Population Health Management through Advanced Risk Stratification Analytics

Transcription

1 Population Health Management through Advanced Risk Stratification Analytics Montana HIMSS Spring 2016 Dan Ulatowski, Advisory Services

2 Goals of the Session Provide a deeper appreciation for the role that a data-rich risk stratification strategy can take in steering a successful population health program. Each effort to cut costs and improve care comes with potentially decreased margins, and therefore we will discuss what it takes to project return on investment of these programs differently from a fee for service care model. Further points of knowledge include (1) how our current adoption of risk stratification progressed, (2) why our industry has opportunity to mature in this area, and (3) how to do more with less by increasing the specificity, and therefore efficacy, of your population health management efforts. Page [ 1 ]

3 Industry Problem Health systems and provider practices are seeing increased pressure to manage their patient population through alternative payment models and performance criteria. The Medicaid Access and CHIP Reauthorization Act will increase the incentive for an organization to pursue such models. However, approaches to defining the population for an accountable care initiatives to improve outcomes or prevent disease exacerbation varies greatly in maturity and approach. We will explore the various models and provide the audience with a greater understanding of how intersecting statistics and care coordination drives patient outcomes. Page [ 2 ]

4 History to Current State Current Example Methods: Adjusted Clinical Groups (ACG) Hierarchical Condition Categories (HCC) Elder Risk Assessment (ERA) Chronic Comorbidity Counts (CCC) MN Tiering Charlson Comorbidity Measure Page [ 3 ]

5 CMS HCC Model ACA Mandated Evaluation The CMS hierarchical condition categories (CMS-HCC) model, implemented in 2004, adjusts Medicare capitation payments to Medicare Advantage health care plans for the health expenditure risk of their enrollees. Its intended use is to pay plans appropriately for their expected relative costs. For example, MA plans that disproportionately enroll the healthy are paid less than they would have been if they had enrolled beneficiaries with the average risk profile, while MA plans that care for the sickest patients are paid proportionately more than if they had enrolled beneficiaries with the average risk profile. For 2011 and periodically thereafter, the Secretary shall evaluate and revise the risk adjustment system under this subparagraph in order to, as accurately as possible, account for higher medical and care coordination costs associated with frailty, individuals with multiple, comorbid chronic conditions, and individuals with a diagnosis of mental illness, and also to account for costs that may be associated with higher concentrations of beneficiaries with those conditions. Page [ 4 ]

6 Landmark Paper by National Quality Forum (2014) There is a large body of evidence that various sociodemographic factors influence outcomes, and thus influence results on outcome performance measures. Sociodemographic Status (SDS) refers to a variety of socioeconomic (e.g., income, education, occupation) and demographic factors (e.g., age, race, ethnicity, primary language). There also is a large body of evidence that there are disparities in health and healthcare related to some sociodemographic factors. Given the evidence, the overarching question addressed in this project is, What, if anything, should be done about sociodemographic factors in relation to outcome performance measurement? Page [ 5 ]

7 Fee for Service to Value Based Care ROI Hurdle Page [ 6 ]

8 Value Based Care Modeling: Leveraging Associations Association as a root thought process for establishing Risk Stratification models. An association is present if probability of occurrence of a variable depends upon one or more variable. (Non-Linear Modeling) (Dictionary of Epidemiology by John M. Last) Synonyms: correlation, statistical dependence, relationship Includes events, characteristics, or variables. Page [ 7 ]

9 Types of Association Direct direct association i.e. not via a known third variable Salt Intake Hypertension Indirect associated through a known third variable Salt Intake Hypertension CAD Page [ 8 ]

10 Determinant Association Applications Type Example Application Absolute Difference AR (Attributable Risk) Primary prevention impact; search for causes PAR (Population Attributable Risk) Efficacy Mean Differences (Continuous Outcome) Primary prevention impact Impact of intervention on recurrences, case fatality Search for Determinants Relative Difference Relative Risk/Rate Search for Causes Relative Odds (ODDS ratio) Search for Causes Page [ 9 ]

11 Value Based Care Justification Attributable Risk (AR) AR is the portion of disease incidence *in the exposed* that is due to the exposure. Therefore = the incidence of a disease *in the exposed* that would be eliminated if the exposure were eliminated Calculation of AR = risk(incidence) in exposed risk(incidence) in nonexposed which provides the risk difference Page [ 10 ]

12 Use of Population Attributable Risk Population Attributable Risk (PAR) This is a similar measure to AR except it is concerned not with the excess rate of disease *in the exposed* but the excess rate of disease *in the population* (compared with the rate of disease in the exposed group) PAR is the proportion of the disease incidence *in the population* (i.e. exposed and non-exposed) that is due to the exposure Therefore it is the disease incidence *in the population* that would be eliminated if the exposure were eliminated Calculation of PAR = risk(incidence) in population risk(incidence) in nonexposed Page [ 11 ]

13 Use of Population Attributable Risk % Population Attributable Risk % (proportion or fraction) PAR% is the proportion of disease incidence *in the population* (i.e. exposed and non-exposed) that is due to the exposure Therefore it is the % of disease incidence *in the population* that would be eliminated if the exposure were eliminated Calculation of PAR% = PAR / risk(incidence) in population When data on disease incidence is not available we can use the RR Calculation of PAR% = prevalence in exposed population x (RR-1) / [1+ prevalence in exposed population (RR-1)] Page [ 12 ]

14 PAR Calculation Page [ 13 ]

15 PAR Calculation Page [ 14 ]

16 PAR Calculation Page [ 15 ]

17 Attributable Risk and Relative Risk Page [ 16 ]

18 Risk of Exposure: PAR Page [ 17 ]

19 Risk Stratification: Image/Demo Page [ 18 ]

20 Test Data Process: CHF Predictive Unit Page [ 19 ]

21 R Statistical Programming Build predictive models in R scripts, which are embedded in SQL stored procedures R has a rich library of predictive analytics algorithms. The stored procedures are executed within the database server, so are the R scripts Call a simple predict() function in R to evaluate the model on validation data, and use another single function to extract performance metrics Page [ 20 ]

22 Adding Determinants to CHF Predictive Model: Behavioral, Social, Geographical Page [ 21 ]

23 Current Trends in Predictive Modeling Toolkits Page [ 22 ]

24 Future Approach to Establishing Analytics Models Page [ 23 ]

25 Page [ 24 ]

26 CHF Model Azure ML Page [ 25 ]

27 About the Presenter Dan Ulatowski Divurgent Advisory Services Mr. Ulatowski is an epidemiologist having practiced at the Mayo Clinic with a track record of improving organizational data processes and strategic initiatives. His current role at Divurgent leverages both his clinical as well as business intelligence experience, serving clients as a trusted advisor for organizations looking to make the next step in accountable care. He has implemented many HIT systems for large IDNs at Epic Systems, as well as smaller FQHCs through boutique firms prior to focusing on strategic ACO and data-driven programs. Dan focuses his industry efforts bringing on quality outcomes and data management closer together, dedicating time to the National Association of Healthcare Quality's Q Essentials solution development program, as well as actively contributing to CMS's VBP and PQRS measure development. Dan holds a degree in Medical Microbiology and Immunology from the University of Wisconsin. He also holds a certificate of program graduation from The Data Warehouse Institute (TWDI), and is an active presenter for his HIMSS chapter. He advises the board of the Washington Healthcare Access Alliance on vendor and technology grants available to their free clinics. Dan.ulatowski@divurgent.com Divurgent 4445 Corporation Lane, Ste. #228 Virginia Beach, VA Divurgent is dedicated to improving healthcare by delivering client-focused solutions resulting in improved patient outcomes and enhanced operational efficiencies for our clients and the communities they serve Page [ 26 ]