COVID-19 and nursing homes: Understanding state-level variation

Nursing homes have borne the
brunt of the COVID-19 pandemic, but recently released data show that the
severity of outbreaks in these facilities has varied substantially across the United
States. Some have argued that policy decisions have driven the variation in
outcomes observed in nursing homes, while a competing theory is that nursing
home outbreaks largely mirror the surrounding area. In this blog post, we
measure how closely outcomes within nursing homes track with outcomes in
non-nursing home populations within states.

District nurses Rebecca McKenzie and Emma Fiello leave the house of 86-year-old patient Margaret Ashton after a home visit in Grimsby, Britain June 9, 2020. Picture taken June 9, 2020. Daniel Leal-Olivas/Pool via REUTERS

From the onset of the COVID-19 pandemic, states have varied in their policy responses, including those affecting nursing homes. New York, for example, initially required facilities to accept patients even if they had been diagnosed with COVID-19 — a decision that some suggest could have triggered greater spread in the state’s nursing homes. Similar orders were issued in states like Michigan and New Jersey. One way to test whether state policies have driven nursing home outcomes is to compare outcomes within nursing homes to those in the general population within a state.

For our analysis, we use data
on state-level fatalities to measure the severity of COVID-19 outbreaks across
states. While this choice avoids challenges associated with testing variation,
it introduces some alternative concerns, particularly differences in
demographics across states that could lead to higher or lower fatalities for a
given infection rate. However, our results are similar if we use the number of
COVID-19 cases instead. This analysis also excludes potentially important
facility-level variation, which we will explore in future blog posts.

We use state-level data on COVID-19 fatalities from the COVID Tracking Project, while data on COVID-19 nursing home deaths come from the Centers for Medicare & Medicaid Services (CMS). As we described in an earlier blog post, we remove from the CMS data all nursing homes where facilities report more deaths than residents. To measure non-nursing home fatalities, we subtract deaths of nursing home residents from the overall state total. To calculate per-capita fatality rates, we adjust the overall state population by the total number of nursing home residents in a state, as reported by the Kaiser Family Foundation.

Results are shown in Figure 1,
which reveals a very high correlation between the average nursing home outcomes
and those of the state more broadly. In most states, overall death rates are
relatively modest, and nursing home outcomes track closely. A simple linear regression
finds that roughly 80 percent of the variation in state-level nursing home
fatality rates can be explained by the state’s overall COVID fatality rates.

Note: State nursing home population are taken from Kaiser Family Foundation. State level data on COVID deaths are taken from the COVID Tracking Project (downloaded June 25). Data on nursing home deaths come from CMS (downloaded June 25). State populations for 2019 are from Census. We exclude observations where facilities report more than 1,000 deaths per 1,000 residents.

It is evident that facilities
in some states, especially Massachusetts and Delaware, have particularly bad
outcomes within nursing homes relative to the population more broadly.
Surprisingly, New York is the one major outlier where outcomes within nursing homes
are relatively good compared to the population more broadly. If the data have
been reported accurately, this raises the question of whether New York’s
controversial policies were as detrimental to nursing home residents as some

With COVID-19 outbreaks now accelerating in a number of states, like Texas, it will be important to track whether these spill into nursing home populations.