Early neurosurgical procedures enhance survival in blunt head injury: propensity score analysis.

BACKGROUND: Studies of trauma systems have identified traumatic brain injury as a frequent cause of death or disability. Due to the heterogeneity of patient presentations, practice variations, and potential for secondary brain injury, the importance of early neurosurgical procedures upon survival remains controversial. Traditional observational outcome studies have been biased because injury severity and clinical prognosis are associated with use of such interventions. OBJECTIVE: We used propensity analysis to investigate the clinical efficacy of early neurosurgical procedures in patients with traumatic brain injury. METHODS: We analyzed a retrospectively identified cohort of 518 consecutive patients (ages 18-65 years) with blunt, traumatic brain injury (head Abbreviated Injury Scale score of >or= 3) presenting to the emergency department of a Level-1 trauma center. The propensity for a neurosurgical procedure (i.e., craniotomy or ventriculostomy) in the first 24 h was determined (based upon demographic, clinical presentation, head computed tomography scan findings, intracranial pressure monitor use, and injury severity). Multivariate logistic regression models for survival were developed using both the propensity for a neurosurgical procedure and actual performance of the procedure. RESULTS: The odds of in-hospital death were substantially less in those patients who received an early neurosurgical procedure (odds ratio [OR] 0.15; 95% confidence interval [CI] 0.05-0.41). The mortality benefit of early neurosurgical intervention persisted after exclusion of patients who died within the first 24 h (OR 0.13; 95% CI 0.04-0.48). CONCLUSIONS: Analysis of observational data after adjustment using the propensity score for a neurosurgical procedure in the first 24 h supports the association of early neurosurgical intervention and patient survival in the setting of significant blunt, traumatic brain injury. Transfer of at-risk head-injured patients to facilities with high-level neurosurgical capabilities seems warranted.

Regional differences in outcomes for hospitalized injured patients.

BACKGROUND: Our goal was to use a hospital population-based data set that was a sample of all injured patients admitted to a hospital in the United States to develop universal measures of outcome and processes of care. METHODS: Patients with a primary discharge diagnosis of injury (ICD-9 800 to 959) in the HCUP/Nationwide Inpatient Sample for the years 1995 to 2000 were used to estimate the annual number of hospitalized injured patients. Using census data, we calculated age- and sex- adjusted average annual incidence rates for four census regions in the United States: Northeast, Midwest, South and West. Outcomes measured were annual rates per million populations of hospitalization rate, death rate, and potentially ineffective care (PIC) rate defined as >28 days of hospitalization ending in death. Length of stay (LOS) was calculated as total number of days annually hospitalized for injury for census regions per million populations. RESULTS: Incidence rates per million populations and 95% confidence intervals for rate of hospitalizations for injury were: Northeast, 5596 (5338-5853); Midwest, 5516 (5316-5716); South, 5639 (5410-5869); West, 5307 (5071-5543). Incidence rates per million populations and 95% confidence intervals for rate of in-hospital deaths were: Northeast, 129 (119-139); Midwest, 131 (122-139); South, 141 (129-152); West, 114 (106-123). Incidence rates per million populations and 95% confidence intervals for rate of PIC were: Northeast, 11 (10-13); Midwest, 5 (4-5); South, 6 (5-7); West, 4 (3-4). Incidence rates per million populations and 95% confidence intervals for hospital days were: Northeast, 34 (32-36); Midwest, 30 (28-31); South, 30 (29-32); West, 26 (24-27). CONCLUSION: Regional differences in outcomes and processes of care for hospitalized injured patients exist and may be influenced by hospital characteristics and region of the country. Research to identify the factors that cause these hospital and regional variations is needed. These observations suggest that to develop a uniform standard for quality of care, it will be essential to have valid and robust hospital population-based measures.