What’s in a Name? Why ED Observation Suffers from Semantic Ambiguity

Anthony Rosania, MD, MHA, FAMIA, FACEP
Associate Professor of Emergency Medicine, Rutgers – New Jersey Medical School

Ask five people in a hospital what “observation” means and you will get five different answers. A nurse may think of a physical location—the observation unit down the hall. A case manager hears a billing status. A hospitalist pictures a specific patient population. An emergency physician recalls a clinical workflow rooted in serial assessment and time-limited decision-making. And a hospital administrator sees a line item on a denial report.

They are all correct. That is the problem.

A Word Doing Too Much Work

In everyday English, “observation” is a deverbal noun—a word derived from the verb “to observe.” In that generic sense, virtually every hospitalized patient is under observation. Physicians observe patients; that is what we do. But in the American healthcare system, the word has accumulated a remarkable number of additional, overlapping, and often contradictory meanings.

Observation is a status. It is a CMS-defined designation that determines whether a patient is classified as an outpatient or an inpatient, with far-reaching implications for reimbursement, cost-sharing, and downstream eligibility for skilled nursing facility coverage. It is a place as well: a dedicated unit, sometimes called an observation unit or clinical decision unit, with its own beds, staffing model, and physical footprint. It is a service: billable under specific CPT codes and facility revenue codes, although at least some of this ambiguity has been resolved as of late on the professional fee side. It is also a time constraint as CMS policy anticipates resolution within 24 hours, and regards anything beyond 48 hours as exceptional. And, perhaps worse of all, it is often seen as a proxy severity marker—implying that the patient’s condition is not quite sick enough for inpatient admission but too complex or uncertain for immediate discharge. Any “Observationalist” worth his salt knows all too well that “observation” patients can be quite sick. And perhaps most importantly, it is a clinical workflow – a way of delivering care. It is a protocolized, endpoint-driven model of care with intellectual roots in emergency medicine.

No single word should carry this many meanings in a system where precision matters. And yet, because we use “observation” to mean all of these things simultaneously, we routinely talk past one another.

How We Got Here

The clinical practice of observation care predates its regulatory identity by decades. Emergency physicians have always used tincture-of-time care—holding patients for serial examinations, reassessment, and risk stratification before deciding on admission or discharge. This was not a billing construct. It was, and remains, an extension of the core work of emergency medicine: managing dynamic, acute, undifferentiated patients whose presenting problems require more time than a standard ED visit but do not necessarily require the full resources of an inpatient stay.

The formalization of this practice began in the 1990s with chest pain units—dedicated, protocolized spaces designed for time-limited cardiac evaluation. Landmark work by Graff, Farkouh, and many others demonstrated that these units improved outcomes, shortened stays, and reduced unnecessary admissions. The model expanded to encompass heart failure, syncope, transient ischemic attack, asthma, and other conditions amenable to structured, short-duration evaluation. This was emergency medicine at its best – and we gave it the name “observation medicine.”

CMS, recognizing that this clinical practice already existed, codified observation as an outpatient hospital service in the late 1980s and 1990s. Dedicated physician E/M codes followed in 2003–2004. This was a case of policy following practice. But something subtle happened in the process: the regulatory tail began to wag the clinical dog. As observation acquired a formal billing identity, the word stopped referring primarily to what clinicians did and started referring primarily to what payers paid for, and increasingly, what they denied.

The Conflation That Hurts Us

Because “observation” simultaneously names a clinical model, a physical space, a patient status, and a revenue category, ED observation units are routinely called to account for problems that fall outside the true scope of observation medicine.

Consider the most common institutional conversation about observation today: denial management. When a hospital experiences a surge in inpatient claim denials, leadership turns to the observation unit and asks whether the hospital is placing too few patients on observation status, or too many. The implicit logic is that the observation unit exists to optimize status assignment. From this vantage point, observation is not a care model. It is a hedge against payer audits.

Observation medicine is not a denial management tool. If anything, choosing observation status over inpatient when clinical criteria support admission is a self-denial.

This framing fundamentally misunderstands what an ED observation unit does. An observation unit in this context is far better described as a clinical decision unit. It exists to provide protocolized, time-limited, endpoint-driven care for patients whose clinical trajectories remain uncertain. The question it answers is clinical: does this patient need to be admitted, or can they safely go home? The question it does not answer, and should not be asked to answer, is whether a given patient’s status designation will survive a retrospective utilization review.

Not All Observation Patients Belong in a CDU

One of the most important distinctions lost in the semantic fog is this: not all patients on observation status are appropriate for a clinical decision unit care model, and not all patients in a clinical decision unit need to be on observation status.

Consider a patient with diabetic ketoacidosis requiring active resuscitation, or a patient with severe alcohol withdrawal who needs close titration of benzodiazepines, or a septic patient still being stabilized. These patients are undeniably emergency department patients. The rise of “bundles” for ED patients, whether they bit for sepsis, DKA, etc., has created a need for being able to manage patients for longer than your typical ED stay. These patients may or may not meet strict inpatient admission criteria at the moment, but the care they require is dynamic, acute, and temporally intensive, making it part and parcel to emergency medicine and ripe for protocolized pathways. These are not observation patients – but they are ideal patients for a Clinical Decision Unit.

Conversely, a patient placed on a well-defined chest pain rule-out protocol may have such a confluence of problems that their chest pain work up is in a large part secondary to their “protocoled” issue. They may require prolonged care, or more intensive resources secondary to CHF, psychosocial issues, functional decline, etc. This renders it unlikely they would be ready for discharge in under 24 hours. That patient may be best suited for observation “status” but is not well suited to the short, intensive, protocolized care of the ED observation model – or the CDU.

When we conflate the care model with the billing status, we lose the ability to match the right patient to the right workflow. Worse, we create perverse incentives to shoehorn complex patients into observation pathways because the status seems convenient, or to avoid the unit entirely because the status seems financially disadvantageous.

Reframing the Conversation

If observation medicine is to continue maturing as a discipline—and it must, because the clinical need it serves is only growing—we need to disentangle the word from its accumulated meanings and be deliberate about which one we intend.

ED observation units should be defined by their clinical model, not their billing category. The organizing principle should be the patient’s clinical needs and the type of care pathway that best serves those needs. Protocol selection should be driven by the presenting problem and its natural history, not by status assignment. The unit’s success should be measured by clinical outcomes—safe discharge rates, return visits, time to disposition—not by denial rates or revenue capture. In addition, Clinical Decision Units may embrace the brief, intensive, triage-based therapy of care bundles applied to patients we know will be staying as inpatient: severe ETOH withdrawal, DKA, and sepsis come to mind.

If the care in question is acute, undifferentiated, time-sensitive, and can be delivered and performed by Emergency Physicians in a protocoled manner – it may belong in a CDU regardless of the patient status.

This means embracing the concept of the clinical decision unit as the more accurate descriptor for what we do. The work of the CDU is fundamentally the work of emergency medicine: dynamic, acute, undifferentiated care focused on the triage of a primary presenting problem, extended over a longer time horizon. Some of these patients will be on observation status. Others will be on emergency department status. Some may even be inpatients. The status is a downstream administrative assignment. The care model is the thing.

Conclusion

What we practice is observation medicine: a subspecialty of emergency medicine focused on the care of dynamic, acute, undifferentiated patients with complex but time-limited presenting problems, over a time horizon longer than the typical emergency department visit. We do this work in clinical decision units. We also do it in emergency departments. The patients we care for may be on observation status, emergency department status, or inpatient status. The status follows the care; it should never define it.

The semantic ambiguity of “observation” is not merely an academic curiosity. It distorts institutional expectations, misdirects operational priorities, and threatens to hollow out a care model that emergency medicine built because patients needed it. The fix is not a new billing code or a revised CMS definition. The fix is clarity of language. When we say observation, we should mean the medicine, not the status, not the place, not the revenue line.