90 Discrepancies in Mortality Metrics Between National Datasets

Abstract

Abstract Introduction The National Burn Repository (NBR) is the recognized standard for data collection models within the academic burn community. Despite its robust capabilities, the NBR poses challenges to access and exploration that can be intimidating to some researchers. As a result, user-friendly alternative databases have grown in popularity. We compared the NBR to two commercially available large datasets using the well-recognized relationship between age and total body surface area (TBSA) on mortality as our metric to assess correlation. Methods We accessed the TriNetX Global Health Research Network and queried the Diamond network (medical clearinghouses) and Research network (41 healthcare organizations (HCO)) for all-cause mortality within 3 years following burn injuries that occurring between 2000 – 2020 using ICD-10 codes (T31-32). We explored the distribution of TBSA and age across these cohorts and compared the variance in the distribution to the 2008–2017 NBR report using one proportion z-tests for each age-TBSA matched sub-group. We also compared demographics and lethal area for 50% mortality (LA50). Results The Diamond network identified 336,965 entries with an all-cause mortality rate of 3.21% and an LA50 of < 10%. Demographics showed 50% male, 81% unknown race and a mean age of 39. In 2016 - 2017, 56,430 entries were reported. The Research network identified 114,778 entries with a mortality rate of 2.54% and an LA50 of < 10%. Demographics showed 61% male, 58% white, 24% unknown and 16% black with a mean age of 37. In 2016 - 2017, 14,164 entries were reported. In comparison, the NBR database reported 185,239 entries with a mortality rate of 2.96% and an LA50 of >70%. Demographics showed 67% male, 59% white and 3.6% unknown with a mean age between 20–29. In 2016 - 2017, 42,402 entries were reported. Comparison of mean mortality between age-TBSA matched subgroups in the Diamond and the Research networks relative to NBR showed correlation among pediatric populations but lacked significance. Conclusions The Diamond and Research networks are large datasets, which appear to be statistically different from the NBR dataset and are derived from different populations (insured patients, academic healthcare organization and accredited academic burn centers). The exact overlap between datasets is unknown, but demographics suggest that they are very different populations. The Figure depicts the relationship between age and TBSA on mortality for each database. Each database is large enough to achieve statistically significant conclusions, but caution should be used when contrasting conclusions between datasets due to the significant degree of divergence.