Observation units (OUs) are a hot topic in Emergency Medicine (EM) today. They are expanding their presence among hospitals, and their footprint within hospitals, every day (Feng 2013). The reason for this is multifactorial, but is almost always a response to cost and policy challenges. First, OUs present reductions in cost for hospitals, as they are cheaper than full inpatient stays and often include shorter length of stays. Second, they are a response to policies from the Center for Medicare and Medicaid Services (CMS) that make observations units attractive alternatives to inpatient stays. Examples of such policies include the “two midnight rule” (such that a hospital only receives reimbursement for inpatient stays if the stay is for two days or longer) and readmission reimbursement policy (where hospitals are effectively paid less if a recently discharged patient gets readmitted).
When first experimented with, OUs were estimated to cause a 25% reduction in cost for a routine chest pain stay (Farkouh 1998). As indications for observation stays have swollen, there has been much excitement that such cost reductions could be seen for a wide variety of diagnosis from chest pain to syncope, to congestive heart failure, transient ischemic attacks, and infections. Notably, the cost savings have always been framed as a lower cost of an observation stay versus a more expensive inpatient stay. However, these estimates often ignore a third disposition: discharge home.
Suggesting that the existence of an OU promotes the use of it (with a corresponding reduction in discharges home), this article questions the conventional wisdom that OUs are cost saving.
How does the availability of OUs effect dispositions from the emergency department for patients presenting with chest pain, including admissions to the hospital and discharges home?
Patients were identified from the National Hospital Ambulatory Medical Care Survey (NHAMCS) between 2007 and 2010. Patients were included in the study if they were over the age of 18 and had a visit coded with a chief complaint of “chest pain.”
This was a cross sectional study. Patients were split into two categories: those presenting to a hospital with an OU and those presenting to a hospital without one. Patients who presented to a hospital with an OU complaining of chest pain and who were ultimately sent to an OU were evaluated to create a propensity score — a statistical matching technique designed to predict which attributes account for disposition. The attributes considered include: Age, Sex, Race, Ethnicity, Insurance, Presenting vital signs, Comorbidities (including heart failure, anemia, hypertension, diabetes, coronary atherosclerosis, cardiac dysrhythmias, chronic obstructive pulmonary disease, and asthma), Metropolitan area, Hospital region, Performance of cardiac enzymes, CT scan, and Administration of aspirin, antiplatelet agents, antianginal agents, anticoagulants, and NSAIDs.
The propensity score was used to match the group of patients presenting with chest pain to an observation hospital with a similar group of patients also presenting with chest pain to a non-observation hospital. A prediction model was then developed, again, based on the attributes listed above, to determine what predicted whether a patient at a non-observation hospital would be admitted vs discharged him.
Finally, this prediction model was applied to the group of patients with chest pain who presented to the observation hospital and were ultimately sent to observation.
Patients were excluded if they had a specific chest pain diagnosis (for example, myocardial infarction, costochondritis, etc), if they were intubated or had CPR performed on them, if they became deceased, or if they ultimately had a disposition other than home, admission, or observation (for example, elopement, hospital transfer, leaving against medical advice, etc).
- Data was available from 1,363 hospitals
- Non-OU hospitals: 457
- OU hospitals: 458
- Unable to determine: 448
- Patients available for analysis after applying inclusion and exclusion criteria
- Non-OU hospitals: n = 1,098
- OU hospitals: n = 227
- Prediction model designated 50.1% of patients as “hospitalization likely” from the observation group
- Prediction model designated 49.9% of patients as “discharge home likely” from the observation group
- Among the “discharge home likely” group, 9.2% were converted from observation to full admission
- The database included 1,363 hospitals of which, data from 915 were used in the study
- The propensity score successfully matched patients in the observation hospital group with the non-observation hospital group, with both groups having very similar characteristics
- Data was objectively obtained from the CDC from a wide diversity of hospitals nationwide
- There may be fundamental differences in observation hospitals vs non-observation hospitals that would prohibit “matching” of the groups even if they appear similar by their attributes (for example, the patient population between hospitals, quality of care at the hospitals)
- It is impossible to match for unmeasured or unreported attributes. Important attributes that went unmeasured in this study are: troponins, EKGs, patient descriptions of their chest pain
- While the NHAMCS database is large, after applying inclusion criteria, exclusion criteria, and applying propensity matching, the final sample size of patients was relatively small
- The NHAMCS survey only designates “observation status” if the observation takes place in a dedicated unit. Hospitals that have their observations units outside of dedicated units would have mistakenly been identified as non-observation hospitals.
“In this study, we estimated that half of ED visits for chest pain that resulted in an observation unit admission were made by patients who may have been discharged home had the observation unit not been available. Increased availability of observation units may result in both decreased hospitalizations and decreased discharges to home.”
We agree with the authors’ conclusions that observations units likely reduce not only admissions but also discharges home.
Potential Impact To Current Practice
This study will likely not affect an individual emergency medicine physician’s practice, but is an important study in the ongoing discussion about the role of OUs. This study appropriately reframes the argument that OUs are cost saving. While OUs may be cost saving compared to inpatient stays, they are certainly more costly than discharges home. Future calculations of cost and savings must take into account the fact that it is not strictly patients who would otherwise be admitted that are being transferred to observation, but also patients that would otherwise have been discharged home.
Notably, the discussion about the role of OUs, this paper included, leaves out the notion of “what is best for the patient.” Mentioned in passing in the paper is that “of patients categorized as discharge home likely, 9.2% were hospitalized after admission to the OU.” This stark observation reminds us that while observation status may not be as cost saving as previously believed, a not-insignificant number of patients transferred to observation ultimately have an outcome that is clinically important enough to require a full inpatient admission. In other words, it is possible that patients who may otherwise have been discharged him, may be well served with a disposition to observation. We urge those who guide hospital policy to remember that beyond the fiscal challenges that guide many policy discussions, there exist real-world patient outcomes that are affected by such policies — it is these outcomes, above all other considerations, that should guide policy making.
As with other medical innovations like novel imaging modalities and cardiac catheterizations, the introduction of OUs is likely susceptible to the “Field of Dreams” phenomenon: “if you build it, they will come.” When implemented, OUs not only cut down on full inpatient admissions, but also likely reduce discharges — patients who otherwise would have been sent home routinely stay in the hospital for observation. Future calculations of the cost-saving nature of OUs must take this into account. Finally, and most important of all, there must be further analysis on how patients’ medical care is affected by the increased utilization of OUs.
Farkouh ME et al. A clinical trial of a chest-pain OU for patients with unstable angina. Chest Pain Evaluation in the Emergency Room (CHEER) Investigators. N Engl J Med. 1998;339:1882-1888. PMID: 9862943
Feng Z et al. Sharp rise in Medicare enrollees being held in hospitals for observation raises concerns about causes and consequences. Health Aff (Millwood). 2012;31:1251-1259. PMID: 22665837