Association of State COVID-19 Vaccine Mandates With Staff Vaccination Coverage and Staffing Shortages in US Nursing Homes

Key Points Question Are state COVID-19 vaccine mandates for US nursing home employees associated with staff vaccination coverage and reported staff shortages? Findings This cohort study of nursing homes in 38 states found that states with a vaccine mandate experienced an increase in staff vaccination coverage compared with facilities in states with no mandate and no worsening of reported staffing shortages following the mandates. Meaning These findings suggest that given the waning vaccine-induced immunity and low booster dose coverage among nursing home staff in many parts of the US, state mandates for booster doses may be warranted to improve and sustain vaccination coverage in nursing homes.


Analysis
Equation 1 summarizes our main event study analytic approach used to estimate the effect of state mandates on staff vaccination coverage and staff shortages, relative to the changes in outcomes experienced in non-mandate states over the same time period. , = ( ) + ( ) + + + , Eq. 1 , represents the outcome of interest for facility in calendar week . is a facility fixed effect and represents calendar week fixed effects. and are indicators for whether facility is in a mandate state with testing opt-out and without a testing opt-out, respectively. Note that both indicators will be zero for facilities in a state without a mandate. For facility in a state with a mandate, ( ) gives the announcement date of the mandate so that − ( ) gives the event time during calendar week . (Note that we truncate event weeks at 6 weeks prior to mandate announcement-i.e., all weeks prior to 6 weeks before mandate announcement are set equal to -6-and 10 weeks following mandate announcement.) The coefficients and respectively represent our estimate of the effect of mandate announcement on event week . We estimate effects for event-times prior to the mandate as well as a means by which to test the plausibility of our method's assumption: that mandate and non-mandate states would follow similar trends in absence of the mandate. We cluster our standard errors at the state level because mandates are determined at the state level.
Equation 2 summarizes the event study model used to separately examine the effects of mandates in Republican-leaning and Democratic-leaning counties. Specifically, we estimate a version of Equation 1 that includes interactions between mandate effect estimates and an indicator for nursing home being located in a Republican-leaning County ( ) and separate time trends for Republican-and Democratic-leaning counties.

Sensitivity Analyses
To examine the robustness of our results we performed a series of sensitivity analyses. First we estimated overall mandate effects using a differences-in-differences (DID) analytic approach instead of an event study approach (Equation 3). , = , + , + + + , Eq. 3 Where , is an indicator for all calendar weeks following mandate announcement in states without a test-out option, , calendar weeks following mandate announcement in states with a test-out option, and and represents facility and calendar week fixed effects respectively.
, and , are the standard DID estimates of interest, representing the mean change in the outcomes of interest after mandate announcement relative the changes observed in non-mandate states over the same time period.
Second, we use the DID framework to estimates separate effects of mandate announcement and mandate enactment.
Third, we extend Equation 3 to formally examine whether estimates mandate effects differ between Republican-and Democrat-leaning counties within the same state. This is accomplished through use of a triple differences (DDD) model that compares relative changes in outcomes in Republican-leaning counties in mandate states (in comparison to Republicancounties in non-mandate states) to the relative changes in Democrat-leaning counties in mandate states (in comparison to Democrat-leaning counties in non-mandate states) following mandate announcement. Equation 4 summarizes this approach with being an indicator for facility being in a county with a majority vote share going to the Republican candidate in the 2020 presidential election.