BACKGROUND: Severity-adjusted mortality is an unequivocal measure of burn care success. Hospitals can be compared on this metric using administrative data because information required for calculating statistically adjusted risk of mortality is routinely collected on hospital admission. METHODS: The New York State Department of Health provided information on all 13,113 thermally injured patients hospitalized at 1 of 194 hospitals between 2004 and 2008. We compared hospital survival rates using a random effects logistic model of mortality that incorporated age and several predictors that were present on admission and captured as International Classification of Diseases-9 codes: burn surface area, inhalation injury, three measures of physiologic compromise, and four medical comorbidities. Hospitals were compared on the adjusted odds of death and the number of excess deaths. RESULTS: Overall mortality was 3.2%. Nine high-volume hospitals (>100 patients/year) cared for 83% of patients with burn injuries. Overall variability of the odds of mortality among these high-volume centers was modest (median odds ratio=1.2) and we found little evidence for differences in the adjusted odds of mortality. A secondary analysis of the 185 low-volume hospitals that cared for 2,235 patients disclosed only 24 deaths. When examined in aggregate, these hospitals had better than predicted risk-adjusted mortality; a logical explanation is judicious case selection. CONCLUSIONS: Administrative hospital discharge data are extensive and comparably enough collected to allow comparison of the performance of burn centers. Risk-adjusted models show that patients have statistically indistinguishable risk-adjusted odds of mortality regardless of which hospital in New York State cared for them.
BACKGROUND: Severity-adjusted mortality is an unequivocal measure of burn care success. Hospitals can be compared on this metric using administrative data because information required for calculating statistically adjusted risk of mortality is routinely collected on hospital admission. METHODS: The New York State Department of Health provided information on all 13,113 thermally injured patients hospitalized at 1 of 194 hospitals between 2004 and 2008. We compared hospital survival rates using a random effects logistic model of mortality that incorporated age and several predictors that were present on admission and captured as International Classification of Diseases-9 codes: burn surface area, inhalation injury, three measures of physiologic compromise, and four medical comorbidities. Hospitals were compared on the adjusted odds of death and the number of excess deaths. RESULTS: Overall mortality was 3.2%. Nine high-volume hospitals (>100 patients/year) cared for 83% of patients with burn injuries. Overall variability of the odds of mortality among these high-volume centers was modest (median odds ratio=1.2) and we found little evidence for differences in the adjusted odds of mortality. A secondary analysis of the 185 low-volume hospitals that cared for 2,235 patients disclosed only 24 deaths. When examined in aggregate, these hospitals had better than predicted risk-adjusted mortality; a logical explanation is judicious case selection. CONCLUSIONS: Administrative hospital discharge data are extensive and comparably enough collected to allow comparison of the performance of burn centers. Risk-adjusted models show that patients have statistically indistinguishable risk-adjusted odds of mortality regardless of which hospital in New York State cared for them.