JC: Emergency Department Delays and Mortality

Estimated reading time: 7 minutes

This paper is recently off the press and has already sparked both interest and debate here at Virchester and in the wider Emergency Medicine community. It tackles the topic of whether delays to patient admission from the ED are associated with an increase in 30-day mortality and, spoiler alert, it concludes that it does.

This study comes at a time of political and media focus on Emergency Care and the headline findings are such that our Royal College of Emergency Medicine (RCEM) president has openly responded to them in a published statement on the college website. The work builds on a recent report published by RCEM in November 2021 titled “crowding and its consequences” which clearly outlines the harms and dangers of ED crowding.

This paper tells a story that is all too familiar and gives findings that many like me will want to believe as proof, to show those who still need it, of the dangers of our current state of play. Working in a crowded Emergency Department, within a hospital at capacity, caring for patients with increasingly long waits for an inpatient bed, it is easy to believe that longer delays lead to an increase in deaths. But whilst we may want to believe this paper, we must not let cognitive bias cloud our evidence appraisal process and we must cast the same critical appraisal eye that we would always use at St. Emlyn’s, so this is an objective (as possible) review.

We are currently corresponding with the authors in relation to some of the thoughts outlined here. The potential for this study to influence policy makers is immediately evident so it is imperative that the right conclusions are drawn.

The abstract is below, but as we always say, read the paper for yourself and come to your own conclusions.

What kind of paper is this?

This is a cross-sectional, comparative, retrospective observational study which means it compares data at a specific point in time looking backward examining exposure in relation to a known outcome. In this case the outcome is all-cause 30-day mortality and the exposures included delay to hospital admission from ED.

Tell me about the patients.

This is big data; the study used routinely collected NHS England Hospital Episode Statistics (HES) data and Office of National Statistics (ONS) mortality data (linked by NHS digital and anonymised prior to being received by the authors).The population is all patients admitted to hospital from all type 1 EDs in England between April 2016 and March 2018. (Type 1 EDs are open 24 hours a day under the supervision of a consultant in Emergency Medicine).

Over this period 26,738,514 ED attendances resulted in 7,472,480 admissions. If patients were admitted multiple times within the study period (~30%), only the first admission was included. The authors also excluded patients that waited over 12 hours (3%) from the mortality analysis alongside those with missing data (0.09%) giving 5,249,891 patients included in the analysis.

What about the intervention?

There were no interventions, this is an observational study with the variables being analysed not under study control.

What about the outcomes?

The primary outcome was death from all causes within 30-days of hospital admission from the ED. This is a hard endpoint and clearly relevant to patients. However, the decision to settle on 30-day mortality over a shorter period that may more directly reflect consequences from a delay to admission from ED, such as 7-days, wasn’t explored and feels arbitrary. Importantly, and as the authors openly acknowledge, neither patient morbidity nor patient experience are examined within the study.

What did they find?

There were 433,962 30-day deaths following hospital admission from the ED which gave a crude mortality rate of 8.7%.

The median age of admitted patients was 55 years old and unsurprisingly comorbidities increased with age. Over 1.8 times as many patients were admitted from the highest decile of deprivation as from the lowest which is a stark reminder of the public health side of socioeconomic deprivation. The most frequent time of arrival to the ED was between 12:00 – 17:59 and the busiest quarter was Jan – March with 28% of ED attendances. The mean length of stay was 291 minutes (just under 5 hours) and the 4-hour breach rate was ~38%.

The headline finding is a linear increase in mortality with increasing delay to admission from ED over 5 hours with a visible inflection point after the 4-hour mark. The linear regression analysis shows a 0.008% increase in mortality per minute between 4 – 12 hours (p < 0.001). The authors report a number needed to harm of 191 for patients waiting for admission between 4-6 hours, 82 for 6-8 hours and 72 for 8-12 hours. 

So what does this mean?

We need to be cautious with how we interpret and use this data. Whilst the study shows an association between delay to admission from ED and 30-day mortality, the methodology does not allow causality to be demonstrated.

There are potential confounders that have not been adjusted for and we cannot know from this work alone whether length of stay in ED is the cause of the increase in mortality observed. The authors have developed their model and used it to predict 30-day mortality and calculate the standardised mortality rate (observed/expected deaths) used to report the main headline finding. Whilst the model adjusts for many confounders (patient characteristics, time of ED attendance and the level of ED crowding), importantly the severity of illness has not been accounted for.

Higher acuity patients will often remain in ED longer receiving value added care or awaiting a scarce high dependency inpatient bed. The authors were limited to available data and we have no doubt that had they been able to design the dataset this would have been factored. We must however consider this in our interpretation of the findings.  Reporting the number needed to harm and inference of a causal relationship is beyond what we can reliably conclude. 

This study is not focussed on a prediction model and has understandably not been published with a TRIPOD checklist regarding how it was developed or derived. Whilst the area under the receiver operating characteristic curve for is 0.862 (excellent), given the significance of the model to the analysis, full reporting in line with the TRIPOD checklist would appease uncertainty around the model calibration.

The notable trend in increasing risk of 30-day mortality with longer delays to admission at 4-hours is very interesting and worth discussing. The authors offer several plausible explanations for the temporal relationship which include crowding, delays to treatment and increased risk of harm with longer stays. However, none have face validity in adequately explaining why this occurs just after the 4-hour mark. It is far more likely that we are observing the consequences of system-wide behaviour linked to the 4-hour performance target. You may even hear in your mind a voice asking, “your patient is on 3 hours 35, what’s your plan?” and words along the line of “well they’ve breached now”.

The 4-hour target has become entrenched within our thinking and decision making and separating the impact of behaviour linked to this from the variable of time is incredibly difficult. If we had a deeply rooted 6-hour target would the same inflection point be seen after this time point? Conclusions about performance at other time-based targets from this data would be counterfactual.

So what does this actually mean?

This study clearly shows that, in the current system with our current 4-hour target, patients admitted to hospital from the ED that are delayed past the 4-hours target are at an increased risk of death (at 30 days). Whether this is due to confounding factors such as acuity of illness or clinician decision or to the delays to medications and substandard care associated with crowding inherent with prolonged length of stay in the ED, is not established.

However, regardless of causality this is an accurate reflection of the system as it stands, and we should see it as such. Patients awaiting hospital admission, who go beyond the 4-hour target, need particular attention. You might be sat with them in resus, but they equally maybe in a majors cubical, or even more out of sight in the waiting room, a victim of exit block and department crowding and deprioritised because they’ve already been seen (albeit a few hours ago) and have a plan.

Our take-away

This is a fantastic piece of work linking big data with national outcome datasets and is amongst the best evidence we have on the 4-hour target. It demonstrates a very clear and thought-provoking association between delay to admission from ED and mortality. Importantly the standardised mortality does not account for confounders including severity of illness and any conclusion of causality cannot be reliably drawn. It is imperative that healthcare workers, managers and policy makers address timely admission. There is a deeper discussion on what appropriate targets should be, what outcome measures we should use to assess performance for Emergency Care and the benefit and harm of time-based targets.

Govind Oliver

Charlie Reynard


  1. Jones S, Moulton C, Swift S, et al Association between delays to patient admission from the emergency department and all-cause 30-day mortality Emergency Medicine Journal Published Online First: 18 January 2022. doi: 10.1136/emermed-2021-211572
  2. RCEM responds to EMJ article that reveals five hour + waits increase risk of death. 18th January 2022
  3. RCEM Acute Insight Series: Crowding and its Consequences. November 2021
  4. Critical Appraisal Nuggets at St Emlyn’s

Cite this article as: Govind Oliver, "JC: Emergency Department Delays and Mortality," in St.Emlyn's, March 9, 2022, https://www.stemlynsblog.org/jc-association-between-delays-to-patient-admission-from-the-ed-and-all-cause-30-day-mortality-st-emlyns/.

Thanks so much for following. Viva la #FOAMed

Scroll to Top