Evidence-Based Optimization of Urban Firefighter First Response to Emergency Medical Services 9-1-1 Incidents

Introduction. Many emergency medical services (EMS) systems dispatch nonparamedic firefighter first responders (FFRs) to selected EMS 9-1-1 calls, intending to deliver time-sensitive interventions such as defibrillation, cardiopulmonary resuscitation (CPR), and bag-mask ventilation prior to arrival of paramedics. Deciding when to send FFRs is complicated because critical cases are rare, paramedics often arrive before FFRs, and lights-and-siren responses by emergency vehicles are associated with the risk of en-route traffic collisions. Objective. To describe a methodology allowing EMS systems to optimize their own FFR programs using local data, and reflecting local medical oversight policy and local risk-benefit opinion. Methods. We constructed a generalized input-output model that retrospectively reviews EMS dispatch and electronic prehospital clinical records to identify a subset of Medical Priority Dispatch System (MPDS) call categories ("determinants") that maximize the opportunities for FFR interventions while minimizing unwarranted responses. Input parameters include local FFR interventions, the local FFR "first-on-scene" rate, and the locally acceptable ratio of risk to benefit. The model uses a receiver-operating characteristic (ROC) curve to identify the optimal mix of response specificity and sensitivity achieved by sending FFRs to progressively more categories of EMS calls while remaining within a defined risk-benefit ratio. The model was applied to a 16-month retrospective sample of 220,358 incidents from a large urban EMS system to compare the model's recommendations with the system's current practices. Results. The model predicts that FFR lights-and-siren responses in the sample could be reduced by 83%, from 93,058 to 16,091 incidents, by confining FFR responses to 27 of 509 MPDS dispatch determinants, representing 7.3% of incidents but 58.9% of all predicted FFR interventions. Of the 93,058 incidents, another 58,275 incidents could be downgraded to safer nonemergency FFR responses and 18,692 responses could be eliminated entirely, improving the specificity of FFR response from 57.8% to 93.0%. Conclusions. This model provides a robust generalized methodology allowing EMS systems to optimize FFR lights-and-siren responses to emergency medical calls. Further validation is warranted to assess the model's generality.