SGEM#503: Waiting is the Hardest Part – Factors Associated with ED LOS

Date: February 13, 2026

Reference: Lang et al. Factors associated with emergency department length of stay in Alberta: a study of patient-, visit-, and facility-level factors using administrative health data. CJEM. 2026 Jan 29.

Guest Skeptic: Dr. Paul Parks is an emergency physician from Medicine Hat, Alberta. He has been the President of the Alberta Medical Association (AMA) Section of Emergency Medicine for many years, the AMA Board of Directors for 9 years, and the Previous President of the Alberta Medical Association. Paul has won the Canadian Association of Emergency Physicians (CAEP) National Teacher of the Year Award and the CAEP Alan Drummond National Advocacy Award.

Case: A 78-year-old man with congestive heart failure (CHF) and chronic obstructive pulmonary disease (COPD) arrives at the emergency department (ED) by ground emergency medical services (EMS) at 15:30 with dyspnea and hypoxia. He’s triaged Canadian Triage and Acuity Scale (CTAS) 2, needs non-invasive ventilation (NIV), diuresis, labs, chest x-ray, and likely admission. The department is packed; multiple admitted patients are boarded in hallway spaces because inpatient beds are unavailable, and nursing assignments are stretched. The patient is placed in the “EMS-PARK” area, which is an extension of the waiting room, and part of a mandatory EMS offload policy. Workup is done while the patient is still technically in the waiting room. The workup and disposition decision happen within a few hours, but transfer to an inpatient bed doesn’t occur until 2-3 days later.

Background: ED length of stay (LOS) can be considered a vital sign of ED operations and the broader acute-care system. When LOS rises, it often signals that the ED is no longer functioning as a short-stay diagnostic and stabilization unit but is serving as a buffer for upstream demand and downstream capacity issues. The consequences are not just operational (hallway beds, delayed assessments, delayed analgesia, delayed imaging), but also human. We covered a study that showed for older patients, one overnight stay in the ED waiting for an inpatient bed was associated with a 4% absolute increase in mortality (SGEM#424). In addition, increasing LOS can lead to clinician burnout and moral injury.

LOS is also tricky because ED crowding is rarely a single-point failure within the ED. Modern crowding frameworks (often summarized as input–throughput–output) remind us that while ED processes matter, some of the most powerful determinants are output constraints. This is especially true when there is access block and inpatient bed scarcity. In other words, you can run an efficient front-end, but if admitted patients cannot be moved to inpatient beds, the system backs up, and ED LOS climbs. As one concrete example of the output challenges many provinces struggle with, in Alberta, 1/3 of our acute hospital capacity, or about 30%, can be occupied by Alternate Level of Care patients. These alternative level of care (ALC) patients have had their acute care needs met, but they cannot be safely discharged from the hospital without specific continuing care resources – home care, assisted living, or long-term care.

We’ve talked about ED crowding on an SGEM Xtra. It covered some of the Zombie Ideas that have been circulating around for decades. The classic one is to blame non-urgent patients for using the ED. They are not responsible for ED crowding. Diverting non-urgent patients away can be dangerous and won’t solve the underlying problem.

CAEP published a position statement on emergency department overcrowding in 2013. CAEP argued for nationally standardized performance benchmarks. The statement also called for system-level solutions to improve flow while recognizing that ED optimization alone cannot solve crowding without hospital-wide and community-wide action. While CAEP’s advocacy has influenced awareness, policy discussion, and accountability framing, significant problems continue into 2026.

Clinical Question: Across Alberta ED visits, what patient-, visit-, and facility-level factors are associated with longer ED length of stay?

Population: ED visits drawn from linked Alberta Health Services administrative data for 14 ED facilities in Alberta, covering May 2022 to March 2023.
Exposures: Factors such as age, deprivation measures, EMS arrival, triage acuity (CTAS), primary care continuity, time/day patterns, and facility-level constraints, including emergency inpatient pressure and hospital occupancy; staffing signals (hours worked per nurse) were also examined.
Comparison:Between levels of each exposure, typically relative to a reference category or per-unit change (hospital occupancy, EMS vs non-EMS arrival, different facility types, weekday vs weekend, etc.).
Outcomes
- Primary Outcome:ED total length of stay (LOS).
- Secondary Outcomes: There were no clearly prespecified secondary outcomes; however, the analysis was stratified by disposition (admitted vs discharged vs other = LWBS, Left AMA, transferred, or died), which functions like a planned subgroup/stratified analysis rather than a distinct secondary endpoint.
Type of Study: This is an observational cross-sectional study using population-based administrative data.

Authors’ Conclusions: “ED length of stay is associated with modifiable factors, including hospital capacity constraints, hours worked per nurse, and healthcare access inequities. Addressing hospital occupancy, optimizing staffing, and improving care coordination across the patient trajectory—such as between the ED, inpatient units, and post-discharge services—may enhance ED efficiency and reduce prolonged stays. Our findings align with established frameworks describing ED overcrowding and support targeted, system-level interventions to improve the efficiency of emergency care.”

Quality Checklist for Observational Studies (Yes/No/Unsure)

Did the study address a clearly focused issue? Yes
Did the authors use an appropriate method to answer their question? Yes
Was the cohort recruited in an acceptable way? Unsure
Was the exposure accurately measured to minimize bias? Unsure
Was the outcome accurately measured to minimize bias? Unsure
Have the authors identified all-important confounding factors? No
Was the follow-up of subjects complete enough? N/A
How precise are the results? Very precise due to a large sample size, resulting in narrow confidence intervals for several of the point estimates.
Do you believe the results? Yes
Can the results be applied to the local population? Unsure
Do the results fit with other available evidence? Yes
Who funded the trial? The authors acknowledge support under the Alberta Atlas of Healthcare Variation initiative.
Did the authors declare any conflicts of interest? Brian R. Holroyd was the Senior Medical Director of the Emergency Strategic Clinical Network of Alberta Health Services at the start of this work. Matthew Pietrosanu was employed by Alberta Health Services for statistical consulting, technical writing, and general advising in the Alberta Atlas of Healthcare Variation initiative, which was expanded to include the preparation of this manuscript.

Results: The dataset included 587,419 ED visits. The median age was 38 years, and 52% were female. Most patients were discharged (68%), with 18% being admitted and 14% left without being seen, left AMA, transferred, or died. The median ED LOS was 3.1 hours overall, and LOS differed substantially by disposition (admitted patients had a much longer median LOS than discharged patients).

Key Result: Facility- and system-level constraints were strongly associated with ED LOS, especially among admitted patients. The more emergency inpatient hours and higher hospital occupancy were associated with longer stays.

Primary Outcome: Across all disposition categories, several patient-level factors were consistently associated with longer ED LOS, including older age, higher material or social deprivation, and arrival by EMS (ground or air). At the visit level, higher triage acuity and certain temporal factors (weekend admissions) were also associated with prolonged LOS, particularly among admitted patients.

However, the largest and most clinically meaningful associations were at the facility level. Measures of hospital capacity strain dominated the results. Higher hospital inpatient occupancy and a greater number of emergency inpatients boarding in the ED were strongly associated with longer LOS, especially for admitted patients. For admitted patients, a one–standard deviation increase in hospital occupancy (approximately 0.11) was associated with a 17% increase in ED LOS, an effect size that dwarfed most patient- and visit-level predictors. This finding strongly supports the concept of access block (outflow from the ED) as the primary driver of prolonged ED stays.

Higher hours worked per nurse were associated with shorter ED LOS in initial models, suggesting a potential staffing effect. However, this association disappeared after accounting for facility-level clustering, indicating that staffing effects may reflect broader organizational or structural differences between hospitals rather than a simple linear relationship with nursing hours.

1) Cross-Sectional Design & Temporality: The biggest design constraint is that this is a cross-sectional observational analysis. Exposures and outcomes are assessed within the same time frame. This means the direction of association can be unclear and may be difficult to determine.

2) Selection Bias: Although the dataset is large, it is not all Alberta EDs. The authors explicitly acknowledge that excluding facilities without staffing/occupancy data introduces possible selection bias and limits generalizability. Within the eligible facilities, the analytic sample also excluded visits with missing information (notably on deprivation and primary care continuity), which can further skew associations if missingness relates to both exposure and LOS. This study was done in 2022-23, and Alberta was still just rolling out its province-wide Connect Care / an Epic electronic information system and charting, and so it could be interesting to revisit this study in a couple of years, as there will likely be a lot less missing data or incomplete information going forward.

3) Unmeasured Confounding: This is the classic Achilles heel of observational studies. Even when investigators adjust for measured covariates, we should be cautious that residual differences between groups may still bias results. In this specific paper, the authors note that important clinical factors, such as presenting complaints and comorbidities, were unavailable. This is the sort of unmeasured severity/complexity factors that can confound relationships between deprivation, EMS arrival, facility type, and LOS.

4) Measurement Error & Misclassification: Administrative data are powerful but imperfect. Measurement error in timestamps, coding, or derived indices can create misclassification bias. This happens when patients/exposures/outcomes are misclassified, diluting or distorting associations. In this data set, several constructs (deprivation indices, “hours worked per nurse,” continuity metrics) are proxies rather than bedside-measured clinical variables. The paper itself signals model sensitivity in at least one area: the association between hours worked per nurse and shorter LOS disappeared after accounting for facility-level nesting, suggesting that some findings may be dependent on modelling choices and clustering structure.

5) External Validity: This study reflects a specific provincial system (Alberta), a defined set of 14 facilities, and a particular operational context (including factors like Connect Care implementation and local bed-management structures). As a result, ED leaders elsewhere should treat effect sizes as context-dependent and focus on whether their own system constraints mirror the exposure patterns studied here.

Comment on the Authors’ Conclusion Compared to the SGEM Conclusion: We generally agree with the authors’ conclusions.

SGEM Bottom Line: In this large observational study, ED LOS (especially for admitted patients) was associated most strongly with system-level output constraints (hospital occupancy/boarding), reinforcing that meaningful LOS improvement requires hospital-wide flow solutions, not just faster ED throughput.

Case Resolution: Our 78-year-old patient with COPD and CHF finally got admitted to the ward after 2 days as an inpatient in the ED. He is the third patient placed in a room originally designed for only two, but he is thankful for the excellent care he receives on the ward. During his stay in the hospital, he witnesses that the ward nurses are also working in challenging circumstances, with patients in the hallways and extra ones jammed into any available space, and they are often working understaffed. But there’s a slight twist: given his worsening COPD and medical condition, he can no longer return to independent living in his own home and requires placement in assisted living. He finds himself one of those ALC patients stuck in the hospital awaiting safe discharge with continuing care supports. He spends weeks as an ALC patient awaiting a safe placement. While stranded in the hospital with a prolonged LOS, he is inspired to write his local politician to tell them about the hell that it is to be a patient stranded for days in an EMS offload area, and then weeks in an overcrowded hospital. He tries to put a human face on his experience, so that he isn’t just another statistic.

Dr. Paul Parks

Clinical Application: This should inform your system-level view of hospital crowding. If your admitted patients are boarding and your hospital occupancy is high, you should expect LOS to worsen regardless of how efficient your front-end triage and physician decision-making are, because output constraints dominate flow. This supports prioritizing interventions that target hospital-wide capacity and transitions, while also recognizing patient-level inequities that may increase complexity and prolong ED stays. Importantly, because associations do not prove causation in observational studies, clinicians should use these findings to guide quality-improvement hypotheses and resource discussions, rather than to assign blame to specific groups.

What Do I Tell the Patient? For those stranded in an ED awaiting an in-patient bed, who are well enough (or to their loved ones at the bedside), maybe consider writing to your provincial politicians to help with hospital output solutions. This is because these delays will increase their LOS and end up costing the system more, and the patients have more morbidity and mortality.

Keener Kontest: Last week’s winner was paramedic Chris Bare. He knew the first ECG task to be successfully automated using AI techniques was the detection and classification of cardiac conditions, such as atrial fibrillation.