RN staffing and patient safety events in Florida hospitals 1996-2004: New variable and longitudinal approaches

Objective: The majority of studies investigating the relationship between nurse staffing and patient outcomes find that worse patient outcomes are associated with lower RN staffing. Yet inconsistencies exist within and between studies, possibly due to data and measurement issues, and because studies have not modeled causation, which requires, among other things, a model of change among the variables. This study applies new patient outcomes measures developed by the Agency for Healthcare Research and quality (AHRQ) in a growth curve model (GCM) in order to assess the impact on patient safety events when RN FTEs are reduced in hospitals, or when the RN to patient days of care (RN/APDC) decreases.

Methods: We use 9 years (1996 through 2004) of patient, hospital characteristic, and nurse staffing administrative data from Florida hospitals. Six AHRQ Patient Safety Indicators (PSIs) (decubitus ulcers, failure to rescue, pneumothorax, infection, post-operative respiratory failure, post-operative sepsis), and a composite measure, “patient safety events” (PSEs) (derived from principle component analysis of the PSIs), are extracted from patient-level administrative data and aggregated to hospital-level rates. In the GCM, PSIs/PSEs and RN FTEs are endogenous and time-varying, and RN/APDC and most control variables are exogenous and time varying. The waves of endogenous and time-varying exogenous variables are correlated with latent change trajectories for those years. Correlations between the time-invariant exogenous variables and the latent change variables are also analyzed. The procedure uses a multivariate system equation modeling (Mplus) with maximum likelihood estimation methods.

Results: We find that some PSIs were significantly related to staffing whereas others were not, or were related counter-intuitively. Change relationships show more significance than static relationships. RN/APDC was a weaker predictor of patient outcomes than RN FTEs. Decubitus ulcers, failure to rescue, infection, post-operative respiratory failure, and post-operative sepsis are significantly related to one or both staffing measures. Of these five, failure to rescue has the strongest relationship to both staffing measures. Causal negative relationships were not found between staffing and either pneumothorax or the composite PSE.

Conclusions: Although our results support those of other studies of nurse staffing and patient outcomes, they do not resolve the issues surrounding those studies. Our inconsistent results could be due to the same issues we spoke of earlier. That is, this study may be limited by its data, measures, or model. We recommend retesting the PSIs in an expanded change model study that uses multi-state HCUP data.