Actionable Taxonomy to Identify Patients with High Healthcare Utilization and Potential Interventions

Background & Unmet Need

• The top 10% of healthcare-utilizing patients account for 50% of healthcare spending
• This high-need, high-cost category of patients is highly heterogenous and difficult to identify with current taxonomy
• Current systems of identifying these high-touch patients rely on claims data, which do not incorporate important aspect of social circumstance
• Moreover, current classifiers limit patients to mutually exclusive groups, giving an incomplete picture of their medical needs
• Unmet Need: A method to classify high-need, high-cost patients into actionable groups which incorporate the complex and numerous determinants of health

Technology Overview

• The Technology: A new taxonomy which integrates both claims data and social determinants of health to identify high-cost patients and potential interventions
• The new groups are classified by differentiated patient attributes, including the presence of chronic conditions, substance use disorders, mental illness, and social vulnerability
• PoC Data: In a cross-sectional study of a Medicare fee-for-service cohort in NYC, patients were sorted into 10 overlapping categories
• The study identified “multiple chronic conditions” as the category with the most high-cost patients, and found that 73% of high-cost patients fell into multiple of these categories
• In another study, the group found that patients with both the highest preventable utilization and highest costs represented only 1.9% of patients but 33% of preventable costs among Medicare patients

Technology Applications

• Analyze patient populations to identify targeted interventions for those with high healthcare utilization
• Inform strategies for improving population health outcomes and healthcare delivery
• Reduce the burden of healthcare on traditional providers by identifying areas of alternative care

Technology Advantages

• Classifiers integrate data from both claims and social health data sources
• Datasets integrate more current data sets as well as longitudinal data
• The taxonomy can help identify opportunities for targeted interventions among patient populations