Type of facility influences lengths of stay of children presenting to high volume emergency departments

Emergency department (ED) crowding is an increasingly common concern for health care systems in developed countries. Crowding is a condition where the demand for services exceeds the ability to provide timely and high quality care [1]. There are many reasons why crowding is an important problem, including but not limited to: delays in delivering time-sensitive therapies such as analgesics [2‐5] and antibiotics, [4, 6‐10] increased departures prior to completion of care, [11‐13] worse patient outcomes [14‐20] and provider/patient dissatisfaction [14, 21, 22]. Research using population-based administrative data conducted in Ontario demonstrated that crowded EDs resulted in increased admissions and deaths [18].

The role of ED crowding in the care of pediatric patients is complicated by the fact that few studies have been performed. Pediatric patients may be seen in mixed EDs – which see both adult and pediatric patients – and the number of dedicated pediatric EDs is limited and restricted to 14 urban/academic centres across Canada. A study conducted in 2017 found a strong association between markers of crowding and pre-defined adverse outcomes in a pediatric ED in Vancouver, British Columbia [23].

There is no single universally accepted ED crowding metric [24] and multiple metrics have been defined that summarize aspects related to how patients arrive in the ED, the processes of care in the ED, and the patient’s disposition from the ED. [25] The Canadian Association of Emergency Physicians (CAEP) has recommended benchmarks for ED performance: [26] time to placement in an ED bed, time to physician initial assessment (PIA), and length of stay (LOS) in the ED (for admitted and discharged patients).

Our objectives were to provide a comprehensive picture of ED care for pediatric patients in one large geographic region and examine differences in crowding metrics in unique categories of EDs.

There were 465,709 children who made 1,124,119 presentations to these high volume EDs during the five year study period. Just over half (51.2%) of presentations were to the academic/teaching EDs that are dedicated pediatric EDs whereas regional EDs and urban EDs each had about 25% of the presentations (Table 1). Presentations were made by slightly more males than females (53.6% vs 46.4%), with a median age of 5.0 (IQR: 1, 12), and were made by children mainly in the most urbanized areas (Calgary and Edmonton). The number of presentations generally increased over time with the largest increases seen in the urban and academic/teaching EDs (Table 1, Fig. 1). Pediatric presentations to the ED were similar across ED categories for month, day of the week, and time. Urban and academic/teaching EDs had presentations with higher acuity than regional EDs (15.0, 12.8, and 7.7% were emergency triage level in the academic/teaching, urban, and regional EDs, respectively). Academic/teaching EDs had the highest proportion of admissions (7.8%) and smallest proportions of LWBS (2.8%) and LAMA (0.3%). Few presentations (n = 213) ended in death.

Two urban EDs opened at 7 am on January 14, 2013, and May 21, 2014, and the remaining 13 EDs were operational for the full duration of the five-year study period. Thus, a total of 596,628 facility-specific hours were used in the calculation of metrics during the study period. Some facilities and hours had missing values because there may have been no children presenting or presentations with non-missing times did not meet the disposition requirements (e.g., PIA based on 304,959 facility-specific hours, ED LOS for discharged based on 62,033 facility-specific hours, and ED LOS for admitted based on 37,574 facility-specific hours).

The metrics based on PIA were similar across the ED categories (Table 2, Fig. 2). The median of the PIA-M metric was 1h7min (IQR 37 min, 1h51min) for the regional EDs, 1h15min (IQR 43 min, 2h2min) for the urban EDs, and 1h8min (IQR 39 min, 1h54min) for the academic/teaching EDs. The PIA-M metric remained relatively stable over time (Additional file 1 Fig. 1). Of 916,641 ED visits where PIA could be calculated, 517,150 (56.4%) exceeded 1 h and 87,952 (9.6%) exceeded 3 h.

For presentations ending in discharge, the median LOS-M metric was 2h21min (IQR 1h34min, 3 h24min). The median LOS-M metric was highest for presentations at academic/teaching EDs (2h31min) and urban EDs (2h27min), and was lowest for regional EDs (2h05min). The median LOS-M metric for discharges remained relatively stable over time (Additional file 1 Fig. 2). The summary statistics for the LOS-A metric were close to the summary statistics for the LOS-M metric. When the 90th percentile is examined, the ED categories differed with median LOS-90 metrics of 3 h46min for the academic/teaching EDs, 2h46min for the urban EDs, and 2h25min for the regional EDs. Of the 996,380 discharges with times and triage level, 185,983 (18.7%) had LOS greater than 4 h (24.2% exceeded 4 h for CTAS 1, 2, or 3; 11.9% exceeded 4 h for CTAS 4 or 5 recommendations [26]).

Presentations ending in admission had longer LOSs than presentations ending in discharge: the median of the LOS-M metric was equal to 6h29min (IQR 4h16min, 9h43min). Academic/teaching EDs had the highest median of the LOS-M metric (7h28min), followed by urban (7h12min) and regional (4h32min) EDs. These metrics generally increased in the most recent years with more variability seen for the urban EDs (Additional file 1 Fig. 3). The summary statistics for the LOS-A metric were very close to the summary statistics for the LOS-M metric. When the 90th percentile was examined, the ED categories differed with median of the LOS-90 metric equal to of 7h54min for the academic/teaching EDs, 7h12min for the urban EDs, and 4h35min for the regional EDs. Of the 63,695 admissions with non-missing times, 23,380 (36.7%) had LOS greater than 8 h recommended [26] as the median LOS for admitted patients.

For the day aggregated metric of LWBS, the median percent LWBS was 0, 4.0, and 3.7% in academic/teaching, urban, and regional EDs, respectively (Table 2, Additional file 1 Fig. 4). Over time, LWBS has been variable with increases seen in more recent years for the academic/teaching EDs (Additional file 1 Fig. 5). Very few presentations ended with LAMA and the median percent LAMA was 0 across all categories of EDs and the mean percent LAMA was 1.2, 0.7, and 0.6% in academic/teaching, urban, and regional EDs, respectively.

This study focused on five years of population-based administrative data on presentations made by children (age < 18 years) to 15 high volume EDs in one Canadian province. High volume EDs are the most likely EDs to experience crowding and we described the crowding metrics PIA, LOS for discharges, LOS for admissions, LWBS, and LAMA that were obtainable from administrative data sources. These metrics focus on the efficiency (i.e., PIA, LOS) and safety (i.e., LWBS, LAMA) domains of medical care [33] and throughput (i.e., PIA, LOS) and output (i.e., LWBS, LAMA) aspects of the conceptual model of ED crowding [25]. Across three hospital categories, the patient demographics were similar: however, the academic/teaching EDs had the highest acuity scores and the highest proportion of admissions. Regional EDs had the lowest acuity scores and the lowest proportion of admissions.

The ED volumes for children have increased substantially over time for the academic/teaching and urban EDs yielding 40 and 47% increases during the 2010/2011 to 2014/2015 study period, respectively. These increases exceeded the 10% increase in the population of children and youth (≤19 years) [34] and the 12% increase in ED presentations for all ages, and all ED facilities during the same period [35, 36]. The disproportionate increases to the academic/teaching and urban EDs may reflect preferential choices for care in urban centres or lack of access to other care options. Urban EDs in the United States have seen an estimated increase in ED presentations during 2005 to 2016 of about 19% for all ages [37] and other jurisdictions have seen increases in ED presentations, [38, 39] many above the rate of population growth [38].

The crowding metrics based on delays in initial assessment were similar across ED category and time. The CAEP benchmark for PIA is a median of 1 h, [26] and 56% of the ED visits exceeded 1 h and the median of the PIA-Ms by hour and ED facility exceeded that recommendation by 11 min. Those results are similar to other Canadian studies where at Ottawa, Ontario’s children’s hospital during four, 2-week study periods during 2011/2012 where the median PIA was 1h30min, [40] in Ontario EDs during 2012 the median PIA as approximately 1 h, [39] and in urban EDs in Vancouver, British Columbia during 2012 the median PIA was approximately 1h10min [39]. In a Columbus, Ohio children’s hospital during 2016/17 the estimated mean PIA was 1h34min [41]. In a Cincinnati, Ohio pediatric teaching hospital during 2003 to 2007, the daily mean PIA was 53.6 m and increases in ED census was associated with increased PIA [42]. Crowding may increase PIA and such increases will increase the ED LOS.

Our study showed that the LOS metrics for both discharged and admitted patients was shortest for regional EDs. Academic/teaching EDs had the longest LOS metrics for both discharged patients (median of LOS-M = 2h31min) and admitted patients (median of LOS-M = 7h28min). The LOS for admitted patients has increased over recent years, bed capacity is low, regionalization of pediatric beds has occurred, and increased demands for inpatient beds for ED patients results in delays from a decision to admit to the time when an inpatient bed is available (so called “access block”). The LOS metrics for both discharged and admitted patients in this study are larger than EDs in Ontario (median LOS just under 2 h), [39] urban EDs in Vancouver (median LOS just under 2h30min), [39] and a Cincinnati pediatric teaching hospital (daily mean LOS was 2h30min) [42]. Another Ohio children’s hospital during 2016/17 had an estimated mean LOS of 4h10min [41] and in the United States national, the estimated median LOS for discharged patients of any age was 2h19min and for admitted patients was 4h22min in 2013 [43]. While the LOS metrics in our study were longer than in other studies, most of the ED presentations met the recommendations for LOS [26]. The longer LOS metrics in academic/teaching EDs compared to regional EDs may be directly related to the higher acuity and admission rates in academic/teaching EDs. Longer LOS in urban and academic/teaching EDs were not primarily a result of increased PIA as the PIA was similar across EDs.

The proportion of LWBS was small (all EDs: 4.1%) and variable, ranging from 2.8% for academic/teaching EDs to 5.7% for urban EDs. These results were similar to other jurisdictions in Canada. In Ontario EDs in 2008, 4.9% of presentations by children ended in LWBS and in urban Vancouver EDs LWBS occurred in 3–4% of presentations [39]. That same study found that LWBS did not change substantially with increasing volumes or acuity, and some centres in both provinces have participated in strategies to reduce ED wait times [39]. In our study, very few patients left against medical advice (all EDs: 0.5%), ranging from 0.3% for academic/teaching EDs to 0.8% for urban EDs. This means that while the average daily LAMA was 1.2% for academic/teaching EDs, it was only 0.6 and 0.7% for the regional and urban EDs, respectively. There was more variability in the daily metrics for academic/teaching EDs in our study that may be because of higher volumes and more variability in crowding during different times of day.

The implications of this study vary based on one’s perspective. For example, for patients and families, emergency pediatric care is a priority and efforts to ensure government ministries are aware and responsive should be a priority. Alternatively, delays in urban Pediatric EDs are impressive and implementing throughput interventions (e.g., fast track, staffing models, rapid assessment zones, etc.) require urgent attention. For the health care system, increasing the availability of primary care providers (e.g., Family Physicians and Pediatricians) and access to same or next day primary pediatric care may reduce ED visits in the larger centres. Clearly, each zone and hospital is unique and solutions need to be based on an understanding of the bottlenecks and evaluated to determine effectiveness.

This robust and comprehensive five-year study demonstrates that ED presentations for children have increased over time across the province, especially in the urban and academic/teaching hospitals. Metrics for PIA were similar across ED category whereas LOS for discharges and admissions varied by ED category. Crowding is more common in urban and academic/teaching EDs. The impact of crowding on outcomes for children presenting to the ED requires further study, especially in crowded academic/teaching hospitals. Moreover, interventions to address ED crowing in these locations are urgently required.