Electronic Health Record (EHR) Data Analysis Capabilities January 2014 Boston Strategic Partners, Inc. 4 Wellington St. Suite 3 Boston, MA 02118 www.bostonsp.com
Boston Strategic Partners is uniquely positioned to provide insights and value to its clients + Boston Strategic Partners is the only life sciences consulting firm with preferred access to the complete Cerner Health Facts EHR database 14+ years of real world EHR data (from 2000 to present day) 45+ years of combined analytics expertise PhD level experience with SAS data management and biostatistics 2
Cerner Health Facts includes EHR data on over 36 million unique patients across 486 facilities in the United States About Health Facts : Health Facts is a de-identified research database sourced by real-world patient care data from electronic health records (EHRs). Cerner aggregates and maps data contributed by participating facilities and employs quality assurance processes to help ensure data integrity. >36M Health Facts includes comprehensive clinical records comprising pharmacy, laboratory, registration and billing data from all patient care locations. All admissions, discharges, medication orders, clinical events, laboratory orders and specimen collections are time-stamped and sequenced. With Health Facts data, researchers can answer challenging research questions and gain deep insight into real-world practice patterns, clinical decision making and care strategies. Additionally, they can track drug usage relative to diagnoses and across geographic region and hospital type. 3
Cerner Health Facts includes the most common and often most costly conditions Just a few of the conditions included in Health Facts Database Acute coronary syndrome (ACS) Asthma Atrial fibrillation (AF) Cardiovascular disease Chronic obstructive pulmonary disease (COPD) Congestive heart failure (CHF) Complicated skin & skin-related structure infection Cystic fibrosis Diabetes Hip and knee surgery Hypertension Intra-abdominal infection (IAI) Pneumonia (including CAP) Sepsis and bacteremia Stroke/TIA Venous thromboembolic event (VTE) And many more Health Facts includes a complete set of both ICD-9 diagnosis and procedure codes >16,000 Diagnosis Codes >4,000 Procedure Codes Health Facts provides time stamped real world data on de-identified patients Each patient s hospital visit defined as an episode of care This data allows BSP to track and evaluate patients longitudinally across their hospital stay including clinical values, diagnosis, procedures, lab/microbiology, medications and outcomes 4
Health Facts data includes detailed information on most aspects of the patient s hospital visit Patient data is grouped together into an Encounter and includes the following information Facility characteristics Patient characteristics ER encounters as admitting source are merged Clinical assessments including height/weight/bmi vital signs respiratory Clinical characteristics, including key comorbid conditions and selected admission labs 1º and 2º diagnoses and procedure codes (ICD-9) Supplemental diagnoses and procedures (outside inpatient encounters) Inpatient pharmacy orders and dispensing data General lab with or without micro (beyond admission labs) SurgiNet (surgical suite) data, where available Key outcomes (in-hospital mortality, LOS, total billed charges/estimated costs, 30-day readmission) Reimbursement/billing charges EHR-level detailed clinical information, not just orders or claims Longitudinal tracking of patients Date-time stamped to minute-wise resolution 5
BSP has the capability to utilize the Cerner Health Facts data to provide unique insights to our clients BSP s Capabilities Prepare complex models utilizing industry-validated and widely accepted software like SAS Leverage scientific /clinical expertise to apply heuristics that exclude garbage data points as well as identify unique trends Identify patient populations / drill-down based on ICD-9 diagnostic, procedure, DRG codes, demographics, care loci / site-of-service, payor Stratify patients not only by administrative codes but also by quantitative (numeric) clinical and lab values, for example, score severity of illness by APACHE II Step through time to look for patterns during the patient stay, through re-admissions, or otherwise follow the patient over time Identify potential loci of care where medications are administered Leverage our in-house technical expertise to efficiently query multiple fact tables that are each on the order of gigabytes+ and 10MM-900MM+ rows If needed, augment the team with other, in-house professionals who can extend the value of the analysis to support the creation of peer-reviewed papers/posters and drive downstream strategic initiatives Provide easy-to-understand project outputs with explanations using level-of-language specific to the intended audience(s) Data Analysis Methodologies Develop a detailed and comprehensive Data Analysis Plan document prior to data analysis Conduct analysis of baseline patient descriptors (t-test, Χ 2 test, distribution histograms) including patient demographics, hospital demographics and comorbidity data Develop propensity score (probability of receiving a specific treatment) models for patient population using logistic regression Match patients based on their propensity score (1:1 and N:1) to control for bias and confounding that is inherent in retrospective clinical data Apply standard clinical guidelines to develop criteria for key outcomes Conduct in-depth analysis of clinical outcomes using multivariate analysis Validate the statistical significance of reported outcomes using odds ratios (ORs) with confidence intervals (CIs), C statistics and receiver operating characteristic (ROC) curves Effects of specific patient characteristics on any outcome, such as age and severity of illness can be estimated by including these parameters explicitly in the outcome models 6
Potential applications of EHR include support for products that are indevelopment as well as evaluation of marketed products In Development Analysis Understand current practices and event rates to guide and expedite clinical trial design Analyze unmet needs by understanding a patient s journey from admission to discharge Understand how health care is delivered today over a longitudinal basis Understanding the economics of healthcare Analyze outcomes On Market Analysis Analyze sequences of events for patients of interest leveraging time stamped data Identify initial treatment prescribed and subsequent modifications to treatment over the length of stay Identify adverse events defined by treatment or labs Interactions between concomitant medications Interactions between comorbid clinical conditions Ability to leverage rich clinical data for severity-of-illness adjustment T O S U P P O R T R&D / Medical Affairs Clinical trial design Patient selection Event rate assessment Comparative effectiveness evidence generation Better understanding of disease and care pathways Improved post-marketing surveillance capability Clinical messaging Meta-analyses and publications Understanding how drugs are actually utilized in real world scenarios (including off-label use) Health Economics Health economic modeling: translation of proposed clinical differentiation into economic outcomes to support pricing: budget impact models and cost effectiveness analyses Improve coverage and payments Commercial/ Marketing Value proposition identification and quantification Prescriber and/or patient segmentation E.g. lines of therapy Forecasting Pricing studies Competitive positioning Life cycle management 7
EHR data mining provides a wealth of information that can help modernize the research process, create more efficient clinical trials and improve the results of marketing efforts EHR-based analysis can improve efficiency, reduce costs and provide a competitive advantage in postmarketing surveillance and clinical trial cycle time Source: Evans, Stemple: Electronic Health Records and the Value of Health IT. Journal of Managed Care Pharmacy, 14:6, S-c, 2008. 8
The average delay from missing enrollment deadlines is about 90 days for which each day of delay costs an estimated $1.3 million in lost sales Example: Slightly altering the inclusion / exclusion criteria could dramatically increase the patient population and in turn help further reduce recruitment cycle times In addition, prior clinical and diagnostic data could help improve clinical trial design through a more comprehensive understanding of disease progression and care pathways Source: 3rd Annual Merging Electronic Health Records and eclinical Technologies: Leveraging EHRs for Clinical Research Considerations for tomorrows technology, Sept 24-25, Annapolis 9
Health Facts data includes detailed data on most aspects of the patients hospital visit 10
Data Facilities Type data includes: US Census Region and Division (all represented) Bed size category Teaching status Urban/rural community setting Part of hospital system Acute care status Cardiac cath lab status Full cath lab Diagnostic cath lab only Statistically derived cost-to-charge ratio available 11
Admissions and Discharge Date Admissions and Discharge data includes: Type of encounter Physician specialty Date/time of admission Admission type (e.g. elective) Admission source Date of discharge Discharge disposition 12
Data data includes: Patient identifier Age at admission Race Gender Payer 13
Data data includes: Height Respiratory rate Weight Smoking status Blood pressure Alcohol use Heart rate Pregnancy status Pulse Allergies Temperature Apgar LOC / Glasgow Coma Scale Symptoms BMI Pain assessment 14
Data data includes: 1º and 2º diagnoses and procedure codes (ICD-9) Supplemental diagnoses and procedures (outside inpatient encounters) Discharge diagnoses, procedures 15
General Labs Data General Labs data includes: Procedure name Specimen source and type Date / time of order, collection, and completion Result and unit of measure Type of result Normal ranges, if applicable Ordering physician specialty Treatment setting 16
Microbiology Susceptibility Data Microbiology Susceptibility data includes: Procedure name Positive isolate name Anti-microbial used for testing Result Specimen source Date / time of order Date / time of collection Date / time of performance Date / time of verification 17
Data Medication data includes: Drug name Drug class Ordering physician specialty Treatment setting (on order) Dose, rout of administration Order frequency Total quantity dispensed Start and discontinue dates 18
Data data includes: Total billed charges Cost to charge ratios 19