Length of Stay as a Predictor of Long-Term Mortality in Patients Surviving a Traumatic Brain Injury: A Nationwide TBI Cohort Study of 153 177 Adults

OBJECTIVE: Our aim was to explore the length of stay (LOS) as a predictor of long-term mortality in patients surviving a traumatic brain injury (TBI). The objectives were to (1) establish TBI length of stay (LOS)-based groups empirically, (2) assess their accuracy in predicting death, and (3) compare the long-term risk of death between the LOS-based groups and non-TBI controls (Controls).

PARTICIPANTS: Patients (≥18.0) with TBIs in Denmark between 1994 and 2018 that survived beyond discharge were (1:5) age-matched with Controls.

DESIGN: Nationwide cohort study. TBI patients and Controls were stratified according to age: 18.0-39.9, 40.0-69.9, and ≥70.0.

MAIN MEASURES: Adjusted binomial regression was used to estimate odd ratios for mortality, and predictive accuracy was assessed using cross-validated area under the receiver operation curve (AUC) values. The Kaplan-Meier method and the adjusted binomial regression models were used to compare the all-cause mortality between LOS-based groups and Controls.

RESULTS: Among 153 177 TBI survivors, LOS-based groupings showed modest stratification of mortality risk, particularly in those under 70.0 years. However, based on the AUC, they did not improve individual-level prediction of death once adjusted for key confounders. Compared with Controls, TBI survivors had persistently excess mortality risk of up to 20 years post-index across all age groups. In older adults, LOS was positively correlated with the number of new comorbidities detected during hospitalization, which likely contributed to the observation that TBI patients with short LOS had better survival than Controls in this age population (surveillance bias).

CONCLUSIONS: LOS-based groups did not predict individual-level risk of death in TBI survivors but offered some short-term risk stratification at a group-level. Its application to group-level mortality outcomes in older adults is complicated by surveillance bias. Persistent excess mortality across all age groups supports the need for long-term follow-up after TBI.