Trauma registries are used to evaluate and improve trauma care, yet potentially miss certain trauma deaths and high-risk patients. We estimated the number of missed deaths and high-risk trauma patients using commonly available sources of trauma data and resulting bias in quality metrics for field trauma triage.METHODS
This was a preplanned secondary analysis of a population-based prospective cohort of injured patients transported by 44 emergency medical services agencies to 28 hospitals in seven Northwest counties from January 1, 2011 to December 31, 2011 and followed through hospitalization. We used a stratified probability sampling design for 17,633 patients, weighted to represent all 53,487 injured patients transported by emergency medical services. We compared patients meeting National Trauma Data Bank (NTDB) criteria (weighted n = 5,883), all injured patients presenting to major trauma centers (weighted n = 16,859), and all admitted patients (weighted n = 18,433), to the full sample. Outcomes included in-hospital mortality, Injury Severity Score (ISS) of 16 or higher, and critical resource use within 24 hours.RESULTS
Among 53,487 injured patients, there were 520 emergency department and in-hospital deaths, 1,745 with ISS of 16 or higher, and 923 requiring early critical resources. Compared to the full cohort, the NTDB cohort missed 62.1% of deaths, 39.2% of patients with ISS of 16 or higher, and 23.8% requiring early critical resources, especially older adults injured by falls and admitted to nontrauma hospitals. The admission cohort missed the fewest patients—23.3% of deaths, 10.5% with an ISS of 16 or higher, and 13.1% requiring early resources. Compared to triage sensitivity in the full cohort (66.2%), sensitivity estimates ranged from 63.6% (all admissions) to 93.4% (NTDB). Compared to triage specificity in the full cohort (87.8%), estimates ranged from 36.4% (NTDB) to 77.3% (all admissions).CONCLUSION
Common sources of trauma data miss substantial numbers of trauma deaths and high-risk trauma patients and can generate biased estimates for trauma system quality metrics.LEVEL OF EVIDENCE
Epidemiologic, level III.