Non-linear relationships between nurse staffing and patients' length of stay in acute care units: Bayesian dependence modelling.

AIMS: This study sought to analyse relationships between nurse staffing and patients' length of stay in acute care units and to determine whether non-linear relationships exist between variables.
BACKGROUND: Healthcare systems are complex and it could be assumed that they comprise non-linear associations. However, current planning and evaluation of nurse staffing are based primary on linear reasoning.
DESIGN: This quantitative study adopted a retrospective longitudinal design.
METHOD: Retrospective register data, consisting of information relating to 35,306 patient episodes and administrative information concerning 381 nurses, were used. Data were collected in 2009 from 20 somatic inpatient units at a university hospital in Finland as a monthly time series of 2008 data and analysed using Bayesian dependency modelling.
RESULTS: Patients' acuity was the most important agent that connected all eleven variables in the dependency network of nurse staffing and short length of stay. Non-linear associations were found between short length of stay and the proportion of Registered Nurses. Skill mix consisting of an average proportion of Registered Nurses (65-80%) was conducive to a short length of stay and predicted a 66% likelihood of short length of stay. Higher and lower percentages of Registered Nurses predicted lower likelihood of short length of stay.
CONCLUSION: Flexible nurse staffing is preferable to fixed staffing to provide patients with shorter length of stay in acute care units. In the present research, the Bayesian method revealed non-linear relationships between nurse staffing and patient and care outcomes.