ED Observation Units: Fix Boarding Before It Breaks You

At 3 a.m. last month I counted nineteen boarders in my department. Nineteen admitted patients parked in ED beds, some for their second calendar day. Meanwhile, the chest pain patient in bed 14 — HEART score of 3, first troponin negative — was occupying an acute bed for the next six hours waiting on a repeat troponin and a treadmill that would not open until 8 a.m. He did not need an inpatient bed. He did not need an acute ED bed either. He needed a third option, and most hospitals still do not have one. After 25 years in emergency medicine and a career spent on the operational side of throughput, I am convinced the ED observation unit is the single most underbuilt asset in American hospitals. In this post I will cover why boarding is partly an observation problem in disguise, how to build a clinical decision unit that actually works, and why the real value of an obs unit is not where you think it is.

Why ED Boarding Is an Observation Problem in Disguise

We talk about ED boarding as if it were purely an inpatient capacity problem — not enough staffed beds upstairs, not enough discharges before noon. That is half true. The other half is that a meaningful slice of the patients we admit never needed a 24-to-48-hour inpatient stay in the first place. They needed 15 hours of protocolized care: serial troponins, a syncope workup, IV antibiotics for cellulitis, fluids and reassessment for dehydration.

The literature backs this up. A 2024 scoping review of ED observation units in JACEP Open (Goodwin et al.) found length of stay was the most commonly reported outcome across 461 articles, with multidisciplinary EDOUs averaging roughly 14 to 15 hours per stay — against the one-to-two-day inpatient alternative for the same conditions. Research abstracts presented at ACEP 2025 in Annals of Emergency Medicine went further, describing a measurable halo effect: an emergency-physician-run observation unit improved hospital-wide throughput, not just ED metrics. When obs-appropriate patients stop consuming inpatient beds, those beds absorb the boarders.

I have watched this play out from both sides. As a medical director, I have seen departments where every low-risk chest pain became a hospitalist admission by default — not because the patient needed it, but because there was nowhere else to put them. Each of those admissions is a bed a genuinely sick patient cannot use, and every one of them shows up later as a boarder in somebody's hallway. Nationally, somewhere between 5 and 10 percent of ED visits are observation-appropriate. Run the math on your annual volume. That is the size of the pressure-relief valve you are choosing not to build.

And remember what boarding actually costs. This is not an inconvenience metric — boarding is associated with increased mortality, more medication errors, longer overall hospital stays, and higher rates of patients leaving without being seen. It burns out nurses faster than any other single operational failure I have witnessed in 25 years. Every obs-appropriate patient you admit by default is a small deposit into that account. If your hospital's answer to boarding is another throughput committee and no observation strategy, you are bailing water without patching the hull.

Building a Clinical Decision Unit That Actually Works

I have seen clinical decision units succeed and I have seen them turn into dumping grounds. The difference is never the square footage. It is the discipline of the design. Here is the framework I use.

Step one: pick your protocols before you pick your beds. Start with the conditions the evidence supports — low-risk chest pain, syncope, TIA, cellulitis, dehydration, asthma and COPD exacerbations, atrial fibrillation with controlled rates, renal colic. Write a protocol for each: inclusion criteria, exclusion criteria, order set, and a hard disposition decision point. No protocol, no placement.

Step two: make exclusions non-negotiable. The fastest way to kill an obs unit is to let it become the landing zone for "not sick enough for the ICU, too complicated to discharge." Hemodynamic instability, need for titratable drips, expected stay over 24 hours, unresolved placement issues — those are admissions, full stop.

Step three: emergency medicine owns it. The Annals data on the halo effect came from an EM-run unit, and that is not a coincidence. EM physicians supervising PAs or NPs, rounding twice daily minimum, with disposition authority that does not require negotiating with three consult services. The moment the unit becomes an orphan service, length of stay balloons.

Step four: set time targets and defend them. Under 24 hours is the ceiling; 15 to 18 hours is the goal. Every patient gets a disposition decision at protocol-defined checkpoints, not "we will see how the morning labs look."

Step five: build the dashboard on day one. Length of stay, conversion-to-inpatient rate, 7-day ED return rate, and left-without-completion. If you cannot measure it, administration will eventually defund it.

One clarification that trips up half the administrators I brief: observation status and an observation unit are not the same thing. Most hospitals already have plenty of patients in observation status — scattered across inpatient floors, managed by whoever admitted them, averaging far longer stays than a dedicated unit produces. A geographic unit with dedicated staff and protocols is what converts a billing designation into an operational asset. If your CFO says "we already do observation," ask where those patients physically are and what their average length of stay is. The answer usually makes the case for you.

I go deeper on each protocol, staffing model, and the financial case in my ED Observation Units guide — this is the condensed brief.

The Counterintuitive Pearl: The Value Is Not in the Obs Patients

Here is the insight that took me years of operational work to fully appreciate: the patients inside your observation unit are not where the unit earns its keep. The return on investment lives everywhere else in the building.

Think of the obs unit as a pressure-relief valve on a closed hydraulic system. The 15-hour chest pain patient who goes to obs instead of telemetry frees a monitored inpatient bed. That bed absorbs an ICU downgrade, which opens an ICU bed, which lets the OR run its afternoon schedule without holding cases. The 2025 Annals abstracts called it a halo effect for a reason — the throughput gain shows up hospital-wide, in units that never see an obs patient. When I make the business case to a C-suite, I do not lead with obs unit revenue. I lead with the inpatient beds recovered per year and what those beds are worth in transfer acceptances and surgical volume the hospital is currently turning away.

The second half of the pearl is knowing when your unit is quietly failing, and the tell is your conversion rate — the percentage of obs patients ultimately admitted. The healthy band is roughly 15 to 20 percent. If your conversion rate is under 5 percent, you are not running an observation unit; you are running an expensive discharge lounge for patients your ED should have sent home directly, and you are adding 12 hours of length of stay to people who never needed it. If it is over 30 percent, your inclusion discipline has collapsed and you are boarding admissions in disguise — the exact disease you built the unit to cure. I check that single number before I look at anything else in an obs unit review, because it tells me in ten seconds whether the unit is a valve or a cul-de-sac.

Watch the conversion rate quarterly. It drifts. It always drifts. Protocols get informally loosened, a new hospitalist group starts using the unit as a soft landing for shaky discharges, a strong PA leaves and disposition discipline leaves with them. The unit that was a valve in January can be a cul-de-sac by June, and nobody will announce it — you have to go looking. Put the conversion rate on the same monthly dashboard as your door-to-doc times and treat a two-quarter trend in either direction as an operational alarm, not a curiosity.

The Bottom Line

Boarding is not going away, and no committee will fix it while observation-appropriate patients keep consuming inpatient beds. The evidence — from the JACEP Open scoping review to the ACEP 2025 throughput data — says a disciplined, EM-owned observation unit shortens stays, protects inpatient capacity, and improves flow across the whole hospital.

Three takeaways. First, protocol discipline beats square footage: no protocol, no placement, no exceptions on exclusions. Second, emergency medicine must own the unit, with disposition authority and time targets of 15 to 18 hours. Third, judge the unit by its conversion rate — 15 to 20 percent is the healthy band, and drift in either direction is your earliest warning.

If you are building or fixing an observation unit, my ED Observation Units ebook walks through every protocol, the staffing model, and the financial case in detail — and its companion, Emergency Department Efficiency, covers the rest of the throughput battle. For the clinical side — how a disciplined obs unit lowers admissions and readmissions — see my earlier article, The Physician's Guide to Reducing Admits and Readmissions.

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.

Sources

1. Goodwin J, et al. "Emergency department observation units: A scoping review." JACEP Open, 2024. https://onlinelibrary.wiley.com/doi/full/10.1002/emp2.13254
2. ACEP 2025 Research Forum abstract (emergency-physician-run observation unit and hospital-wide throughput). Annals of Emergency Medicine, 2025. https://www.annemergmed.com/article/S0196-0644%2825%2900720-6/fulltext