Nursing-home level differences in mortality and morbidity

Nursing home quality is a major public concern in many developed countries, which in the Netherlands recently resulted in the government’s pledge to spend an additional 2 billion euro per year to improve nursing home staffing ratios. We study how this quality varies by exploiting Dutch administrative data. Prior research from the US shows large, persistent differences in quality but has two major limitations: these studies i) lack data on objective measures to judge nursing home quality and ii) cannot fully rule out confounding by variation in patient characteristics.

We solve the first problem using Dutch administrative data from the period 2009-2013 containing information on two ultimate outcome measures: patient mortality and morbidity. The second problem is mitigated because, when applying for nursing home care, elderly in the Netherlands are subjected to a uniform assessment of nursing home eligibility, which yields detailed information on their needs. This key information is complemented by data from other administrative sources on the health and healthcare use of nursing home resident prior to their admission. We select relevant variables from this extensive data using machine learning techniques.

We find substantial differences across nursing homes for both outcomes. We also find that a nursing home’s performance on mortality is not related to that on hospital admission. Using administrative measures of final outcomes to estimate differences in performance across nursing homes has potential. At the same time, caution seems warranted in evaluating the performance of individual nursing homes based on these outcomes, and most likely on any quality measure.