Underreporting of Early Nursing Home COVID-19 Cases and Deaths in Federal Data

The initial US epicenter of the COVID-19 pandemic was a nursing home in Kirkland, Washington, in late February 2020.¹ This event was a foreshadowing of the catastrophic and disproportionate toll COVID-19 would take on nursing home residents over the course of the pandemic. However, measuring the true impact of COVID-19 in nursing homes is challenging owing to inconsistent data collection in the early months of the pandemic.

In a new study, Shen and colleagues² examine differences in COVID-19 cases and deaths among nursing home residents reported for those early months from 2 sources: the Centers for Disease Control and Prevention’s National Healthcare Safety Network (NHSN) system,³ and data they compiled from 20 state health departments that tracked and reported nursing home COVID-19 data as of late May 2020. Systematic national data collection did not begin until the week ending May 24, 2020, when the federal government initiated mandatory reporting for nursing homes via the NHSN system. However, while nursing homes were required to report weekly incident cases and deaths from that point on, they were given the option of whether to report cumulative cases and deaths that had occurred prior to the week ending May 24.

The findings from Shen and colleagues² suggest that the NHSN data significantly underrepresent COVID-19 cases and deaths among nursing home residents for the pre–May 24 period. Across the 20 states with available data, they observed that a mean of 44% of COVID-19 cases and 40% of COVID-19 deaths counted by state health departments for that period went unreported in the federal NHSN data. In adjusted analyses accounting for facility characteristics and COVID-19 burden within individual states, they extrapolated their findings for the entire country to estimate that 68 613 cases and 16 623 deaths were omitted from the NHSN data, representing 11.6% of total cases and 14.0% of total deaths across all US nursing home residents in 2020.

The first key takeaway from these findings is that future research using the NHSN data must account for this significant data limitation. Cumulative measures of COVID-19 prevalence or mortality are likely to be flawed without correcting for underreporting, and may particularly bias estimates for nursing homes that had their most severe outbreaks during the initial US wave of the pandemic. A large body of research has already developed examining various facility- and community-level factors associated with COVID-19 outcomes in nursing homes using the NHSN data. The study by Shen and colleagues² provides important context for evaluating the rigor of these studies. The authors also made their corrected facility-level estimates of cases and deaths publicly available for other researchers to use.⁴

The second, more humbling conclusion from these findings is that the true toll of COVID-19 on nursing home residents may never be known. Given the delay in federal data collection, the most accurate data sources for nursing home cases and deaths during the initial US wave of the pandemic are state health departments. Yet only approximately half of US states collected and publicly released nursing home COVID-19 data during Spring 2020, and these states varied widely in the amount and quality of information reported. The estimates by Shen and colleagues² help to quantify the amount of underreporting in the federal NHSN data but they are just that: estimates. These data limitations are only further compounded by the early shortages of testing resources that prevented timely and accurate identification of COVID-19 cases. As such, although current measures of COVID-19 prevalence and mortality reflect a devastating public health crisis for the nursing home population, it is likely that even more lives were impacted that will never be fully captured in existing data.