Moving toward more Real-Time and Efficient Value- Based Care Analytics
By Doug Thompson, Director of Advanced Analytics, Rush Health
Doug Thompson, Director of Advanced Analytics, Rush Health
Value-based care (VBC) contracts are designed to incentivize providers to manage healthcare costs while maximizing healthcare quality for a specific set of individuals (“members”). VBC is currently an area of active experimentation by healthcare providers and payers. Many VBC models include a mixture of shared savings, global and partial capitation, episode-based payments and bonuses around specific quality and utilization metrics.
Effective analytics are essential for success in VBC. The analytics needed include provider performance scorecards, member risk stratification, intervention outcome evaluations, forecasting, quality dashboards and algorithms to improve diagnosis coding accuracy. Payers that create VBC contracts typically supply their provider partners with data and reports based on healthcare claims data, to highlight opportunities and provide performance feedback. The data and reports supplied by payers often have limitations including time lags (latency inherent in healthcare claims data), lack of personal ID numbers that tie across data sources and absence of important clinical details in the data (e.g., lab test results).
"Because the mix of VBC contracts changes over time, the ability to create or modify custom analytics to meet current needs is crucial to success"
Providers participating in VBC contracts have the opportunity to make analytics more “real time” by supplementing healthcare claims data with electronic medical records (EMR). EMR data has many strengths. The data has minimal latency (typically, EMR data in analytic warehouses is refreshed daily), EMRs have excellent patient coverage (providers strive to keep accurate EMRs for all patients) and EMR data include a great amount of clinical detail. While parts of the EMR data are unstructured, much of the most useful information in EMR data is structured and relatively easy to use for analytic purposes (e.g., detailed service descriptions and dates, service location, lab test results, patient vitals, providers with various roles in the encounter).
EMR data is excellent in forecasting and prediction, early utilization monitoring, physician quality dashboards and identifying high-risk patients, among other analytic purposes.
Despite the strengths of EMR data, there are some challenges to incorporating EMR data in VBC analytics. Payers typically supply providers with lists of members participating in a VBC program. However, these member lists often include sparse member information which is subject to change and inconsistent formatting, and member ID numbers (when available) may not tie to member ID numbers available in the EMR data. Sometimes it is necessary to match EMR data to member lists using cleaned data on name, gender, date of birth, primary insurance type and other member information as available. Because such matching can be complex, it important to test different matching algorithms and to carefully validate the results.
A second and more significant challenge to incorporating EMR data in VBC analytics is absence of records on services from other providers in the EMR data. Unlike healthcare claims data, which include all services that a member receives from any provider, EMR data typically only include services that the member receives from one provider, or the set of providers who work for a single health system or group. There is typically no EMR data available for services rendered by other providers. It could be argued that EMR data, which represent services received from the VBC provider’s own office, group or health system, are most important because they are most easily impacted by the provider’s actions. That is, it is easier for a provider to influence the healthcare of their own patients than it is to influence the healthcare of patients seeing other providers. Following the old adage that the best way to deal with missing data is not to have any, the best way to overcome gaps in EMR data is to perform excellent care coordination and strong primary care, which often decreases the need for members to get services from many different providers, in turn leading to more complete available EMR data on the services that members receive.
Due to challenges such as those mentioned above with combining healthcare claims with EMR data for VBC analytics, many vendors offer tools and support to providers in this area. While vendor solutions might be right for some organizations, they can be costly and inflexible with steep implementation fees for adding new data sources or other change orders.
Presence Health is the largest Catholic health system in Illinois, with 11 hospitals and over 150 sites of service. Currently, Presence Health provides care coordination and other services for members in a wide variety of VBC arrangements. At Presence, our experience is that it is feasible (as well as cost-efficient) to create customized VBC analytics, incorporating both healthcare claims and EMR data, using primarily internal resources. Important components of being able to do this included the following: 1) availability of EMR data (Epic and Meditech) structured to be consumed for analytic purposes and refreshed daily; 2) availability of sufficient server space to process and combine healthcare claims and EMR data; 3) programming and analytic software (especially SAS and SQL) to process the data and create analytics and reports; and 4) staff with programming skills and experience creating custom analytics using a variety of healthcare data.
We have found that in addition to minimizing costs of vendor consulting and tools, this setup increases ability to customize analytics to meet current needs and quickly created new analytics as needed. Because the mix of VBC contracts changes over time and the terms of the contracts change frequently (often annually), the ability to quickly and cost-efficiently create or modify custom analytics to meet current needs is crucial to success.