BACKGROUND: Surgical mortality varies widely across hospitals, but the degree of temporal variation within individual hospitals remains unexplored and may reflect unsafe care. OBJECTIVES: To add a longitudinal dimension to large-scale profiling efforts for interpreting surgical mortality variations over time within individual hospitals. DESIGN: Longitudinal analysis of the French nationwide hospital database using statistical process control methodology. SUBJECTS: A total of 9,474,879 inpatient stays linked with open surgery from 2006 through 2010 in 699 hospitals. MEASURES: For each hospital, a control chart was designed to monitor inpatient mortality within 30 days of admission and mortality trend was determined. Aggregated funnel plots were also used for comparisons across hospitals. RESULTS: Over 20 successive quarters, 52 hospitals (7.4%) experienced the detection of at least 1 potential safety issue reflected by a substantial increase in mortality momentarily. Mortality variation was higher among these institutions compared with other hospitals (7.4 vs. 5.0 small variation signals, P<0.001). Also, over the 5-year period, 119 (17.0%) hospitals reduced and 36 (5.2%) increased their mortality rate. Hospitals with improved outcomes had better control of mortality variation over time than those with deteriorating trends (5.2 vs. 6.3 signals, P=0.04). Funnel plots did not match with hospitals experiencing mortality variations over time. CONCLUSIONS: Dynamic monitoring of outcomes within every hospital may detect safety issues earlier than traditional benchmarking and guide efforts to improve the value of surgical care nationwide.
BACKGROUND: Surgical mortality varies widely across hospitals, but the degree of temporal variation within individual hospitals remains unexplored and may reflect unsafe care. OBJECTIVES: To add a longitudinal dimension to large-scale profiling efforts for interpreting surgical mortality variations over time within individual hospitals. DESIGN: Longitudinal analysis of the French nationwide hospital database using statistical process control methodology. SUBJECTS: A total of 9,474,879 inpatient stays linked with open surgery from 2006 through 2010 in 699 hospitals. MEASURES: For each hospital, a control chart was designed to monitor inpatient mortality within 30 days of admission and mortality trend was determined. Aggregated funnel plots were also used for comparisons across hospitals. RESULTS: Over 20 successive quarters, 52 hospitals (7.4%) experienced the detection of at least 1 potential safety issue reflected by a substantial increase in mortality momentarily. Mortality variation was higher among these institutions compared with other hospitals (7.4 vs. 5.0 small variation signals, P<0.001). Also, over the 5-year period, 119 (17.0%) hospitals reduced and 36 (5.2%) increased their mortality rate. Hospitals with improved outcomes had better control of mortality variation over time than those with deteriorating trends (5.2 vs. 6.3 signals, P=0.04). Funnel plots did not match with hospitals experiencing mortality variations over time. CONCLUSIONS: Dynamic monitoring of outcomes within every hospital may detect safety issues earlier than traditional benchmarking and guide efforts to improve the value of surgical care nationwide.