Monitoring hospital trauma mortality using statistical process control methods.

Abstract

BACKGROUND: We sought to develop a simple and effective way to monitor trends in trauma mortality, using objective clinical categories and methods of statistical process control. STUDY DESIGN: Control charts and Pareto analysis were applied to trauma mortality data at the Maine Medical Center. We collected data prospectively on patients who died in our hospital after acute injury during 1985-1996 (and retrospectively for 1975-1984) to identify cases requiring medical quality review. We excluded from this study patients older than 80 years, those whose Glasgow Coma Scale motor component was never > 3 at any time after admission, and those with pathologic fractures, carcinomatosis, high quadriplegia, or severe burns. The remaining deaths were classified as resulting from inability to resuscitate (mostly hemorrhage), neurologic deterioration, or organ failure. The annual numbers in each of these categories were evaluated under the hypothesis of stationary Poisson processes with mean values equal to those seen from 1975-1984. RESULTS: After the exclusions, annual mortality from trauma has remained within control limits consistent with the Poisson model. Death from neurologic deterioration has shown a trend consistent with significant improvement in the process mean. Transient peaks in the other categories did not exceed control limits, but Pareto analysis prompted detailed studies of aortic and liver trauma. CONCLUSIONS: Process control methodology is easy to apply and potentially useful in monitoring hospital trauma mortality.