Mukamel DB, Saliba D, Ladd H, Konetzka RT.
Objective
Nursing homes make staffing decisions in conjunction with choosing quality goals, potentially leading to endogeneity bias between staffing and quality. We use instrumental variables (IVs) to explore it.
Design
Retrospective statistical analysis of 2017–2019 Payroll-Based Journal, Minimum Data Set, Nursing Home Care Compare, and Long-Term Care Focus.
Settings and Participants
A total of 11,261 nursing homes nationally.
Methods
We estimated separate models for each of 6 quality measures as dependent variables, and registered nurses (RNs), certified nurse assistants (CNAs), and licensed practical nurses (LPNs) as independent variables, including other control variables associated with quality. The models were estimated using both ordinary least squares (OLS) and 2-stage least squares (2SLS) methods, the latter accounting for endogeneity. The IVs were defined as the average staffing of competing nursing homes in the same market as the index facility.
Results
Estimated coefficients for the quality measures in the 2SLS models were up to 5 times larger than in the OLS models. The 2SLS estimates for antipsychotic medications use increased with higher RN staffing (0.279 [0.004 to 0.553]) and decreased with higher CNAs (−0.125 [−0.198 to −0.052]). Hospitalizations decreased with more RNs (−1.328 [−1.673 to −0.983]) and LPN staffing (−0.483 [−0.755 to −0.211] and increased with CNA (0.201 [0.109 to 0.293]) staffing. Emergency room visits decreased with higher RNs (−1.098 [−1.500 to −0.696]) and increased with CNAs (0.191 [0.084 to 0.298]). Long-stay activities of daily living (−0.313 [−0.416 to −0.209]) and short-stay functioning (−0.481 [−0.598 to −0.364]) improved only with higher CNA staffing and pressure sores improved only with increased RN staffing (−0.436 [−0.836 to −0.035]).
Conclusions and Implications
Our findings demonstrate the importance of accounting for endogeneity in studies of staffing and quality. Endogeneity changes conclusions about significance, direction, and magnitude of the relationship between staffing and specific quality measures. These findings highlight the need to further study and understand the nuanced relationship between different staffing types and different health outcomes such as the difference between the relationship of RN and CNA hours per resident day to antipsychotic quality measures.