The nation depends upon its trauma centers for everyday emergencies as well as potential disasters. The objective of this proposal is to provide information and techniques for ensuring the quality of trauma care in the United States by improving the databases, statistical methods, and reporting systems used to assess trauma center performance. The American College of Surgeons (ACS) has created a National Trauma Data Bank (NTDB), now combining detailed data from more than a million patients in more than 500 hospital registries. This provides a new and unique data source for this project and other trauma outcomes research. However, because the NTDB is a convenience sample biased toward larger hospitals and sicker patients, contemporary population- based data from the AHRQ's Nationwide Inpatient Sample (NIS) will also be utilized for this project. Hospital mortality after admission is expected to vary systematically among institutions in either database, even after adjustments for injury type and severity, comorbidity, and age. Optimal combinations of patient covariates will be sought to explain this primary outcome. The residual variability will be evaluated in order to identify hospital characteristics associated with improved survival and individual institutions with exceptional results. Other outcomes (length of stay, discharge to long-term care) will also be evaluated. Multilevel (hierarchical) models will be used for statistical analysis. These models have become more widely applied to health care research and are now a standard way to account for clustered data structures (e.g., patients within hospitals). They reduce spurious outlier identification resulting from small sample sizes, while at the same time providing more valid estimates of the independent effects of patient and hospital characteristics. Estimations and summary statistics using the new databases, variables, and statistical models will be compared with results from traditional methods. This project is designed to produce a credible and verifiable data management structure and analytic methodology that can be used for the continuing evaluation of trauma care by the ACS, AHRQ, and others. ? ? ?
| Clark, David E; Hannan, Edward L; Raudenbush, Stephen W (2010) Using a hierarchical model to estimate risk-adjusted mortality for hospitals not included in the reference sample. Health Serv Res 45:577-87 |
| Clark, David E; Hannan, Edward L; Wu, Chuntao (2010) Predicting risk-adjusted mortality for trauma patients: logistic versus multilevel logistic models. J Am Coll Surg 211:224-31 |
