Hospital nursing units in the USA: government ownership, Magnet designation, unit population age group and higher skill mix are associated with lower nursing turnover
- Department of Community Health, Outcomes, and Systems, University of Alabama at Birmingham School of Nursing, Birmingham, Alabama, USA
- Correspondence to
: Dr Patricia A Patrician
Department of Community Health, Outcomes, and Systems, University of Alabama School of Nursing, 1720 Second Ave South, NB 324, Birmingham, AL 35294, USA;
Commentary on: Staggs VS, Dunton N. Hospital and unit characteristics associated with nursing turnover include skill mix but not staffing level: an observational cross-sectional study. Int J Nurs Stud 2012;49:1138–45.
Implications for practice and research
Registered nurse (RN) and total nurse turnover are the organisational variables that should be measured and tracked routinely.
Working in highly specialised areas is associated with lower turnover, therefore future research should explore characteristics of these areas and how they may be applied to less specialised areas of nursing practice.
Future studies should explore determinants of low turnover in government hospitals.
Replacing nurse vacancies in hospitals is a substantial expense and hospitals are wise to explore strategies that minimise nursing turnover. The US hospitals that attracted and retained a high-quality RN workforce despite severe nursing shortages in the early 1980s were termed ‘magnet hospitals’. Magnet hospital certification by the American Nurses Credentialing Center (ANCC) began in the mid-1990s, using similar criteria to that of the original magnet hospitals. Three decades of studies have documented superior patient and nurse outcomes of Magnet hospitals, one being nurses’ intent to stay employed in the organisation.1 However, actual turnover has not been routinely measured in studies comparing ANCC-designated Magnet hospitals from others without this distinction.2
Nursing turnover was measured using data from 1884 units in 306 hospitals that participate in the National Database of Nursing Quality Indicators (NDNQI), a voluntary database established by the American Nurses Association to track and compare nursing sensitive indicators. The outcome variables, RN turnover and total nursing turnover, were obtained from the participating hospitals’ reports of staff separations for any reason. From these data, unit average monthly and annualised RN and total nursing turnover were calculated. Both hospital (Magnet status, ownership, teaching status, locale and hospital size) and unit level (total nursing care hours per patient day, RN skill mix, unit staff size, age group of population cared for and service line) independent variables were included. State was included to control for regional economic/unemployment differences. Hierarchical Poisson regression was performed in SAS version 9.2 with the generalized linear mixed model (GLIMMIX) procedure.
RN turnover was predicted by government hospital ownership, Magnet status, RN skill mix, population age group of unit and state. Total nursing turnover was predicted by all of the above plus unit service line. The strongest predictors were government ownership followed by population age group. The effect of skill mix was modest. Total staffing was not predictive of turnover in the multivariate model.
It is commendable that the authors explored a phenomenon we usually assume to be true: that there is less turnover in Magnet hospitals. However, there are several issues which should be considered for future research on nursing turnover. First, this study could have been strengthened by excluding involuntary resignations, which were higher in non-government hospitals. In addition, the turnover numbers were based upon definitions of ‘permanent staff’ at the hospitals. The authors explain that this may be problematic in that hospitals may not report separations of staff nurses who are still on probationary status because they do not consider these nurses ‘permanent’. NDNQI may want to consider more precise definitions of who or what should be considered in the numerators and denominators of the indicators, leaving nothing to the interpretation of the reporting hospitals.
The description of the dependent variable and the ‘exposure’ variable was not clear. It seems that the researchers in this case use turnover frequency (instead of turnover rate) as the response in the Poisson model. The author has stated that “the sum of the monthly unit RN or total nursing staff counts was included in the model as an exposure variable, making the unit-level turnover rate (number of separations divided by number of staff) the dependent variable.” If the total number of staff is included in the log scale as an offset term, then the response may be interpreted as ratio, but there was no such indication.
The lack of a turnover-staffing association may be related to the inability to measure staffing requirement. The staffing-outcomes body of research suffers from the inability to precisely measure how many and what types of nurses are actually needed to provide the care required of patients on a particular unit.3 Until researchers can come up with measures for workload, total staffing associations may remain a puzzle.