Predictors of survival and length of intensive care unit stay in children with severe traumatic brain injury

Introduction: We hypothesized that clinical variables under the physician's control may predict outcome for children with severe traumatic brain injury (TBI). Methods: We identified children aged 3 weeks - 17 years diagnosed with non-penetrating TBI and Glasgow Coma Score (GCS)<9. The first 72 hours of hospitalization were analyzed in detail. Thirty candidate predictors of outcome were evaluated including severity scores (Pediatric Trauma Score, GCS), demographics, vital signs, laboratory values, CT scans and neuro-resuscitative medications. In particular, maximum and minimum systolic blood pressures during the first 72 hours were analyzed. Survival and PICU length of stay (LOS) were study endpoints. Data were analyzed by chi-square, Kruskal-Wallis test, and multiple logistic regression to determine odds ratio (OR). Results: From 1991-1995, 752 pediatric patients suffered TBI. 147 patients had an admission GCS< 9. Of these, 116 (79%) survived. Median age was 6.3 years. 94% of nonsurvivors (29/31) vs. 41% of survivors (41/116) had an admission GCS ≤ 4. Univariate predictors of survival included: higher systolic blood pressure (SBP) and temperatures on admission through 48 hours. All non -survivors had an abnormal first CT scan. Nonsurvivors were significantly more likely to have received mannitol, pentobarbital, or vasopressors. A multiple logistic regression model revealed GCS at 6 hours (OR =3.6, 95% CI= 1.2 - 11.9) and maximum systolic blood pressure of greater than 140 mm Hg (OR = 36.4, 95%CI = 2.0-662.0) to be independent predictors of survival. For survivors, mean PICU LOS was 4.6 days. Long (>5 days) LOS was associated with phenytoin, pentobarbitol, vasopressor, and especially mannitol use as well as edema on admission CT scan (p=. 015). Multiple logistic regression determined that mannitol use was a strong predictor of long PICU stay independent of GCS, ICP, CT scan findings, or other neuro-resuscitative drugs (OR= 8.94, 95% CI = 2.16-37.02). Conclusions: Maximum systolic blood pressure greater than 140 mm Hg is associated with survival in pediatric TBI. Mannitol administration does not increase survival and it independently increases the likelihood of prolonged ICU stay in TBI survivors. We propose that mannitol use in TBI be reevaluated.