C. K. Zogg1, J. W. Scott1, L. L. Wolf1, O. A. Olufajo1, A. J. Rios Diaz1, D. Metcalfe1, M. Chaudhary1, A. A. Shah1,2, J. M. Havens1, S. L. Nitzschke1, Z. Cooper1, A. Salim1, A. H. Haider1 2Division Of General Surgery, Mayo Clinic,Scottsdale, AZ, USA 1Center For Surgery And Public Health, Harvard Medical School & Harvard School Of Public Health, Department Of Surgery, Brigham And Women’s Hospital,Boston, MA, USA
Introduction: Efforts to benchmark/improve trauma outcomes frequently rely on data obtained from the National Trauma Data Bank (NTDB), Nationwide Inpatient Sample (NIS), and Nationwide Emergency Department Sample (NEDS). Despite their widespread use, limited understanding is available of how these large databases compare. The objective of the study was to present a direct comparison and stratified, risk-adjusted assessment of trauma patients captured in NTDB, NIS, and NEDS.
Methods: Data (2007-2011) were abstracted for adult (18-64y) patients with primary trauma diagnoses (800-959). Patients with late effects of injury (905-909), superficial injuries (910-924), and foreign bodies (930-939) were excluded, as were those with non-blunt/penetrating injuries. Differences in patient (primary payer, sex, race/ethnicity, age, ISS, intent, mechanism, head AIS, CCI, year) and hospital (teaching status, trauma center level, annual trauma volume, region) factors were compared using descriptive statistics. Database-stratified, multilevel-logistic models assessed risk-adjusted differences in mortality, major morbidity, and extended LOS (>75th percentile) for trauma patients overall and among cohorts restricted to TBI, pneumothorax/hemothorax, blunt splenic injury, and pelvic fractures. In an effort to test the similarity of conclusions among the most comparable groups of patients, differences in trauma outcomes were further compared between ED/inpatients in NTDB and NEDS using coarsened-exact matching techniques.
Results: ED/inpatient combined totals of 1,578,487 (NTDB), 539,200 (NIS), and 5,069,811 (NEDS) patients were extracted. Patient encounters in NTDB were predominately managed at high-volume (86.7% in the top two trauma volume quintiles), teaching (87.3%), L1-L2 trauma (62.2%) centers. Encounters in NIS and NEDS were more varied: 1.5% (effect dominated by ED presentations) and 31.9% high trauma volume, 62.5% and 33.4% teaching, and 18.0% L1-L2 (only reported in NEDS) centers. Differences in patient factors, while present, were less pronounced. Risk-adjusted regression demonstrated qualitatively similar conclusions for most variables albeit notably stronger associations related to lack of insurance, Black race, age, and ISS in NTDB. Associations with self-inflicted injury, severe head injury (AIS>3), and teaching status were more pronounced in NIS and NEDS.
Conclusion: This methodological assessment of large trauma databases points to important differences underlying trauma populations used in outcomes research. While largely comparable for predictions related to demographic factors, discrepancies at the hospital level need to be carefully weighed. Researchers are encouraged to make use of the wealth of information provided by these large population resources while remaining cognizant of the generalizability and limitations of each.