Objective: The purpose of this presentation is to share the findings of one study that increases our understanding of the effects of nursing-related hospital variables on 30-day mortality for hospitalized patients. A ‘30-Day Mortality Model’ was hypothesized, tested, and refined to explain relationships between predictor variables and 30-day mortality.
Design: The study employed a retrospective comparative design. The unit of analysis was the individual acute care hospital.
Population, Sample, Setting, Year of Study: The population of interest is Canadian acute care hospitals. The sample consisted of 75 Ontario, Canada teaching and community hospitals in operation during the fiscal year of April 1, 1998 through March 31, 1999. The study was undertaken throughout January 2000 until August 2001. To develop risk-adjusted 30-day mortality rates for each hospital, 46,941 discharged patients from the 75 sample Ontario hospitals were included in the study. Only patients with a ‘most responsible diagnosis’ of acute myocardial infarction, stroke, pneumonia, or septicemia were included. To develop several of the predictor variables, 3,988 registered nurse respondents who completed the Ontario Registered Nurse Survey of Hospital Characteristics were included from the 75 sample hospitals.
Concepts & Variables: The dependent variable, a weighted and risk-adjusted 30-day mortality rate, was developed for each hospital using the general formula of observed hospital deaths divided by predicted hospital deaths. Hospital deaths were predicted through four logistic regression models. Hospital-level predictor variables included: dose of nursing staff, nursing skill mix, amount of professional role support for nurses, nurse experience, nurse capacity to work, condition of the nursing practice environment, physician expertise, and hospital type / location.
Methods: Data from multiple sources, including large administrative health databases, were accessed to answer the research question and included the following: the Discharge Abstract Database 1998-99 from the Ontario Ministry of Health and Long-Term Care, the Ontario Hospital Reporting System 1998-99, the Ontario Registered Nurse Survey of Hospital Characteristics, the Ontario Registered Persons Database, and Statistics Canada Census 1996 Population Files. Multiple regression models were used to test the hypothesized relationships between the predictor variables and 30-day mortality rates.
Findings: In the final multiple regression model, 32 percent of variance among risk-adjusted mortality rates in sample hospitals was explained. The results provide evidence of an association between a richer nurse skill mix and lower 30-day mortality rates in sample hospitals. Evidence is also found to support a relationship between more years of registered nurse experience and lower 30-day mortality rates. One surprising finding is that reduced nurse capacity to work (higher numbers of reported missed shifts) is associated with lower 30-day mortality. None of the other 30-day mortality model predictors added to the explanation of variation in risk-adjusted 30-day mortality rates among sample hospitals.
Back to Nursing and Patient Outcomes Research: Canadian Perspectives
Back to The Advancing Nursing Practice Excellence: State of the Science