|
Abstract
|
INTRODUCTION: Adjusting for confounding variables is critical for objective comparison of outcomes. The explanatory power of variables used in adjusted models for injury and their relative utility across age groups has not been well-defined. This study aimed to assess the explanatory power of covariates commonly adjusted in injury research and their relative performance across age groups.
METHODS: Inpatients 18-100 y (2017-2022) were selected from 90 hospital trauma registries. Patients were grouped into sequential 5-y age blocks. Mortality was defined as the proportion of patients "expired + hospice". Dominance analysis was used to determine the average contribution (McFadden's R(2)) for covariates commonly included in multivariable logistic regressions.
RESULTS: Three hundred seventeen-thousand one hundred thirty-six patients were included (51.1% male, mean age: 63, mean injury severity score [ISS]: 9.8, mean Glasgow Coma Scale: 14.3, 93.5% blunt). Total explanatory power (McFadden's R(2)) for mortality was highest in youngest age group (52.7% in 18-24 group) and decreased with age, with the lowest R(2) (19.6%) in 95-100 group. Regardless of age, the Glasgow Coma Scale was the most important covariate (R(2) ranging from 9.0% to 20.4%). At age 18-24 y, ISS was a more dominant contributor than Elixhauser Score, but beyond 55 y, Elixhauser Score became more dominant than ISS.
CONCLUSIONS: The explanatory power of adjustment models including common covariates is limited and varies significantly across age groups, decreasing linearly with age. Adjusting for outcomes using these covariates may limit objective comparisons especially for older adults. Additional research is needed to identify covariates that enhance the explanatory power of adjustment models to allow for more objective comparisons across all ages.
Language: en |