Data-driven Investigation of U.S. Nursing Homes

Recent initiatives by the CMS (Centers for Medicare and Medicaid Services) have bolstered the accessibility of data and evaluations regarding the performance of the over 20,000 long-term care facilities for the elderly, or nursing homes, across the U.S. In particular, the CMS has sought to utilize both public ratings and financial incentives to improve the quality of care in U.S. nursing homes.

At the same time, data science and analytics have revolutionized the delivery of services, especially in healthcare. Our proposed research aims to develop and use data analytics to advance U.S. nursing home quality of care in two main areas of interest.

First, policymakers and academics have long advocated for public reporting in order to drive market mechanisms for improving care quality levels. Yet, to date, little evidence convincingly supports that public reporting improves consumer choices or moves demand. Our aim is to help close this gap between practice and promise by studying the optimal design of policies that employ inspections and disclosures to regulate the marketplace for long-term care. In particular, how should inspection efforts be allocated in order to generate the most valuable information regarding nursing home quality of care? Relatedly, what mix of performance-based incentives and disclosures most efficiently raises nursing homes’ quality of care?

Second, nursing homes have faced a nationwide staffing crisis over the past twenty years. We examine the empirical linkages between short-staffing and patient abuse or malfeasance in nursing homes, with a special focus on the abuse of antipsychotic drugs.

These efforts are united by the need for novel data assembly and collection. We have already begun construction of a dataset covering virtually all U.S. nursing homes from 2012 to 2020. The data include critical agency data obtained from CMS through Freedom of Information Act requests, and we plan to utilize the requested funding to assemble key data regarding nursing homes’ individual characteristics and operations. A mix of statistical analyses, including causal inference techniques, and analytical methods, such as Markov decision models, will be used to analyze the dataset and produce actionable insights.