Is overtriage associated with increased mortality? Insights from a simulation model of mass casualty trauma care

To examine the relationship between overtriage and critical mortality after a mass casualty incident (MCI) using a simulation model of trauma system response.

We created a discrete event simulation model of trauma system management of MCIs involving individual patient triage and treatment. Model variables include triage performance, treatment capability, treatment time, and time-dependent mortality of critically injured patients. We model triage as a variable selection process applied to a hypothetical population of critically and noncritically injured patients. Treatment capability is represented by staffed emergency department trauma bays with associated staffed operating rooms that are recycled after each use. We estimated critical and noncritical patient treatment times and time-dependent mortality rates from the trauma literature.

In this simulation model, overtriage, the proportion of noncritical patients among all of those labeled as critical, has a positive, negative, or variable association with critical mortality depending on its etiology (ie, related to changes in triage sensitivity or to changes in the prevalence and total number of critical patients). In all of the modeled scenarios, the ratio of critical patients to treatment capability has a greater impact on critical mortality than overtriage level or time-dependent mortality assumption.

Increasing overtriage may have positive, negative, or mixed effects on critical mortality in this trauma system simulation model. These results, which contrast with prior analyses describing a positive linear relationship between overtriage and mortality, highlight the need for alternative metrics to describe trauma system response after MCIs. We explore using the relative number of critical patients to available and staffed treatment units, or the critical surge to capability ratio, which exhibits a consistent and nonlinear association with critical mortality in this model.