Triaging patients with serious head injury: results of a simulation evaluating strategies to bypass hospitals without neurosurgical facilities

Abstract

Objectives: to inform the debate on whether seriously head-injured adult patients should be transported directly to the regional neurosurgical unit or indirectly after evaluation and stabilisation at the nearest hospital. Design: a simulation model was constructed to compare triage strategies and to identify those that predicted the maximum survivors. In each strategy, an estimate of the patient's condition in the field was used to determine the receiving hospital. The model used data from previous publications and local ambulance service and hospital databases. In the absence of valid data, expert clinical estimates were made and subjected to sensitivity analyses. Setting: an area in the North West Midlands of UK, covered by six acute hospitals including one with a regional neurosurgical unit. Outcome measure: the number of survivors predicted by each triage strategy. Results: five strategies were identified which consistently predicted the highest number of survivors. Compared with current policy it was predicted that in the North West Midlands, ten lives per year could be saved (6 per million total population per year). The results from sensitivity analyses did not alter these successful policies. Conclusion: the successful strategies should be considered as potential improvements to be introduced into clinical practice.

Introduction

There has been considerable debate about whether patients with serious head trauma should be transported directly to a neurosurgical unit or indirectly after initial assessment and stabilisation at the nearest hospital, which may lack neurosurgical capability [1], [2], [3], [4], [5], [6], [7], [8], [9], [10].

Most of the earlier studies have taken a qualitative approach with no clear consensus on the best practice. The main benefit of direct transportation to the neurosurgical unit is that neurosurgical intervention can take place faster. A survival advantage has been attributed to early definitive care, [11], [12], [13], [14], [15] but if patients deteriorate during a longer journey to the neurosurgical unit, the consequences are potentially fatal. A major benefit of being taken directly to the nearest hospital is that intubation and ventilation can be performed more quickly, reducing the risk of hypoxic secondary insult. In addition, earlier treatment of other injuries, which themselves contribute directly to mortality, may reduce the risk of a secondary insult from hypotension. These benefits may be partly offset by the need for secondary transfer, which has been associated with detrimental health affects [16], [17], [18], [19], [20].

In 1988, a report from the Royal College of Surgeons of England [21] concluded that a scientific model was required to evaluate the process of trauma care, using injury severity measures. The aim of our study was to use mathematical simulation techniques to predict and compare patient outcome following different triage strategies based on the severity of the injuries suffered.

A modelling approach was undertaken for the following reasons.

• To evaluate the strategies within a short time period and limited costs. The model simulated a time period for which the costs of an observational study would be prohibitive. The advent of new technologies during an observational study may also render the original question redundant.

• To ensure that comparisons between triage strategies are valid by using the same cohort of patients. In an observational study, patients with different severities of head and other injuries may bias the results.

• To allow sensitivity analyses to be performed. Sensitivity analyses determine the influence of each variable upon the final outcome measure. This allows an insight of which variables are the most important and also where additional expenditure may be focussed in order to produce the maximum benefit.

• To not unnecessarily risk fatalities. Preliminary modelling work allows early rejection of strategies with poor predicted outcomes.

The authors chose a simulation approach rather than a logistic regression approach due to the large number of cases within the Keele database where data upon variables that were considered important were not available or not routinely recorded.