Investigators have used a variety of operational definitions of nursing hours of care in measuring nurse staffing for health services research. However, little is known about which approach is best for nurse staffing measurement.Objective:
To examine whether various nursing hours measures yield different model estimations when predicting patient outcomes and to determine the best method to measure nurse staffing based on the model estimations.Data Sources/Setting:
We analyzed data from the University HealthSystem Consortium for 2005. The sample comprised 208 hospital-quarter observations from 54 hospitals, representing information on 971 adult-care units and about 1 million inpatient discharges.Methods:
We compared regression models using different combinations of staffing measures based on productive/nonproductive and direct-care/indirect-care hours. Akaike Information Criterion and Bayesian Information Criterion were used in the assessment of staffing measure performance.Results:
The models that included the staffing measure calculated from productive hours by direct-care providers were best, in general. However, the Akaike Information Criterion and Bayesian Information Criterion differences between models were small, indicating that distinguishing nonproductive and indirect-care hours from productive direct-care hours does not substantially affect the approximation of the relationship between nurse staffing and patient outcomes.Conclusions:
This study is the first to explicitly evaluate various measures of nurse staffing. Productive hours by direct-care providers are the strongest measure related to patient outcomes and thus should be preferred in research on nurse staffing and patient outcomes.