UnitedHealth Premium Physician Designation Program Detailed Methodology

Transcription

1 UnitedHealth Premium Physician Designation Program Detailed Methodology Resources Phone: Toll-free, Website: UnitedHealthcareOnline.com > UnitedHealth Premium > Premium Methodology Table of Contents Overview Physician Eligibility Markets Included in Assessment Data Sufficiency Requirements Data Used for Assessment Specialties Included of Clinical Quality s Defining Cost Efficiency Episodes Claims Grouping Software (Groupers) Attribution Case-Mix Adjustments Severity Adjustments Designation Criteria Practice Improvement Programs Applying the Quality Recognition Benefit Teaching Mission - Academic Credit Applying the Group Methodology for Quality and Cost Efficiency Medical Group Definition Medical Group Methodology Process Medical Group Methodology Notification Public Designation Displays Quality Overview Generating National Rates for Each Quality EBM Case-Mix and Severity Adjustment EBM Attribution Attribution Methods with Examples PCP Imputation Process PEG Quality s PEG Quality Case-Mix and Severity Adjustment... 9 PEG Quality Attribution Special Case: PEG Redo Procedures Quality Outcome Statistical Methods Additional Credit s Practice Improvement Program Credit Creating the 75 th Percentile Performance Rate Interpreting the Chi-Square Test Results Quality Individual Outcome Example Quality Individual Outcome Example Procedure Quality s Cost Efficiency Overview ETG Episodes ETG Case-Mix and Severity Adjustment ETG Episode Attribution PEG Episodes PEG Cost Efficiency Case-Mix and Severity Adjustment PEG Cost Efficiency Episode Attribution Further Processing of Benchmark Sets Creating a Comparison Superset Efficiency Outcome Statistical Methods Interpreting the Wilcoxon Rank-Sum Test Results.. 18 Cost Efficiency Outcome Example Important Program Information Table 1: and Procedure Quality ment Table 2: Summary of Clinical s Table 3: Practice Improvement Programs by. 69 American Board of Internal Medicine (ABIM) Practice Improvement Modules Table 4: Recognition Programs by NCQA Individual Recognition (not group-level).. 71 Bridges to Excellence (BTE) Table 5: Acronyms and Abbreviations

2 Overview The UnitedHealth Premium physician designation program uses clinical information from health care claims to assist physicians in their continuous practice improvement and to help consumers in making more informed and personally appropriate choices for their medical care. The program uses evidence-based, medical society, and national industry standards with a transparent methodology and robust data sources to evaluate physicians across 21 specialties to advance safe, timely, effective, efficient, equitable and patient-centered care. The program supports practice improvement and provides physicians with access to information on how their clinical practice compares with national and specialty-specific measures for quality, and with local cost efficiency benchmarks in the same geography. Evaluation for quality compares a physician s observed practice to the UnitedHealthcare national rate among other physicians who are responsible for the same interventions. Evaluation for cost efficiency compares a physician s observed episode costs to the risk-adjusted costs of their peers in the same specialty and market. Quality is the primary measurement, demonstrating our commitment to evidencebased practice. The quality designation is separate from the cost efficiency designation. Physicians must first be designated for quality in order to be designated for cost efficiency. Physicians who meet the quality designation criteria will receive the quality designation regardless of their cost efficiency evaluation. Quality and cost efficiency evaluations both incorporate adjustments for case mix and severity of illness where appropriate. The designation process is described in the following image, showing a fictional Dr. Pratt being evaluated for quality and cost efficiency. First, from claims data, we identify all of Dr. Pratt s patients for the conditions measured in the program. We then compare the actual number of times the treatment his patients received was consistent with evidence-based measures compared to the benchmark number based on the UnitedHealthcare national rate for each measure. Using statistical testing, we determine if the quality criteria are met. 2

3 The criteria used to measure physician practices are based on the following aspects of care: Preventive care cancer screening and other indicated screening interventions Appropriate care appropriate use of medications and diagnostic tests Chronic disease care monitoring for control, progression, and complications Patient safety avoiding duplicate testing or adverse drug interactions, and monitoring safety Sequencing of care diagnostic tests and procedures, treatment, and monitoring Effectiveness of procedures lack of failed therapy and complications Each year we review and enhance our program, including the methodology used to evaluate quality and cost efficiency for the Premium physician designation program. This document describes the methodology for the 2012 version of the Premium program. Physician Eligibility Individual physicians are eligible for the UnitedHealth Premium program designation if they are contracted with UnitedHealthcare, credentialed by UnitedHealthcare, practice in a specialty and geographic location that are included in the Premium program, are board certified in their primary specialty, and have an unencumbered license at the time of the designation. Physicians who are being tracked for potential fraud and abuse are not eligible for Premium designation. Markets Included in Assessment UnitedHealthcare defines markets based on geographic service areas that have an average of 130,000 or more UnitedHealthcare enrollees. Markets may include adjacent states as long as they are consistent with the Metropolitan Statistical Area (MSA) market definition. MSAs are defined by the United States Office of Management and Budget. The Premium program is available in 41 states and 148 markets across the United States. In many situations the Premium program combines two or more markets to achieve sufficient sample size for the Cost Efficiency assessment. A complete list of markets is available upon request. Data Sufficiency Requirements Sufficient data for the quality assessment requires a minimum of five unique patients and 20 quality measure opportunities across all conditions or procedures. Opportunities are the number of times a measurement criterion could have been met. For cost efficiency, a minimum of 10 medical cases (episodes of care) or 10 procedure or surgical cases (procedure episodes) is required. Cardiologists who provide medical care as well as perform procedures such as cardiac catheterization and percutaneous revascularization can meet the cost efficiency minimum through a combination of medical and procedure episodes. Data Used for Assessment UnitedHealthcare relies primarily on paid claims data to assess the quality and cost efficiency of care. Paid claims data are commonly used by many types of organizations (e.g., health plans, academia, regulatory agencies, public health, health service research, specialty societies, and others) to analyze and understand many aspects of health care delivery. The data are readily available, comprehensive, and can provide detailed information about the type, quantity, and cost of services that patients receive. The measures used in the Premium program have been designed specifically for use with administrative claims data. Many of the National Quality Forum s NQF- Endorsed standards have been in use for decades. While no data are 100% accurate, we continually validate and improve the data set through internal work, comments from our advisory committees and scientific advisory boards, and feedback from groups and individual physicians. The Premium program counts several non-claims-based programs towards quality designation for the specialties appropriate to each program. These include National Committee for Quality Assurance (NCQA) recognition programs, Bridges to Excellence (BTE) programs, and American Board of Internal Medicine (ABIM) Practice Improvement Modules. The Premium program uses claims for patients enrolled in UnitedHealthcare commercial fee-for-service products only. Claims from patients whom we can identify as receiving hospice services are not considered in the assessment of physician performance. The data used in the designation cycle includes commercial claims that were incurred and paid from January 1, 2009 through February 29, The program includes paid claims data from members who were disenrolled from UnitedHealthcare on the end date of the data collection window as long as those members had a sufficient window of coverage to satisfy the criteria for the 3

4 particular measure. However, HEDIS -based NCQA quality measures use data only from members enrolled as of December 31, Such measures include Breast Cancer Screening, Cervical Cancer Screening, Chlamydia Screening, and Appropriate Testing for Children With Pharyngitis. Specialties Included Primary Care Areas Family Medicine Internal Medicine Obstetrics and Gynecology Pediatrics Other Areas Allergy Cardiology Cardiology - Electrophysiology Cardiology - Interventional Endocrinology Infectious Disease Nephrology Neurology Neurosurgery - Spine Orthopaedics - General Orthopaedics - Hand Orthopaedics - Foot/Ankle Orthopaedics - Hip/Knee Orthopaedics - Shoulder/Elbow Orthopaedics - Spine Pulmonology Rheumatology of Clinical Quality s The clinical quality measures consist first of the National Quality Forum s NQF-Endorsed standards when available for the conditions being evaluated, and in accordance with the principles of the Consumer-Purchaser Disclosure Project s Patient Charter. Consistent with the Patient Charter, those measures are supplemented with others as necessary to evaluate clinically important conditions and specialties. Additional measures are developed using published literature and information from organizations such as the following: The AQA Alliance (formerly the Ambulatory Care Quality Alliance) The National Committee for Quality Assurance (NCQA) societies relevant to a specific disease and clinical condition Government agencies Other national expert panels Defining Cost Efficiency Episodes The evaluation of physicians for cost efficiency compares observed cost for episodes of care to expected cost for episodes of care, with adjustments for the patient s severity of illness and the physician's case mix. An episode of care is a grouping of services provided to an individual patient within a given time period surrounding a given illness (or group of similar illnesses). Claims Grouping Software (Groupers) The Premium program uses four different software programs to collect, or group, claims data into quality measures and episodes of care: Symmetry EBM Connect EBM Connect is an Optum tool that assesses quality measures related to preventive care and medical conditions. Each instance of an EBM Connect quality measure, applied to a patient, is referred to hereafter as an EBM measure. Symmetry Episode Treatment Groups (ETG ) ETG is an Optum tool that creates condition-oriented episodes of care, referred to hereafter as ETG episodes. Symmetry Procedure Episode Group (PEG ) PEG is an Optum tool that creates procedure-based episodes. Additional software then creates quality measures related to sequencing of care around procedures, their complications, and repeat studies or related repeat surgery/procedures. Each instance of a procedural episodebased quality measure, applied to a patient, is referred to as a PEG measure. The major procedure performed during the episode is referred to hereafter as the anchor procedure. Related interventions are the targets. If more than one major procedure occurs in a PEG episode, clinical logic determines the anchor. PEG episodes are also used for cost efficiency analysis of procedure related episodes, referred to hereafter as PEG episodes. Examples of PEG anchor and target logic: For an outpatient diagnostic cardiac catheterization, the catheterization is the anchor. Myocardial perfusion imaging (SPECT MPI) performed close in time, on the same patient, would be a target. 4

5 If a diagnostic catheterization is performed soon before placement of a drug-eluting stent, the stent is the anchor and the diagnostic catheterization is a target. 3M All Patient Refined Diagnosis Related Groups (APR DRG) APR DRG software from 3M creates severity of illness levels for risk adjusting inpatient PEG quality measures and inpatient PEG episode costs. Attribution Rules of attribution are used to assign a member s care and episode cost to the responsible physician. The software used in the UnitedHealth Premium program groups the attendant information in the paid claims stream diagnosis, procedures, and allowed claims to a specific member. A physician is then assigned to that member for that service and episode. Episodes are built around patients. Each episode is assigned (attributed) to the one physician who generated the highest percentage of services in an episode. To make sure there was significant involvement on the part of the physician to whom the episode was attributed, we discard all episodes where the highest percentage of involvement was less than 30%. In practice, about 90% of episodes used for the Premium program are managed by one physician. Radiologists, pathologists, anesthesiologists, and other specialties that are not included in the Premium program are not recognized as the responsible physician. Case-Mix Adjustments Case-mix adjustment means correcting a measure for differences in the types of cases each physician treats. Some recommended interventions among quality measures may be more or less difficult to accomplish. Some conditions require more resources in general than others (e.g., pneumonia vs. sinusitis, or knee replacement vs. knee arthroscopy with meniscectomy). The Premium program uses case-mix adjustment for both quality and for cost efficiency measurement. For quality assessment, the program uses a separate benchmark for each measure based on the national rate at which it is accomplished. For the cost efficiency assessment, the Premium program divides cases among categories to perform case-mix adjustment. Severity Adjustments Severity adjustment means correcting for patient-level differences that might affect costs within a given condition and quality measures related to inpatient procedures. The Premium program uses severity adjustment for certain quality measures and for cost efficiency measurement. Designation Criteria The designation process starts with the physician s individual quality and cost efficiency outcomes. The initial individual outcome is determined by comparing a physician s own results to that of their peer group. The process concludes with the public designation (two, one, or no stars) that is displayed online and in reports given to physicians. The designation rules determine a physician s assessment based on the physician s individual outcomes, participation in qualifying recognition programs, or the assessment results for their specialty in an affiliated group practice, when applicable. 1. If the physician does not meet the quality criteria because the physician does not have enough data for assessment, the physician can meet quality criteria through non claimsbased methods including National Committee for Quality Assurance (NCQA) recognition programs, Bridges to Excellence (BTE) programs, and American Board of Internal Medicine (ABIM) Practice Improvement Modules. 2. If the physician does not receive quality recognition through a qualifying recognition program, does not have enough data for assessment, and is affiliated with a group practice, the physician may benefit through application of the group assessment result for their specialty. 3. A physician must be board certified in his/her primary specialty in order to receive the quality designation. If a physician is not board certified in his/her primary specialty, there will be no public designation display. This indicates that the physician is not designated. 4. Quality is assessed first. If the physician does not receive the quality designation, the physician is not eligible for the cost efficiency designation. 5. Physicians who do not have enough data for the cost efficiency assessment or who do not receive the cost efficiency designation on their own, can benefit if their specialty in their affiliated group meets the cost efficiency criteria. 6. The final designation results are displayed publicly in the UnitedHealthcare physician directories unless there are other reasons that they should not be displayed (e.g., the physician is under investigation for fraud). 5

6 Practice Improvement Programs Based on their participation in qualifying American Board of Internal Medicine s (ABIM) Practice Improvement Modules (PIMs), the Premium program provides physicians an opportunity to receive additional credit toward their quality result. Internists and related subspecialists affiliated with the ABIM can complete PIMs. Physicians are encouraged to visit abim.org to learn more about the ABIM s Maintenance of Certification programs. Please see Table 3 at the end of this document for a list of ABIM PIMs by specialty. If a physician is eligible for ABIM practice improvement credit, the larger of 10% of opportunities or 25 opportunities, is added to the physician result as "100% compliant" for those opportunities. Applying the Quality Recognition Benefit Physicians who do not have enough data for quality assessment may receive the quality designation by other methods. If a physician is board certified in their primary specialty and is otherwise eligible for the quality designation, they can receive the quality designation for individual recognition through an NCQA recognition program appropriate to their specialty. They also can receive quality designation for individual certification in a BTE program appropriate to their specialty. Please see Table 4 at the end of this document for the Recognition Programs by. Teaching Mission Academic Credit The Premium program recognizes the teaching mission of academic medical group practices through an academic credit. Academic credit will be awarded to all members of a group with full-time faculty who have primary teaching responsibilities, and are affiliated with and admit 100 percent of their patients to a medical center that is a sponsor or participating institution of a program accredited by the Accreditation Council for Graduate Medical Education (ACGME). A ten percent credit is applied to the cost efficiency score of the affiliated physicians. Applying the Group Methodology for Quality and Cost Efficiency Only physicians who are eligible for designation are included in the determination of the medical group s assessment results. Physicians who are represented by the following two situations may benefit from group methodology: 1. Physicians who do not have enough data by themselves to assess quality and/or cost efficiency. 2. Physicians who have met the quality criteria on their own, but not the cost efficiency designation, may receive the cost efficiency designation through the group methodology. Physicians who benefit from group methodology, based on the assessment result for their specialty in an affiliated medical group, must still meet all other program criteria to be designated. Medical group methodology is an extension of the UnitedHealth Premium designation program that accommodates physicians who, individually, may not have enough experience on their own with UnitedHealthcare patients to be considered for designation or physicians who have met the quality criteria but not the cost efficiency criteria. Group-level measurement is designed to support physicians within a group who share similar practice patterns and common care protocols. Medical Group Definition A medical group is defined by the Premium program as a group having one or more individual physicians under a single UnitedHealthcare contract. The group must be a single legal entity (e.g., partnership, PC, LLC). The group can consist of a single specialty or multiple specialties. More complex entities (e.g., market, regional, state, or national clinical specialty societies, IPAs, PHOs) are generally not considered a medical group and are not eligible for the medical group methodology. A group of legal entities may be assessed as a medical group only if they all share a common contract with UnitedHealthcare. Medical group methodology is used only if the group is contracted with UnitedHealthcare using standard participation agreements and is in good standing with UnitedHealthcare at the time of the group assessment. Medical Group Methodology Process Each specialty within a medical group is assessed for quality and cost efficiency using the same methodology applied to individual physicians. Each physician s attributed patients, quality measures and cost efficiency episodes are aggregated with all the physicians in the group in the same Premium specialty. If the aggregated data, evaluated as if a single physician, meet the program s criteria, the specialty group is then assigned a qualityonly or a quality and cost efficiency assessment result. In situations where a physician is affiliated with more than one group, the highest assessment result is used. The possible specialty group assessment results are: Quality Only Quality and Cost Efficiency Does Not Meet Criteria Not Enough Data to Assess 6

7 Physicians may benefit by the group methodology if they do not have enough data for quality and/or cost efficiency assessment. If the group methodology generates a result of Does Not Meet Criteria, the physician remains at Not Enough Data to Assess. Physicians may also benefit from group methodology if they do not have enough data to assess cost efficiency or do not meet the cost efficiency criteria. Note: In this situation, only physicians who have received quality designation on their own or as a result of group methodology may benefit from this methodology. Medical Group Methodology Notification We communicate the assessment results for the medical group to the practice administrator at the same time we announce the individual physician designation results. It is important to note that our directories do not display a medical group result. Rather, only individual physician results are displayed, even when the physician benefits from the group methodology. The directories do not differentiate whether the individual physician s designation is a result of their own individual assessment or is a result of the use of the medical group methodology. Public Designation Displays The following designation results are displayed publicly in UnitedHealthcare s physician directories for use by members when making health care choices and by physicians when making referrals. (Quality and cost efficiency criteria met) (Quality criteria met) Not enough data to assess cost efficiency (Quality criteria met and not enough data to assess cost efficiency) Not enough data to assess Not evaluated Quality Overview A physician s quality individual outcome is determined by comparing the number of times their patients received recommended care with a benchmark number based on the UnitedHealthcare national rate of the same recommended care for each quality measure. Physicians must perform at a level that meets or exceeds the equivalent of the 75 th percentile performance for all physicians measured in order to meet the quality criteria. All quality measures are based on evidence-based, medical society, or national industry standards. This aspect of the Premium designation program only incorporates those clinical measures that can be assessed from paid claims data. We also focus on measures that are actionable by a physician. Specialties for which there are not sufficient measures to evaluate quality are not included in the assessment. The quality measures are based on nationally recognized and established evidence-based performance measurements from organizations such as the National Quality Forum (NQF ), the AQA Alliance, the National Committee for Quality Assurance, and the American College of Cardiology, as well as measures developed by national expert panels. The flow chart on this page describes the quality evaluation process. First, measures that evaluate adherence to evidencebased practices are mapped to specific specialties. For example, asthma is mapped to allergists and primary care physicians. Then, we analyze the amount of involvement of each physician with each patient s care to determine if the involvement was significant ( attribution ). If we are analyzing a procedure, the procedure is attributed to the physician who performed it. If we are analyzing a course of treatment over time for an 7

8 ongoing condition, one or more physicians who saw the patient may be attributed the measure for the patient s care. Quality measures for inpatient procedures are risk adjusted by 3M APR DRG severity of illness level. Then, for each physician, we compare the number of instances observed to the number generated by applying the national rate for each measure to the physician s opportunities for that measure. If the sample size is adequate, the chi-square test is applied to determine statistical significance, an individual quality outcome is assigned, and a quality designation is determined. A quality designation means the physician met or exceeded the equivalent of the 75 th percentile rate of recommended care for his or her patient mix and sample size. Generating National Rates for Each Quality A national rate is generated for each EBM measure, and for each PEG measure at the APR DRG and severity level. The rate is determined from the UnitedHealthcare commercial feefor-service product data set used in the Premium program. All patients qualifying for the measure are identified. Only instances where the patient s measure was attributed to at least one physician are counted. The number of interventions accomplished is summed and is divided by the number of patients, producing a national average rate for each measure. EBM Case-Mix and Severity Adjustment The quality assessment is case-mix adjusted at the measure level by using a benchmark result based on the national average for each measure. For example, one might expect diabetics to obtain a retinal exam at a lower rate than a blood test due to the higher effort required. Similarly, the rate of adverse reactions to one class of medications might be different from the rate of adverse reactions to another medication. Using a benchmark rate for each measure takes these differences into account. Process measures do not need severity adjustment, and therefore EBM measures apply regardless of the severity of a particular patient s condition. For example, medical evidence indicates that patients with diabetes should receive HbA1c testing at regular intervals, and patients with pharyngitis should be tested for group A streptococcus before being treated with antibiotics. All diabetics should receive HbA1c testing whether or not they have complications, and all patients with pharyngitis should be tested for group A streptococcus before antibiotic treatment regardless of the severity of their pharyngitis. Many EBM measures adjust for severity through specific clinical exclusions. Beta blocker use after heart attack is an example. Patients with a history of asthma, heart block greater than first degree, and other contraindications to beta blockers are excluded during processing of the measure. For EBM measures, the Premium program uses rule-specific exclusions to adjust for patient severity rather than numerical severity adjustments. An EBM measure that relates to pharmacy is only included in the quality evaluation when the claim record is for a patient who has a pharmacy benefit. Furthermore, the patient s pharmacy claim information for the period specified in the measure must be in the measurement data set. EBM Attribution The Premium program applies three general principles for attributing EBM measures to physicians. First, measures are applied to appropriate specialties. Second, attribution requires significant involvement in the care of the patient, which varies by attribution rule (See "PCP Imputation Process" at the end of this section). Third, attribution is allowed to multiple physicians when appropriate. Significant involvement varies by physician role. Primary care physicians (PCPs) are determined ( imputed ) from a pattern of visits, and include physicians from the disciplines of internal medicine, family medicine, pediatrics and obstetricsgynecology, as applicable to a given measure. Certain measures are attributed to both the imputed PCP(s) and specialists involved in the patient s care as appropriate. Multiple attribution is possible if more than one physician provides care to a patient based on the relevant attribution rules (see attribution methods and examples below). For instance, because it defines a preventive measure, the EBM measure "Cervical Cancer Screening" could be attributed to both an internal medicine physician and an obstetriciangynecologist (through the imputed PCP process). However, acute care conditions, such as pharyngitis, and measures related to services that should not have occurred, are attributed to only one physician. Attribution Methods with Examples Preventive Care Example: Breast cancer screening Attribute to imputed internal medicine or family medicine physician, and imputed obstetrician-gynecologist, if the physician saw the patient in the most recent calendar year available. Chronic Disease Care Example: Diabetes Attribute to imputed primary care physician if the physician saw the patient in the most recent calendar year available and attribute to any physician seeing the patient twice in the 8

9 two most recent calendar years available (e.g., patient with diabetes), including once in the most recent calendar year available. Exception: Hypertension and hyperlipidemia only require one visit, in the most recent calendar year available. Acute Conditions Example: Antibiotic use in bronchitis Attribute to the physician who saw the patient for the condition when only one physician was involved in the care. Pregnancy Management Example: Check urine culture in pregnant women Attribute to any family medicine physician and any obstetrician who had two visits; or an imputed obstetrician who had one visit and one pregnancy-related procedure in the patient s first two trimesters. Global s (not specific to a condition) Medication Safety Monitoring Attribute to imputed primary care physician and any physician who saw the patient twice in the two most recent calendar years available for any condition; and attribute only if the physician has seen the patient in the most recent calendar year available. PCP Imputation Process PCP imputation entails a step-by-step process to select the most probable primary care giver. The process is described below and stops when an imputed PCP is determined. PCP Imputation Method - Allowed specialties: Medical PCP imputation: Internal Medicine (IM), Family Practice (FP), Pediatrics (PD) Gynecological PCP imputation: Obstetrics and Gynecology (OB) 1. Most recent physical examination or assessment performed by a physician limited to the allowed specialties shown above. 2. Physician with an allowed specialty and who performed the largest number of evaluation and management type visits (the most recent visit breaks any ties). 3. Physician with an allowed specialty and who performed the largest number of prenatal, postpartum, or antepartum visits, or routine obstetrical care services (the most recent service breaks any ties). A given rule can have both an imputed Internal Medicine, Family Medicine, or Pediatrics PCP as well as an imputed Obstetrics and Gynecology PCP, if applicable (e.g., breast cancer screening). If the process does not yield an imputed PCP, then the measure is only attributed by other methods (e.g., see descriptions under Chronic Disease Care or Global s). PEG Quality s PEG measures identify surgical procedures and the services that comprise the episodes around them. These episodes are associated with a major therapeutic procedure. PEG measures are based on the anchor procedure and associated services. Premium designation uses PEG measures for physicians who perform cardiac, orthopaedic, and spine procedures. PEG Quality Case-Mix and Severity Adjustment Each PEG quality measure has a national benchmark rate for a given anchor procedure (the major procedure performed during the episode) and subprocedure, if applicable. Inpatient procedures are further divided into unique APR DRG combinations. Having a different benchmark for each combination of anchor procedure and APR DRG accomplishes case-mix adjustment. Inpatient procedure quality measures are further subdivided into APR DRG severity of illness levels, each with its own benchmark. Outpatient procedures are assigned a single severity level because they tend to be the lower-severity level procedures, or are performed on patients with lower burdens of illness. Because PEG quality measures include outcome measures such as rate of complication, it is appropriate to adjust for patient severity. Separate benchmarks create separate expected results and therefore adjust the PEG for patient severity. In this way, physicians are not penalized for treating patients with a higher burden of illness. PEG Quality Attribution PEG quality measures are attributed to the physician who performed the surgical anchor procedure associated with the quality measure. Complications and redos are assessed first, with separate benchmarks for each combination of anchor procedure, APR DRG and severity level for inpatient procedures. If physicians meet or exceed expected performance on complications and redo procedures, or they do not have enough data to assess performance on complications and redo procedures first, then they are evaluated across all attributed measures. These physicians may also be attributed EBM measures that have clinical relevance to the anchor procedures. For example, EBM measures may also be attributed to cardiologists who perform procedures that are appropriate to their specialty. Cardiologists have the potential to be evaluated on lipid 9

10 testing, statin and beta-blocker utilization, as well as quality measures related to PEG anchor procedures they perform. EBM measures are attributed using the rules defined for EBM measures. Special Case: PEG Redo Procedures PEG episodes that represent redos of other PEG anchors count in the quality assessment of the other anchor only. For example, a total knee replacement that is a redo of an arthroscopic knee meniscectomy would only be used as part of the arthroscopic knee meniscectomy episode. No quality measures are applied to the redo procedures themselves, and they do not count towards establishing national quality rates or cost efficiency benchmarks. Quality Outcome Statistical Methods Once case-mix and severity adjusted observed and expected result for all quality measures are attributed to a physician, the observed and expected results for each measure are summed to form total observed and expected results. The chi-square test is a standard statistical test used to determine whether the frequency distribution of observed events is consistent with a theoretical or expected distribution. The data that comes from the quality measurements are binomial (i.e., yes or no), and the result of the measurement is a proportion, or rate. We use the chi-square test for assessing quality of care because it allows us to combine the observed and expected results of many measures, and assess how well the observed rate fits the expected results. As it relates to the Premium program, the chi-square test is used to determine whether or not the physician s actual proportion of quality measures accomplished is consistent with the expected proportion accomplished. If the physician s actual result is not consistent with the expected result, then the difference between the two is considered to be statistically significant. Additional Credit s There are three situations where a national rate for a given quality measure cannot be set: s that depend exclusively on CPT II codes: Not all physicians submit the codes. An example is use of antiplatelet therapy in patients with coronary artery disease. s related to behavioral health: Patients may have a benefit plan that carves out behavioral health. Colorectal cancer screening: The time frame for the measure goes beyond the claims data time frame used in the Premium program. Nonetheless, these measures represent important clinical interventions and these measures are used for additional credit (the measure only counts when it is met). These additional credit measures add to data sufficiency and improve the observed results rate. Additional credit measures are added before applying the chi-square test. Practice Improvement Program Credit Based on their participation in a qualifying American Board of Internal Medicine s (ABIM) Practice Improvement Modules (PIMs), the Premium program provides physicians an opportunity to receive additional credit toward their quality result. If a physician is eligible for ABIM practice improvement credit, the larger of 10% of opportunities or 25 opportunities, is added to the physician result as "100% compliant" for those opportunities. Creating the 75 th Percentile Performance Rate The 75 th percentile performance rate for quality is a benchmark for comparing the physician's quality measure result rates to national rates, after adjusting for the physician's mix of patients and the number of eligible quality measures in the physician's panel (sample size). The 75 th percentile performance rate is based on the 50 th percentile expected rate that is raised to a 75 th percentile rate by adding a factor based on the standard deviation of the physician's expected compliance rate. The standard deviation (SD) is described by the following formula: SD = [n x (p) x (1-p)] This is where n is the number of opportunities and p is the physician s expected result rate. One SD above the physician s ratio would represent the 84 th percentile (50 th percentile for the average and 34 th percentile for the first standard deviation above). The 75 th percentile is of a SD above the physician s average. That fraction of a SD is added to the expected number to generate the equivalent of a 75 th percentile performance for the physician s case mix and sample size. An example of the process is included at the end of this section. Interpreting the Chi-Square Test Results If the chi-square test shows no statistically significant difference from the 75 th percentile equivalent performance, then the physician receives an individual quality outcome of Meets Expectations. We use a 98% confidence level (p < 0.02), and no statistical difference at the p < 0.02 level (chi-square value was ). If the chi-square test shows statistically different performance, it could be because the physician's performance is either higher or lower than expected, as shown in the following example. 10

11 If the chi-square test shows a statistically significant difference and the results are lower than the 75 th percentile equivalent performance then the physician receives an outcome of Below Expectations. If the chi-square test shows a statistically significant difference and the results are higher than the 75 th percentile equivalent performance, then the physician receives an individual quality outcome of Exceeds Expectations. Both outcomes of Meets Expectations and Exceeds Expectations will result in a quality designation for an individual physician if all other criteria are met. Quality Individual Outcome Example The process to establish a physician s quality outcome begins by identifying all quality measures attributed to that physician. For each unique measure, we establish two counts the number of opportunities the physician had to meet the compliance criteria for the measure, and the number of times the physician's results actually met that criteria. From there, we can also determine the number of times the physician's results did not meet the measure criteria. The following table shows measures, opportunities, and observed and expected results for a hypothetical physician: Note: The following example is based on EBM measures, but the same principle applies to PEG quality measures. Quality Opportunities Observed Results Expected Results (Average) Diabetes: HbA1c in last 12 months Coronary Artery Disease: Lipid Rx in last 12 months Hypertension: Kidney function test Totals Rate 83.63% 70.91% In this example, the recommended interventions were accomplished in 46 of 55 opportunities or at an 83.63% rate. Compared with an expected (average) rate of 39 of 55 opportunities or at a 70.91% rate, the physician is performing at an above-average rate. However, this comparison with the national rate is a comparison with an average rate. To meet criteria for the Premium program, physician performance will be compared against the 75 th percentile rate. The following process is used to shift the expected proportion of rules accomplished from the 50 th percentile value (average rate) to a 75 th percentile equivalent performance, basing the calculations on the properties of binomial distributions: The expected (average) measure results rate (70.91%) would represent the center of the performance distribution for the physician being evaluated. Since each physician has his or her own expected value based on case-mix adjustment, the performance distribution varies from physician to physician. Next, a standard deviation is calculated based on a binomial distribution. A portion 1 of that standard deviation is added to the physician expected result rate to create a 75 th percentile equivalent for the physician. 1 Because one full standard deviation above the mean is approximately the 84 th percentile on a normally distributed curve, we only add a portion of the full standard deviation when shifting from the 50 th to the 75 th percentile. 11

12 The following calculations describe this process, using an expected result rate of.7091: Physician Expected Result Rate:.7091 Physician Actual Opportunities: 55 Standard Deviation of the : Proportional Standard Deviation Coefficient for 75 th Percentile: Portion of Standard Deviation to move to 75 th Percentile: x.6745 = Expected (Average) Adjusted to 75 th Percentile: = (Expected 50 th Percentile + Portion of Standard Deviation) This adjustment is shown in the following table. With the adjustment made to the 75 th percentile, the physician is expected to have a completed count of rather than 39 in order to receive the quality designation. Observed Results Met Observed Results Not Met Physician Actual Opportunities 46 9 Expected at the 50 th Percentile (Average) Expected (Average) Adjusted to the 75 th Percentile The chi-square calculation is then performed on the following values: Observed Results Met Observed Results Not Met Physician Actual Opportunities 46 9 Expected (Average) Values Adjusted to the 75 th Percentile The calculated chi-square statistic for the above physician is Values greater than are statistically significant at the 98% confidence (p < 0.02) level. Therefore, this physician is not statistically different from expected. The physician s quality outcome is Meets Expectations. The chi-square statistic is always a positive number, which means that a high chi-square value shows that the observed rate is statistically different from the expected rate at a 98% confidence level (i.e., a two-tailed test at p < 0.02) but does not give information in terms of the direction of the difference (high or low). To determine the direction of the difference, the observed rate is directly compared to the expected rate. If the chi-square statistic shows a statistically significant difference and the observed rate is less than the expected rate, the outcome is Below Expectations (at p < 0.01 or at the 99% confidence level). If the chi-square is statistically significant and the observed rate is more than the expected rate, the outcome is Exceeds Expectations (at p < 0.01 or at the 99% confidence level). Both outcomes of Meets Expectations and Exceeds Expectations will result in a quality designation for an individual physician if all other criteria are met. 12

13 Quality Individual Outcome Example Procedure Quality s The following table shows measures, opportunities, and observed and expected results for a hypothetical physician performing orthopaedic procedures, and illustrates case-mix and severity adjustment. In this case, the physician has already been determined not to have a significantly high number of complications and redo procedures. Quality Opportunities Observed Results Expected (Average) Results Total Knee Replacement Severity Total Knee Replacement Severity Knee Arthroscopy with Menisectomy Totals Rate 83.64% 67.27% The remainder of the calculations proceeds as described above. 13

14 Cost Efficiency Overview The AQA Alliance has defined efficiency of care as being the measured cost of care associated with a specific level of quality of care (AQA Principles of Efficiency s, April 2006). Consistent with this principle, the physician must first be designated for quality in order to be designated for cost efficiency. Cost efficiency analysis is based on total cost, which is a combination of resource utilization, resource mix, and unit cost, as collected into ETG episodes. Episodes include all services delivered to a patient, including those of other physicians or clinicians and related to a specific procedure or treatment of a condition. Episodes include dollars paid to the physician for direct services as well as facility costs and ancillary services which the software logic determined were related (e.g., medications, diagnostic tests). Using software, we categorize episodes as Episode Treatment Groups (ETG ) or Procedure Episode Groups (PEG ) episodes. Physicians must perform at a level that meets or exceeds the median performance for all physicians (measured in the same specialty for the same types of episodes in the same geographic area) in order to meet the cost efficiency criteria. Complete ETG episodes are attributed to the physician who was responsible for at least 30% of the total costs. The responsible physician must be in a Premium-evaluated specialty that typically manages the care of patients for a given type of episode. Surgeries and certain other procedures among cardiologists, orthopaedic surgeons, and spine surgeons are assessed for cost efficiency through analysis of procedural episodes constructed by PEG, which aggregates paid claims into procedure-based ETG episodes. The unit price of each discrete clinical service, the choice of diagnostic or therapeutic modality, facility costs, and the volume and mix of services used in the episode influence the cost of an episode of care. The cost efficiency evaluation process is described in the flowchart on this page. ETG and PEG software generate ETG and PEG episodes and allow for case-mix and severity adjustments. Episodes are attributed to a single responsible physician. The physician s actual episode costs are put into sets of episodes from other physicians according to the same types of cases and severity levels. The physician s episodes within each set are evaluated against their peer group episodes by ordering the episodes from the lowest to highest cost. The low outliers are removed and the costs are converted into percentiles to allow comparison across different types of cases. Then sets of comparable episodes for all peer group physicians are combined and ranked from lowest to highest percentile. If the sample size is adequate, the Wilcoxon rank-sum test is applied to determine the statistical significance of the individual physician s ranks compared with their peer s median rank, an individual cost efficiency outcome is assigned, and a cost efficiency designation is determined. ETG Episodes Acute and chronic condition-based cases are assessed for cost efficiency through analysis of episodes constructed by Symmetry Episode Treatment Groups (ETG) 2, an Optum tool that aggregates paid claims into condition-based ETG episodes. The unit price of each discrete clinical service, the choice of diagnostic or therapeutic modality, facility costs, and the volume and mix of services used in the episode influence the cost of an ETG episode. The Premium program uses actual allowed costs in order to make actual cost differences transparent to consumers. Costs are determined from the beginning of the episode through completion. An episode ends when there has been no face-to-face encounter for a given period of time, which varies from ETG to ETG. Chronic disease episodes such as hypertension and diabetes, which do not have a clear beginning or end, are typically assessed using a specified time interval, such as a one-year episode. Episodes are classified as complete (fully contained within the data window used for evaluation) or incomplete. Each episode is assigned (attributed) to the physician who generated the highest percentage of services in an episode. About 90% of episodes used for Premium designation are managed by one 2 ETG can refer to the grouper software itself, and also to each of the approximately 450 clinical conditions (ETGs) into which episodes are classified. 14

15 physician. To make sure there was significant involvement on the part of the physician to whom the episode was attributed, all episodes are discarded where the percentage of involvement by at least one physician was less than 30%. The total episode cost includes dollars paid to the physician for direct services, as well as for ancillary and other services for which the grouper's clinical logic determined were related (e.g., medications, diagnostic tests). The responsible physician must be in a specialty that typically manages the care of patients for a given condition s episodes. The Premium program only uses a subset of all available ETGs. ETGs that are not in the typical scope of practice of a given specialty are removed (e.g., bone fractures treated by cardiologists). In addition, because UnitedHealthcare encourages preventive care, the ETGs for routine exams and vaccinations are not included in the cost efficiency analysis. ETGs are also excluded when any service within them is part of a PEG episode used in the Premium program. Some patients have a pharmacy benefit where UnitedHealthcare has the claims data and some do not. We split groups of ETGs into two categories accordingly. In addition, ETG episodes used for cost efficiency evaluation must meet the following criteria: Episodes must be completely contained within the data window (complete episodes). Episodes used are limited to a two-year date span (beginning between 1/1/2010 and 12/31/2011, and complete by 2/29/2012). Episodes must be attributed to eligible physicians in Premium specialties. Episodes must involve eligible members (e.g., hospice patients are not evaluated). Responsible provider must have a minimum 30% of total cost of the episode. Additional information on ETGs is available online at ETG Case-Mix and Severity Adjustment Case-mix and severity adjustments are accomplished by comparing the same type of episodes through the creation of comparable sets of episodes. For a given specialty and market, a set consists of all episodes that share the following characteristics: Condition Treatment Indicator: This shows whether or not there was surgery (e.g., for obstetrics-gynecology conditions, such as post-menopausal bleeding) or active treatment (e.g., thyroid malignancy with or without radioactive iodine therapy). Severity Level: This varies from one to four. These are generated by the ETG grouper for the approximately 120 conditions in which burden of illness was found empirically to influence costs. Pharmacy Benefit: This adjusts for the difference in whether pharmacy costs are included. The entire set of episodes forms the benchmark that allows comparison of an individual physician s episode costs to comparable episodes for their peer groups. By matching the same types of episodes through the above criteria we are able to ensure that only episodes with the same types of cases and patient factors are being compared. The division of episode sets into different base conditions and treatment indicators accomplishes case-mix adjustment. The further division by severity level accomplishes severity adjustment. The pharmacy benefit is included to adjust for otherwise equivalent episodes for patients where there is pharmacy claims information. Those episodes will have higher costs than equivalent episodes for patients without pharmacy claims. One condition might have many subsets of comparable episodes. For example, in a given specialty and market, there could be eight separate sets for diabetes episodes (one for each of four severity levels, each with and without pharmacy benefit). ETG Episode Attribution Complete ETG episodes are attributed to the physician who generated the highest costs, and who was responsible for at least 30% of the total costs. Episodes include all services delivered to a patient, including those of other physicians or clinicians. Episodes include dollars paid to the physician for direct services, as well as for facility costs and ancillary services which the grouper's clinical logic determined were related (e.g., medications, diagnostic tests). The responsible physician must be in a Premium-evaluated specialty that manages the care of patients during an episode. There is no multiple-attribution for cost efficiency. 15

16 PEG Episodes Surgeries and certain other procedures among cardiologists, orthopaedic surgeons, and spine surgeons are assessed for cost efficiency through analysis of episodes constructed by Symmetry Procedure Episode Groups (PEG) 3, an Optum tool that aggregates paid claims into procedure based ETG episodes. The unit price of each discrete clinical service, the choice of diagnostic or therapeutic modality, facility costs, and the volume and mix of services used in the episode influence the cost of an ETG episode. The Premium program uses actual allowed costs in order to make actual cost differences transparent to consumers. PEG creates a PEG episode by identifying an anchor procedure as well as claims that have a temporal and/or clinical relationship to the anchor procedure. Each PEG has a set of specific services that are strongly clinically related to the anchor. These are called targets. For example, an MRI of the knee is considered a target procedure for knee procedures anchors. First, PEG episodes include all the services that are target procedures from 14 days before the anchor procedure through 180 days after. PEG also includes all services (except pharmacy) from related ETGs, from 14 days before the anchor through 42 days after the anchor procedure. The Premium program only uses a subset of all available PEGs. Only those PEGs within the typical scope of practice of a given specialty are used (for example, cardiac catheterization for cardiologists, or joint replacement for orthopaedists). We exclude PEGs for procedures that are only performed as an inpatient (e.g., fusion, lumbar), but where inpatient site of service cannot be verified. We also exclude PEGs if we cannot identify the physician who performed the anchor procedure. Because pharmacy costs make up only a very small percent of PEG episode costs, we do not use pharmacy claim costs in PEG episodes. In addition, PEG episodes used for cost efficiency evaluation must meet the following criteria: Episodes must be completely contained within the data window (complete episodes). Episodes must not abut other PEG episodes in the same procedure category (e.g., among cardiology PEGs within cardiologists, and among orthopaedic PEGs within orthopaedists). PEG episodes whose anchors are redos of other episodes are not used for cost efficiency analysis. PEG episodes are limited to a two-year date span (begin between 1/1/2010 and 12/31/2011, and complete by 2/29/2012). PEG episodes need to be attributed to eligible physicians in Premium specialties. PEG episodes must only involve eligible members (e.g., hospice patients are not evaluated). Additional information on PEGs is available online at PEG Cost Efficiency Case-Mix and Severity Adjustment Case-mix and severity adjustments are accomplished by comparing the same type of episodes through the creation of comparable sets of episodes. For a given specialty and market, a set consists of all episodes that share the following characteristics: Procedure (Anchor) Subanchor (if applicable) APR DRG (inpatient) roll-up category For example, APR DRGs are combined for inpatient PEG episodes with chest pain and angina. Severity of illness level from APR DRG. (There are up to four inpatient severity levels; outpatient episodes are assigned a severity level 1.) The entire set of episodes form the benchmark that allows comparison of an individual physician s episode costs to comparable episodes for the physician s peer groups. By matching the same types of episodes through the above criteria we are able to ensure that only episodes with the same types of procedures and patient factors are being compared. Case-mix adjustment is accomplished by dividing sets into different anchors and subanchors, and there is a unique benchmark set for each anchor, subanchor, and appropriate APR DRG combination. The further division by severity generates a different benchmark set according to each procedure s severity, accomplishing severity adjustment. 3 PEG can refer to the grouper software itself, and also to each of the procedures (PEGs) into which episodes are classified. 16

17 A given procedure might have many subsets of comparable episodes. For example, in a given specialty and market there could be multiple sets for bypass surgery based on the APR DRGs with and without cardiac catheterization and up to four severity levels, generating potentially eight separate treatment sets. PEG Cost Efficiency Episode Attribution PEG episodes are associated with a major therapeutic procedure, or anchor. The physician who submitted the claim for performing the procedure is the responsible physician assigned to the episode. In rare cases in which multiple claims occur on the same day for the same type of procedure with different physicians of equally relevant specialties, the physician with the largest total paid claim cost for the relevant type of procedure on that day is assigned responsibility for the procedure episode. There is no multiple-attribution for cost efficiency. Further processing of Benchmark Sets Episodes for a given condition (ETG episodes) or anchor procedure (PEG episodes) are separated into sets of comparable episodes according to subcategories and severity as applicable. Instead of generating a single average cost value for each of these sets and using that as the market-specialty expected cost, the entire set is retained as the benchmark. These benchmark sets undergo the following processing steps before being used in cost efficiency evaluation: Removing Low Outliers Within each treatment set, we remove all episodes with a cost of less than $ Generally these represent unusual episodes, or potentially, errors. The remaining episodes are ordered from lowest to highest cost and converted into percentiles and all episodes with costs at or below the 5 th percentile are removed. Addressing High Outliers From this point on, only the percentile value ( to ) is used in assessing cost efficiency. This process converts all episodes into a uniform scale, and also means that outliers are capped, since no episode can have a value over 100 ( th percentile). The effect of high outliers is further mitigated by ranking the percentiles and using the corresponding non-parametric statistical test (i.e., a test that does not assume normally distributed data). Sets of comparable episodes must contain at least 50 ETG episodes or 20 PEG episodes, from at least two physicians. Sets that do not meet these criteria are not used further. Creating a Comparison Superset A given physician will treat many conditions among patients with a variety of illness severity. Cardiologists may treat conditions or perform procedures, generating ETG or PEG episodes. Generating cost percentiles within the benchmark sets allows us to assess where physicians episode costs tend to rank compared to other physicians in their specialty and market. Starting with the episodes attributed to a given physician, all sets of comparable episodes into which the attributed episodes fall are identified. The percentiles (originally corresponding to the relative cost of each episode within the comparison set) are combined into one superset of episodes. The superset is then re-ordered by the percentiles of each episode, lowest to highest. The sorted episodes are then assigned a rank from lowest (1) to highest (equal to the total number of episodes in the superset). Physicians in larger markets, treating a larger variety of episodes, could have supersets with tens of thousands of episodes. The physician s own episodes will be spread out among the others in the superset according to their percentiles compared to similar cases and severities. We then use a statistical test to assess whether the rankings are statistically different from the expected ranking of a random episode in the superset, which would be the median rank. To have enough data to evaluate cost efficiency, a physician s comparison superset must have at least ten episodes, and there have to be at least ten other episodes in the superset. Efficiency Outcome Statistical Methods The previous steps generate a case-mix and severity-adjusted superset of episodes containing the physician s observed episode ranks, and the expected ranks for their episodes (the median rank of the superset) within their specialty and market. We then apply the Wilcoxon rank-sum test to determine if the physician s episode ranks are statistically different from the median. The Wilcoxon rank-sum test is a standard test used to determine if a set of rankings is different from the median (expected) rank. The test compares a physician s episode ranks by summing them to generate an observed rank-sum. The expected rank sum is simply the median rank times the number of episodes attributed to the physician. The difference between the expected and observed rank sum is expressed as a number of standard deviations (Z-score). Higher Z-scores indicate higher cost rankings (or lower cost efficiency), and lower Z-scores indicate lower rankings (or higher cost efficiency). 17

18 Interpreting the Wilcoxon Rank-Sum Test Results If the Wilcoxon rank-sum test shows no statistically significant difference from the 50 th percentile, then the physician receives an individual cost efficiency outcome of Meets Expectations. We use a 95% confidence level (p < 0.05), and no statistical difference at the p < 0.05 level (Z-score produced by the Wilcoxon test was and ). If the Wilcoxon rank-sum test shows statistically different performance, it could be because the physician's performance is either higher or lower than expected. If the Wilcoxon test shows lower cost ranks that are statistically significantly different from the 50 th percentile performance, the physician receives an outcome of Exceeds Expectations. If the Wilcoxon test shows higher cost ranks that are statistically significantly different from the 50 th percentile performance, the physician receives an individual quality outcome of Below Expectations. Both outcomes of Meets Expectations and Exceeds Expectations will result in a cost efficiency designation for an individual physician if all other criteria are met. Cost Efficiency Outcome Example The process of assessing cost efficiency for an individual physician begins by grouping all of the physician s episodes into sets of comparable episodes including other physicians within the same market, specialty, and risk-adjustment level. The list is then arranged from lowest episode cost to highest, and the ordered episode costs are converted to percentiles. Separate ordered lists of episodes with percentiles are created for each of the physician s unique combinations of episodes and episode attributes such as severity. In this example, low outlier episodes have already been removed from the sets. For illustrative purposes, this example contains fewer than the actual required minimum number of episodes. Note: The following example describes two different conditions for the fictional Dr. Smith, but the same principle applies to procedure episodes. Treatment Set 1 Diabetes Severity 2 with pharmacy benefit Episode Attribution Episode Costs Percentile Physician 1 $2, % Physician 2 $2, % Physician 2 $2, % Physician 3 $2, % Dr. Smith $4, % Dr. Smith $4, % Treatment Set 2 Hypertension Severity 1 with pharmacy benefit Episode Attribution Episode Costs Percentile Physician 1 $1, % Physician 3 $1, % Physician 4 $1, % Dr. Smith $1, % Dr. Smith $1, % Dr. Smith $1, % Physician 4 $2, % The ordered episode lists with percentiles are then combined to create a single list to create Dr. Smith s comparison superset (Combined Cost Set). The Combined Cost Set includes all of the episodes and the corresponding percentile ranks from the original comparison sets for both the physician being assessed and the physician s peer group. The episode percentile ranks in the larger list are then sorted by percentile from smallest to largest. The sorted percentiles are converted to a numerical rank (from 1 to N, with N being the number of episodes in the Combined Cost Set). 18

19 Dr. Smith s Combined Cost Set Episode Attribution Diabetes Severity 2 with pharmacy benefit Episode Percentiles Hypertension Severity 1 with pharmacy benefit Episode Percentiles Rank Physician % 1 Physician % 2 Physician % 3 Physician % 4 Physician % 5 Physician % 6 Dr. Smith 50.0% 7 Physician % 8 Dr. Smith 62.5% 9 Dr. Smith 71.4% 10 Dr. Smith 75.0% 11 Dr. Smith 85.7% 12 Physician % 13 The number of cases Dr. Smith treats within one set of comparable episodes will contribute to his rank sum, as compared to the number of comparable episodes treated by all his specialty-market peers, achieving case-mix adjustment. The percentiles were originally determined within specific severity levels, creating severity adjustment. By inspection it is apparent that Dr. Smith s episodes tend to be higher cost ranks than those of other physicians in this set. The Wilcoxon rank-sum test confirms that they are statistically significantly higher at a 95% confidence (p < 0.05) level. Expected (Median) Rank for Episode Superset 4 : (5+8+1))/2 = 7 Expected Rank Sum for Dr. Smith 5 : 5 x 7 = 35 Observed Rank Sum for Dr. Smith: = 49 Standard Deviation of the 7 : (((5x8) (5+8+1))/12) = 6.83 Z-score 8 : (49 35)/6.83 = 2.05 The Z-score represents the distance between the physician s value and the expected value in numbers of standard deviations. A negative Z-score means that the value is below the expected value. A positive Z-score means that the value is above the expected value. A Z-score with absolute value greater than means statistical significance at the 95% confidence (p < 0.05) level. Technically, it means that either a high or low score could be different from the median rank (it is a two-tailed test). Since only high scores do not meet or exceed expectations, we are statistically confident that a score in the upper tail is different from the expected value at the 97.5% confidence (p < 0.025) level. For Dr. Smith, the positive Z-score greater than means that his observed cost ranks are higher than expected and that he will receive a result of Below Expectations for the cost efficiency outcome. 4 Standard formula for median 5 Five episodes at the median rank 6 See table above for Dr. Smith s ranks. 7 Standard deviation formula used in the Wilcoxon rank-sum test 8 Z-score calculation for Dr. Smith = (Observed sum of ranks for Dr. Smith) (Expected rank sum for the entire set) / (Standard deviation for the set) 19

20 Important Program Information The information from the Premium physician designation program is not an endorsement of a particular physician or health care professional's suitability for the health care needs of any particular member. UnitedHealthcare does not practice medicine nor provide health care services. Physicians are solely responsible for medical judgments and treatments supplied. The quality and/or cost efficiency designation of a physician does not guarantee the quality of health care services members will receive from a doctor and does not guarantee the outcome of any health care services members will receive. Likewise, the fact that a physician may not be designated by this program does not mean that the physician does not provide quality health care services. All physicians in the UnitedHealthcare Network have met certain minimum credentialing requirements. Regardless of whether a physician has received a designation, members have access to all physicians in the UnitedHealthcare Network, as further described under the member's benefit plan. UnitedHealthcare informs members that designations are intended only as a guide when choosing a physician and should not be the sole factor in selecting a physician. As with all programs that evaluate performance based on analysis of a sample, there is a risk of error. There is a risk of error in the claims data used in the evaluation, in the calculations used in the evaluation, and in the way the program determined that an individual physician was responsible for the treatment of the patient's condition. Physicians have the opportunity to review and reconsider this data. UnitedHealthcare uses statistical testing to compare a physician s results to expected or normative results. There is a risk of error in statistical tests when applied to the data and a result based on statistical testing is not a guarantee of correct inference or classification. We inform members that it is important that they consider many factors and information from as many sources as possible when selecting a physician. We also inform our members that they may wish to discuss designations with a physician before choosing him or her, or confer with their current physician for advice on selecting other physicians. The information contained in this Detailed Methodology is subject to change. Insurance coverage provided by UnitedHealthcare Insurance Company or its affiliates. Health plan coverage provided by UnitedHealthcare of Colorado, Inc., UnitedHealthcare of Oregon, Inc., UnitedHealthcare of Texas, Inc., and UnitedHealthcare of Washington, Inc. or other affiliates. Administrative services provided by United HealthCare Services, Inc. or its affiliates. 20