How Air Medical Programs Should Think About Decision Support Technology

The FDA refined what counts as a regulated clinical decision support tool. Air medical programs operate at the edge of that definition every shift.

Flight crewmember rendering aid to a polytrauma victim while in flight.

The helicopter lifts off the pad at 0210 with a STEMI patient bound for a percutaneous coronary intervention center ninety miles away. The flight crew is reading a 12-lead transmitted from the referring emergency department on a tablet that has been integrated into the medical kit for eighteen months. The software flags the tracing as an anterior STEMI and recommends a heparin bolus dose based on the patient’s weight and renal function. The flight nurse looks at the recommendation, looks at the tracing, looks at the basic metabolic panel the referring ED ran ninety minutes earlier, and administers a different dose. She documents the reasoning. The patient does well. Nobody asks whether the tablet was a regulated medical device or whether the nurse’s override pattern is being tracked anywhere outside her own chart.

Multiply that scene across every air medical program in the country and you have one of the most important and least examined clinical decision support deployments in American medicine. Air medical operates at the edge of every regulatory framework that touches clinical AI — FDA software oversight, state medical practice acts, scope of practice for flight crews, NTSB safety culture obligations, and the medical direction structure that physicians like me are responsible for. Most air medical programs have not built the governance infrastructure to handle the volume and complexity of decision support technology that is already in the airframe.

What the FDA Actually Regulates and What It Does Not

The FDA’s stance on software-based clinical decision support, articulated in the 2022 final guidance on Clinical Decision Support Software and reaffirmed in the agency’s ongoing Device Software Functions Including Mobile Medical Applications framework, is risk-based. The agency intends to enforce device regulation on software that drives or directly informs a clinical action when the clinician cannot reasonably independently review the basis for the recommendation. The agency exercises enforcement discretion for software that supports a clinician’s independent review and judgment. The boundary is not bright. It depends on the clinical context, the speed at which a decision must be made, and the practical ability of the clinician to verify the tool’s reasoning against the underlying data.

Air medical operates almost entirely in the gray zone. The flight nurse and flight paramedic frequently cannot independently review the basis for a software recommendation in the time available, especially during the high-acuity early phase of a transport. They are operating in a noisy, vibrating, dimly lit cabin with a patient who is hemodynamically unstable and limited diagnostic capacity. The decision support tool that recommends a pressor selection or an antiarrhythmic dose is not in the same regulatory category as the same tool in an outpatient clinic where the physician has fifteen minutes to look up the package insert.

Most air medical programs have not done the analysis. They have deployed decision support tools that arrived bundled with the cardiac monitor, the ventilator interface, the transport-specific clinical reference applications. The procurement decision happened at the device level. The regulatory and governance analysis at the system level did not. That is the gap the medical director and the air medical leadership need to close, and the closing has to happen before the next NTSB-driven safety review forces the issue.

The Safety Culture Lens

The National Transportation Safety Board has driven a generation of operational safety improvement in air medical transport. Crew resource management, simulator training, weather minima, fatigue management, and standardized communication protocols are now the norm because the NTSB demanded them after a series of fatal accidents in the 2000s. The clinical safety culture has not caught up to the operational safety culture, and decision support technology is where the gap shows most clearly.

The same crew that runs a five-minute pre-flight risk assessment for weather, weight and balance, and fuel state often runs no pre-flight risk assessment for the clinical decision support tools they will rely on during the transport. There is no checklist for which tools are active, which version of the software is loaded, which patient-specific data has been entered, or which alerts are configured. There is no after-action review of the cases in which the decision support tool produced a recommendation that the crew overrode. The clinical decision data is not feeding the program’s quality improvement infrastructure the way the operational data is.

That has to change, and air medical leadership is positioned to drive the change because the program already has the safety culture muscle. Crew resource management techniques apply directly to clinical AI: structured communication when the tool and the clinician disagree, explicit calling-out of the override, debrief at the end of the mission. Standardized pre-flight checks expand naturally to the clinical decision support stack: which tools are loaded, which version, which configuration. The infrastructure exists. It needs to be extended to clinical AI.

Boarding, Throughput, and the Air Medical Mission

There is a downstream operational lens that air medical leaders should not ignore. A May 2026 systematic review in Health Affairs Scholar synthesized the evidence on emergency department boarding and confirmed what frontline clinicians have been saying for a decade: boarding increases morbidity, mortality, length of stay, and staff burnout, and the harms compound the longer the patient remains in the ED waiting for an inpatient bed. The downstream effect on air medical is significant. The referring ED that calls for transport is often boarding fifteen other patients. The receiving facility is often boarding ten. The transport itself does not solve the boarding problem; it relocates it.

Decision support technology in the air medical environment is sometimes positioned by vendors as a throughput tool — “faster, better, more accurate decisions in the airframe.” That framing is incomplete. The bottleneck in most air medical transports is not the decision support in the cabin. It is the boarding pressure at the destination and the limited inpatient capacity in the region. Air medical leadership should evaluate decision support tools against the question of where they actually improve the patient’s outcome, not against the vendor’s marketing claim. Sometimes the right answer is a slower, better-informed decision on the ground before the patient lifts off. Sometimes the right answer is a different destination facility. Decision support technology should serve that judgment, not preempt it.

For the operational pattern of how air medical programs should manage the scenarios where decision support outputs need to be triaged against multiple competing patient flows, the systems thinking I have written about in the context of mass casualty triage applies directly. See the GMOC blog post on AI in mass casualty incident triage at globalmedopscommand.com/blog/ai-mass-casualty-incident-triage-2026 for the framework on human override structure and command-and-control of AI-augmented decisions.

What This Means for the Air Medical Director

The air medical director should run, this quarter, a comprehensive inventory of every decision support tool deployed in every airframe in the program. Most directors I know would be surprised at the number. Cardiac monitor algorithms, ventilator weaning protocols, transport-specific pharmacology references, integrated electronic health record decision support, telemedicine consultation tools, weather and routing decision support. The inventory is the precondition for everything else.

Once the inventory exists, each tool gets classified against the FDA risk framework. Which tools drive or directly inform a clinical action in a way that the flight crew cannot reasonably independently review in the time available? Those tools are the high-risk tools, and they require formal governance — vendor disclosure of the algorithm’s basis, post-deployment monitoring of the override patterns, peer review of the divergent cases, and crew training that meets the same rigor as crew resource management training meets for operational decisions.

Override documentation has to become standard. Every time the flight crew overrides a decision support recommendation, the override and the reasoning enter the chart and the program’s quality improvement database. The database is the program’s defense against the inevitable post-incident review, and it is also the program’s best tool for evaluating the decision support vendor against ground truth. The data is generated by the missions the program already flies. The infrastructure to capture it is a single field in the run sheet.

Simulation is the third element. The clinical scenarios that the flight crew encounters are well-known, the failure modes of the decision support tools are increasingly well-characterized, and the training environment for combining them is the simulator. Air medical programs already run simulator-based training for cabin emergencies, patient deteriorations, and crew resource management. The same training infrastructure can run scenarios in which the decision support tool produces a borderline or wrong recommendation and the crew has to manage the disagreement. For the simulation curriculum framework that supports this kind of crew-level AI training, see emsmedsim.globalmedopscommand.com.

Dr. Chet’s Take

I have served as the medical director for an air medical program for years. The most important conversation I have with the flight crew is not about a specific patient. It is about the structure of authority when the technology and the clinician disagree. The flight nurse who overrode the heparin dose at 0210 made the right call. She made it because she had the experience, the data, and the institutional support to disagree with the tablet. She also made it in an environment where the program had not formally defined what “disagree with the tablet” looks like in policy, training, or documentation. The program got lucky. The program will not always get lucky.

Air medical leadership has a window. The FDA framework is clear enough to act on. The NTSB safety culture infrastructure is mature enough to extend. The clinical decision support tools are mature enough to demand governance. The medical directors who build the governance framework in the next twelve to eighteen months will have a defensible posture when the inevitable review comes. The medical directors who do not will be answering questions about why their flight crew was relying on a tool the program never formally evaluated.

Resources

If you lead an air medical program, an EMS service, or an emergency department that interfaces with transport, my book Emergency Department Efficiency Playbook covers the operational frameworks that connect ED throughput, transport, and inpatient capacity. For the broader AI risk-and-readiness framing, How to Avoid Becoming an AI Casualty walks through the same risk-classification approach applied to the broader clinical AI landscape. The full course on AI integration in emergency medicine is at courses.globalmedopscommand.com/store.

If you’re an emergency physician (or any clinician treating patients daily) trying to understand how AI will actually impact your clinical practice — not just the hype — I put together a free practical guide. You can download it here: AI in EM Survival Guide.

— — — — — — — — — — — — — — — — — — — —

Dr. Chester “Chet” Shermer, MD, FACEP is a Professor of Emergency Medicine, TeleHealth, HEMS and Critical Care Transport, and State Surgeon for the Army National Guard. He is the founder of Global MedOps Command and creator of the course AI in Emergency Medicine: Becoming AI Bulletproof.

His books — Emergency Department Efficiency Playbook, How to Avoid Becoming an AI Casualty, and The Emergency Medicine Observation Unit — are available on Amazon, Gumroad, and Kajabi.

Connect: globalmedopscommand.com | LinkedIn

Read more on the GMOC blog at globalmedopscommand.com/blog.

Sources

1. U.S. Food and Drug Administration, “Device Software Functions Including Mobile Medical Applications,” https://www.fda.gov/medical-devices/digital-health-center-excellence/device-software-functions-including-mobile-medical-applications

2. Health Affairs Scholar, “Patient and staff safety implications of emergency department boarding: a systematic review,” May 2026, https://academic.oup.com/healthaffairsscholar/article/4/5/qxag084/8586715?searchresult=1

3. World Health Organization, “New WHO discussion paper sets out opportunities and risks of AI in evidence-informed health policy,” June 2026, https://www.who.int/news/item/02-06-2026-new-who-discussion-paper-sets-out-opportunities-and-risks-of-ai-in-evidence-informed-health-policy