Predicting nursing turnover with catastrophe theory

AIM: This paper is a report of a study comparing an innovative nonlinear model and a traditional linear model for accuracy in prediction of nursing turnover. BACKGROUND: An international, sustained nursing shortage creates a need to target accurately the staff population at risk for turnover. Existing linear methodology is cumbersome with the number of variables needed, while producing inadequate results. Nonlinear modelling methods offer increased simplicity and accuracy in predictability. METHODS: A correlational survey with a longitudinal cohort prospective study was carried out in 2005-2006 with a convenience sample of 1033 Registered Nurses from the Midwest region of the United States of America. At time 1, 756 usable questionnaires were returned and 496 at time 2. Data analysis included analyses of a cusp catastrophe model, a cube-shaped four-dimensional figure with a top that provided a down-turning slope area (the catastrophe/cusp zone). This fluid, dynamic cusp version employed the smallest number of control and dependent variables. RESULTS: The exceedingly small turnover sample preempted the use of the computerized program Cuspfit; a proven quasi-quantitative methodology demonstrated 80.4% predictability in the cusp catastrophe model overall and 53.6% correct predictions of actual terminations, particularly in nurses with <5 years of nursing experience. Additional accurate predictions were obtained with the use of a time-staged model. Organizational commitment and anticipated turnover were accurate predictor variables; job tension was not. CONCLUSION: Catastrophe models are useful in predicting nursing turnover. Future nursing researchers should act on this evidence to benefit forthcoming studies and the profession.