Revealing Within-Agency Variation in Medicare Home Health Quality Ratings

TL;DR

This study shows that average quality ratings for home health agencies can hide big differences in care for different patient groups, which could help improve consumer choices and provider performance.

Contribution

The study introduces a method to reveal hidden variation in home health quality ratings across patient subgroups using existing data.

Findings

Agencies with average to high overall ratings can have significant variation in care outcomes across patient subgroups.

Disaggregated ratings could improve usability for marginalized groups and support data-driven improvements.

Existing administrative data can be used to generate more detailed quality information.

Abstract

The Centers for Medicare & Medicaid Services (CMS) publicly reports Home Health Agency (HHA) quality ratings using a composite 5-star system based on seven assessment- and claims-based quality indicators. These ratings are intended to guide consumer decision-making and encourage performance improvement among providers. However, because the ratings reflect agency-level averages, they may mask important variation in the quality of care delivered to different patient subgroups. This study examines within-agency variation in reported performance using national Medicare administrative, assessment, and claims data. We applied CMS’s published star rating methodology to simulate overall and group-specific quality ratings and assess the reliability and informativeness of disaggregated quality ratings. Our findings show that agencies with average to high overall ratings (3–5 stars) can exhibit substantial variation in care outcomes across patient subgroups. This variability is obscured in the reported rating, limiting its utility for some consumers and stakeholders, typically those from marginalized groups. By identifying patterns of intra-agency variation, this analysis highlights opportunities to enhance the information provided by publicly reported quality measures to increase usability for all patient groups. These refinements could better equip consumers to make informed choices and support data-driven improvement efforts among providers. This work demonstrates the feasibility of generating more detailed quality information using existing administrative data sources and methods. These findings provide timely insights into how quality measurement tools can evolve to capture meaningful variation across different patient populations—enabling a clearer understanding of what works best, for whom, and under what conditions.