Factors affecting nursing turnover: A multi-level predictive model

Abstract

The purpose of this study was to test three models which explain the effects of perceived participation, administrative support, ability to deliver quality care, job satisfaction and organizational commitment on nursing turnover intentions and unit level turnover rate within a professional practice model. The models to be tested were stated at the individual nurse level and the work group level. This study used secondary analysis of a data set obtained at the University of Maryland Medical System in 1993. The data set included responses from 863 nurses from 52 work groups. Models were tested at the individual and group levels using the LISREL structural equation model program. The model of variables at the individual level was tested in a random sample of 350 nurses from the data set, and was modified by dropping two latent variables which were highly collinear with the two independent variables. This revised model fit the data well and cross validated in a second random sample of 350 nurses from the data set. Empirical support was provided for the aggregation of general job satisfaction, organizational commitment and turnover intent at the work group levels. Homology between models at the individual and group levels was supported by path coefficients which were similar in size and direction, and by ratios between correlations at the two levels which were close to one. Although fit indices provided mixed support for the fit of the aggregate variable model, which may have been due to small sample size, relationships were similar to the individual model. Adding turnover rate to the model produced an aggregate level model which explained 26% of the variance in unit turnover rate, although fit indices provided mixed support for model fit. Results from this study support the existence of similar relationships among factors which affect nursing turnover at the individual nurse and work group levels. Implications are suggested for interventions, which can be targeted at individual nurses, which may affect the work group outcome of turnover. Implications for the design of multilevel studies of organizational outcomes are also suggested.