aVisits may have met more than 1 exclusion criterion.
Fee C, Burstin H, Maselli JH, Hsia RY. Association of Emergency Department Length of Stay With Safety-Net Status. JAMA. 2012;307(5):476-482. doi:10.1001/jama.2012.41
Author Affiliations: Departments of Emergency Medicine (Drs Fee and Hsia) and Medicine (Ms Maselli), University of California, San Francisco; and National Quality Forum, Washington, DC (Dr Burstin).
Context Performance measures, particularly pay for performance, may have unintended consequences for safety-net institutions caring for disproportionate shares of Medicaid or uninsured patients.
Objective To describe emergency department (ED) compliance with proposed length-of-stay measures for admissions (8 hours or 480 minutes) and discharges, transfers, and observations (4 hours or 240 minutes) by safety-net status.
Design, Setting, and Participants The 2008 National Hospital Ambulatory Medical Care Survey (NHAMCS) ED data were stratified by safety-net status (Centers for Disease Control and Prevention definition) and disposition (admission, discharge, observation, transfer). The 2008 NHAMCS is a national probability sample of 396 hospitals (90.2% unweighted response rate) and 34 134 patient records. Visits were excluded for patients younger than 18 years, missing length-of-stay data or dispositions of missing, other, left against medical advice, or dead on arrival. Median and 90th percentile ED lengths of stay were calculated for each disposition and admission/discharge subcategories (critical care, psychiatric, routine) stratified by safety-net status. Multivariable analyses determined associations with length-of-stay measure compliance.
Main Outcome Measures Emergency Department length-of-stay measure compliance by disposition and safety-net status.
Results Of the 72.1% ED visits (N = 24 719) included in the analysis, 42.3% were to safety-net EDs and 57.7% were to non–safety-net EDs. The median length of stay for safety-net was 269 minutes (interquartile range [IQR], 178-397 minutes) for admission vs 281 minutes (IQR, 178-401 minutes) for non–safety-net EDs; 156 minutes (IQR, 95-239 minutes) for discharge vs 148 minutes (IQR, 88-238 minutes); 355 minutes (IQR, 221-675 minutes) for observations vs 298 minutes (IQR, 195-440 minutes); and 235 minutes (IQR, 155-378 minutes) for transfers vs 239 minutes (IQR, 142-368 minutes). Safety-net status was not independently associated with compliance with ED length-of-stay measures; the odds ratio was 0.83 for admissions (95% CI, 0.52-1.34); 1.03 for discharges (95% CI, 0.83-1.27); 1.05 for observations (95% CI, 0.57-1.95), 1.30 for transfers (95% CI, 0.70-2.45]); or subcategories except for psychiatric discharges (1.67, [95% CI, 1.02-2.74]).
Conclusion Compliance with proposed ED length-of-stay measures for admissions, discharges, transfers, and observations did not differ significantly between safety-net and non–safety-net hospitals.
Performance measures and pay-for-performance schemes aim to improve quality of care in all arenas of health care, including the emergency department (ED). Performance measures established by the Centers for Medicare & Medicaid Services (CMS) and The Joint Commission are among the most widely distributed and well known. In January 2009, the Department of Health and Human Services contracted with the National Quality Forum to vet quality and efficiency measures for use in reporting on and improving health care quality. When selecting new measures for implementation, CMS is encouraged to choose from among measures approved by the National Quality Forum.
One of the main concerns has been the potential for unintended consequences of such measures on facilities that provide care to vulnerable populations. Such consequences are of particular concern to EDs. Although all EDs must, by law, provide care to any patient presenting to their doors, those identified as safety-net EDs provide a disproportionate share of services to patients with Medicaid and the uninsured. The number of EDs qualifying as safety-net providers has increased from 43% in 2000 to 63% in 2007.1
In 2008, the National Quality Forum approved 2 quality measures related to ED length of stay: the median time from arrival to ED departure for admitted patients and for discharged patients.2 Although these measures do not stipulate specific acceptable timeframes for ED length of stay or proscribe a given percentage of ED patients that must meet these goals, other organizations have suggested a median or 90th percentile less than 4 hours for discharged patients and less than 8 hours for those admitted to the hospital.3,4 If these measures are tied to pay for performance, chronically underfunded safety-net EDs could be at risk of further reductions in funding, which could only exacerbate the lack of resources available in those settings.
This study examines the performance of US EDs with respect to length of stay targets of 4 hours for patients discharged to home, transferred to another hospital, or admitted to observation and 8 hours for those admitted to an inpatient bed. We hypothesize that safety-net EDs perform worse on the ED length-of-stay measures than non–safety-net EDs as measured by medians and 90th percentiles.
We analyzed data from the 2008 National Hospital Ambulatory Medical Care Survey (NHAMCS), an annual national probability sample survey of visits to EDs of noninstitutional general and short-stay hospitals conducted by the Centers for Disease Control and Prevention (CDC) National Center for Health Statistics.5 The NHAMCS data are derived through a multistage estimation procedure that produces unbiased estimates.6 The 2008 NHAMCS data set was obtained from 431 of 463 emergency service areas (93.1% unweighted ED response rate) and a total of 34 134 patient visits. This is a publicly available data set with no patient identifiers; therefore, this study was exempt from review by the institutional review board of the University of California, San Francisco.
We examined ED lengths of stay for all adult ED visits from 2008 stratified by disposition and hospital safety-net status. Adults were defined as individuals 18 years or older. Emergency department length of stay is defined as the interval between time of ED arrival and time of ED departure (either admission or discharge). In the 2008 NHAMCS ED visit survey tool, individual ED visits could have multiple dispositions. For example, a single ED visit could have a disposition of admission to the hospital and admission to observation. To create mutually exclusive categories, we assigned ED visits with a single disposition of admission, discharge, observation, transfer, left without being seen, and died in the ED according to the following hierarchy: (1) patients younger than 18 years, missing ED length of stay data, those with no answer to disposition, dead on arrival, left against medical advice, and disposition of “other” were excluded; (2) those with any disposition of “transfer” were considered transferred; (3) of the remaining, those with any disposition of admission to observation were considered admitted to observation; (4) of the remaining, those with any disposition of admission to the hospital were considered admitted to the hospital; (5) of the remaining, those with any disposition of left before medical screening examination or left after medical screening examination were considered as left without being seen; (6) of the remaining, those with any disposition of having died in the ED were considered to have died in the ED; (7) of the remaining, those with any disposition of “no follow-up planned,” “return if needed,” “return or refer to physician or clinic for follow-up,” or “refer to social services” were considered “discharged.”
Admitted and discharged patients were grouped into categories designated by the National Quality Forum's National Voluntary Consensus Standards for Emergency Care, which was conducted under contract from CMS: critical care, psychiatric, and routine (noncritical care, nonpsychiatric) admissions.2 We identified psychiatric visits as those with a primary International Classification of Diseases, Ninth Revision, code meeting those established by the Healthcare Cost and Utilization Project Mental Health and Substance Abuse Clinical Classifications Software.7
Safety-net status was determined according to the CDC definition of more than 30% of total ED visits with Medicaid as the expected source of payment, more than 30% of total ED visits with self-pay or no charge as the expected pay source (considered uninsured), or a combined Medicaid and uninsured patient pool greater than 40% of the total ED visits.8
We obtained the following demographic and presenting characteristics: patient age, sex, race/ethnicity, triage acuity, and clinician type (attending physician, resident/intern physician, nurse practitioner or physician assistant, or other or missing). For the 2008 NHAMCS survey, race/ethnicity was entered by hospital personnel according to each hospital's usual practice. Data abstractors were instructed not to ask patients for this information and, for those cases for whom the race/ethnicity was not known or obvious, to enter what they thought was most appropriate. The National Center for Health Statistics replaced missing values with imputed values randomly assigned from patient records with similar characteristics. Safety-net EDs provide care for a disproportionate volume of patients with Medicaid and who are uninsured who are themselves overrepresented by racial and ethnic minorities. We therefore included race/ethnicity in our analyses to investigate whether either was independently associated with compliance with the proposed ED length-of-stay targets. Triage acuity is defined by the immediacy with which a patient should be seen (immediate, 1-14 minutes, 15-60 minutes, >1-2 hours, >2-24 hours, no triage, or unknown). Additionally, we obtained hospital ED characteristics including hospital setting (rural, urban), hospital ownership type (nonprofit, government, proprietary), and location (Northeast, Midwest, South, West).
We present unweighted and weighted characteristics of ED visits in descriptive terms. Emergency department length of stay data are presented as medians (interquartile range) and 90th percentiles stratified by ED safety-net status and patient disposition. Ninety-five percent confidence intervals and P values were calculated using standard methods accounting for the complex survey design and sampling weights.
We chose to analyze the data with respect to the suggested ED length of stay goals of a median or 90th percentile less than 4 hours for discharged patients and less than 8 hours for those admitted to the hospital as suggested by the Accreditation Council for Graduate Medical Education and Ontario Ministry of Health, respectively.3,4 We created separate bivariate models to assess the association between ED safety-net status and performance on ED length of stay goals for patients admitted to the hospital (stratified by all admissions, critical care admissions, psychiatric admissions, and routine admissions [all noncritical care and psychiatric admissions]), discharged (stratified by all discharges, psychiatric discharges, and routine discharges [all nonpsychiatric discharges]), admitted to an observation unit, transferred to another hospital, and left without being seen.
To determine independent associations with compliance with the proposed ED length of stay targets, we developed multivariable models stratified by disposition type. All predictors except patient insurance (because this was factored into the ED safety-net status) were included in the multivariable models. Results are presented as odds ratios with 95% confidence intervals. All analyses were performed using SAS version 9.2 (SAS Institute Inc) and Sudaan, version 10.0 (RTI International) to account for the complex sampling design and the patient weights.
Patients who left without being seen could potentially lower the median length of stay and result in harm, depending on the characteristics of the patients who left. Similarly, patients who died in the ED may also lower the median length of stay. We determined the number of unweighted observations, weighted percentage of visits, and ED lengths of stay among patients who left without being seen and among those who died in the ED, stratified by ED safety-net status. To explore the effect these 2 patient groups have on median length of stay and on compliance with the proposed length of stay target for admitted patients (<8 hours), we constructed a multivariable model, incorporating both patient groups into the admitted patient group.
Of the 2008 NHAMCS data set, 27.9% of the weighted visits were excluded, leaving 72.1% for analysis. Of the latter, 42.3% were seen in safety-net and 57.7% in non–safety-net EDs (Figure).
Overall, patients going to safety-net EDs were more likely to be young and minority than those treated
at non–safety-net EDs (Table 1; eTables 1-4 for demographics by analysis category, available at http://www.jama.com). They were less likely to need emergent or urgent care in both admitted and discharged populations.
For admitted patients, the median ED length of stay was 269 minutes (interquartile range [IQR], 178-397 minutes) for safety-net EDs vs 281 minutes (IQR, 178-401 minutes) for non–safety-net EDs. Critical care admissions accounted for 12.5% of all admissions to safety-net EDs and 13.2% in non–safety-net EDs in 2008. The median ED length of stay for critical care admissions was 236 minutes (IQR, 149-371 minutes) in safety-net EDs vs 248 minutes (IQR, 157-346 minutes) for non–safety-net EDs. Psychiatric admissions accounted for 3.9% of all admissions to safety-net EDs and 3.2% of non–safety-net EDs, with the median ED length of stay of 253 minutes (IQR, 172-506 minutes) for safety-net EDs vs 290 minutes (IQR, 173-579 minutes) for non–safety-net EDs. For discharged patients, the median ED length of stay was 156 minutes (IQR, 95- 239 minutes) for safety-net EDs vs 148 minutes (IQR, 88-238 minutes) for non–safety-net EDs. Complete results of the median and 90th percentile ED lengths of stay stratified by patient disposition safety-net status are presented in Table 2. Median lengths of stay for admitted, discharged, transferred or observed patients or patients who died in the ED were similar between safety-net and non–safety-net EDs. The mean ED lengths of stay among those who left without being seen in non–safety-net EDs were shorter (97 vs 120 minutes).
Results of the bivariable (ED length of stay by patient disposition and ED safety-net status alone) and multivariable models for odds of failing to comply with a target ED length of stay of less than 8 hours for admissions are shown in Table 3 (eTables 5-7 depict the results of the bivariable and multivariable models for subcategories of admitted patients, available at http://www.jama.com). Results of the bivariable (ED length of stay by patient disposition and ED safety-net status alone) and multivariable models for odds of failing to comply with the proposed ED length of stay of less than 4 hours for discharges are shown in Table 4 (eTables 8
and 9 depict the results of the bivariable and multivariable models for subcategories of routine and psychiatric
discharges, respectively). eTables 10 and 11 present the analyses of patients admitted to observation and transferred to another hospital.
Emergency department safety-net status is not independently associated with ED length of stay for patients admitted, discharged, transferred, or admitted to observation. This was true not only for all ED admissions and discharges but for all the subcategories tested with the exception of psychiatric discharges (OR, 1.67 [95% CI, 1.02-2.74]). Nonwhite race is independently associated with longer ED length of stay among admissions, a finding consistent with prior reports.9 Male sex is independently associated with shorter ED length of stay for psychiatric admissions and nonpsychiatric discharges. Lower triage acuities are independently associated with prolonged ED length of stay among admissions and with shorter ED length of stays among those discharged. Clinician type (nurse practitioner, physician assistant, or resident) is independently associated with prolonged ED lengths of stay for admitted (resident only), discharged patients (nurse practitioner, physician assistant, and resident), and transferred (resident only) patients.
eTable 12 shows the results of the sensitivity analysis in which we incorporated patients who left without being seen or had died in the ED into the admission category. This did not significantly alter the outcome (OR for length of stay >8 hours in safety-net EDs, 0.83 [95% CI, 0.52-1.33] relative to non–safety-net EDs, which is nearly identical to the results of the original model).
Although concerns have been raised that performance measures, particularly those linked to payment, may ultimately penalize safety-net institutions that are already underfunded and that care for a disproportionate volume of patients with poorer health care status, our findings suggest that those concerns about ED length of stay will not penalize safety-net institutions.10 Our results show that both safety-net and non–safety-net EDs perform well on the ED length of stay goals that have been proposed, with median ED lengths of stay for both ED types well under 8 hours for admissions and under 4 hours for discharges.3,4
Beyond this, however, we find that evaluating median length of stay alone provides only limited information; the 90th percentile results are more revealing. Both safety-net and non–safety-net hospitals demonstrate poor performance with the ED length of stay goals when the 90th percentile is used. Lengths of stay among routine and critical care admissions at safety-net hospitals had a 90th percentile of nearly 10 hours each, and more than 15 hours for psychiatric admissions. The 90th percentile ED lengths of stay for these same dispositions were only slightly better at non–safety-net EDs. The 90th percentile ED lengths of stay among nonpsychiatric discharges approached 6 hours regardless of safety-net status. The 90th percentile ED length of stay for psychiatric discharges was more than 22 hours for safety-net EDs and 15 hours for non–safety-net hospitals.
Previous studies demonstrate that the median ED length of stay has increased over time, approximately 3.5% per year.11 Our findings about the 90th percentile ED length of stay are particularly concerning, given that this measure is often seen as a surrogate marker for crowding.12 It is plausible that ED length of stay for patients with certain psychiatric conditions, for example, is skewed by an abundance of intoxicated patients requiring time to sober before being able to be safely discharged. Although this may be true, prior research has demonstrated significantly longer ED lengths of stay for psychiatric admissions as a result of a lack of psychiatric inpatient beds.13 In general, it is now widely accepted that ED boarding (the practice of admitted patients remaining in the ED due to lack of an available staffed inpatient bed), alternatively known as access block, plays the largest role in crowding in the ED.14 In other words, ED crowding is the result of hospital crowding. Emegency department crowding has been associated with adverse effects such as the timeliness and quality of care, patient satisfaction, and increased rates of medication errors in both pediatric and adult populations.15- 24
Prolonged ED lengths of stay may be the consequence of poor throughput secondary to ED inefficiencies or the result of lack of output (ie, no inpatient bed available for an admitted patient to move to). Although currently there is no accepted ED length of stay target in the United States, Graff et al25 suggested that 2 hours is “best practice.” Emergency department throughput targets of 4 to 8 hours are currently being tested in Canada, New Zealand, and Australia.3,26,27 The ultimate cause of poor performance on ED throughput measures may differ among nations and individual institutions, and thus solutions to this problem may differ. Lessons learned from the implementation of perhaps the most aggressive attempt to regulate ED throughput may be particularly valuable. In 2005, British EDs were mandated to have 98% of their patients leave within 4 hours of arrival (either discharged or in an inpatient hospital bed). Weber et al,28 in a qualitative study of the implementation of the “four hour rule,” found that success was dependent on a collaborative approach between the ED and hospital leadership. Viewing the mandate as an ED rule rather than a hospital rule only encouraged conflict among staff. Additionally, focusing on the target rather than on the patient potentially places patients at risk.
Our study has several limitations. Data collection for the NHAMCS survey is conducted by the US Census Bureau. The NHAMCS surveyors attempt to safeguard against the introduction of errors at this stage by requiring hospital staff to perform the actual visit sampling and data collection from the medical record. The NHAMCS field staff conducts completeness checks on site before forwarding data, and clerical staff edits the data in an attempt to reduce errors. The inclusion of self-reported data fields, such as insurance status, and variables with high nonresponse rates, such as race/ethnicity, may introduce inaccuracies. The NHAMCS analysts use imputation in case of missing variables that could contribute to inaccurate data as well.
This study includes the 2008 data for US ED visits, the latest available data. It is unclear what effect, if any, the current recession and resultant increase in uninsured and Medicaid populations will have on ED visit volume or length of stay and their distribution among safety-net and non–safety-net EDs. Additionally, effects of the sweeping health care reform currently under way remain to be seen. Analysis of ED visit rates following health care reform in Massachusetts has been mixed.29
A critical piece of the implementation of payment rules based on performance metrics is careful consideration of its financial effect on safety-net institutions. Our findings show that compliance with proposed ED length of stay measures for admitted, discharged, transferred, and observed patients do not differ between safety-net and non–safety-net hospitals and could be a useful measure for assessing throughput across these institutions.
Corresponding Author: Christopher Fee, MD, Department of Emergency Medicine, University of California, San Francisco, 505 Parnassus Ave, PO Box 0208, San Francisco, CA 94143 (firstname.lastname@example.org).
Author Contributions: Dr Fee had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs Fee and Hsia contributed equally to this work as senior authors.
Study concept and design: Fee, Hsia, Burstin.
Acquisition of data: Hsia, Maselli.
Analysis and interpretation of data: Maselli, Fee, Hsia, Burstin.
Drafting of the manuscript: Fee, Hsia.
Critical revision of the manuscript for important intellectual content: Fee, Hsia, Burstin, Maselli.
Statistical analysis: Maselli, Fee, Hsia.
Obtained funding: Hsia.
Study supervision: Fee, Hsia.
Conflict of Interest Disclosures: The authors have completed and submitted the ICMJE form for Disclosure of Potential Conflicts of Interest. Dr Fee has reported that he received consultancy fees from Google and honoraria from the University of California, San Francisco Office of Continuing Medical Education for organizing and speaking at an annual CME course. Dr Burstin is a paid employee of the National Quality Forum. No other disclosures were reported.
Funding/Support: This work was supported by grant KL2 RR024130 (Dr Hsia) from the NIH/NCRR/OD UCSF-CTSI and by the Robert Wood Johnson Foundation Physician Faculty Scholars Program (Dr Hsia).
Role of the Sponsors: The funding agencies had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.