Emergency Department (ED) crowding occurs when the demand for ED care is higher than the available resources, and has been associated with increased morbidity, mortality and decreased quality of ED care (1, 2). Historically, crowding has been a minor problem in Swedish EDs (3), but recent reports have found an association between increased length of stay or high occupancy rate and increased mortality (2, 4, 5). There are several combined measures based on staff assessment of crowding, like the national emergency department overcrowding score (NEDOCS), international crowding measure in emergency departments (ICMED) and Emergency Department Work Index (EDWIN). However, these models have not been evaluated against patient mortality and are not readily generalisable to Swedish EDs due to differences in ED operations between health care systems (6).

We have previously derived and validated the modified Skåne Emergency Department Assessment of Patient Load model (mSEAL) to measure crowding based on workload in Swedish EDs (6). The model includes the variables Patient Hours and Time to Physician. Patient hours is defined as the sum of time in hours that all patients spent in the ED during the previous hour, divided by the ED census during the same hour. Time to Physician is the average time in hours, from registration at the ED to first physician contact, for patients waiting for a physician during the previous hour. The mSEAL score is calculated as 1.485 + (9.715*Patient Hours) + (0.177*Time to Physician), with results ranging from 1 (no crowding) to 6 (extreme crowding). The mSEAL model correlated well with staff-assessed crowding in the derivation and validation study, with test characteristics comparable to that of occupancy rate (6). However, the model’s ability to identify situations where crowding affects patient outcomes is still unclear.

Occupancy rate (OR) has been promoted as an easy-to-use metric for crowding that performs equal to more complex models (7), like the mSEAL, but its ability to predict mortality related to crowding has varied in recent studies. Jo et al. (8) and Af Ugglas et al. (2) found an association between high OR and increased mortality in two Korean hospitals and in a Swedish county, respectively. In contrast, Jones et al. (9) and Derose et al. (10) found no association between OR and mortality in large studies from New Zealand and the US, respectively.

It is reasonable to assume that a substantial proportion of the negative effects of crowding are mediated through high workload, and that a metric based on staff perception may provide additional information compared to operational measures like OR. The aim of the present study was to investigate if ED crowding, as measured by mSEAL based on staff assessment or OR, is associated with increased all-cause short-term mortality.

A total of 172,336 visits were made to the study EDs during the study period. After excluding duplicates (n = 3,424), very long lengths of stay (LOS) likely to be incorrect (>3 days, n = 2) and incorrect mortality data (n = 10), 168,900 visits made by 122,833 patients were included in the final analysis (Figure 1). There were considerable differences in hospital admissions, 30 day mortality and LOS between the included EDs (Table 1).

Table 2 shows that LOS and unadjusted mortality for all three follow-up periods seemed to increase with increased crowding. The reference group had higher mortality than the lowest crowding group and was similar to the >90% groups. Mortality was generally higher in the reference group compared to patients arriving at hours when the mSEAL score could not be calculated (missing).

The Cox proportional hazard model adjusted for age, sex, hospital admission, acuity, chief complaint and EMS arrival is described in Table 3. High levels of crowding, defined as a mSEAL score over the 95th percentile for ED and hour of day, was associated with a non-significant increase in HR for 7-day mortality of 1.04 (95% CI 0.96–1.13). Similarly, an occupancy rate above the 95th percentile at arrival was associated with a non-significant increase HR for 7-day mortality of 1.03 (0.95–1.13). The HR for one- and 30-day mortality for mSEAL and occupancy rate above the 95th percentile were also increased, but all results were not statistically significant with confidence intervals included one. When left-truncating the data, i.e., removing all subsequent visits by the patients during the follow-up period, the statistical power decreased and 95% confidence intervals widened for all models (Supplementary Tables S1). In our cohort, 4.6% of patients who died within 7 days of a visit had >1 visit during these 7 days. Of patients who did not die, 7.5% had >1 visit within 7 days.

There were large differences in average mSEAL score between the study EDs for each exposure group, as shown in Table 4. The highest average reference and 95th percentile scores (3.79 and 5.17) were observed at Lund, and the lowest at Motala (3.03) and Norrköping (4.25), respectively. Similarly, there were large differences in occupancy rate at the 85th and 95th percentile for each hour of day and ED where the 95th percentile occupancy rate was consistently 2–4 times higher in the ED with the highest occupancy rate (Lund) compared to the ED with the lowest occupancy rate (Linköping) (Supplementary Tables S2, S3).

In this study, crowding defined as an mSEAL score above the 95th percentile had no statistically significant association with increased HR for 7-day mortality adjusted for age, sex, arrival by EMS, acuity, chief complaint and admission to hospital, compared with non-crowding (score < 85th percentile). There were positive hazard ratios for death in the highest crowding group (>95th percentile) for both mSEAL and occupancy rate, but the 95% CI included one, and the results were therefore not significant at the p < 0.05 level.

Crowding occurs when the demand for ED care is higher than the available resources with negative effects for patients and providers (24). Although the definition is universal, the extent of the problem and its consequences may vary between institutions, being more prominent in urban and academic EDs (25, 26). A recent study by Af Ugglas et al. (27) found variable association between crowding and mortality within Sweden and a previous study by our group showed considerable differences in resource availability in Sweden, specifically treatment beds and staffing in relation to patient volume (15). Our study included a mix of rural, urban and academic centers which may contribute to the lack of association between crowding and mortality. Although initial analysis indicated variation in mortality between sites, the differences were small given the limited overall absolute mortality and further subgroup analysis was not performed.

The average mSEAL score of the different crowding groups varied substantially between the EDs; Lund ED had the highest average score of 5.17 at the 95th percentile group and Norrköping ED the lowest at 4.25. There was also a large variation in baseline occupancy rate (Supplementary Table S2). Previous studies from our group and others (15, 28–30) have shown that crowding may indeed vary between EDs and that scores need to be adjusted to the individual site. In this study, we also found large intra-day variations at all EDs for both mSEAL scores and occupancy rate (Supplementary Tables S2, S3). By stratifying the scores by ED and hour of day we accounted for these differences, making them comparable between EDs and likely interchangeable as markers of crowding.

When left-truncating the data (counting only the first visit during the follow-up period), the HR point estimates decreased and the uncertainty of the results increased, likely due to less statistical power. Truncation will potentially remove bias caused by multiple ED visits before death that may systematically skew the results if patients that are more likely to die visit the ED more often than other patients. This has been shown during the last year of life (31–33), but is unknown for short-term follow-up. It was not the case in our study where patients with repeated visits within the follow-up period had lower mortality compared to the cohort, limiting the possible confounding. Furthermore, truncation may increase the risk of missing a true crowding effect on mortality since crowding may appear at an excluded ED visit. ED visit data were left truncated in the study by Af Ugglas et al. (2), but not in the study by Jones et al. (9).

In this multicenter study in Sweden, high ED crowding measured by mSEAL or occupancy rate was not significantly associated with an increase in short-term mortality.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Ethics committee of Lund University, permit number 2016/69. The ethics committee waived the requirement of written informed consent for participation.

JW, DW, and UE conceived the study and designed the trial. UE obtained the ethical permit. JW coordinated data collection and managed the data, with the assistance of DW and UE. DW and UE obtained the research funding. JW drafted the manuscript and takes responsibility for the paper as a whole. All authors contributed to the article and approved the submitted version.

This work was supported by two grants from Region Östergötland to author DW (LIO-532001 and LIO-700271), and from Region Skåne to author UE. The funding bodies had no role in study design, data collection, data analysis or writing of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.