal agents Antianginal agents Beta-blockers Beta-blockers Calcium channel Calcium channel blockers blockers Osteoporosis END02 ENDx010 1 inpatient or ER 1 inpatient or ER diagnosis or 2 diagnosis or 2 outpatient diagnoses outpatient diagnoses plus 2 prescription with less than 2 fills (within 120 days prescription fills of each other) in the (within 120 days of drug class: each other) in the Osteoporosis drug class: Hormones Osteoporosis Hormones Parkinson s disease NUR06 Anticholinergic 1 inpatient or ER 1 inpatient or ER antiparkinson agents diagnosis or 2 diagnosis or 2 Dopaminergic outpatient diagnoses outpatient diagnoses antiparkinson agents plus 2 prescription with less than 2 fills (within 120 days prescription fills of each other) in at (within 120 days of least one of the drug each other) in a classes: single drug class: Anticholinergic Anticholinergic antiparkinson agents antiparkinson agents Dopaminergic Dopaminergic antiparkinson agents antiparkinsonism agents 71

72 Chapter 4: Diagnosis+Pharmacy-based Markers Condition Diagnostic Criteria Pharmacy Criteria Treatment Criteria Untreated Criteria Persistent asthma ALL04, ALL05 RESx040 1 inpatient or ER 1 inpatient or ER principal diagnosis or principal diagnosis or 4 outpatient 4 outpatient diagnoses plus 2 diagnoses with less prescription fills than 2 prescription (within 120 days of fills (within 120 days each other) in at of each other) in a least one of the drug single drug class: classes, or 4 Long and shortprescription fills in acting Adrenergic any of the drug bronchodilators classes with at least Immunosuppressive 2 prescription fills monoclonal (within 120 days of antibodies each other) in one of Inhaled the drug classes corticosteroids other than Leukotriene Leukotriene modifiers modifiers: Mast cell stabilizers Long and shortacting Adrenergic Methylxanthines bronchodilators Immunosuppressive monoclonal antibodies Inhaled corticosteroids Leukotriene modifiers Mast cell stabilizers Methylxanthines Rheumatoid arthritis RHU05 Disease-modifying 1 inpatient or ER 1 inpatient or ER anti-rheumatic drugs diagnosis or 2 diagnosis or 2 (DMARDs) outpatient diagnoses outpatient diagnoses Immunologic agents plus 2 prescription with less than 2 fills (within 120 days prescription fills of each other) in at (within 120 days of least one of the drug each other) in a classes: single drug class: Disease-modifying Disease-modifying anti-rheumatic drugs anti-rheumatic drugs (DMARDs) (DMARDs) Immunologic agents Immunologic agents 72

73 Chapter 4: Diagnosis+Pharmacy-based Markers Condition Diagnostic Criteria Pharmacy Criteria Treatment Criteria Untreated Criteria Schizophrenia PSY07 N/A 1 inpatient or ER 1 inpatient or ER diagnosis or 2 diagnosis or 2 outpatient diagnoses outpatient diagnoses plus 2 prescription with less than 2 fills (within 120 days prescription fills of each other) in the (within 120 days of drug class: each other) in the Anti-psychotics drug class: Anti-psychotics Seizure disorders NUR07 NURx050 1 inpatient or ER 1 inpatient or ER diagnosis or 2 diagnosis or 2 outpatient diagnoses outpatient diagnoses plus 2 prescription with less than 2 fills (within 120 days prescription fills of each other) in the (within 120 days of drug class: each other) in the Anti-convulsants drug class: Anti-convulsants Age-related macular EYE15 Anti-angiogenic N/A N/A degeneration ophthalmic agents COPD RES04 N/A N/A N/A Chronic Renal Failure REN01, REN06 GURx020 N/A N/A Low back pain MUS14 N/A N/A N/A *HAART represents a multi-drug cocktail that can be delivered in a number of configurations. 1. HAART (a three drug combination filled with a single prescription) 2. Any NRTI combination drug including either zidovudine, abacavir or tenofovir with an additional NNRTI, entry inhibitor, integrase inhibitor, or protease inhibitor 3. A NNRTI plus a ritonavir-boosted protease inhibitor or nelfinavir 4. Three of more drugs from two or more different drug classes (Entry Inhibitors, Integrase Inhibitors, NNRTIs, NRTIs, Protease Inhibitors) Pharmacy Adherence Significance of Pharmacy Adherence for Effective Care There is a considerable literature on pharmacy adherence and a substantial body of evidence that high levels of medication adherence yield improved therapeutic outcomes and more cost effective treatment. Medication adherence represents an important dimension of effective disease management. Medication adherence is one dimension of patient compliance with therapy. Optimally, medication adherence means taking the correct drugs, for the correct indications, at the correct times, at the correct dose, and under the proper conditions for safe and effective use (storage, shelf life, avoiding substances that could affect efficacy, etc.) 73

74 Chapter 4: Diagnosis+Pharmacy-based Markers The only way to accurately assess medication adherence on all of these dimensions is through direct patient observation, a highly impractical approach. Researchers have developed a number of less intrusive strategies for assessing medication adherence including patient logs, periodic assessments of the patient s medication supply, and electronic sensors in the medication bottles that record when the bottle was opened. These approaches constitute indicators of medication adherence, but none fully address all of the dimensions of medication adherence defined above. A significant amount of medication adherence research relies on information that is reported on pharmacy claims. Claims data provide an indicator of medication possession. The pharmacy claim captures the date that the medication prescription was filled and the number of days supply. There is substantial literature on medication possession that detail a number of widely used metrics. For an excellent review, see Hess et al. (Annals of Pharmacotherapy, 40: , 2006). The ACG System pharmacy adherence markers were developed in line with these existing methodologies. Development of Possession/Adherence Markers Generally, measurement strategies have targeted specific possession events (i.e., gaps) or average time of possession expressed as a ratio (supply over prescribing period). The ACG System employs both of these strategies as they address different dimensions of adherence. Gaps in filling medication prescriptions capture acute occurrences and may represent a therapeutically significant event that could be overlooked if only averages were considered. Gaps, if captured on a timely basis, are amenable to care management intervention. Averages represent how well medication is supplied over a span of time (i.e., one year). They are a good summary indicator of possession that can assist in assessing the overall compliance of a patient, but are less amenable to immediate clinical action. Both approaches follow the same conceptual model. The following figure depicts a gap event: Figure 8. Gap in Medication Possession The analysis begins with a prescription for a targeted medication during the time period of interest. The prescription covers a span of time that is defined by the associated days supply. The interval between the end of the days supply of the first prescription and the onset of a new prescription for the same medication represents a gap. A common approach to gap measurement is to include a grace period at the end of the days supply to ensure that the reported gap is clinically significant. Thus, a patient could be reported as having no gaps in possession, but actually have a number of small gaps that fall within the grace period. In this example, the grace period is 15 days. In most instances, a 15-day grace period has been employed within the ACG System. A 30-day grace period is used when the impact of a longer gap is considered to be minimal or it is necessary to accommodate attributes of dispensing. 74

75 Chapter 4: Diagnosis+Pharmacy-based Markers Certain events which can complicate the measurement of gaps and potentially introduce false positives are addressed as follows. The first scenario in which a patient renews a prescription before their current supply has been completely exhausted, is presented in the following figure. Figure 9. Gap in Medication Possession Following Oversupply Under this circumstance there is still supply on hand when the prescription is refilled. This surplus is added to the days supply of the refill. The assessment for a gap considers both the surplus and the 15 day grace period. The second scenario addresses supply on hand if a patient is hospitalized. Customarily, the hospital pharmacy is responsible for all prescriptions during the period of hospitalization, as shown in the following figure. Figure 10. Gap in Medication Possession Following Hospitalization If a hospitalization occurs during the prescribing period, the supply on hand when hospitalized is presumed to still be available after discharge. This remaining supply plus the 15-day grace period must be exhausted before a gap is counted. Pharmacy Adherence Defined Medication possession can have different implications for various conditions and medications. The majority of drugs are given acutely and in these situations, the concept of medication adherence has little relevance. What distinguishes the ACG System pharmacy adherence analyses from other approaches commonly used in the literature or provided by pharmacy benefits managers, is the effort dedicated to identifying conditions and drugs that are intended to be administered chronically. The ACG System 75

76 Chapter 4: Diagnosis+Pharmacy-based Markers measures adherence for 17 conditions where the chronic administration of medication is, in most instances, appropriate. These 17 Conditions that Require Chronic Administration of Medication are: Bipolar Disorder Congestive Heart Failure Depression Diabetes Disorders of Lipid Metabolism Glaucoma Human Immunodeficiency Virus Hypertension Hypothyroidism Immunosuppression/Transplant Ischemic Heart Disease Osteoporosis Parkinson s Disease Persistent Asthma Rheumatoid Arthritis Schizophrenia Seizure Disorders Each targeted condition is associated with one or more target drug classes identified by the Johns Hopkins clinician advisors as a subset of drugs that once started, should be given continuously. The resultant condition-drug class pairings are presented in the Condition-Drug Class Pairings table. Note: The measurement of adherence is confined to only these condition-drug class pairings. It is possible for patients to take multiple ingredients within the same drug class and/or temporarily substitute one ingredient for another within the drug class. To prevent the appearance of oversupply when multiple ingredients are prescribed, calculations are performed at the ingredient (not drug class) level. Condition-Drug Class Pairings Condition Bipolar Disorder Bipolar Disorder Congestive Heart Failure Congestive Heart Failure Congestive Heart Failure Congestive Heart Failure Congestive Heart Failure Congestive Heart Failure Depression Drug Category Anti-convulsants Anti-psychotics ACEI/ARB Aldosterone receptor blockers Beta-blockers Diuretics Inotropic agents Vasodilators Anti-depressants 76

77 Chapter 4: Diagnosis+Pharmacy-based Markers Condition Diabetes Diabetes Diabetes Diabetes Diabetes Diabetes Diabetes Glaucoma Human Immunodeficiency Virus Disorders of Lipid Metabolism Disorders of Lipid Metabolism Disorders of Lipid Metabolism Disorders of Lipid Metabolism Disorders of Lipid Metabolism Hypertension Hypertension Hypertension Hypertension Hypertension Hypertension Hypertension Hypothyroidism Ischemic Heart Disease Ischemic Heart Disease Ischemic Heart Disease Osteoporosis Parkinson's Disease Parkinson's Disease Persistent Asthma Persistent Asthma Persistent Asthma Drug Category Insulins Meglitinides Miscellaneous antidiabetic agents Non-Sulfonylureas Other Anti-Hyperglycemic Agents Sulfonylureas Thiazolidinediones Ophthalmic glaucoma agents HAART* (see below) Bile acid sequestrants Cholesterol absorption inhibitors Fibric acid derivatives HMG-CoA reductase inhibitors Miscellaneous antihyperlipidemic agents ACEI/ARB Aldosterone receptor blockers Anti-adrenergic agents Beta-blockers Calcium channel blockers Diuretics Vasodilators Thyroid drugs Antianginal agents Beta-blockers Calcium channel blockers Hormones Anticholinergic antiparkinson agents Dopaminergic antiparkinsonism agents Adrenergic bronchodilators Immunosuppressive monoclonal antibodies Inhaled corticosteroids 77

78 Chapter 4: Diagnosis+Pharmacy-based Markers Condition Persistent Asthma Persistent Asthma Persistent Asthma Rheumatoid Arthritis Rheumatoid Arthritis Schizophrenia Seizure Disorder Immunosuppression/Transplant Drug Category Leukotriene modifiers Mast cell stabilizers Methylxanthines Disease-modifying anti-rheumatic drugs (DMARDs) Immunologic agents Anti-psychotics Anti-convulsants Immunologic agents *HAART represents a multi-drug cocktail that can be delivered in a number of configurations: 1. HAART (a three drug combination filled with a single prescription) 2. Any NRTI combination drug including either Zidovudine, Abacavir or Tenofovir with an additional NNRTI, entry inhibitor, Integrase inhibitor or Protease inhibitor 3. A NNRTI plus a Ritonavir-boosted Protease inhibitor or Nelfinavir 4. Three of more drugs from two or more different drug classes (Entry Inhibitors, Integrase Inhibitors, NNRTIs, NRTIs, Protease Inhibitors) To reduce the number of false positives in identifying patients with a given condition, and to avoid misidentification based on drugs that are used to treat more than one condition, the majority of the conditions assessed for a TRT assignment will require an inpatient diagnosis, an emergency department diagnosis or a minimum of two outpatient diagnoses for a condition assignment. Alternately, when medication usage patterns are clear, a minimum of two prescriptions on different dates of service within an applicable drug class can be used to identify a condition (see conditions marked with an arrow in the preceding Conditions that Require Chronic Administration of Medication list). The general rules for identifying patients as treated (i.e., TRT) and including them in the adherence calculations are as follows: For Diabetes, HIV, Disorders of Lipid Metabolism, and Hypothyroidism, patients may be identified as treated with a minimum of two prescriptions within 120-days of each other (but with different fill dates) in an applicable drug class. For Persistent Asthma, there must be a minimum of two prescriptions within 120-days of each other (but with different fill dates) in an applicable drug class (other than Leukotriene Modifiers) plus any one of the following criteria: An inpatient discharge with a PRINCIPAL diagnosis of asthma An ER visit with a PRINCIPAL diagnosis of asthma At least four outpatient visits with asthma listed as any diagnosis Two additional asthma medication prescribing events for a total of four asthma prescriptions in the observation period All other conditions will be satisfied with a minimum of two prescriptions within 120-days of each other (but with different fill dates) in an applicable drug class plus either a single inpatient diagnosis, single emergency department diagnosis or a minimum of two outpatient diagnoses. 78

79 Chapter 4: Diagnosis+Pharmacy-based Markers Detailed condition marker definitions are included in Definitions of Condition Markers on page 68. Note: Pharmacy adherence calculations are ONLY performed for members with a TRT condition marker. There are situations, however, where a member may receive a TRT condition marker, but pharmacy adherence will NOT be assessed. Some examples include: The prescribed active ingredient falls within a drug class that can be used to identify that the condition is present, but is not eligible for adherence calculations, i.e., short-acting adrenergic bronchodilators for persistent asthma or short-acting insulins for diabetes. The patient did not have at least 120 days supply within a targeted drug class during the observation period, the minimum amount required for the possession calculations. The patient had two prescriptions within the same drug class, but not for the same active ingredient. Five markers have been adopted for inclusion in the pharmacy adherence module. Each marker represents a slightly different perspective on adherence. Note that gaps identify significant intervals during the observation period where there is no medication supply and indicate individual problematic events, while averages provide a perspective on adherence over time. The averaging and gaps approaches are intended to complement each other. It is possible to have an acceptable average and still experience a major gap in medication supply. Averages can be used to measure adherence at a population level. Pharmacy Adherence Module Markers Marker Definition Application/Interpretation Number of Gaps Count of occurrences where the To be significant, the gap must time interval between the end of extend beyond a grace period supply of one prescription and the (currently 15 or 30 days depending onset of the next prescription for on the condition and drug class the same medication (active pair). ingredient) is more than the grace period. Medication Possession Ratio (MPR) Total number of days for which The Medication Possession Ratio medication is dispensed (excluding (MPR) represents an average final prescription) divided by the medication possession over the total number of days between the prescribing period. In common first and last prescription. If a patient usage, an MPR of.80 or higher is is on multiple medications for a considered good adherence. The single condition, the days supply and MPR can exceed 1.0 if there is prescribing days are totaled across excess supply. In general, this all and averaged. measure is more sensitive to large gaps than to frequent gaps. Continuous Single-Interval Measure Ratio of days supply to days until the The Continuous Single Interval of Medication Availability (CSA) next prescription averaged across all Measure of Medication Acquisition prescriptions. (CSA), also an average, considers adherence over discrete prescribing events. This approach equally weights each prescribing event. This metric may also exceed 1.0. In general, this measure is more sensitive to frequent gaps than to large gaps. 79

80 Chapter 4: Diagnosis+Pharmacy-based Markers Marker Definition Application/Interpretation Proportion of Days Covered (PDC) Ratio of days supply divided by days The Proportion of Days Covered between first prescription fill and (PDC) averages possession over the end of observation period. observation period. While MPR and CSA measure the distance between fills, PDC will consider the time period from the first prescription through the end of the observation period, regardless of the timing of the last prescription fill. This metric has a maximum value of 1.0. Untreated No current evidence of treatment The untreated marker indicates with a designated class of instances where although chronic pharmaceuticals (see Definitions of drug administration may be Condition Markers on page 68). warranted for a condition, there is The untreated marker can have the no evidence of such through the following values: pharmacy claims. Non-treatment Y The patient has no represents another potential care prescriptions in a targeted drug management issue. It should be class. noted, however, that there are situations where prescribing can P - The patient is classified as occur, but not be captured in the potentially untreated based upon pharmacy claims. Additional research one of the following criteria: may be necessary before trying to The patient has only one draw conclusions from this metric. prescription in a targeted drug class. The patient has more than one prescription in a targeted drug class but all prescriptions occurred on a single fill date. The patient has more than one prescription from multiple targeted drug classes but does not have two or more from a single targeted drug class. The patient has more than one prescription in a targeted drug class but there are not two prescriptions within 120 days of each other. 80

81 Chapter 4: Diagnosis+Pharmacy-based Markers Marker Definition Application/Interpretation D The patient meets treatment criteria, but it appears all medications within an appropriate chronic medication drug class have been discontinued, and the patient is now likely untreated based on the fact that greater than 120 days have elapsed between the supply end date and the observation period end date. N The patient meets the treatment criteria. If none of the previous values apply to a particular member, the untreated column will be blank indicating that there was insufficient evidence to confirm the presence of the specific condition. Pharmacy adherence calculations can be complicated by the multitude of ways that patients access medications and by the patient's eligibility (or lack thereof) for prescription benefits. Examples: Pharmacy benefits outsourced or absent Enrollment gaps or discontinuation of eligibility Samples provided in the physician's office Discount drug programs Medications administered in the physician's office (e.g., injections to treat rheumatoid arthritis) Psychotherapy as a replacement or adjunct to medication therapy for behavioral health conditions (Psychotherapy Service Marker on page 92) Be mindful of these factors when drawing conclusions about pharmacy adherence. Medication Possession Ratio (MPR) Medication possession ratio is calculated for patients taking at least one chronic medication within the condition-drug class pairings listed in the Condition-Drug Class Pairings table. For each condition, there is a marker with the naming convention of the specific condition appended with _MPR. This marker contains the calculation of the ratio of days supply to total prescribing days. If a person is taking multiple medications (active ingredients) for a single condition, the days supply and total prescribing days are totaled across all of the medications and then averaged. The MPR Example 1 table, MPR Example 2 table, and MPR Example 3 table provide examples of MPR calculations. MPR Example 1 Supply Available Upon Days Exceeding Grace Rx_fill_date Days_supply Refill Period 1/15/

82 Chapter 4: Diagnosis+Pharmacy-based Markers Supply Available Upon Days Exceeding Grace Rx_fill_date Days_supply Refill Period 2/10/ /2/ /19/ MPR=90 days supply/(5/19/2014 1/15/2014) = 0.73 MPR Example 2 Supply Available Upon Days Exceeding Grace Rx_fill_date Days_supply Refill Period 1/15/ /10/ /18/ /19/ MPR=90 days supply/(4/19/2014 1/15/2014) = 0.96 MPR Example 3 Supply Available Upon Days Exceeding Grace Rx_fill_date Days_supply Refill Period 1/15/ /10/ /10/ /06/ MPR=90 days supply/(4/06/2014 1/15/2014) = 1.11 In general, this marker is more sensitive to large gaps than to frequent gaps. MPR can be greater than 1.0 if the member consistently refills prior to exhausting supply on hand. Additional patient-level detail at the active ingredient/drug class level for each prescription evaluated in the adherence calculations is available in the Pharmacy Spans Export File. Refer to Appendix A of the Installation and Usage Guide for more information on this file. Continuous, Single-interval Measure of Medication Availability (CSA) The Continuous, Single-interval Measure of Medication Availability is calculated for patients taking at least one chronic medication within the condition-drug class pairings listed in the Condition-Drug Class Pairings table. For each condition, there is a marker with the naming convention of the specific condition appended with _CSA. The CSA Example 1 table, CSA Example 2 table, and CSA Example 3 table provide examples of CSA calculations. 82

83 Chapter 4: Diagnosis+Pharmacy-based Markers CSA Example 1 Supply Available Upon Days Exceeding Grace Rx_fill_date Days_supply Refill Period 1/15/ /10/ /2/ /19/ CSA = ((30/(2/10/2014 1/15/2014)) + (30/(4/2/2014 2/10/2014)) + (30/(5/19/2014 4/2/2014)))/3 = 0.79 CSA Example 2 Supply Available Upon Days Exceeding Grace Rx_fill_date Days_supply Refill Period 1/15/ /10/ /18/ /19/ CSA = ((30/(2/10/2014 1/15/2014)) + (30/(3/18/2014 2/10/2014)) + (30/(4/19/2014 3/18/2014)))/3 = 0.97 CSA Example 3 Supply Available Upon Days Exceeding Grace Rx_fill_date Days_supply Refill Period 1/15/ /10/ /10/ /06/ CSA = ((30/(2/10/2014 1/15/2014)) + (30/(3/10/2014 2/10/2014)) + (30/(4/6/2014 3/10/2014)))/3 = 1.11 The CSA, since it is an average, shows more sensitivity to extremes and thus is more suited to identifying frequent gaps in medication possession. CSA can be greater than 1.0 if the member consistently refills prior to the exhausting supply on hand. Additional patient-level detail at the active ingredient/drug class level is available in the Pharmacy Spans Export File. Refer to Appendix A of the Installation and Usage Guide for more information on this file. Proportion of Days Covered (PDC) The proportion of days covered, available at the drug class level only, indicates the rate of possession from the first dispensing event in the period through the observation period end date. The ACG System 83

84 Chapter 4: Diagnosis+Pharmacy-based Markers does not consider length of eligibility, so exercise caution when interpreting PDC. PDC will not exceed 1.0. The PDC Example 1 table, PDC Example 2 table, and PDC Example 3 table provide examples of PDC calculations. PDC Example 1 rx_supply_available_upon rx_fill_date rx_refill_date rx_days_supply _refill rx_final_supply_end_date 1/21/2014 3/19/ /19/2014 5/3/ /3/2014 5/24/ /24/2014 5/25/ /25/2014 7/5/ /5/2014 7/20/ /20/2014 8/16/ /16/2014 9/13/ /13/ /14/ /14/ /17/ /17/ /18/ /31/2014 PDC=11 fills x 30 days supply plus 0 days supply used from final prescription due to prior oversupply / (Observation End Date 12/31/2014 First Fill date 1/21/2014) = 330 / 344 = 0.96 PDC Example 2 rx_fill_date rx_refill_date rx_days_supply rx_final_supply_end_date rx_final_supply_used 2/9/2014 4/17/ /17/2014 6/28/ /6/ PDC=2 fills x 90 days supply plus 1 fill x 71 days supply used / (Observation End Date 12/31/2014 First Fill date 2/9/2014) = 251 / 325 = 0.77 PDC Example 3 rx_fill_date rx_refill_date rx_days_supply rx_final_supply_end_date rx_final_supply_end_date 7/20/2014 8/17/ /18/2014 8/17/2014 9/14/ /17/2014 9/14/ /12/ /17/ /12/ /9/ /16/ /9/ /7/ /16/ /7/ /24/ /15/ /31/

85 Chapter 4: Diagnosis+Pharmacy-based Markers PDC=6 fills x 30 days supply plus 0 days supply used from final prescription due to prior oversupply / (Observation End Date 12/31/2014 First Fill date 7/20/2014) = 180 /164 = 1.10 capped to 1.0 Reporting of Possession With the exception of PDC, reporting of possession is summarized at the condition level within the Patients and ACG Results file. A number of conditions have only one condition-drug class pair (e.g., osteoporosis), while others have multiple pairs (e.g., persistent asthma). If a condition is associated with multiple drug classes, all gaps associated with these drugs are reported. Further, if a gap occurs for a combination drug, the gap is reported for each component that falls into a separate drug class. Specific drugs and prescribing periods are stored and made available in the Pharmacy Spans Export File which is described in more detail in Appendix A of the Installation and Usage Guide. In addition, reporting of possession (including PDC) is available at the drug class level within the Comprehensive Patient Clinical Profile Report and the related Drug Class Summary Export which is described in more detail in Appendix A of the Installation and Usage Guide. The maximum gap (largest gap, in days, for a medication within the specific drug class) is reported within this analysis. Gaps may differ greatly in terms of duration and the duration has potential implications for care management. The intervention for someone with an extended gap in possession may be different than for someone experiencing a one-time shorter gap. In some cases a span may not (N) be included in the adherence calculations, may be included in possession (P) calculations only, or may be included in gaps (G) calculations only. Examples include: N: The drug class is not eligible for adherence calculations, i.e., short-acting adrenergic bronchodilators or short-acting insulins. N: There is an extended gap (greater than 60 days exceeding the grace period). N: The required minimum prescribing period (120-day supply) at the drug class level was not met AND there were spans with at least one day exceeding the grace period, but the patient has a fill for a different ingredient within the same drug class (potential substitution). G: The required minimum prescribing period (120-day supply) at the drug class level was not met. P: Spans with at least one day exceeding the grace period, but the patient has a fill for a different ingredient within the same drug class (potential substitution). Excluded spans are flagged in the Rx Eligible for Adherence column within the Pharmacy Spans Export File, which is described in more detail in Appendix A of the Installation and Usage Guide. Note: There must be at least one eligible span in order for adherence to be calculated. End-of-Period Possession The pharmacy adherence methodology assesses gaps and possession ratios between the first prescription fill and the last prescription refill. When assessing the last fill, it is helpful to differentiate patients that have supply through the full observation period from patients that are without supply at the end of the observation period. If a member is without supply at the end of the period, further investigation may be needed to determine if the member has stopped therapy, if the member has started an alternate therapy or if the member is non-adherent with the current therapy. For each ingredient that is evaluated, the supply of medications is tracked through the Pharmacy Spans Export File. By comparing the supply end date to the observation period end date, possession at the end of the period can be calculated. The Drug Class Summary Export File captures an End of Period Possession field (Y/N) to indicate if there is still supply available as of the observation end date. 85

86 Chapter 4: Diagnosis+Pharmacy-based Markers If more than 120 days have elapsed between the supply end date and the observation period end date for all drug classes within a given condition, the individual is likely now untreated and the untreated marker will be set to D for discontinued. The patient will still have the condition marker set to TRT if all other criteria are met. Validation and Testing Pharmacy adherence has strong face validity among clinicians. If patients fail to comply with therapy then there are likely to be consequences downstream. For some conditions these will be felt immediately while for others years may pass before serious adverse events are experienced. As part of the validation of the pharmacy adherence markers, the relationship between gap counts and several important outcomes was assessed. The purpose of this validation exercise was to determine whether or not there was an association between numbers of gaps and either concurrent or prospective total cost. Internal consistency of the markers was also assessed by looking at the correlation between gaps and possession (i.e., a higher number of gaps resulted in a lower MPR). The following table suggests some validation analyses. The example evaluates patients with osteoporosis who were prescribed hormones, and the data shown are for persons in the highest two Resource Utilization Band (RUB) categories (RUBs 4 and 5). Validation Analyses Maximum Gap Duration DEPENDENT EFFECTS No Gap <7 days 8-29 Days 30+ Days Number of Cases 2, ,035 1,167 Median Total Cost $7,789 $7,656 $8,640 $8,094 Year One Median Pharmacy $2,787 $2,671 $2,653 $2,426 Cost Year One Median Total Cost $7,178 $6,863 $7,319 $7,907 Year Two Median Pharmacy $2,842 $2,724 $2,625 $2,645 Cost Year Two Median MPR Number of Gaps DEPENDENT EFFECTS No Gap One Gap Two Gaps 3+ Gaps Number of Cases 2,844 1, Median Total Cost $7,789 $8,154 $8,255 $8,416 Year One Median Pharmacy $2,787 $2,529 $2,514 $2,729 Cost Year One Median Total Cost $7,178 $7,345 $7,322 $8,517 Year Two 86

87 Chapter 4: Diagnosis+Pharmacy-based Markers Number of Gaps DEPENDENT EFFECTS No Gap One Gap Two Gaps 3+ Gaps Median Pharmacy $2,842 $2,637 $2,622 $2,910 Cost Year Two Median MPR Source: IMS Health Payer Solutions, Watertown, MA; national cross-section of managed care plans; population of 4,740,000 commercially insured lives (less than 65 years old) and population of 257,404 Medicare beneficiaries (65 years and older), These data exemplify some of the complexity involved in gap-related data. Patients with many gaps and gaps of long duration show a different pattern than patients with moderate gaps and gaps of moderate duration. These high gap patients may reflect treatment uncertainty, with factors that contra-indicate the use of the particular medication (hormones), or some other complicating factor. Their year two total costs are higher than the other groups, suggesting a need for monitoring. Persons familiar with the use of the MPR will know that a value of 0.8 is widely accepted as a good level of medication possession. The previous table shows that an MPR of 0.8 can be present along with potential gap-related issues that impact the costs of care. MPR alone should not be used to determine appropriate medication possession. Medication possession is a complex issue that will require a close partnership between analysts and clinicians. There are subtle distinctions in the significance of gaps depending upon the condition. Further, gaps may potentially be indicative of good management. For persons with persistent asthma, gaps in adrenergic bronchodilators could mean that the asthma is under good control and over-supply could mean the converse, poor control. Because of the subtleties in interpreting gaps, it is highly recommended that ACG System results be shared with one or more clinical reviewers before identifying which condition-drug class pairs are targeted for potential intervention. 87

88 Chapter 5: Utilization and Resource Use Markers Chapter 5: Utilization and Resource Use Markers Utilization markers identify unique markers of risk and are specifically used in the hospitalization risk models. There are a variety of utilization measures that are usually captured from dates of service, place of service, type of bill, diagnosis codes, revenue codes and procedure codes. As there are variances in coding systems, the following are intended to fully explain how the software calculates the associated utilization variables. Optionally, these variables may be manually calculated and provided on the patient file (see "Patient Data Format" in the Installation and Usage Guide). Business Rules for Utilization Markers All Cause Inpatient Hospitalization Count The intent is to count acute care inpatient hospitalization stays, regardless of cause. To count inpatient visits, count as one event each inpatient confinement with either: (a) IP in place of service, or (b) room and board revenue codes on contiguous dates of service and either the number 21 in the place of service or 11x, 12x or 18x in the type of bill. The sum of these events represents the number of inpatient visits. The criteria for identifying an inpatient confinement are: Inpatient hospital place of service (IP); or Inpatient hospital place of service code (21) or type of bill 11x, 12x or 18x, along with a revenue code (three-digit version) for room and board: 100 to 219. If interim claims are submitted, identify a single inpatient confinement for claims with contiguous begin and end dates of service. If a patient is transferred within or between facilities without a gap in service dates, count as a single inpatient visit. Inpatient Hospitalization Count The intent is to count unanticipated acute care inpatient hospitalization stays. To count inpatient visits, count as one event each inpatient confinement with either: (a) IP in place of service, or (b) room and board revenue codes on contiguous dates of service and either the number 21 in the place of service or 11x, 12x or 18x in the type of bill. The sum of these events represents the number of inpatient visits. The criteria for identifying an inpatient confinement are: Inpatient hospital place of service (IP); or Inpatient hospital place of service code (21) or type of bill 11x, 12x or 18x, along with a revenue code (three-digit version) for room and board: 100 to 219. If interim claims are submitted, identify a single inpatient confinement for claims with contiguous begin and end dates of service. If a patient is transferred within or between facilities without a gap in service dates, count as a single inpatient visit. Exclude from the All Cause Inpatient Hospitalization count, inpatient admissions with a primary discharge diagnosis of: 88

89 Chapter 5: Utilization and Resource Use Markers ICD-9-CM Pregnancy and delivery: Newborns: V30-V39 Injury: E000-E999 (except E870-E876 and E878-E879) ICD-10 Pregnancy and delivery: O00-O99, Z37 Newborns: Z38 Injury: S00-T98 (except T80-T88) V01-Y98 (except Y40-Y84) Inpatient Hospitalization Days Inpatient hospitalization days count the days between the minimum service begin date and the maximum service end date associated with each inpatient confinement in the inpatient hospitalization count. For confinements with the same service begin and end date, count a minimum of one inpatient day. Unplanned Inpatient Hospitalization Count The unplanned inpatient hospitalization count is a subset of the inpatient hospitalization count with exclusions for planned admissions. A planned admission is defined as either: A definitively planned procedure, identified by either ICD procedure code or diagnosis code. Definitively planned procedures includes items such as rehabilitation services, chemotherapy and transplants. A potentially planned procedure, identified by ICD procedure code. Potentially planned procedures are more common surgical procedures, such as hip replacements, cardiovascular procedures and other inpatient surgical treatments without evidence of acute complications such as infections, burns or injuries. Any inpatient hospitalization that cannot be identified as definitively or potentially planned is considered unplanned. Readmission 30 Day Count The readmission 30 day count is a subset of the all cause inpatient hospitalization count. The presence of this count indicates, regardless of cause, the identified admission occurred within 30 days of a previous inpatient hospitalization. Unplanned Readmission 30 Day Count The unplanned readmission 30 day count is a subset of the inpatient hospitalization count. The presence indicates the identified admission meets the unplanned inpatient hospitalization criteria and occurred within 30 days of a previous inpatient hospitalization. 89

90 Chapter 5: Utilization and Resource Use Markers Emergency Visit Count The intent is to count emergency room visits that are not precursors to subsequent inpatient hospital stays in the same time period. Those emergency room visits that are immediately followed by an inpatient hospital stay are absorbed by a hospitalization. To count emergency room visits, count as one event each patient/day in which any record exists for emergency room. The sum of these events represents the count of emergency room visits. The criteria for identifying an emergency room record are any of the following: Revenue code (three-digit version) for emergency room 450, 452, 459 CPT code for evaluation and management in emergency room , 99282, 99283, 99284, 99285, Emergency room place of service code (23) with a CPT code for surgery to Emergency room place of service (ED) Outpatient Visit Count The intent is to count instances where patients receive ambulatory care in outpatient settings. To count outpatient visits, count as one event each unique record of a patient id, provider id and date of service. The sum of these events represents the number of outpatient visits. The criteria for identifying an outpatient service record are the following: Physician office (11), outpatient hospital (22), and other (24, 25, 26, 50, 53, 60, 62, 65, 71, 72) place of service codes; type of bill 13x or 14x; or outpatient (OP) place of service code. Dialysis Service Marker The dialysis service marker identifies the presence of (outpatient) dialysis services using procedure codes for patients who have chronic renal failure. The default for this marker is 0. Flag with a 1, any patients with a diagnosis for chronic renal failure recorded at any time during the observation period that also have at least one ICD, CPT, or HCPCS procedure code indicating dialysis. Diagnosis Codes for Chronic Renal Failure ICD-9-CM: 585, 586, V451, V56, V560, V561, V562, V568, 587, 7925, V5631, V5632, 45821, 5851, 5852, 5853, 5854, 5855, 5856, 5859 ICD-10: N18.x, N19, Z49.x, Z99.2 ICD Procedure Codes for Dialysis ICD-9-CM: 38.95, 39.27, 39.42, 39.95, ICD-10-PCS: 3E1M39Z, 5A1D00Z, 5A1D60Z CPT Codes for Dialysis 90918, 90919, 90920, 90921, 90922, 90923, 90924, 90925, 90935, 90937, 90939, 90940, 90945, 90947, 90989, 90993, 90997, 90999, 93990, CPT codes copyright 2014 American Medical Association. All rights reserved. CPT is a registered trademark of the AMA. 90

91 Chapter 5: Utilization and Resource Use Markers HCPCS Codes for Dialysis G0308, G0309, G0310, G0311, G0312, G0313, G0314, G0315, G0316, G0317, G0318, G0319, G0320, G0321, G0322, G0323, G0324, G0325, G0326, G0327, G0365 Nursing Service Marker The nursing service marker identifies the presence of nursing home services through procedure codes. The default for this marker is 0. Flag with a 1 any patients with an instance of a procedure code indicating nursing services. CPT Codes for Nursing Services , Major Procedure Marker The intent is to identify patients with major procedures that occur in inpatient hospital settings. The major procedure marker identifies the presence of major procedure codes. The default for this marker is 0. Flag with a 1 any patient with a procedure code for a major procedure on a claim indicating an inpatient place of service code (21 or IP). There are approximately 3,000 procedure codes on the major procedure list. Due to the number of major procedure codes, this list is not included within the body of the Technical Reference Guide. Please contact your account representative for more details. Cancer Treatment Marker The goal of the dichotomous marker for cancer treatment is to differentiate the active treatment phase from the diagnostic and remission phases since the use of resources is different. To be identified as some with active cancer treatment a patient must have at least one EDC in the malignancy series (MAL) and any of the procedure (CPT or ICD-9-CM) or revenue codes. CPT ICD Procedure ICD-9-CM: ICD-10-PCS: Given the large number of qualifying codes, this list in not included within the body of the Technical Reference Guide. Please contact your account representative for more details. 91

92 Chapter 5: Utilization and Resource Use Markers Revenue Code 028x Mechanical Ventilation Marker The dichotomous mechanical ventilation marker identifies the presence of mechanical ventilation through tracheostomy or ventilator support. Mechanical ventilation carries many potential complications including lung injury, infection and respiratory distress syndrome. Mechanical ventilation is triggered by the presence of selected procedure or diagnosis codes. ICD-9-CM V44.0 V46.11 V46.12 ICD-10 Z93.0 Z99.11 Z99.12 CPT/HCPCS A4629 C5281-C5282 S8189 A7520-A7527 C A4481 D A4605 K A4608-A4610 K A4621-A4626 L A7501 L A7504-A7506 L C1039 L8511-L C1146 S8180-S ICD Procedure ICD-9-CM: ICD-10-PCS: 5A1935Z, 5A1945Z, 5A1955Z Psychotherapy Service Marker The psychotherapy marker identifies the presence of ambulatory therapeutic interactions or treatments to address problems that are psychological in nature. The focus is on ambulatory treatment as a proxy for ongoing management and not crisis-related interventions. This marker is a complement to the pharmacy adherence module where psychotherapy may supplement or replace medication use in some patients. 92

93 Chapter 5: Utilization and Resource Use Markers There must be at least two instances of a qualifying service on two different dates of service within 120 days of each other to set the marker to 1. A qualifying service is an individual medical services record with either a procedure code or a revenue code in the range below excluding inpatient, emergency department and urgent care services with either a place of service codes 20, 21, 23, 31, 51, IP, UC, or ED or a type of bill 11x, 12x, or 18x or revenue code 45x. CPT-4 Codes Revenue Codes Resource Bands The software incorporates both prior total cost and prior pharmacy cost bands into the ACG predictive models. They are a useful adjunct to analysts wishing to stratify their populations. Possible values include: 0 0 or no costs percentile percentile percentile percentile percentile percentile percentile percentile percentile 93

94 Chapter 6: Coordination Chapter 6: Coordination The Johns Hopkins ACG System introduced ACG Coordination Markers in order to identify populations that are at risk for poorly coordinated care. Used by themselves or in conjunction with other risk markers, this set of markers adds another dimension so as to enhance the clinical screening process. The basic premise behind the creation of ACG Coordination Markers is that individuals receiving poorly coordinated care have worse clinical outcomes and have higher medical expenses than individuals who are being provided coordinated care. Assessing Care Coordination Care coordination has traditionally been assessed using instruments that have been unable to use administrative claims information. These coordination assessment approaches have been survey based, and have targeted the patient's, family member s, or provider's perceptions of care collaboration. Chart reviews have also been used to assess the information exchanged between physicians, teamwork processes, and performance. Finally, resources and structures that support care coordination have also been used to indirectly evaluate coordination. Surveys, chart reviews, and resource information are not readily available for coordination assessment, making it challenging to incorporate this dimension of care into the clinical screening process. Using only administrative claims information, ACG Coordination Markers are able to assess whether an individual is at risk for receiving poorly coordinated care. Five patient markers make up ACG Coordination Markers: Management Visit Count: The number of face-to-face physician visits with an eligible specialty. Majority Source of Care (MSOC): An assessment of the level of participation of those providers that provided care to each patient. Unique Provider Count: A count of the number of unique providers that provided care to the patient. Specialty Count: A count of the number of specialty types (not the same as number of specialists seen) that provided care to the patient. Generalist Seen: A marker indicating a generalist s participation in an individual s care. A coordination risk measure combines these markers to determine whether a person has a "likely," "possible," or "unlikely" coordination issue. A care density ratio quantifies patient sharing based on outpatient face-to-face visits with eligible physicians. Each of the ACG Coordination Markers limit their evaluation to outpatient face-to-face physician visits as defined by procedure code. CPT Codes defining Face to Face Physician Visit CPT Codes copyright 2014 American Medical Association. All rights reserved. CPT is a registered trademark of the AMA. 94

95 Chapter 6: Coordination The Management Visit Count, Majority Source of Care (MSOC), Unique Provider Count, and Specialty Count further consider only those visits provided by eligible specialties (i.e., specialties that could reasonably manage the overall care for a patient). Examples of eligible specialties are provided in the following list of Eligible Specialties: Allergy & Immunology Neuro-musculoskeletal Oral & Maxillofacial Colon & Rectal Surgery Medicine Surgery Family Medicine Nuclear Medicine Orthopedic Surgery Internal Medicine Obstetrics & Otolaryngology Neurological Surgery Gynecology Pain Medicine Ophthalmology 95

96 Chapter 6: Coordination Pediatrics Radiology Advanced Practice Physical Medicine & Surgery Midwife Rehabilitation Thoracic Surgery Clinical Nurse Specialist Plastic Surgery (Cardiothoracic Vascular Nurse Practitioner Preventive Medicine Surgery) Physician Assistant Psychiatry and Transplant Surgery Neurology Urology Face-to-face visits with provider specialties that are not eligible are not considered as part of the Management Visit Count, Majority Source of Care (MSOC), Unique Provider Count, and Specialty Count markers. Examples of specialties that are not considered eligible are provided in the following list of Specialties not Eligible: Ambulance Services Departments, Hospitals, Speech, Language and Agencies Assisted living facilities) Hearing Service Dental Providers Laboratories Providers Dermatologists Managed Care Suppliers of Durable Organizations Medical Equipment & Dietary & Nutritional Pharmacy Service Supplies Service Providers Providers Therapists (Physical Facilities (Clinics, Podiatrists Respiratory) Ambulatory Health Centers, Emergency Psychologists Technicians Social Service Providers A subset of provider specialties defines generalists. Patients with at least one outpatient face-to-face visit to a generalist will have the Generalist Seen marker set to "Y". Patients with no outpatient face-to-face visits to a generalist will have the Generalist Seen marker set to "N". The following list of Generalist Specialties provides the list of specialties considered generalists for this marker: Family Medicine Pediatricians Obstetrics and Internal Medicine Preventive Medicine Gynecology Geriatricians Nurse Practitioners Management Visit Count The Management Visit Count represents the total number of face-to-face visits with an eligible specialty. Majority Source of Care (MSOC) The Majority Source of Care marker will determine the percent of the outpatient face-to-face visits provided by eligible physicians that saw the member most over the measurement period. For each patient, the ACG System determines the following: The percentage of visits provided from the largest source. The Management Visit Count is the denominator for this percentage. The provider ID(s) that is responsible for the majority source of care. In the event that more than one provider is assigned the same MSOC percentage, the ACG System will identify all the providers as being the majority source of care. 96

97 Chapter 6: Coordination The following figure shows two patients. The patient on the left saw three eligible providers and received eight of his ten outpatient face-to-face visits from the geriatrician. In this case, the geriatrician is identified as the MSOC provider with a score of 0.80 (80%.) The patient on the right saw four eligible providers and received four of her outpatient services from the endocrinologist. The endocrinologist provider is identified as the MSOC with a score of 0.40 (40%.) Figure 11. Majority Source of Care The following figure shows the characteristics of Majority Source of Care assignments in a large commercial managed care population. Most (38%) of the patients in this population see a single physician for all of their care. This would be expected for a relatively healthy working age population. Roughly 40% of patients have a physician who handles at least 50% of their visits but not all of their visits. This leaves a relatively small segment whose care is distributed more evenly among a group of physicians. These patients are likely to be those who have more co-morbidities, are likely to be more expensive, and who are more prone to coordination problems. 97

98 Chapter 6: Coordination Figure 12. How Majority Source of Care is Distributed in a Large Commercial Managed Care Population Unique Provider Count The Unique Provider Count marker determines the number of unique eligible providers that imparted outpatient care over the measurement period for any condition. In the following figure the patient on the left saw three eligible providers giving him a unique provider count of 3. The patient on the right saw four eligible providers and received a unique provider count of 4. Figure 13. Unique Provider Count 98

99 Chapter 6: Coordination The prevalence of different levels of unique providers seen is shown in the following figure. Consistent with Figure 12: How Majority Source of Care is Distributed in a Large Commercial Managed Care Population, most patients in this relatively healthy population see only one provider. The distribution has a very long tail with nearly 1% seeing nine or more unique physicians. The large the number of unique providers involved the greater the likelihood of coordination issues. Figure 14. Distribution of Unique Provider Count Within A Large Commercial Managed Care Population Specialty Count The Specialty Count determines the number of eligible specialty types that provided outpatient care over the measurement period for any condition. In the following figure, the patient on the left saw two distinct specialty types. Geriatricians and Internists are both considered Generalists and, thus, are only counted once. The patient on the right saw four eligible providers. This patient received a specialty count of 4. 99

100 Chapter 6: Coordination Figure 15. Specialty Count The distribution of Specialty Count in a large commercial managed care population is shown in the following figure. The results are consistent with what was presented in Figure 14: Distribution of Unique Provider Count Within A Large Commercial Managed Care Population. Nearly two percent saw six or more specialties and represent the population segment most likely to experience coordination issues. Figure 16. Distribution of Specialty Count in a Large Commercial Managed Care Population 100

101 Chapter 6: Coordination Generalist Seen The Generalist Seen marker determines whether or not a patient has been provided outpatient care by a generalist in the measurement period. In the following figure, the patient on the left saw a generalist for outpatient care. The patient on the right did not see a generalist. Analysis of a large commercial managed care population shows that 76% of all patients see a generalist in a year. The remainder may be at risk for coordination issues. Figure 17. Generalist Seen Risk of Poor Coordination The ACG Coordination Markers can be used together to provide a comprehensive picture of coordination of care. The ACG System assigns patients into three levels indicating the risk for coordination issues based on the grid shown in the following table. 101

102 Chapter 6: Coordination Coordination Risk Coordination Unique Provider Majority Source Risk Count Specialty Count of Care Generalist Seen Likely LCI >=7 >=5 <=28 N coordination Issue LCI >=7 >=5 <=28 Y LCI >=7 >=5 >28 N LCI >=7 <5 <=28 N LCI >=7 <5 >28 N LCI >=7 <5 >28 Y LCI >=7 >=5 >28 Y Possible PCI 2-6 >=5 <=28 N Coordination Issue PCI >=7 <5 <=28 Y PCI 2-6 >=5 <=28 Y PCI 2-6 >=5 >28 N PCI 2-6 <5 <=28 N PCI 2-6 >=5 >28 Y PCI 2-6 <5 >28 N Unlikely UCI 2-6 <5 <=28 Y Coordination Issue UCI 2-6 <5 >28 Y UCI <= A high unique provider count is the major risk factor for being a potential coordination issue with one exception. The following figure considers only members with high morbidity levels (in Resource Utilization Bands 4 and 5) to control for illness burden. There is a dramatic impact on year one and year two median total costs for members at high risk for poor coordination. Persons who presented a coordination risk carried higher costs into the following year. 102

103 Chapter 6: Coordination Figure 18. Impact of Poor Coordination on High Morbidity Users Coordination is also expected to influence aspects of the quality of care provided. Considering members of a large commercial population with asthma, the following figure shows the percentage of emergency department (ED) visits over a one year period. Patients at the highest level of risk for coordination issues experience twice the percentage of ED visits during the measurement period. Figure 19. Impact of Poor Coordination on ED Use by Asthma Patients 103

104 Chapter 6: Coordination Care Density Myriad formal and informal referral relationships among clinicians lead to varying levels of patient sharing. Higher levels of patient sharing are often credited with higher intensity of care, better care coordination, and greater potential cost opportunities. The care density ratio is a patient-level measure of the extent of patient sharing among the physicians that a patient saw during face-to-face encounters over a period of time. The numerator is the total number of instances of patient sharing among a patient s eligible physicians. Every relevant instance of patient sharing contributes to the numerator, whereby the contributions of instance range between 1 and 10 maximally. The denominator is the total number of unique pairs of the patient s physicians. A patient who visited at least two physicians receives a care density ratio. Figure 20. Care Density Ratios The distribution of the care density ratio typically varies by the unique provider count, whether a patient saw a generalist, and a patient s resource utilization band (RUB). Among patients with complex morbidity who seek care from multiple physicians one often finds a higher frequency of low care density ratios, compared to healthier patients who seek care from a small number of physicians. Hence the lower and upper quartile care density cutoff values vary among patients. The care density ratio distributions are characterized by three care density quantiles, the bottom quartile (LOW), the middle half (MID), and the top quartile (TOP). The ACG System outputs the care density bottom and top quartile cutoff values. A finding from research is that greater patient sharing and higher care density are associated with lower resource utilization. As a consequence one may form expectations about cost savings relative to the top quartile of care density that is achieved by members. The expected savings associated with high care density are greater for members in the bottom quartile of care density compared to patients in the middle half. The care density cost saving ratio considers the same factors that have an impact on the care density ratio and additionally the patient s current care density quartile. The ACG System applies the care density cost saving ratio to the patient s total cost. The result of this calculation is presented as a care density cost saving range. This range represents the cost opportunity for a similar patient with an assumed high care density compared to the actual lower level of care density. 104

105 Chapter 6: Coordination Conclusion The ACG Coordination Markers provide a meaningful way to identify members at risk for poor coordination. The coordination risk marker provides a means for identifying those members most at risk of coordination issues. Members with poorly coordinated care are more likely to have excess utilization as a result of redundant testing, potentially harmful drug-disease interactions and overall lower quality care. The care density ratio provides a means for quantifying the strength of possible relationships among a member's network of physicians. A network with greater density is expected to have greater potential cost savings. These markers complement the ACG System risk scores and additional system markers in further defining the at risk population. 105

106 Chapter 7: Risk Modeling Chapter 7: Risk Modeling The innovation that the ACG risk models present is their unique clinical logic coupled with their excellent statistical performance. Each model employs risk factors that were created using clinical frameworks that make sense to health professionals and medical managers. This chapter describes the application of the diagnosis and pharmacy based markers in the determination of patient risk. The chapter includes an overview of the methods used and validation of the various scores produced by the ACG System. Elements of a Risk Model: Five Key Dimensions The structure of a risk model can be characterized in five dimensions: conceptual basis; reference population; risk factors; statistical approach; and, modeled outcomes. Each of these dimensions is discussed in more detail below in the context of the ACG System. Conceptual Basis The ACG System s risk models are anchored in the same fundamental insight about morbidity that led to the creation of ACGs. This insight, conceived by Dr. Barbara Starfield in the 1970s, is that morbidity clusters and multi-morbidity are the best indicators of health needs. Clinicians continue to shape the ACG System to ensure that the models retain clinical cogency as improvements are made to explanatory power. On very large datasets, it is possible to make improvements in predictive power by just adding new variables, but there is an art to developing well-constructed risk factors to support a parsimonious model. In the application of risk models, it is important to define models that are less prone to strategic manipulation. Again, statistical improvements may be gained by increasing the number of variables; however, this makes the model susceptible to potential up-coding. The ACG System is robust and less likely to appraise large shifts in risk assessment on the basis of a single diagnosis code. Modeled Outcomes The ACG System risk models can be used to predict or explain costs (annual total healthcare or annual pharmacy costs) and to predict the likelihood of certain sentinel events (the likelihood of being high cost, likelihood of inpatient hospitalization, etc). Optimal performance of ACG System risk models occurs when the model is specifically calibrated to the outcome. A specialized model will be a better fit than attempting to use a model focused on a different outcome, e.g., using a model built to explain total cost to explain hospitalization, and vice versa. Some users of the ACG System have built well-performing models on additional model outcomes; e.g. mortality 9 ; however, one of the criteria for models presented in the ACG System is availability of reference data discussed in the following. 9 Austin PC, van Walraven C, Wodchis WP, Newman A, Anderson GM. (2011) Using the Johns Hopkins aggregated diagnosis groups (ADGs) to predict mortality in a general adult population cohort in Ontario, Canada. Med Care 49:

107 Chapter 7: Risk Modeling Reference Population In order to provide a reliable standard model, the ACG System derives its risk models from a large, representative reference population. The reference population is used as a control for a range of factors known to affect modeled health risks such as age, gender, and morbidity burden. A very large longitudinal claims database, licensed from PharMetrics, a unit of IMS, Watertown, MA, was used for the ACG modeling efforts. The database is generally representative of the U.S. commercially insured population. Multiple populations were used for modeling: (1) commercial (all ages to 65 years old) and (2) Medicare Advantage (United States federal program that pays for certain health care expenses for people aged 65 or older through commercial managed care organizations). The ACG System also contains an all-age model that selectively assigns risk scores based on the model appropriate to each patient: children (0-17 years of age), adults (18-64 years of age), and the elderly (65 and over years of age). There are also reference weights for specific population subgroups such as Medicaid TANF (a health care support program for children and families in the U.S.). The selection of reference population is made when processing the ACG data file with the selection of Risk Assessment Variables. Various subsets of data for the period were used for development and validation, depending on the model requirements. The population was restricted to include enrollees that had both pharmacy and medical eligibility. Additional eligibility requirements vary by model, but no less than six months of eligibility is required for defining risk factors. Risk Factors As described in the previous chapters, the ACG System derives a wide range of risk factors depending on available data. The ACG System models make use of the comprehensive array of metrics available within the entire ACG risk adjustment system. The risk factors that constitute the ACG System models are shown in the following figure, and are discussed in detail following. The full list of variables used in the models are available in Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models on page

108 Chapter 7: Risk Modeling Figure 21. Risk Factors in the ACG System Models Age and gender are included to assess age-related and gender-based health needs. Overall disease burden is measured using the ACG categorization of morbidity burden. For more information, see Patterns of Morbidity Adjusted Clinical Groups (ACGs) on page 17. For predictive models, 27 individual ACGs that represent the most resource intensive morbidity groups are included; for concurrent models, 30 individual ACGs are utilized. All other ACGs are incorporated into the concurrent and prospective models as one of two groups. Individual ACG Categories Included in the Risk Models on page 129 and ACGs Included in Three Resource Groups on page 130 are derived from ACG category alone. Disease markers. High-impact chronic conditions were selected for inclusion in the predictive model because when they are present they have a high impact on resource consumption. They are common chronic conditions that are associated with greater than average resource consumption, uncommon diseases with high impact on both cost and health, complications of chronic disease that signify high disease severity (e.g., diabetic retinopathy), or conditions that are a major biological influences on health status (e.g., transplant status, malignancy). Only conditions for which the evidence linking health care to outcomes is strong were included in the predictive model. A sub-set of the ACG System s Expanded Diagnostic Clusters (EDCs) is used to identify the high impact conditions in the model. The Concurrent Risk models also include EDC conditions that are acute in nature. For more information, see Disease Markers - Expanded Diagnosis Clusters (EDCs) on page

109 Chapter 7: Risk Modeling Pregnancy without delivery is included in predictive models because of the high use of resources associated with delivery that will occur in the subsequent assessment period. For more information, see Pregnancy without Delivery on page 59. This marker is not used in concurrent models because the delivery is a future event. Concurrent models incorporate pregnancy using individual ACG categories. Hospital dominant morbidity types are based on diagnoses that, based on expert clinical review, would be associated with a high likelihood of hospitalization in the next year. All these diagnoses are setting-neutral, i.e., they can be given in any inpatient or outpatient face-to-face encounter with a health professional. The variable is a count of the number of morbidity types (i.e., ADGs) with at least one hospital dominant diagnosis. For more information, see Hospital Dominant Morbidity Types on page 55. Frailty is a dichotomous (on/off) variable that indicates whether an patient has a diagnosis falling within any clusters that represent medical problems associated with frailty. For more information, see Frailty Conditions on page 56. Rx-defined morbidity groups. All Rx-MGs are included in the predictive model, either independently or as a complement to the diagnosis-based markers. For more information, see Rx-Defined Morbidity Groups (Rx- MGs) on page 60. The Concurrent Risk models may optionally consider pharmacy data when available. Due to the circularity in defining morbidity based on prescribing, it is strongly recommended against using pharmacy data to contribute to risk for any profiling or payment applications. The option is available to assess potential biases in the diagnosis data that might be reflected in the pharmacy data. Active Ingredient Count is considered in the predictive model when an individual has had medications representing 14 or greater individual active ingredients. For more information, see Active Ingredient Count on page 66. The Concurrent Risk models may optionally consider pharmacy data when available. Due to the circularity in defining morbidity based on prescribing, it is strongly recommended against using pharmacy data to contribute to risk for any profiling or payment applications. The option is available to assess potential biases in the diagnosis data that might be reflected in the pharmacy data. Costs Percentiles can be added to the predictive model as a measure of demand and need for services not captured by the treated morbidity information available in diagnostic codes. The model uses either prior pharmacy costs or total health care costs as presented in the pharmacy cost band or total cost band (see Resource Bands on page 93) depending on the outcome of interest. Cost percentiles are not utilized in the concurrent model. Utilization Markers are considered in the hospitalization and readmission models only. The complement of utilization markers includes inpatient hospitalization, emergency visits, outpatient visits, dialysis service, nursing service, major procedure, psychotherapy service, active cancer treatment, mechanical ventilation and CAL-SSA. The readmission model also includes inpatient days and unplanned readmissions within 30 days. For more information, see Predictive Models for Hospitalization on page 123. Combining these risk factors is clinically logical because the morbidity information obtained from diagnostic codes complements that which is obtained from medication codes. In effect, the combined model provides the most complete clinical picture of the patient s health needs and demand for care that is possible using administrative data. While a model calibrated to look at diagnosis, pharmacy and prior cost information may yield the highest statistical performance in theory, missing input data is a practical challenge that health care organizations face. Rather than enforce the use of all risk factors, a number of model variants have been developed to 109

110 Chapter 7: Risk Modeling adapt to available data. The ACG System will apply logic to determine the model that will make use of the most available data. This model will be identified in the summary statistics. Statistical Approach Three types of models are commonly used depending on the outcome variable: Actuarial cells are categorical variables that can directly explain variation in an outcome variable (e.g., Age-gender or ACG categories are examples of actuarial cells used for explaining total cost). Linear regression models are used for continuous outcome variables (e.g., total cost or pharmacy cost). Logistic regression models are used for binary/dichotomous outcome variables (e.g., probability of high cost, probability of inpatient hospitalization or probability of readmission). Actuarial cells provide an easily calibrated model for explaining the relative risk of an outcome. Deriving a risk score from an actuarial cell requires only basic arithmetic. In use in the ACG System, the actuarial cell determination of relative risk is calculated as the mean cost of all patients in an actuarial cell divided by the mean cost of all patients in the population. This method is also referred to as indirect standardization because comparisons are made by extrapolating a comparison of each person to the average. This method is employed for age-gender concurrent risk, local ACG concurrent risk and reference ACG concurrent risk. For more information, see Patterns of Morbidity Adjusted Clinical Groups (ACGs) on page 17. Linear Regression using ordinary least squares parameter estimation is widely used and, when calculated against a substantially large population, provides fairly stable estimates. The individual predictions have an intuitive appeal, because they can readily be translated into monetary values. While cost data do not fulfill the assumptions for these models (which are best suited for data following a normal distribution with common variance across the variable range), the models are sufficiently robust that some departures from the assumptions can be tolerated. Our work suggests, for example, that generalized linear models using a log link function and a gamma distribution of error terms provide PM estimates that are no better, and in many cases inferior to, standard linear regression methods. Thus, ACG System uses a standard linear regression approach with ordinary least squares estimation in order to provide risk factor weights (the beta coefficients from the regression equations) with intuitive, real-dollar meaning. Logistic Regression applied to dichotomous/binary outcomes events such as whether someone is hospitalized or is in a high-cost group (e.g., in the top five percent of the highest cost persons in the population). The output from logistic regression estimates the probability that the modeled event will occur. The probability is expressed as a decimal number from 0 to 1. Logistic regression assumes a specific functional form between predictors and the outcome (a logit relationship) that differs from typical linear models of health care resource use. Concurrent Cost Models A concurrent risk model, also commonly referred to as a retrospective model, is an explanatory model where the risk factors are derived from the same time period as the outcome variable. In the ACG System, concurrent models are used to explain historical costs. For example, a concurrent risk model may use patient experience from 2014 to derive risk factors (e.g., age-gender or ACG) in order to explain variation in 2014 total expenditures. Concurrent risk models are often applied in the context of performance assessment in order to derive the expected level of resource utilization. Learn more about the application of concurrent cost models in the Applications Guide. 110

111 Chapter 7: Risk Modeling Local Age-Gender Concurrent Risk There is an occasional need to compare a risk score with some reference standard. Scores based solely on age and gender are often used for this purpose. The software computes the reference standard using a locally-derived, concurrently calibrated measure of risk using the following age groups: 0-4 years 5-11 years years years years years years years years years years 85+ years The computational method uses an actuarial cell approach; refer to Statistical Approach on page 110. Patients are distributed into different risk cells based on their age and gender. The local age-gender concurrent risk is calculated as the mean cost of all patients in the age-gender cell divided by the mean cost of all patients in the population. Risk scores greater than 1.0 indicate that, on average, patients in that age-gender cell cost more than average while risk scores less than 1.0 indicate that patients in the age-gender cell cost less than average. ACG Concurrent Risk An ACG concurrent risk is an assessment of the relative resource use for individuals in the ACG applying an actuarial cell approach. The ACG concurrent risk is calculated as the mean cost of all patients in an ACG divided by the mean cost of all patients in the population. A fixed set of ACG concurrent risk weights derived from external reference data is available in the software output file as Reference ACG Concurrent Risk. Separate sets of weights exist for under 65 working age populations, over age 65 Medicare eligible populations (United States federal program that pays for certain health care expenses for people aged 65 or older), and all age populations. The weights used are determined by the Risk Assessment Variables selected during processing. The weights produced by the software are relative weights, i.e., relative to a population mean, and are standardized to a mean of 1.0. An individual weight is associated with each ACG and all individuals assigned the same ACG category will have the same ACG Concurrent Risk. In the case of an elderly reference set, the weights of some ACGs (e.g., those associated with pediatrics, pregnancy or newborns) may be extrapolated from an under 65 population. If the all age reference is selected, different weights will be applied to children (0-17), adults (0-64) and elderly (65+). The softwaresupplied weights may be considered a national reference or benchmark for comparisons with locally calibrated ACG-weights. In some instances (e.g., for those with limited or no cost data), these weights may also be used as a reasonable proxy for local cost data. The software-supplied Reference ACG Concurrent Risk is supplied in two forms: unscaled and rescaled. Unscaled Reference ACG Concurrent Risk represents the values of the Reference ACG Concurrent Risk 111

112 Chapter 7: Risk Modeling applied to a population of interest. If the mean of the Unscaled ACG Concurrent Risk is greater than 1.0 it indicates the rating population (the population to which the weights are being applied) is sicker than the reference population (the national reference database). If the mean is less than 1.0, it indicates the rating population is healthier. To ensure that dollars in the system are not over-estimated or under-estimated, the ACG System also makes available a Rescaled Reference ACG Concurrent Risk that mathematically manipulates the Unscaled Reference ACG Concurrent Risk to have a mean of 1.0 in the local population. The steps for performing this manually are discussed in the following "Rescale and Assign Monetary Values Process" section. Our experience indicates that ACG concurrent risk weights, especially when expressed as relative values, have remarkable stability. Where differences in ACG concurrent risk are present across healthcare organizations, it is almost universally attributable to differences in covered services reflected by different benefit levels and cost structures. If local cost data are available, the ACG Software also calculates Local ACG Concurrent Risk. These local weights more accurately reflect local benefit levels and area practice patterns. In general it is recommended that the reference population (on which the weights are developed) should be as similar as possible to the assessment population to which the weights are applied. Rescale and Assign Monetary Values Process The rescaling process consists of the following steps: Step 1: Compute population mean risk Compute a separate grand mean for each of the weights (either concurrent or predicted risk) generated for the client s population (the observations represent individuals). The mean for this example is shown at the bottom of Column B in the table following. Step 2: Apply weighting factor Divide each individual s risk by the rescaling factor (i.e., the mean) computed in Step 1. The result is the rescaled relative risk (Column C). Step 3: Compute population mean cost For the same population on which the weights were based, compute the mean cost for the current data year. For this example, the mean cost was 1, Step 4: Compute cost Multiply the rescaled relative weights generated for each member of the population (Column C) by the average population cost generated from Step 3 to calculate an estimated individual cost (Column D). A Member B Relative Weight C Rescaled Weight D Estimated Cost , ,

113 Chapter 7: Risk Modeling A Member B Relative Weight C Rescaled Weight D Estimated Cost , , Mean , The rescaling factor functions as a summary case-mix index for understanding how the rating population (e.g., local population) compares to the reference population. The interpretation of this factor is analogous to how one interprets both relative risks and profiling indicators. If the rescaling factor is greater than 1.0 (as it was in the example), then the client s population is sicker. If the factor is less than 1.0, then the client s population is healthier than the reference population. Concurrent Risk (Regression-based) Virtually since the inception of the ACG System, the ACG Concurrent Risk approach has been the sole means for concurrent adjustment of costs. Because of the more granular refinement offered by using linear regression, a regression-based concurrent risk score based on numerous risk factors with concurrent cost as the outcome, is now also provided. While the ACG Concurrent Risk is the same for all members within an ACG, there is actually a continuum of risk within that cell that can be detected by linear regression. The regression-based concurrent risk model uses the same family of risk factors used in the diagnosis-based predictive model plus a complement of acute conditions to account for potentially costly conditions of limited duration. The model uses standard coefficient weights derived from the selected reference population. These weights are then applied to each patient based on the specific risk factors identified by the ACG System. The weights are additive at the patient level as in the following example. Predictive Modeling Factor Coefficient Demographic Age: Gender: F Diagnosis-based Markers 3 Hospital Dominant Morbidity Types ACG ADGs, 3 Major END09 - Type 1 Diabetes w/complication CAR14 - Hypertension w/o complications Concurrent Risk (regression-based) = sum of coefficients As with the Reference ACG Concurrent Risk, the Concurrent Risk is supplied in two forms: unscaled and rescaled. Unscaled Concurrent Risk represents the risk weights of the Concurrent Risk regression model applied to a population of interest. If the mean of the Unscaled Concurrent Risk is greater than 1.0 it indicates the rating population (the population to which the weights are being applied) is sicker than the reference population (the national reference database). If the mean is less than 1.0, it indicates the rating population is healthier. To ensure that dollars in the system are not over-estimated or under- estimated, the ACG System also makes available a Rescaled Concurrent Risk that mathematically manipulates the Unscaled Concurrent Risk to have a mean of 1.0 in the local population. The steps for performing this 113

114 Chapter 7: Risk Modeling manually are discussed in Rescale and Assign Monetary Values Process; refer to Rescale and Assign Monetary Values Process on page 112. Concurrent Model Performance Assessment and Validation In creating regression models it is customary to divide the data randomly into two parts, a test dataset and a validation dataset. The models are designed using the test dataset, and then validated on the other dataset to avoid statistical over-fitting of the model. The typical statistics used to validate concurrent cost models are Adjusted R-Squared as a measure of individual variation explained and Expected to Actual ratios as a measure of population variation explained. R-Squared The conventional measure of model performance is the R-Squared statistic. This statistic measures how well the model fits the data and has become a standard measure of performance, especially among underwriters and actuaries who must price products across a range of populations. The R-squared value, expressed as a decimal number between 0 and 1, is interpreted as how much variance around the mean is explained by the model. A perfect model would explain all the variation and the R-squared would be 1. This, of course, is a rare occurrence and, in fact, suggests that there might be some design defect in the model. Outlier values can diminish the R-squared performance and are sometimes truncated. Model performance can also be improved by focusing only on those persons who have high values of the modeled outcome where variation is also high. A good fit will occur although the model may actually not work as well for persons who are not in the highest group. The following table provides the R-squared values for various concurrent model options presented in the ACG System using the all age reference population. R-Squared Performance for ACG System Concurrent Risk Models R-Squared Modeling Total Cost R-Squared Modeling Total Cost without Truncation Truncated at $250,000 Local Age-Gender Risk Local ACG Concurrent Risk Reference ACG Concurrent Risk Concurrent Risk (regression-based) Source: PharMetrics, Inc., a unit of IMS, Watertown, MA; national cross-section of managed care plans; population of 3,310,540 commercially insured lives (less than 65 years old) and population of 501,987 Medicare beneficiaries (65 years and older). Data for 2010 were used for model development, and data for 2009 were used for model validation. Expected to Actual Cost Ratios The credibility of models does not solely rest on regression fit, which is a measure of performance at the individual level. To determine how well models perform at estimating cost across the full range of risk, we also recommend looking at persons in different cost cohorts (we use quintiles). Expected to Actual Cost Ratios are another means for assessing risk models. The ratio is modeled outcome to what is actually observed. In the case of costs, one might array the population into different cost quintiles and consider the expected to actual cost ratios for each of these quintiles. The goal is to 114

115 Chapter 7: Risk Modeling approach 1.0 as closely as possible, especially in the high cost quintile. Ratios less than 1.0 mean that the model under-estimates costs while ratios over 1.0 suggest that the model is over-estimating costs. The following table shows these ratios for the various concurrent model options presented in the ACG System using the all age reference population. Expected to Actual Cost Ratios by Cost Quintile for ACG System Concurrent Risk Models Local ACG Reference ACG Concurrent Risk Local Age-Gender Concurrent Risk Cost Concurrent Risk Cost (regression-based) Risk Cost Ratio Ratio Ratio Cost Ratio Top 1% Top 5% Top 20% Mid-High Mid Low-Mid Bottom 20% Source: PharMetrics, Inc., a unit of IMS, Watertown, MA; national cross-section of managed care plans; population of 3,310,540 commercially insured lives (less than 65 years old) and population of 501,987 Medicare beneficiaries (65 years and older). Data for 2010 were used for model development, and data for 2009 were used for model validation. There is a tendency for all of the models to somewhat overestimate costs among the lowest cost quintile and somewhat underestimate costs in the highest cost quintile. This effect is most pronounced in the age and gender model. The various morbidity-based ACG System concurrent models all substantially outperform age and gender. Deciding between ACG Concurrent Risk or Regression-based Concurrent Risk The ACG method of determining concurrent risk using the mean costs within an ACG category remains a mainstay of risk adjustment. The method is conceptually simple and can allow for further calibration on local cost data or other outcomes, such as visits. The ACG method can even be used for relatively small populations by aggregating similar ACG categories into Resource Utilization Bands (RUBs). This method has been validated in numerous settings and remains both stable and resistant to strategic manipulation. The ACG method is not without challenges. In particular, using ACG categories as a basis of risk adjustment does not account for within-cell variation. When a biased population is evaluated, this variation can be distributed disproportionately. For example, if a primary care practice based at a tertiary care center receives referrals for complex cases, the caseload is likely to look quite different (i.e., the patients are likely to have higher health service needs) than a community-based primary care practice, even after adjusting for differences in ACG category. A regression-based approach to concurrent risk may improve the risk assessment for biased populations by creating a more refined prediction for each individual member. This improves statistical performance, specifically for higher risk patients. The improvement in statistical performance for high risk patients is offset by a tendency for regression-based models to over-predict low risk patients. 115

116 Chapter 7: Risk Modeling While statistical performance is improved, there are trade-offs to the regression-based approach as well. Regression-based models can be more difficult to explain and with unique scores for individual patients may be interpreted with a false sense of precision. The model is built on an external population and may not reflect the local practice patterns or cost structures of the population being evaluated. Likewise, a regression model explaining total cost does not readily translate to assessing other outcomes, such as visits. Because each new risk factor discovered will increase the risk score for the patient, regression-based approaches are potentially more susceptible to strategic manipulation. Ultimately, the decision to use an ACG categorical approach to risk adjustment or a regression-based approach to risk adjustment must consider the audience and intended use. If statistical performance for high risk members is critically important, then the regression-based approach may be most appropriate. However, for many applications, especially with the introduction of incentives, the ACG categorical approach may be equally effective without creating opportunities for strategic manipulation. Many of the perceived weaknesses of the ACG categorical approach may be overcome with thoughtful application of the risk adjustment process. In the context of performance assessment, please see Chapter 4 of the Applications Guide for more specific considerations when profiling. One of the unique strengths of the ACG model is the ability to derive local calibrations. For more information, see Local Calibration of ACG Risk Models on page 122. Prospective Cost Models A prospective risk model is a predictive model where the risk factors are derived from the period prior to the assessment of outcome variable. In the ACG System, prospective models are used to predict both the anticipated cost level and to predict those members that are likely to experience high costs. For example, a prospective risk model may use patient experience from 2014 to derive risk factors in order to explain anticipated 2015 total expenditures. Prospective risk models are often applied in the context of population stratification and high risk case identification or in the setting of premium rates as part of the underwriting process. Learn more about the application of prospective cost models in the Applications Guide. Predicted Total/Pharmacy Cost Risk The predicted risk outputs from the ACG System represent the relative risk for a specific outcome, either total cost in the 12 months subsequent to the observation period or pharmacy cost in the 12 months subsequent to the observation period. The predicted risk is the result of a linear regression utilizing risk factors derived from either diagnosis data, pharmacy data, or both, and either exclusive or inclusive of prior cost. The model uses standard coefficient weights derived from the selected reference population. These weights are then applied to each patient based on the specific risk factors identified by the ACG System. The weights are additive at the patient level as in the following example. Predictive Modeling Factor Coefficient Demographic Age: Gender: F

117 Chapter 7: Risk Modeling Predictive Modeling Factor Coefficient Diagnosis-based Markers 2 Hospital Dominant Morbidity Types ACG ADGs, 3 Major END09 - Type 1 Diabetes w/complication CAR15 - Hypertension w/ complications REN01 Chronic Renal Failure Pharmacy-based Markers CARx030 Cardiovascular/High Blood Pressure INFx010 Infections/Acute Major Cost Percentile Markers Total Cost 51-75th percentile Unscaled Predicted Risk = sum of coefficients As with the concurrent models, the Predicted Risk (total cost or pharmacy cost) is supplied in two forms: unscaled and rescaled. Unscaled Predicted Risk represents the risk weights of the Predicted Risk regression model applied to a population of interest. If the mean of the Unscaled Predicted Risk is greater than 1.0 it indicates the rating population (the population to which the weights are being applied) is sicker than the reference population (the national reference database). If the mean is less than 1.0, it indicates the rating population is healthier. To ensure that dollars in the system are not over-estimated or underestimated, the ACG System also makes available a Rescaled Predicted Risk that mathematically manipulates the Unscaled Predicted Risk to have a mean of 1.0 in the local population. The steps for performing this manually are discussed in Rescale and Assign Monetary Values Process (see Rescale and Assign Monetary Values Process on page 112). Rank Probability High Total/Pharmacy Cost While predicted risk is useful in providing a cost estimate, differences in the mean across populations can change the shape of the predicted risk distribution. This creates a challenge when trying to establish a threshold for predicting high cost members. The rank probability of high total cost and rank probability of high pharmacy cost represent transformations of the predicted risk score into a standardized distribution. The predicted risk is sorted and arrayed into 1000 quantile groups. The rank probability is an estimate of the likelihood of the individual having costs in the top 5% of the population in the subsequent period given the quantile group the predicted risk falls into. Because the methodology is looking at 1000 quantiles, it is possible for many individuals to have the same rank probability score. The output is suppressed for populations less than 100. The result will always yield a uniform distribution, regardless of the underlying morbidity of the population. This can be useful when trying to manage the top 1-5% of members. Reference Probability High Total/Pharmacy Cost The reference probability high total cost and reference probability high pharmacy cost are derived from a logistic regression where the dependent variable is patients in the top 5% of costs in the subsequent period. The model takes the form of a logistic regression, meaning model coefficients from the selected reference population are added and the final sum transformed to a probability as in the example following. 117

118 Chapter 7: Risk Modeling Predictive Modeling Factor Coefficient Demographic Age: Gender: F Diagnosis-based Markers 2 Hospital Dominant Morbidity Types ACG ADGs, 3 Major END09 - Type 1 Diabetes w/complication CAR15 - Hypertension w/ complications REN01 Chronic Renal Failure Pharmacy-based Markers CARx030 Cardiovascular/High Blood Pressure INFx010 Infections/Acute Major Cost Percentile Markers Total Cost percentile Sum of coefficients Probability of high total cost = As noted above, the output from logistic regression estimates the probability that the modeled event will occur. The probability is expressed as a decimal number from 0 to 1. Although a cost model that predicts a continuous outcome and one that predicts a dichotomous outcome may use the same set of predictor variables, it is possible for there to be slight discrepancies between who gets identified as the highest risk because of the differences in how the outcomes are defined. The logistic model is predicting those at high risk while the linear model is predicting cost broadly. These discrepancies have been noted in the past and, as a consequence, the rank probability score is based on the distribution of linear risk scores indicating the likelihood that the individual will be high risk relative to the study population while the reference probability score indicates the likelihood that the individual will be high risk relative to the reference population. Prospective Model Performance Assessment and Validation R-Squared The R-squared value is a measure of model performance, expressed as a decimal number between 0 and 1, and interpreted as how much variance around the mean is explained by the model. For more information, see Concurrent Model Performance Assessment and Validation on page 114. The following table provides the R-squared values for the total cost and pharmacy cost predictive model options presented in the ACG System using the all age reference population and diagnosis, pharmacy and prior cost as model inputs. R-Squared Total R-Squared Rx R-Squared Total Cost truncated to R-Squared Total Cost Cost truncated to Cost 250K truncated to 100K R-Squared Rx Cost 50K

119 Chapter 7: Risk Modeling Source: PharMetrics, Inc., a unit of IMS, Watertown, MA; national cross-section of managed care plans; population of 3,310,540 commercially insured lives (less than 65 years old) and population of 501,987 Medicare beneficiaries (65 years and older). Data for were used for model development, and data for were used for model validation. Predictive Ratios Predictive Ratios, the ratio of predicted to actual cost, are another means for assessing predictive risk models. In the case of costs, one might array the population into different cost quintiles and consider the predicted to actual cost ratios for each of these quintiles. The goal is to approach 1.0 as closely as possible, especially in the high cost quintile. Ratios less than 1.0 mean that the model under-estimates costs while ratios over 1.0 suggest that the model is over-estimating costs. The following table shows these ratios for the predictive model presented in the ACG System using the all age reference population and diagnosis, pharmacy and prior cost as model inputs. Expected to Actual Cost Ratios by Cost Quintile for ACG System Predictive Risk Models Quantile Predictive Ratio Top 1% 0.24 Top 5% 0.38 Top 20% 0.60 Mid-High 1.78 Mid 2.90 Low-Mid 5.70 Bottom 20% Source: PharMetrics, Inc., a unit of IMS, Watertown, MA; national cross-section of managed care plans; population of 3,310,540 commercially insured lives (less than 65 years old) and population of 501,987 Medicare beneficiaries (65 years and older). Data for were used for model development, and data for were used for model validation. Sensitivity and Positive Predictive Value For logistic regression, performance is assessed based upon how well the occurrence of the modeled event matches the actual data. Some measures include: Sensitivity Specificity Positive Predictive Value (PPV) Negative Predictive Value (NPV) The sensitivity is the number of predictions with a true outcome identified by the model at a specific cutpoint (true positives) as a percent of all positive outcomes (true positives plus false negatives). Conversely, the specificity represents the number of negative predictions identified by the model (true negatives) as a percent of all negative outcomes (false positives plus true negatives). The PPV represents the ratio of positive predictions identified by the model (true positives) to all positive outcomes at a specific cut-point (true positives plus false positives). Finally, NPV is the proportion of negative predictions (true negatives) as a percent of all negative outcomes (false positives plus true negatives). 119

120 Chapter 7: Risk Modeling A classification accuracy measure can also be calculated as the proportion of true predictions (true positives + true negatives) to all predictions. Prediction Outcome Positive Negative Positive True Positive False Positive Negative False Negative True Negative Sensitivity is very salient when you are concerned about the prevalence of an event in the target population. PPV is useful when you want to intervene on someone with the identified event (e.g, a disease). It expresses the likelihood that this person actually has the targeted event. The following table shows these ratios for the rank probability high total cost and reference probability high total cost models presented in the ACG System using the all age reference population and diagnosis, pharmacy and prior cost as model inputs. Patients with year 2 costs in the top 5% of the population were considered a positive outcome. Patients with a prediction in the top 5% of the population were considered a positive prediction. Classification Accuracy of the ACG System Probability High Total Cost Models Rank Probability High Total Cost Reference Probability High Total Cost Sensitivity 39.8% 40.1% Specificity 96.8% 96.8% Positive Predictive Value 39.8% 40.1% Negative Predictive Value 96.8% 96.8% Classification Accuracy 94.0% 94.0% Source: PharMetrics, Inc., a unit of IMS, Watertown, MA; national cross-section of managed care plans; population of 3,310,540 commercially insured lives (less than 65 years old) and population of 501,987 Medicare beneficiaries (65 years and older). Data for were used for model development, and data for were used for model validation. Additional Probability Models Probability of Unexpected Pharmacy Cost/High Risk Unexpected Pharmacy Cost Traditional predictive modeling focuses on using diagnoses and/or pharmacy information to predict future expenditures. Current models work well and have provided a means of identifying individuals at risk for future high pharmacy expenditures. However, a limitation of the current approach is that these models, by definition, have tended to identify a multi-morbid population whose high use of pharmacy services may well be justified; or, more specifically, their disease profile warrants the high use of medications. The High Risk for Unexpected Pharmacy Cost model is specifically designed to address this shortcoming and is focused on predicting the subset of the multi-morbid population who are consuming drugs above and beyond what might be anticipated based on their diagnosis-based morbidity burden. 120

121 Chapter 7: Risk Modeling The probability of unexpected pharmacy cost is a numerical probability score representing the result of applying weights from a logistic regression model on the development data set predicting individuals with moderate or high morbidity with unusually large pharmacy expenditures. The markers in the model are the same as the predictive model using diagnosis and pharmacy inputs without prior cost (see Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models on page 129). In the development data set individuals were deemed to have moderate or multi-morbidity if they were assigned to a moderate or higher resource utilization band and they were considered to have unusually large pharmacy expenditures if their standardized pharmacy expenditures were more than 1.75 standard deviations from their predicted amount based on diagnosis information. High Risk for Unexpected Pharmacy Cost is a binary flag that indicates individuals with a Probability of Unexpected Pharmacy Cost greater than 0.4. Objectives for success for the High Risk for Unexpected Pharmacy Cost Model included validation statistics similar to those detailed for the traditional predictive models; however, to warrant inclusion of an additional model in the ACG System, the High Risk for Unexpected Pharmacy Cost Model needed to identify a different group of individuals than traditional predictive modeling methods. The following table summarizes the overlap of the High Risk for Unexpected Pharmacy Cost Model and Predicted Risk from diagnosis inputs to identify future high pharmacy users with moderate morbidity burden or higher. This table indicates that the overlap between the two models was only 16%. In short, the High Risk for Unexpected Pharmacy Cost Model is identifying a new group of patients not identified by prior methods. Overlap of the High Risk for Unexpected Pharmacy Cost Model and Diagnosis-Based Predicted Risk (Top 1.5% of Scores) Identified by the High Identified by Diagnosis-Based Identified by both the High Total Risk for Unexpected Predicted Risk only Risk for Unexpected Pharmacy Cost Model Pharmacy Cost Model and only Diagnosis-based Predicted Risk 20,722 20,709 7,822 49,253 Total 42.07% 42.05% 15.88% % There are a variety of reasons why this population subgroup may be of interest and explanations as to why pharmacy utilization might be higher than expected. Sometimes pharmacy costs are higher because of the use of an expensive medication. Other times unanticipated high drug costs may be related to quality issues (i.e., poor or uncoordinated care). Unexpected high pharmacy utilization can also be a result of both cost and quality issues. These can range from poor data (i.e., missing diagnosis codes), to poor or inadequate care and/or gaps in care. Other times it is simply a matter of polypharmacy and/or patient abuse. The hope is that this model can help case managers to: Better identify those at risk for future high pharmacy utilization, and Provide insights that might allow for providing better coordination or refining the patient s drug regimen. Probability of Persistent High User A common phenomenon associated with predictive modeling is that someone who is high cost one year will have lower costs the next year (i.e., regression to the mean). Relatively few studies have characterized and/or developed predictive models to identify persistent high cost users, people with high medical costs consistently across multiple consecutive time periods; these prior studies suggest that people with chronic diseases (such as diabetes) and psycho-social conditions (such as mental illness) are 121

122 Chapter 7: Risk Modeling more likely to be a persistent high cost user. These patients do not experience regression to the mean and do represent an ongoing cost burden to organizations responsible for their care. The persistent high user model uses a logistic regression to predict individuals likely to be in the top 20% of users in the population for four consecutive half year periods following the assessment of risk factors. Model performance was measured using sensitivity and positive predictive value (PPV) (see Sensitivity and Positive Predictive Value on page 119). The specialized persistent high user model outperforms the more generic high cost model in identifying this unique set of members that do not regress to the mean over multiple years. Sensitivity and Positive Predictive Value for the Persistent High User Model compared to Rank Probability High Total Cost True Persistent High Users True High Users (top (top 20% in four 6.14% in Year 2 to keep consecutive 6 month same proportion Threshold Indicators periods of years 2 and 3) identified) % of Total Sample 6.14% 6.14% Top 5% of Probability Sensitivity 49.7% 37.8% Persistent High User PPV 61.1% 37.3% Top 5% of Rank Sensitivity 39.0% 36.7% Probability High Total Cost PPV 47.9% 36.6% Local Calibration of ACG Risk Models The standard model outputs provided in the ACG System are based on a large reference database based on commercial managed care plans in the U.S. We recommend that users attempt local calibration if they feel that their population is likely to differ significantly from the reference population and outcome data is available. To develop a locally-based score would involve fitting a regression to local data using the variables included within the ACG Models. A listing of the predictor variables (the independent variables) is provided in Appendix A: Variables Necessary to Locally Calibrate the ACG Predictive Models on page 129. Using these variables and local cost data, an experienced analyst could develop a new set of scores that are customized for the local population. Custom models should be based on populations of no fewer than 100,000 individuals. 122

123 Chapter 8: Predictive Models for Hospitalization Chapter 8: Predictive Models for Hospitalization Inpatient care is part of the healthcare system in every country. The ability to predict hospitalizations with accuracy improves the allocation of healthcare resources in budgetary systems, and in systems that link financial incentives to efficiency and quality of care. The value of a model-based screening strategy is to identify high risk individuals for population health interventions. Typology of Acute Care Hospitalizations Acute care hospitalizations are clinical events that are potentially predictable. For example, admissions that are attributed to congestive heart failure can have a medical history leading up to the event. Other hospitalizations are potentially predictable based on prior medical history and a set of utilization markers. Unplanned hospitalizations may also have an external cause such as poisonings and injuries suffered in accidents. These hospitalizations are all a focus of the ACG predictive modeling. Admissions for acute illness or for complications of care are unplanned. Planned admissions are those in which a planned inpatient procedure occurs. A set of surgical and medical inpatient procedures that are considered to be planned is used by the U.S. Centers for Medicare and Medicaid Services (CMS) for quality assessment purposes 10. We have adapted the CMS definition of planned admission to include childbirth. The following figure gives an overview over a guiding typology of hospitalizations. Figure 22. Typology of Acute Care Inpatient Hospitalization How the ACG System Predicts Hospitalization The ACG System uses a tiered approach to making predictions about hospitalizations. The predictions about future hospitalization consider whether a person had a hospitalization in the prior time period. Persons who have had a prior hospitalization empirically have an increased likelihood of hospitalization in a subsequent time period. In the absence of a prior hospitalization, a secondary criterion for making predictions about hospitalization relates to the age of a person. Older persons are thought to have an 10 Hospital-wide All-cause Unplanned Readmission (HWR) Measure, see 123

124 Chapter 8: Predictive Models for Hospitalization increased likelihood of hospitalization compared to younger individuals. The ACG System uses age 55 as the threshold for separating two hospitalization risk groups. The hospitalization prediction model differs from the Hospital Dominant Morbidity Types marker. Hospital Dominant Morbidity Types represent a small subset of diagnoses associated with high rates of admission in the following 12 months. The hospital prediction model identifies a larger pool of patients at risk for hospitalization. Predictive Models for Hospitalization include Utilization Measures The ACG predictive model framework has been amended to improve the prediction of hospitalization events. The framework for predicting hospitalization includes several utilization markers which represent a plausible set of valid predictors. For example, emergency care and inpatient hospitalization can be precursors to future inpatient hospitalization. Several utilization measures support the calculation of hospitalization risk. These measures use diagnosis codes, dates of service, place of service, procedures, and revenue codes in the Medical Services Input File. Alternatively, users may provide these markers in the Patient Input File. Dialysis Service The dialysis service marker indicates that a patient with chronic renal failure has received a dialysis service during the observation period. For more information, see Dialysis Service Marker on page 90. Nursing Service The nursing service marker indicates that a patient has used services in a nursing home, domiciliary, rest home, assisted living, or custodial care facility during the observation period. For more information, see Nursing Service Marker on page 91. Active Cancer Treatment The active cancer treatment marker indicates that a malignancy EDC was present and chemotherapy and/or radiation therapy was performed during the observation period. This marker is useful for measuring hospitalization risk, because hospitalization rates for cancer patients are high during the active treatment phase. For more information, see Cancer Treatment Marker on page 91. Psychotherapy The psychotherapy marker identifies the presence of outpatient therapeutic treatments of mental health conditions. The focus is on ambulatory treatment as a proxy for ongoing management. This marker is useful for measuring hospitalization risk, because hospitalization rates for mental health patients are high during active psychotherapy. For more information, see Psychotherapy Service Marker on page 92. Mechanical Ventilation The mechanical ventilation marker identifies the presence of mechanical ventilation through tracheostomy or ventilator support. Mechanical ventilation carries many potential complications that lead to hospitalization, including lung injury, infection and respiratory distress syndromes. For more information, see Mechanical Ventilation Marker on page

125 Chapter 8: Predictive Models for Hospitalization Compassionate Care Allowances (CAL-SSA) The CAL-SSA marker identifies the presence of a compassionate allowance condition defined by the U.S. Social Security Administration as an impairment that invariably meets disability standards. This marker is useful for measuring hospitalization risk, because hospitalization rates for patients with disabilities are high. For more information, see Compassionate Care Allowances (CAL-SSA) on page 58. Emergency Visit Count This count is an integer count of emergency room visits that did not lead to a subsequent acute care inpatient hospitalization during the observation period. For more information, see Emergency Visit Count on page 90. Inpatient Hospitalization Count The inpatient hospitalization count is a count of inpatient confinements during the observation period. The count of inpatient hospitalizations excludes admissions with a primary diagnosis for pregnancy, delivery, newborns, and injuries. Transfers made within and between providers count as a single hospitalization event. For more information, see Inpatient Hospitalization Count on page 88. Inpatient Hospitalization Days Count The inpatient hospitalization days count is a count of inpatient days associated with each inpatient confinement in the inpatient hospitalization count. This count is a proxy for the intensity of inpatient care, and useful in the context of assessing readmission risk. For more information, see Inpatient Hospitalization Days on page 89. Unplanned Readmission 30 Day Count The unplanned 30 day readmission count is the number of unplanned acute care inpatient hospitalizations within 30 days of a previous inpatient hospitalization. For more information, see Unplanned Readmission 30 Day Count on page 89. This count excludes any planned readmission based on inpatient procedures that are likely to have been planned in advance unless there is contradicting diagnosis of an acute condition or complication of care. Outpatient Visit Count The outpatient visit count is a count of outpatient encounters in the physician office and selected outpatient facilities. Visits are counted a maximum of once per date of service and provider. Higher rates of outpatient encounters are associated with higher rates of hospitalization. For more information, see Outpatient Visit Count on page 90. Likelihood of Hospitalization There are five predictive model outputs related to the likelihood of hospitalization. These models are intended to be used for the indicated outcome. A value of providing multiple model outputs is greater sensitivity of each model calibrated to a particular outcome, as compared to using a single model. Probability scores indicating the likelihood of a future hospitalization event are generated. A discussion of the model outputs follows. 125

126 Chapter 8: Predictive Models for Hospitalization Probability IP Hospitalization Score The probability IP hospitalization score is the probability score for an acute care inpatient hospital admission within the 12 months subsequent to the observation period. Probability IP Hospitalization Six Months Score The probability IP hospitalization six months score is the probability score for an acute care inpatient hospital admission within the six months subsequent to the observation period. Probability ICU Hospitalization Score The probability ICU hospitalization score is the probability score for an Intensive Care Unit or Critical Care Unit admission within the 12 months subsequent to the observation period. Probability Injury Hospitalization Score The probability injury hospitalization score is the probability score for an injury-related admission within the 12 months subsequent to the observation period. Probability Extended Hospitalization Score The probability extended hospitalization score is the probability score for being admitted to an acute care hospital for 12 or more days (across one or more admissions) within the 12 months subsequent to the observation period. Empiric Validation of the Likelihood of Hospitalization Model A longitudinal healthcare claims database was used for the ACG model development and validation (see Reference Population on page 107). The database includes commercial and Medicare managed care plans. Data for the period were used for model development, and data for the period were used for model validation. Persons included in the analysis had at least six months of enrollment in the first year and one or more months of enrollment in the second year. The validation database included 3.5 million persons under age 65 and nearly half a million Medicare-eligible (a U.S. government program that pays for healthcare expenses for people for people aged 65 and older) beneficiaries aged 65 or older. Sensitivity and Positive Predictive Value Model performance is measured by how well true cases of admission are prospectively identified and false positives are avoided. The focus is on two key indicators, sensitivity and positive predictive value: Sensitivity = True Cases Identified/True Cases in Population Positive Predictive Value (PPV) = True Cases Identified/(True Cases Identified + False Positives) Sensitivity measures how well the model predicts which individuals are hospitalized. Positive predictive value measures the likelihood that a particular patient is hospitalized in the next time period. The following table illustrates, using sensitivity and positive predictive value measures, that the ACG models are strong performers compared to using prior cost alone. The top 5% of hospitalization probability scores identify cases. 126

127 Chapter 8: Predictive Models for Hospitalization PPV and Sensitivity for Predictive Hospitalization Models Predictive Model Positive Predictive Value Sensitivity IP Hospitalization with prior cost and diagnosis 33.3% 21.2% and pharmacy data input IP Hospitalization with prior cost and diagnosis 32.6% 20.8% data input Prior cost only % 14.2% C-Statistic The C-Statistic is a measure of model fit. A C-Statistic of 0.5 indicates that true cases are indistinguishable from false positives. A C-Statistic of 0.7 is widely accepted as a threshold for good model performance. C-Statistics for models that predict hospitalization are summarized in the following table. The table presents results for each of the models and population groups included in the tiered approach to making predictions about hospitalization. The predictive model is IP Hospitalization with predictors that are based on prior cost and diagnosis and pharmacy data. C-Statistics for Predictive Hospitalization Models Persons Aged less than 55 Persons Aged 55 or older Persons with Prior without Prior without Prior Predictive Model Hospitalization Hospitalization Hospitalization IP Hospitalization IP Hospitalization Six Months ICU Hospitalization Injury Hospitalization Extended Hospitalization The C-Statistics for predicting hospitalization are above 0.7 for most outcomes and groups. Plausible patterns emerge from the empiric validation. First, ACG System models better predict hospitalization events for persons who had a hospitalization within the previous 12 months, compared to persons who did not. Second, the ACG System better predicts specialized hospitalization events, compared to general hospitalization events. This finding validates the fact that optimal performance of an ACG System predictive model occurs when the weighting of risk factors employs the outcome that is being predicted. Likelihood of Unplanned Readmission within 30 Days Hospital financing increasingly considers the efficiency and quality of inpatient care. One area of interest in this context is the rate at which patients are readmitted within a short period after previously having been discharged. Higher than expected 30 day readmission rates may be perceived as a signal of lower quality of care. 11 The prior cost only model serves as a benchmark to the IP Hospitalization model. 127

128 Chapter 8: Predictive Models for Hospitalization A tool for identifying individuals at risk for readmission is valuable, because it enables a prioritization of discharge interventions among patients who are leaving a hospital. Planned admissions may entail a comprehensive discharge program that includes patient education, medication reconciliation, scheduled outpatient visits, and healthcare worker outreach or monitoring. These interventions are intended to reduce readmission risk. The unplanned 30 day readmission model is calibrated so that scores are available at the time of admission to the hospital. Patients are not required to have a prior hospitalization in order to receive a risk score for unplanned readmission. This enables healthcare delivery systems to anticipate discharge planning needs for patients at risk of readmission in the event of an admission. Probability Unplanned 30 Day Readmission Score The probability unplanned 30 day readmission score is the probability score for an acute care inpatient hospital readmission within that time frame in the event of an admission. Empiric Validation of the Readmission Model The C-Statistics for predicting 30 day unplanned readmission was.735 for the predictive model with diagnosis and pharmacy data input. The following table illustrates, using sensitivity and positive predictive value measures, the relationship between identifying larger groups of patients at risk of readmission with high sensitivity, and smaller groups of patients with improved positive predictive values. Patients at Risk of Unplanned 30 Day Readmission Sensitivity Positive Predictive Value Top 20% Scorers in Population 52% 16% Top 10% 39% 24% Top 5% 29% 36% Top 2% 15% 49% Top 1% 8% 53% 128

129 Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models Introduction These tables indicate the variables used in the ACG System standard models. The ACG System standard models have been developed using a reference population provided by PharMetrics, a unit of IMS in Watertown, MA. Modest performance improvements may be attained when predictive models are calibrated against a local population. The ACG software facilitates local calibration of predictive models with the export of the independent variables as model markers. For reference, the independent variables used in the ACG System standard models are presented in the following tables. U.S. and international copyright law protect these assignment algorithms. It is an infringement of copyright law to develop any derivative product based on the grouping algorithm or other information presented in this document without the written permission of The Johns Hopkins University. Copyright The Johns Hopkins University, 1995, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2007, 2008, 2009, 2011, and All Rights Reserved. Individual ACG Categories Included in the Risk Models ACG Description 0400* Acute Major 0500* Likely to Recur, w/o Allergies 0600* Likely to Recur, w/ Allergies Other ADG Combinations, Age 1-17, 1+ Major ADGs Other ADG Combinations, Age 18-44, 2+ Major ADGs Other ADG Combinations, Age >44, 1 Major ADGs Other ADG Combinations, Age >44, 2+ Major ADGs Other ADG Combinations, Age 1-5, No Major ADGs Other ADG Combinations, Age 1-5, 1+ Major ADGs Other ADG Combinations, Age 6-17, No Major ADGs Other ADG Combinations, Age 6-17, 1+ Major ADGs Other ADG Combinations, Male, Age 18-34, 2+ Major ADGs Other ADG Combinations, Female, Age 18-34, 2+ Major ADGs Other ADG Combinations, Age >34, 0-1 Major ADGs 129

130 Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models ACG Description Other ADG Combinations, Age >34, 2 Major ADGs Other ADG Combinations, Age >34, 3 Major ADGs Other ADG Combinations, Age >34, 4+ Major ADGs Other ADG Combinations, Age 1-17 No Major ADGs Other ADG Combinations, Age 1-17, 1 Major ADGs Other ADG Combinations, Age 1-17, 2+ Major ADGs Other ADG Combinations, Age 18+, 0-1 Major ADGs Other ADG Combinations, Age 18+, 2 Major ADGs Other ADG Combinations, Age 18+, 3 Major ADGs Other ADG Combinations, Age 18+, 4+ Major ADGs 5320 Infants: 0-5 ADGs, 1+ Major ADGs 5321 Infants: 0-5 ADGs, 1+ Major ADGs, low birth weight 5322 Infants: 0-5 ADGs, 1+ Major ADGs, normal birth weight 5330 Infants: 6+ ADGs, No Major ADGs 5331 Infants: 6+ ADGs, No Major ADGs, low birth weight 5332 Infants: 6+ ADGs, No Major ADGs, normal birth weight 5340 Infants: 6+ ADGs, 1+ Major ADG 5341 Infants: 6+ ADGs, 1+ Major ADG, low birth weight 5342 Infants: 6+ ADGs, 1+ Major ADG, normal birth weight * - This subset of ACG categories is used in the concurrent cost models only. ACGs Included in Three Resource Groups ACG Resource Groups Prospective Group Concurrent Group ACG RUB Level 1 ACG ACG (included in Intercept) ACG RUB Level 2 ACG ACG ACG RUB Level 3 ACG ACG , 5311,

131 Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models Pregnancy Without Delivery Delivered All non-delivered pregnancy related ACGs (1712, 1722, 1732, 1742, 1752, 1762 and 1772). This marker further evaluates pregnancy ACGs to exclude terminated pregnancies identified by the Expanded Diagnosis Cluster (EDC) FRE14 and to include pregnancies subsequent to a delivery. All delivered pregnancy related ACGs (1711, 1721, 1731, 1741, 1751, 1761 and This marker is used in concurrent cost models only. EDCs Included in the Risk Models EDC ADM02* ADM03 ADM05* ALL04 ALL05 ALL06 CAR01* CAR03 CAR04 CAR05 CAR06 CAR07 CAR08* CAR09 CAR10 CAR12 CAR13 CAR14 CAR15 DEN04* END02 END03* Description Surgical Aftercare Transplant Status Administrative concerns and non-specific laboratory abnormalities Asthma, w/o status asthmaticus Asthma, with status asthmaticus Disorders of the Immune System Cardiovascular Signs and Symptoms Ischemic heart disease (excluding acute myocardial infarction) Congenital heart disease Congestive heart failure Cardiac valve disorders Cardiomyopathy Heart Murmur Cardiac arrhythmia Generalized atherosclerosis Acute Myocardial Infarction Cardiac arrest, shock Hypertension, w/o major complications Hypertension, with major complications Stomatitis Osteoporosis Short Stature 131

132 Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models EDC END06 END07 END08 END09 EYE03 EYE08* EYE11* EYE12* EYE13 EYE15 FRE03 FRE05 FRE12 GAS02 GAS04 GAS05 GAS06 GAS09* GAS10 GAS11 GAS12 GSI02* GSI04* GSI06* GSI07* GSI08 GSU02* GSU03* GSU04* GSU05* GSU08* Description Type 2 Diabetes, w/o Complication Type 2 Diabetes, w/ Complication Type 1 Diabetes, w/o Complication Type 1 Diabetes, w/ Complication Retinal disorders (excluding diabetic retinopathy) Glaucoma Strabismus, amblyopia Traumatic injuries of eye Diabetic Retinopathy Age-related Macular Degeneration Endometriosis Female Infertility Utero-vaginal prolapse Inflammatory bowel disease Acute hepatitis Chronic liver disease Peptic ulcer disease Irritable Bowel Syndrome Diverticular disease of colon Acute pancreatitis Chronic pancreatitis Chest Pain Syncope Debility and undue fatigue Lymphadenopathy Edema Appendicitis Benign and unspecified neoplasm Cholelithiasis, cholecystitis External abdominal hernias, hydroceles Varicose veins of lower extremities 132

133 Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models EDC GSU09* GSU10* GSU11 GSU13 GSU14 GSU15* GTC01 GTC02* GUR04 GUR09* HEM01 HEM02* HEM03 HEM05 HEM06 HEM07 HEM09 INF01* INF04 INF07* INF08 MAL01* MAL02 MAL03 MAL04 MAL05* MAL06 MAL07 MAL08 MAL09 MAL10 Description Nonfungal infections of skin and subcutaneous tissue Abdominal pain Peripheral vascular disease Aortic aneurysm Gastrointestinal Obstruction/Perforation Alimentary or excretory surgical openings Chromosomal anomalies Inherited metabolic disorders Prostatic hypertrophy Renal calculi Other hemolytic anemias Iron deficiency, other anemias Thrombophlebitis Aplastic anemia Deep vein thrombosis Hemophilia, coagulation Disorder Sickle Cell Disease Tuberculosis infection HIV, AIDS Lyme disease Septicemia Malignant neoplasms of the skin Low impact malignant neoplasms High impact malignant neoplasms Malignant neoplasms, breast Malignant neoplasms, cervix, uterus Malignant neoplasms, ovary Malignant neoplasms, esophagus Malignant neoplasms, kidney Malignant neoplasms, liver and biliary tract Malignant neoplasms, lung 133

134 Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models EDC MAL11 MAL12 MAL13 MAL14 MAL15 MAL16 MAL18 MUS03 MUS06* MUS09* MUS10 MUS11* MUS13* MUS14 MUS16 NUR03 NUR05 NUR06 NUR07 NUR08 NUR09 NUR11 NUR12 NUR15 NUR16 NUR17 NUR18 NUR19 NUR20* NUR22* NUR23 Description Malignant neoplasms, lymphomas Malignant neoplasms, colorectal Malignant neoplasms, pancreas Malignant neoplasms, prostate Malignant neoplasms, stomach Acute Leukemia Malignant neoplasms, bladder Degenerative joint disease Kyphoscoliosis Joint disorders, trauma related Fracture of neck of femur (hip) Congenital anomalies of limbs, hands and feet Cervical pain syndromes Low back pain Amputation Status Peripheral neuropathy, neuritis Cerebrovascular disease Parkinson's disease Seizure disorder Multiple sclerosis Muscular dystrophy Dementia and delirium Quadriplegia and Paraplegia Head Injury Spinal Cord Injury/Disorders Paralytic Syndromes, Other Cerebral Palsy Developmental disorder Central nervous system infections Migraines Organic Brain Syndrome 134

135 Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models EDC NUR24 NUR25 NUR26 NUT02 PSY01 PSY02 PSY03 PSY05 PSY07 PSY08 PSY09 PSY10* PSY12 PSY13 PSY14 PSY15 PSY16 PSY17 PSY18 PSY19 PSY20 REC01 REC03 REC04 REN01 REN02 REN03 REN04 REN06 RES02 RES03 Description Dementia Delirium Autism Spectrum Disorders Nutritional deficiencies Anxiety, neuroses Substance use Tobacco abuse Attention deficit disorder Schizophrenia and affective psychosis Personality disorders Depression Psychologic signs and symptoms Bipolar Disorder Adjustment disorder Psychosocial disorders of childhood Eating disorder Impulse control Psycho-physiologic and somatoform disorders Psychosexual Sleep disorders of nonorganic origin Major depression Cleft lip and palate Chronic ulcer of the skin Burns--2nd and 3rd degree Chronic renal failure Fluid/electrolyte disturbances Acute renal failure Nephritis/Nephrosis End stage renal disease Acute lower respiratory tract infection Cystic fibrosis 135

136 Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models EDC RES04 RES06* RES08 RES09 RES12 RES13 RES14 RHU01 RHU02* RHU03* RHU05 SKN18* TOX02 TOX03* TOX04 Description Emphysema, chronic bronchitis, COPD Sleep apnea Pulmonary embolism Tracheostomy Acute respiratory failure Chronic respiratory failure Aspiration and bacterial pneumonias Autoimmune and connective tissue diseases Gout Arthropathy Rheumatoid Arthritis Benign neoplasm of skin and subcutaneous tissues Adverse effects of medicinal agents Adverse effects from medical/surgical procedures Complications of Mechanical Devices * - This subset of EDC categories is used in the concurrent cost models only. Rx-Defined Morbidity Groups (Rx-MGs) Included in the Risk Models Rx-MG ALLx010 ALLx030 ALLx040 ALLx050 CARx010 CARx020 CARx030 CARx040 CARx050 EARx010 ENDx010 Description Allergy/Immunology/Acute Minor Allergy/Immunology/Chronic Inflammatory Allergy/Immunology/Immune Disorders Allergy/Immunology/Transplant Cardiovascular/Chronic Medical Cardiovascular/Congestive Heart Failure Cardiovascular/High Blood Pressure Cardiovascular/Hyperlipidemia Cardiovascular/Vascular Disorders Ears, Nose, Throat/Acute Minor Endocrine/Bone Disorders 136

137 Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models Rx-MG ENDx020 ENDx030 ENDx040 ENDx050 ENDx060 ENDx070 EYEx010 EYEx020 EYEx030 FREx010 FREx020 FREx030 GASx010 GASx020 GASx030 GASx040 GASx050 GASx060 GSIx010 GSIx020 GSIx030 GSIx040 GURx010 GURx020 HEMx010 INFx010 INFx020 INFx030 INFx040 INFx050 MALx010 Description Endocrine/Chronic Medical Endocrine/Diabetes With Insulin Endocrine/Diabetes Without Insulin Endocrine/Thyroid Disorders Endocrine/Growth Problems Endocrine/Weight Control Eye/Acute Minor: Curative Eye/Acute Minor: Palliative Eye/Glaucoma Female Reproductive/Hormone Regulation Female Reproductive/Infertility Female Reproductive/Pregnancy and Delivery Gastrointestinal/Hepatic/Acute Minor Gastrointestinal/Hepatic/Chronic Liver Disease Gastrointestinal/Hepatic/Chronic Stable Gastrointestinal/Hepatic/Inflammatory Bowel Disease Gastrointestinal/Hepatic/Pancreatic Disorder Gastrointestinal/Hepatic/Peptic Disease General Signs and Symptoms/Nausea and Vomiting General Signs and Symptoms/Pain General Signs and Symptoms/Pain and Inflammation General Signs and Symptons/Severe Pain Genito-Urinary/Acute Minor Genito-Urinary/Chronic Renal Failure Hematologic/Coagulation Disorders Infections/Acute Major Infections/Acute Minor Infections/HIV/AIDS Infections/Tuberculosis Infections/Severe Acute Major Malignancies 137

138 Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models Rx-MG Description MUSx010 Musculoskeletal/Gout MUSx020 Musculoskeletal/Inflammatory Conditions NURx010 Neurologic/Alzheimer's Disease NURx020 Neurologic/Chronic Medical NURx030 Neurologic/Migraine Headache NURx040 Neurologic/Parkinsons Disease NURx050 Neurologic/Seizure Disorder PSYx010 Psychosocial/Attention Deficit Hyperactivity Disorder PSYx020 Psychosocial/Addiction PSYx030 Psychosocial/Anxiety PSYx040 Psychosocial/Depression PSYx050 Psychosocial/Acute Minor PSYx060 Psychosocial/Chronic Unstable PSYx070 Psychosocial/Sleep Disorders PSYx080 Psychosocial/Tobacco Cessation PSYx090 Psychosocial/Bipolar Disorder RESx010 Respiratory/Acute Minor RESx020 Respiratory/Chronic Medical RESx030 Respiratory/Cystic Fibrosis RESx040 Respiratory/Airway Hyperactivity SKNx010 Skin/Acne SKNx020 Skin/Acute and Recurrent SKNx030 Skin/Chronic Medical TOXx010 Toxic Effects/Adverse Effects/Acute Major Special Population Markers Special Population Marker HOSDOM0 (included in Intercept) HOSDOM1 HOSDOM2 HOSDOM3+ Use in the Predictive Model Boolean indicator for 0 HOSDOM Boolean indicator for 1 HOSDOM indicator Boolean indicator for 2 HOSDOM indicators Boolean indicator for 3 or more HOSDOM indicators 138

139 Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models Special Population Marker Frailty Marker CAL-SSA* Active Ingredient Count Use in the Predictive Model Boolean indicator Boolean indicator Boolean indicator for 14 or more Active Ingredients * - CAL-SSA is used in the Hospitalization Prediction Models only. Demographic Markers Demographic Marker Use in the Predictive Model Ten age groups 0-4; 5-11; 12-17; 18-34; 35-44; 45-54; (Included in the Intercept); 70-74; 75-79; 80-84; 85+ Female Boolean indicator Optional Prior Cost Markers Prior Cost Markers* Total Expense, Non-users (Included in the Intercept) Total Expense, 1-10th Percentile Total Expense, 11-25th Percentile Total Expense, 26-50th Percentile Total Expense, 51-75th Percentile Total Expense, 76-91st Percentile Total Expense, 91-93rd Percentile Total Expense, 94-95th Percentile Total Expense, 96-97th Percentile Total Expense, 98-99th Percentile Pharmacy Expense, Non-users (Included in the Intercept) Pharmacy Expense, 1-10th Percentile Pharmacy Expense, 11-25th Percentile Pharmacy Expense, 26-50th Percentile Pharmacy Expense, 51-75th Percentile Pharmacy Expense, 76-91st Percentile Pharmacy Expense, 91-93rd Percentile Pharmacy Expense, 94-95th Percentile Pharmacy Expense, 96-97th Percentile Pharmacy Expense, 98-99th Percentile * Either pharmacy expense markers or total expense markers will be used depending on the model variant, but pharmacy expense and total expense will not be used simultaneously. Prior pharmacy expense is preferred when the predicting pharmacy expense and prior total expense is preferred when predicting total expense. 139

140 Appendix A: Variables Necessary to Locally Calibrate the ACG System Risk Models Utilization Markers Note: These are used only in the Hospitalization and 30 Day Unplanned Readmission Prediction Models. Dialysis Service Nursing Service Major Procedure Psychotherapy Active Cancer Treatment Active Cancer Treatment Inpatient Hospitalization Count Emergency Visit Count Outpatient Visit Count Prior Hospitalization Markers Note: These are used only in the 30 Day Unplanned Readmission Prediction Models. Inpatient Hospitalization Days Unplanned 30 day readmission count 140

141 Appendix B: Acknowledgements Appendix B: Acknowledgements Documentation Production Staff Dedication Support Senior Editor: Jonathan P. Weiner, Dr. P.H. Managing Editor: Chad Abrams, M.A. Production assistance provided by: David Bodycombe Sc.D., and Klaus Lemke, Ph.D. Additional production assistance and original content provided by Lisa Kabasakalian, Meg McGinn, Robert Palmer, Amy Salls, and Joseph Steuer of DST Health Solutions, LLC. The ACG Team gratefully acknowledges the support provided by our corporate partner in helping to move this publication forward. This guide is dedicated to the memory of Barbara Starfield, M.D., MPH, whose ideas permeate the ACG System and who continues to inspire the attainment of the best possible primary health care throughout the world. If users have questions regarding the software and its application, they are advised to contact the organization from which they obtained the ACG software. Questions about grants of rights or comments, criticisms, or corrections related to this document should be directed to the Johns Hopkins ACG team. Such communication is encouraged. Website: Third-Party Library Acknowledgements This product includes software developed by the following companies: This product includes software developed by Health + Technologies, Inc. Copyright Health + Technologies, Inc. All rights reserved. This product includes software developed by SafeNet, Inc. Copyright 2010 SafeNet, Inc. All rights reserved. This product includes the Java Runtime Environment from Oracle Copyright 2014 Oracle America, Inc. All rights reserved. The license for the Java Runtime is contained in the LICENSE file in the jre directory of this software. REQUIRED NOTICE REGARDING COMMERCIAL FEATURES IN JAVA 141

142 Appendix B: Acknowledgements Use of the Commercial Features for any commercial or production purpose requires a separate license from Oracle. Commercial Features means those features identified Table 1-1 (Commercial Features In Java SE Product Editions) of the Java SE documentation accessible at This product includes the following open source: This product includes software developed by The Apache Software Foundation Copyright The Apache Software Foundation This product includes software developed by the JDOM Project Copyright Jason Hunter & Brett McLaughlin. This product includes the open source itext library Copyright 1999, 2000, 2001, 2002 by Bruno Lowagie. with modifications made by JasperReports Copyright 2014 JasperSoft Corporation, Inc. The license for these libraries is contained in the ITEXT_LICENSE.txt file distributed in the lib directory of this software. This product includes the open source JasperReports library Copyright 2014 JasperSoft Corporation, Inc. The license for these libraries is contained in the JASPERREPORTS_LICENSE.txt file distributed in the lib directory of this software. This product includes the open source JFreeChart and JCommon libraries Copyright , by Object Refinery Limited and Contributors. The license for these libraries is contained in the JFREECHART_LICENSE.txt file distributed in the lib directory of this software. This product includes the open source JIDE Common Layer library Copyright JIDE Software, Inc, all rights reserved. This product includes JGoodies open source libraries Copyright JGoodies Karsten Lentzsch. All rights reserved. This product includes the open source DejaVu Fonts distributed by JasperReports Copyright 2003 by Bitstream, Inc. All Rights Reserved. Bitstream Vera is a trademark of Bitstream, Copyright 2006 by Tavmjong Bah. All Rights Reserved. The license for these fonts are contained in the JASPER_DEJAVU_FONTS_LICENSE.txt file distributed in the lib directory of this software. 142

143 Index Index A cancer 91 ACG care coordination categories 129 assessing 94 chronic illness examples 29 generalist seen 101 clinical classification of pregnancy/delivery 33 majority source of care 96 clinically-oriented examples 29, 32 management visit count 96 decision tree 21 risk of poor coordination 101 delivery 33 specialty count 99 diabetes example 31 unique provider count 98 hospitalization prediction 123 care density 104 hypertension example 29 categorical variables 6 infants clinical classification 35 chronic condition count marker infants examples 34 distribution 54 predictive model variables 129 EDCs considered 50 pregnancy 33 overview 49 pregnancy/delivery with complications example 33 chronic illness RUB categories 36 diabetes example 31 RUB levels 130 examples 29 terminal groups formation 21 hypertension example 29 active cancer treatment utilization marker 124 clinical aspects active ingredient count variable 66 duration 13 ADG etiology 13 clusters 18 expected need for specialiy care 14 collapsed 18, 20 severity 13 diagnostic certainty 14 clinically-oriented pregnancy example 32 duration 14 collapsed ADG 18 EDC differences 48 common ACG system applications 6 etiology 14 compassionate allowances condition utilization expected need for specialty care 14 marker 125 major 16 condition markers severity 14 criteria 67 subgroups 18 definitions 68 applications 17 continuous single-interval measure of medication availability 82 B coordination markers for population identification 94 cost models birth weight 59 ACG concurrent risk 111 C outcomes 106 CSA 82 CADG c-statistic 127 collapsed 18 combinations 19 D MAC assignments 20 decision tree CAL-SSA

144 Index ACG 21 hospitalization 123 MAC-12 pregnant women 23 hospitalization likelihood MAC-24 multiple ADG categories 27 model outputs 125, 127 MAC-26 infants 25 probability extended hospitalization score 126 delivered 58 probability ICU hospitalization score 126 delivered pregnancy variables 131 probability injury hospitalization score 126 delivery ACGs 33 probability IP hospitalization score 126 demographic markers 139 probability IP hospitalization six months score 126 diagnostic certainty 14, 47 probability unplanned 30 day readmission dialysis service utilization marker 124 score 128 duration 13, 14 I E ICD-9-CM otitis media EDC 42 EDC ICD-10 otitis media EDC 44 ADG differences 48 ICD mapping to ADG group 10 applications 49 impact conditions 66 chronic condition count marker 50 impact EDCs 48 development 41 infants diabetes complications examples 46 ACG 35 diagnostic certainty 47 examples 34 MEDC types 47 infants decision tree 25 otitis media ICD-9-CM codes 42 inpatient days count utilization marker 125 otitis media ICD-10 codes 44 inpatient hospitalization count utilization marker 125 use of 42 input files 6 variables 131 eitology 14 emergency visit counts utilization marker 125 local calibration of PM scores 122 empiric validation low birth weight 59 c-statistic 127 likelihood of hospitlization model 126 positive predictive value 126 MAC readmission model 128 assigned collapsed ADG 20 sensitivity 126 combinations 19 end-of-period possession 85 MAC etiology 13 MAC expanded diagnostic clusters 41 MAC expected need for speciality care 14 major ADG 16 expected need for specialty care 14 majority source of care 96 F management visit count 96 mapping ICD to ADG set 10 frailty conditions markers 56 markers demographic 139 G optional prior cost 139 generalist seen 101 special population 138 GS 101 utilization 140 mechanical ventilation 92 H hospital dominant conditions markers 55 L M mechanical ventilation utilization marker 124 MEDC types

145 Index medication-defined morbidity ACG use 123 active ingredient count variable 66 empiric validation 126, 128 NDC-to-Rx-MG assignment methodology 65 overview 123 overview 60 utilization markers 124 Rx-MG morbidity taxonomy 62 predictive modeling medication possession ratio 81 ACG categories 129 morbidity groups conceptual bassis 106 Rx-defined 60 EDC variables 131 MPR 81 elements of 106 MSCO 96 impact conditions 66 multiple ADG categories decision tree 27 impact EDCs 48 N positive predictive value 119 ratios 119 NDC-to-Rx-MG assignment methodology 65 resource bands 93 non-delivery pregnancy variables 131 r-squared 114, 118 nursing service utilization marker 124 Rx-defined morbidity groups 136 sensitivity 119 O statistical approach 110 variables 129 optional prior cost markers 139 pregnancy outpatient visit count utilization marker 125 ACGs 33 with complications examples 33 P without delivery markers 59 PDC 83 pregnant 58 pharmacy adherence pregnant women decision tree 23 continuous single-interval measure of medication Probability availability 82 extended hospitalization score 126 defined 75 ICU hospitalization score 126 effective care significance 73 injury hospitalization score 126 end of period possession 85 IP hospitalization score 126 medication possession ratio 81 IP hospitalization six months score 126 possession/adherence marker development 74 persistent high user 121 proportion of days covered 83 unexpected pharmacy cost 120 reporting of possession 85 unplanned 30 day readmission score 128 testing 86 proportion of days covered 83 validation 86 prospective cost models 116 pharmacy cost 117 psychotherapy 92 pharmacy cost risk 116 psychotherapy utilization marker 124 population markers chronic condition count 49 R condition 67 rank probability 117 frailty conditions 56 reference population 107 hospital dominant conditions 55 reference probability 117 pregnancy without delivery 59 reporting of possession 85 populations rescaling 112 assign dollar value 112 resource bands 93 rescaling 112 resource utilization bands 35 positive predictive value 126 risk predicting hospitalization concurrent

146 Index risk adjustment pyramid 6 unplanned readmissions within 30 days count risk factors 107 utilization marker 125 risk of poor coordination 101 utilization markers RUB active cancer treatment 124 ACG categories 36 compassionate allowances condition 125 ACG variables 130 dialysis service 124 overview 35 emergency visit count 125 Rx-defined morbidity groups 60 inpatient days count 125 Rx-defined morbidity group variables 136 inpatient hospitalization count 125 Rx-MG list of 140 clinical basis 60 mechanical ventilation 124 morbidity taxonomy 62 nursing service 124 S outpatient visit count 125 psychotherapy 124 SC 99 unplanned readmissions within 30 days count 125 sensitivity 126 use of 124 severity 13, 14 special population markers 138 V specialty count 99 variables statistical performance ACG categories 129 positive predictive value 119 ACG RUB levels 130 predictive ratios 119 delivered pregnancy 131 r-squared 114, 118 demographic markers 139 sensitivity 119 EDC 131 T non-delivery pregnancy 131 optional prior cost markers 139 terminal groups 21 predictive model 129 total risk 116 Rx-defined morbidity groups 136 typology 123 special population markers 138 utilization markers 140 U unique provider count 98 W weights