AI Won’t Fix Your Boarding Crisis

Every hospital executive who has watched an emergency department collapse under the weight of boarded patients has heard the same pitch from at least one vendor: artificial intelligence will optimize your throughput. The algorithms will predict admissions, streamline bed assignments, identify bottlenecks before they cascade, and restore your department to its designed capacity.

It sounds reasonable. It is also fundamentally wrong about where the problem lives.

The Throughput Illusion

Throughput is a systems engineering term that hospital operations borrowed without fully understanding its implications. In manufacturing, throughput measures the rate at which a system converts inputs to outputs. In an emergency department, the concept maps imperfectly because the “output” — a dispositioned patient who has either been discharged or admitted to an inpatient bed — depends on downstream capacity that the ED does not control.

AI-powered throughput tools can do genuinely useful things. Predictive models can forecast admission likelihood within the first thirty minutes of a patient’s arrival, allowing the admitting team to begin planning before the formal request comes through. Natural language processing can flag incomplete orders or pending results that are holding up disposition decisions. Dashboard analytics can surface real-time bottleneck data that would take a charge nurse twenty minutes to compile manually.

None of that matters if there are no inpatient beds to receive the patients.

This is the core deception in the AI-throughput sales pitch: it treats the emergency department as a self-contained system whose efficiency can be optimized internally. The reality is that ED boarding is an output problem, not a process problem. The department is functioning exactly as designed — it is the hospital that has failed to build sufficient downstream capacity, or more commonly, failed to discharge inpatients with the same urgency it uses to fill beds.

Boarding Is Not an ED Problem

The data on this is unambiguous. Boarded patients — admitted patients held in the ED because no inpatient bed is available — degrade every measurable performance metric. Door-to-provider times increase. Left-without-being-seen rates climb. Time-sensitive interventions for new arrivals get delayed because nursing ratios are consumed by patients who should be on a med-surg floor.

The American College of Emergency Physicians has stated repeatedly that boarding is a hospital-wide systems failure, not an emergency department workflow issue. When an ED physician has completed the evaluation, made the admission decision, and written the orders, that patient’s continued presence in the emergency department represents a failure of inpatient operations to accept a transfer of care.

AI can quantify this failure with impressive precision. It can show you exactly how many boarded patients are in your department at any moment, how long they have been waiting, which units have the slowest turnover, and which attending services are consistently the last to accept transfers. What AI cannot do is compel the hospitalist service to round earlier, force environmental services to turn rooms faster, or make the C-suite allocate capital for additional beds.

That is a leadership problem. Algorithms do not solve leadership problems.

Where AI Actually Helps — and Where It Doesn’t

There are legitimate applications of AI in ED operations that deserve investment. Predictive admission modeling gives the admitting team a meaningful head start. Automated bed management systems — when integrated with real-time census data from every unit in the hospital — can reduce the lag between a bed becoming available and a boarded patient being moved. Machine learning applied to historical flow data can help administrators understand seasonal surge patterns and staff accordingly.

These are useful tools. They are also incremental improvements to a system whose fundamental constraint is physical capacity and institutional will.

The danger comes when hospital leadership uses AI adoption as a substitute for the harder work of restructuring inpatient discharge processes, investing in observation unit capacity, or confronting the cultural norm that treats the emergency department as an overflow ward. I have watched health systems spend seven figures on AI throughput platforms while refusing to fund a single additional observation unit bed — a unit that, when properly utilized, can decompress an ED faster than any algorithm.

If your hospital has an AI-powered bed management system and a thirty-patient boarding crisis at 2 AM, you do not have a technology gap. You have a resource allocation failure that your leadership team has chosen not to address.

What You Should Be Doing Now

Start by separating the problems. AI can improve internal ED processes — order completion, disposition decision support, predictive modeling for resource needs. Accept that utility and invest in it where the evidence supports it. But do not allow anyone — vendors, administrators, or board members — to frame AI as the solution to boarding.

Demand that your hospital track boarding hours as a system-level metric, not an ED metric. When boarding data lives exclusively in the emergency department’s quality dashboard, it becomes the emergency department’s problem to solve. It is not. Every boarded hour should appear on the report card of the inpatient unit that failed to accept the transfer, the discharge planning team that did not expedite the outgoing patient, and the CMO who oversees the system.

Build or expand your observation unit. This is the single most effective structural intervention for ED throughput that most hospitals underutilize. An observation unit with clear admission criteria, time-limited protocols, and dedicated staffing can absorb a significant percentage of the patients who currently board in your ED waiting for inpatient beds. I wrote the Emergency Department Efficiency Playbook and The Emergency Medicine Observation Unit specifically because these operational tools are the interventions that actually move the needle — not vendor dashboards.

Finally, be honest with your administration about what AI can and cannot do. If your CMO believes that the new throughput platform will reduce boarding by forty percent, correct that expectation before it becomes the reason they defer the capacity investment your department actually needs.

Dr. Chet’s Take

I have worked in several departments where boarding was the default state of operations. Not the exception — the baseline. Thirty patients on the board, twelve of them admitted and waiting, three in hallway beds, and the ambulances still coming. That is not a technology problem. That is a command failure at the institutional level. Thankfully, my current institution is attacking this problem from multiple angles with modest success.  

In military medical operations, when your aid station is overrun, you do not optimize the triage algorithm. You call for additional assets, establish alternative collection points, and push the problem upstream to the echelon that has the resources to solve it. The emergency department is no different. The ED is the enhanced battalion aid station. The hospital is the next higher echelon of care. When the aid station is overwhelmed because the echelon above refuses to accept patients, no amount of internal optimization changes the outcome.

AI is a tool. A good one, in the right context. But it is not a substitute for institutional courage, and it will never replace the decision to invest in the beds, the staffing, and the discharge infrastructure that your emergency department has needed for years.

Stop letting vendors sell you software when what you need is leadership.

———

AI Won’t Wait. Neither Should You.

If you are an emergency physician navigating the intersection of AI adoption and operational reality, my course covers the strategic framework you need. The boarding crisis is just one example of how AI gets deployed without addressing the actual problem. Understanding where AI fits — and where it does not — is the skill that separates physicians who lead from physicians who get led.

--> 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 AI Survival guide. You can download it here.

👉 Link → Free AI in EM Survival Guide

Additionally, if you want to delve deeper in to AI and the intersection with medicine, click to learn more about the course I created: AI in Emergency Medicine: Becoming AI Bulletproof

———

— 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 the creator of AI in Emergency Medicine: Becoming AI Bulletproof.