Early-Care Variables May Distort the Apparent Benefit of Neurosurgery-Capable Hospitals

A letter published in Neurocritical Care raises a methodological challenge to a study suggesting that patients with moderate or severe traumatic brain injury in South Africa had lower mortality when their initial treatment occurred at a neurosurgery-capable facility. Dilek Kuzukıran Kocataş, writing as a letter to the editor, argues that the reported odds ratio of 0.51 may not represent the overall effect of the initial facility type, because some of the variables used to adjust the comparison could themselves have been influenced by care at the first receiving hospital.

The original study by Bhaumik and colleagues examined the association between initial care at a neurosurgery-capable facility and traumatic brain injury-related mortality in a trauma system with limited specialist resources. The authors reported that initial care at such a facility was associated with lower adjusted odds of traumatic brain injury-related mortality, with an odds ratio of 0.51 and a 95% confidence interval of 0.34 to 0.75. This finding emerged despite greater injury severity and higher unadjusted mortality in the group initially presenting to the neurosurgery-capable facility. The letter acknowledges that the study addresses an important clinical question for trauma systems where specialist capacity is scarce.

The concern centers on the inverse probability of treatment weighting analysis. The propensity score model included transport from the scene by emergency medical services, intubation within 24 hours of injury, New Injury Severity Score, minimum Glasgow Coma Scale score, polytrauma, hypotension, and blood product administration. Intubation and blood product administration included interventions performed at any time during the first 24 hours after injury. That interval may encompass both prehospital treatment and care delivered after arrival at the first facility. It is therefore unclear how many of these interventions preceded arrival and how many occurred after care at the initial facility had begun. Similarly, hypotension recorded at the first facility may reflect either the patient’s condition on arrival or physiology observed during early resuscitation.

This temporal distinction matters because the clinical question concerns the type of initial receiving facility. Care at a neurosurgery-capable facility may influence decisions regarding airway management, transfusion, resuscitation, imaging, and subsequent intervention. If intubation or blood product administration occurred after arrival, these variables may not be baseline confounders of the association between facility type and mortality. Instead, they may have arisen after initial presentation and could have been influenced by care at that facility. Adjustment for such variables could block part of a pathway through which the initial facility affects survival. Conditioning on an intervention influenced by both injury severity and care at the receiving facility could also introduce, rather than reduce, bias.

The magnitude of the shift between crude and adjusted estimates underscores why the letter’s author considers clarification necessary. Traumatic brain injury-related mortality was higher among patients initially presenting to the neurosurgery-capable facility than among those presenting elsewhere, at 37.1% versus 24.6%. In contrast, the weighted and adjusted analysis indicated lower odds of mortality. The crude comparison should not be interpreted causally because patients presenting to the neurosurgery-capable facility had more severe injuries. Nevertheless, the extent of this change makes it important to show how the estimate evolves when baseline characteristics and early-care variables are introduced into the analysis.

The letter proposes a specific analytical remedy. The interpretation could be clarified by distinguishing variables determined before arrival at the initial facility from those measured or initiated after facility care began. The analysis could then be repeated using only covariates that could not have been influenced by treatment at the first receiving facility. A staged presentation of the unadjusted estimate, an estimate adjusted for prearrival characteristics, and the published fully adjusted estimate would show how the inclusion of early-care variables affects the association. If the intended target was an association conditional on early treatment rather than the overall effect of initial facility type, this objective should be stated explicitly.

Additional information about the weighting procedure would also aid interpretation, according to the letter. Transport from the scene by emergency medical services differed markedly between groups, occurring in 93.8% of patients presenting initially to the neurosurgery-capable facility and 51.0% of those presenting elsewhere. Reporting post-weighting covariate balance, propensity score overlap, the distribution of weights, and the effective sample size would help establish whether the adjusted comparison was supported by patients with sufficiently similar probabilities of presenting to either facility type. Standardized mortality risks or risk differences would also convey the potential implications for regional triage more directly than an odds ratio alone.

The letter’s conclusion is measured rather than dismissive. Direct access to neurosurgical and neurocritical care may plausibly improve outcomes after severe brain injury, and the study is described as providing valuable evidence from a resource-constrained trauma system. Until the timing of the covariates is clarified, however, the odds ratio of 0.51 should not automatically be interpreted as the overall effect of initial facility type on traumatic brain injury-related mortality. It may instead represent an association conditional on early-care variables whose temporal position has not been fully established.

An analysis restricted to covariates determined before care at the initial facility would, in the letter’s view, more directly indicate how strongly the findings support direct transport to a neurosurgery-capable facility. For trauma systems weighing whether to route patients with suspected severe brain injury toward specialist centers, the distinction between an overall effect of facility type and an association conditional on early treatment is not a statistical nicety but the difference between two different answers to the triage question the study set out to address.

Subject of Research: Methodological critique of adjusted mortality estimates for initial treatment at a neurosurgery-capable facility after traumatic brain injury

Article Title: Adjustment for Early Care and the Estimated Effect of Initial Treatment at a Neurosurgery-Capable Facility

Article References: Kuzukıran Kocataş, D. (2026). Adjustment for Early Care and the Estimated Effect of Initial Treatment at a Neurosurgery-Capable Facility. Neurocritical Care. https://doi.org/10.1007/s12028-026-02712-y

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02712-y

Keywords: traumatic brain injury, neurosurgery-capable facility, IPTW, confounding, propensity score, mortality, South Africa, trauma systems, triage, odds ratio, early care, methodology