Abstract
Methods: Four years of panel data (2002-2006) from all general acute care units and intensive care units (ICUs) in Department of Veterans Affairs (VA) hospitals were used. Using average patient length of stay (LOS) as the outcome, fixed effects, multivariate regressions controlling for known confounding variables were estimated with robust standard errors clustered by unit. A cost-benefit calculation was computed.
Results: 161 ICUs (7,148 monthly observations) and 266 acute care units (11,637 monthly observations). The mean nursing hours per patient day (HPPD) and LOS varied by unit (HPPD: ICU 18.2 vs acute care 7.9; LOS: ICU 3.0 days vs acute care 4.9 days). A one hour increase in HPPD was associated with LOS reductions of 1.6% (95% CI 1.3-1.9, p<0.001) in ICUs and 3.1% (95% CI 2.6-3.5,p<0.001) in acute care units. Both were approximately linear across the entire range of staffing levels. Increasing the HPPD by 1 hour per day would save $350 per ICU and $160 per acute care. An increase in contract nurses from 0% to 10% of nurse staffing was associated with 3.3% (95% CI 0.8-5.8, p<.0.001) and 3.1% (95% CI 2.1-4.1, p<0.001) longer LOS for ICUs and acute care units, respectively. We also found that failure to control for unobserved heterogeneity results in a doubling of the estimated HPPD effect.
Conclusions: Hospitals should minimize the use of contract nurses while maintaining nurse staffing levels.