David W Harless1, Barbara A Mark. 1. Department of Economics, Virginia Commonwealth University, Richmond, VA 23284, USA. dwharles@vcu.edu
Abstract
BACKGROUND: Studies of the impact of registered nurse (RN) staffing on hospital quality of care for hospital inpatients often rely on data sources that do not distinguish inpatient from outpatient staffing, thus requiring imputation of staffing level. As a result, estimates of the impact of staffing on quality may be biased. OBJECTIVE: To estimate the impact of changes in RN staffing on changes in quality of care with direct measurement of staffing levels. RESEARCH DESIGN: Longitudinal regression analysis of California general acute care hospitals where inpatient staffing is measured directly. SUBJECTS: Estimation sample reflects outcomes for 11,945,276 adult inpatients at 283 hospitals from 1996 to 2001. MEASURES: Patient outcomes are in-hospital mortality ratio and surgical failure-to-rescue ratio after nurse-sensitive complications with risk adjustment through calculation of the expected number of adverse outcomes using the Medstat disease staging algorithm. Staffing levels were measured as the number of full-time equivalent nurses per 1000 inpatient days. RESULTS: Estimates suggest that changes in RN staffing were associated with reductions in mortality and failure to rescue. At 2.97 RN full-time equivalents per 1000 inpatient days, a 1-unit increase in staffing was associated with a 0.043 decrease in the mortality ratio (P < 0.05), and the estimated effect was smaller at hospitals with higher staffing levels. Estimates for failure to rescue ratio were statistically significant only at higher staffing levels. CONCLUSIONS: Results are compared with those from similar studies, including studies using imputation of inpatient staffing, and are found to be consistent with attenuation bias induced by imputation.
BACKGROUND: Studies of the impact of registered nurse (RN) staffing on hospital quality of care for hospital inpatients often rely on data sources that do not distinguish inpatient from outpatient staffing, thus requiring imputation of staffing level. As a result, estimates of the impact of staffing on quality may be biased. OBJECTIVE: To estimate the impact of changes in RN staffing on changes in quality of care with direct measurement of staffing levels. RESEARCH DESIGN: Longitudinal regression analysis of California general acute care hospitals where inpatient staffing is measured directly. SUBJECTS: Estimation sample reflects outcomes for 11,945,276 adult inpatients at 283 hospitals from 1996 to 2001. MEASURES: Patient outcomes are in-hospital mortality ratio and surgical failure-to-rescue ratio after nurse-sensitive complications with risk adjustment through calculation of the expected number of adverse outcomes using the Medstat disease staging algorithm. Staffing levels were measured as the number of full-time equivalent nurses per 1000 inpatient days. RESULTS: Estimates suggest that changes in RN staffing were associated with reductions in mortality and failure to rescue. At 2.97 RN full-time equivalents per 1000 inpatient days, a 1-unit increase in staffing was associated with a 0.043 decrease in the mortality ratio (P < 0.05), and the estimated effect was smaller at hospitals with higher staffing levels. Estimates for failure to rescue ratio were statistically significant only at higher staffing levels. CONCLUSIONS: Results are compared with those from similar studies, including studies using imputation of inpatient staffing, and are found to be consistent with attenuation bias induced by imputation.
Authors: Matthew D McHugh; Margo Brooks Carthon; Douglas M Sloane; Evan Wu; Lesly Kelly; Linda H Aiken Journal: Milbank Q Date: 2012-03 Impact factor: 4.911