Providers are experts at creating relationships in an individual patient setting, but how can those powerful relationships be extended across thousands of patients in a population? Value-based care requires providers to be responsible for the quality of care they deliver, as well as the overall health and cost of their entire patient population. Balancing the demands of value-based care with an understanding of unique patient needs is best accomplished through data analytics.
The following five pillars explore critical aspects of data analytics, including risk identification and stratification. These pillars extend the reach of physicians, and drive stronger patient relationships and measureable outcomes for providers developing population health management strategies in value-based arrangements.
1. Build a data architecture and aggregate claims data
Just as physicians need to know the patients visiting their office, provider organizations need to have a deep understanding of their entire patient population. Creating relationships with your population starts with data architecture. Build a patient index with all individuals attributed to the practice, and verify that all patients have been assigned to a primary care provider.
Once the necessary architecture has been established, aggregate a comprehensive claims data set for the patient index that includes all healthcare utilization across the care continuum (including care outside your organization). Physicians using claims data have a 360-degree view of patient care, including in- and out-of-network utilization.
2. Understand patient and population risk
Now that you know who your patients are, you need to diagnose their risks. Predictive modeling and risk analytics are powerful tools that can help manage existing patients or identify new patients at risk of adverse health outcomes. Provider organizations have numerous risk model options from expert vendors as well as the public domain.
Consider the characteristics of your attributed patients when selecting risk models. Commercially insured patients have a different profile from a Medicare population, just as the drivers of cost and utilization are different for employed versus retired patients. The best predictive models are tuned to the profiles of commercial, Medicare, and Medicaid patients, and also include risk adjustment and benchmarking. With the right models, you can compare the burden of disease within your patients to regional and national standards. You can answer such questions as, “How do hospitalization rates compare for different medical practices within my population?” or “How costly are my diabetic patients compared with the regional benchmark?”
3. Stratify the population
You’ve identified your patients and their risk levels, but how do you organize care based on this data? Stratification ensures that providers effectively align patients for ongoing care. While prospective risk models help identify patients most likely to require hospitalization and utilize services in the coming year, patient stratification provides more granular segmentation of the population according to risk, disease, quality metrics, and other factors.
A strong combination of risk modeling, condition diagnoses, and utilization patterns helps create actionable patient panels and subgroups that can be managed through targeted programs such as case management or home care interventions. Stratification allows providers to easily monitor risk trends, utilizations, quality metrics, and outcomes specific to each cohort. By breaking high-risk patients into smaller groups that share characteristics, the creation of management programs will flow naturally.
4. Evaluate and use additional sources of data
In-office care planning is based on patient data as well as personal dialog with the patient. Similarly, provider organizations managing large populations will have a more complete understanding of each patient by integrating claims data with additional sources such as lab results or health risk assessment (HRA) surveys. The inclusion of additional data sources helps capture information unique to each patient, such as motivation to improve health or willingness to engage with a physician.
Additional data sources can also flag significant risks that may slip past claims data. For example, obesity is highly predictive of poor health outcomes, yet numerous studies have shown that, in many populations, fewer than half of obese patients are coded by physicians. HRA, patient activation scores, lab, and biometric data sources can help identify patients with significant risks or care gaps.
5. Monitor the impact of your initiatives
Once data aggregation, risk assessment, and stratification are implemented, the analytic process has just begun. It is crucial to rigorously track cost, quality, and utilization across all interventions and readjust programs based on outcomes. Evaluate the impact of care management programs and other population health management efforts through ongoing analytics. Constantly build on your knowledge as you determine which interventions lead to improvements in both process measures and patient outcomes. Use what you have learned to modify your interventions so patient outcomes continue to improve.
Effective provider-patient relationships start with an understanding of each patient’s needs. Risk identification and stratification enable providers to continuously and effectively uncover these needs across large populations where face-to-face care is not always possible.
