Association Between Nurse Staffing and Hospital Mortality: Methodological considerations
*Luwis Diya, Biostatistical Center, Catholic University of Leuven
Emmanuel Lesaffre, Biostatistical Center, Catholic University of Leuven
Walter Sermeus, Centre for Health Services and Nursing Research, Catholic University of Leuven
Koen Van den Heede, Centre for Health Services and Nursing Research, Catholic University of Leuven
Arthur Vleugels, Centre for Health Services and Nursing Research, Catholic University of Leuven
Keywords: Quality of care, random effects, confounding
Quality of care assessment is most often based on data from administrative databases or surveys, e.g. by recording adverse outcomes in hospitals. The goals of this study are to determine the relationship between nurse staffing and mortality in Belgian hospitals and to explore the most appropriate statistical approach to estimate this relationship. We modeled mortality as a function of staffing variables using a clustered discrete time survival model. The staffing variables were not significant and attenuated as compared to neglecting clustering. Thus, there seems to be confounding between the hospital random effects and the staffing variables. This confounding might mask the importance of the quality of care determinants. This can be treated as a problem of an omitted variable bias (Bafumi and Gelman, 2006; Nehaus and McCulloch, 2006).