Independent determinants of prolonged emergency department length of stay in a tertiary care centre: a prospective cohort study

A prolonged ED length of stay (LOS) keeps doctors and nurses longer occupied with one patient, and decreases the effective capacity which contributes to overcrowding. Emergency department (ED) overcrowding is associated with worse patient outcomes and satisfaction among health workers and patients [1‐3]. Crowding is also a problem in the Netherlands; 68% of the ED managers experienced crowding several times a week or even daily [4]. In the UK the 4-h rule was introduced to restrict the ED work-up time [5]. Also in our hospital in the Netherlands crowding is an issue affecting satisfaction of patients and medical personal. Therefore, an ED LOS of 4 h was adopted as an important cut-off point to indicate prolonged LOS.

Reduction of ED LOS may contribute to reduction of ED overcrowding [2, 6]. Despite the fact that numerous ED and patient characteristics have been associated with prolonged ED LOS, a recent systematic review demonstrated that previous studies were inappropriate to decide which factors need to be targeted to optimize ED logistics [7]. Several issues arise in these studies. First, almost all have been retrospective studies while real-time measurements are especially important when studying time delays in ED logistics. Secondly, often only a few factors have been measured while all potential bottlenecks should be taken into account to understand which factors should be targeted and can actually be modified in clinical practice. For example, age has frequently been associated with ED LOS [7‐9]. However, this could be caused by age-related differences in comorbidity and disease severity. Comorbidity has been associated with a higher number of ED visits, hospital admissions, readmissions and health care costs [10‐13]. Comorbidity increases, in combination with multiple presenting complaints, the complexity of care and could contribute to a prolonged ED LOS. Typically, patients of a tertiary care centre have more comorbidities and presenting complaints.

Finally, previous studies mostly originate from the USA or Canada [7]. The European ED setting is quite different with regard to the number of ED visits and ED LOS. In addition, EDs are staffed by both ED physicians and other specialists (as opposed to only ED physicians), and the general practitioner has an important role as gate keeper in referral of patients to the hospital [14].

Successful solutions for reduction of ED LOS in the European health-care setting can only be developed if ED patient flows are prospectively studied and independent modifiable ‘bottlenecks’ are identified by using multivariable prediction modelling [15].

The purpose of the present study was therefore two-fold: first, to identify independent determinants of prolonged ED LOS (including all important patient, doctor and ED (management) factors). Secondly, to measure (in real time) the absolute time delays associated with the modifiable independent determinants of prolonged ED LOS, and the delays associated with decision-making and ED discharge.

This was a prospective observational cohort study including ED patients of the Leiden University Medical Centre (LUMC), a Dutch tertiary care centre with ~ 26.000 ED visits each year. Data were collected from 16 December 2014 to 11 February 2015.

In the Netherlands approximately 84 hospitals have EDs, the highest concentration in the western part of the Netherlands. Patients are either referred to the ED by a general practitioner (GP) or are self-referred (which is accepted). Furthermore, the Netherlands is divided in regional ambulance services. In the region of the LUMC the ambulance service “Hollands Midden” is responsible for most of the ambulance transports. This district covers 875 km2 with 760.000 citizens and is accountable for ~ 60,000 transports per year. In 95% of the rides ambulances arrive at the scene within 15 min of the dispatch call. The Dutch ED setting is characterized by the presence of both ED physicians and other specialists in the ED. The ED is staffed 24/7 by ED physicians, who are responsible for self-referred patients, trauma and critical care and patients who are directly referred to an ED physician.

The study was approved by the medical ethics committee of the LUMC, who waived the need for individual informed consent because of the purely observational character of the study (Protocol number P14.288).

All consecutive patients presenting to the ED between 10 a.m. and 10 p.m. were included during 30 randomly chosen days, including weekends, in a 9 week time period [16]. Approximately 70% of all ED patients per 24 h arrive in the selected time period_. It is important to note that this way of sampling does not create selection bias [17].

In Fig. 1 the patient flow through the ED is depicted schematically.

Demographic variables, referral status, type of arrival (by ambulance or own transport), triage category, treating specialty, comorbidity, type and number of presenting complaints, diagnostic testing, consultations of other specialties in the ED, number of patients present in the ED at the time of ED registration and final disposition, were prospectively registered by an observer (DvdV) in a standardized case report file in SPSS (SPSS V.23.0, IBM, New York, USA). Transit times (ED bed placement time, ED departure time and time of physician arrival, consultation request and arrival of the consulted physician, time of physically leaving the ED) were measured in real-time.

In the digital hospital information system (Chipsoft, Amsterdam) information about the ED registration time, the day of discharge and times of diagnostic testing (times when laboratory analysis was started or radiology imaging was requested and times when these were finished) were obtained. In the final conclusion of the medical file the total number of presenting complaints or problems were quantified. The triage category and triage complaint were registered according to the Manchester Triage System (MTS).

Comorbidity was assessed in two ways: the Charlson comorbidity index (CCI) was calculated and comorbidities were registered by an organ-based method [18]. For exact definitions and scoring system for the comorbidities see online Additional file 1.

ED LOS was calculated by subtracting the ED registration time from the time that the patient physically left the ED. If the ED departure time was after 10 pm, the time registered in the digital hospital information system was used. Waiting time was calculated by subtracting the ED registration time from the time a patient was placed in an ED treatment room. Time until seen by a physician was calculated by subtracting ED room entrance time from the time the physician had arrived.

The primary outcome measure of the present study was a total ED LOS of more than 4 h.

During the inclusion period a total of 2050 patients visited the ED, of which 1434 patients (70%) arrived between 10 am en 10 pm. The patient flow through the study is visualized in Fig. 2. Of the included patients, 1146 patients (80%) had an ED LOS < 4 h and 288 patients (20%) had an ED LOS > 4 h. Patient characteristics are summarized in Table 1. Patients with an ED LOS > 4 h were on average older (58 (21) versus 45 (26)) and had more comorbidities (2 (1–3) versus 1 (0–2). In addition, the number of blood testing (93% versus 57%), urine testing (40% versus 16%), radiology imaging (76% versus 41%), consultations (53% versus 19%) and admissions to a ward (51% versus 19%) was higher in patients with an ED LOS > 4 compared to patients with an ED LOS < 4 h.

In Table 2, the independent determinants for ED LOS > 4 h are shown. The adjusted ORs (95%-CI) were 3.45 (1.95–6.11) for blood testing, 1.79 (1.21–2.63) for urine testing, 3.02 (2.13–4.30) for radiology imaging, 5.92 (4.08–8.60) for consultations in the ED, 2.10 (1.21–3.66) for internal medicine and 2.24 (1.06–4.73) for neurology as treating physician, and 3.66 (1.08–12.41) for headache and 2.91 (1.49–5.71) for chestpain as presenting complaint. Age, number of comorbidities, and admission to a ward were not an independent determinant of prolonged ED LOS. Patients who were admitted to an ICU or MCU had a much shorter ED LOS; adjusted OR of 0.26 (0.10–0.66). The number of patients present in the ED at the time of ED registration was merely a determinant if 15 patients or more were present in the ED at the time of ED registration; Adjusted OR of 2.21 (1.30–3.75). In Fig. 3, ED LOS is depicted as a function of number of patients present in the ED at the time of ED registration in the 15 days with the lowest number of patients (3A) and the 15 days with the highest number per day (3B). Even during the busiest days ED LOS barely increased with increasing number of patients present in the ED.

The several time components, which contribute to the total ED LOS, are illustrated in Fig. 1.

The absolute time delays of diagnostic testing and consultations are shown in Table 3. In the total patient cohort the median ED LOS was 156 min (98–225). In 1123 patients (78%) at least one consultation or diagnostic test was requested, which increased the total ED LOS to 177 min (129–242). The median time between patient registration and requesting diagnostic testing or consultations ranged between 23 (13%) to 82 (46%) minutes and the median time between requesting and having the results ranged between 26 (15%) to 69 (39%) minutes of the total ED LOS. The median time between finishing diagnostic testing or consultation and ED discharge was 48 (27%) minutes of the total ED LOS.

Although our study has several strengths like the prospective design (with real time measurements of time delays) our study has also some limitations. Firstly, ED crowding was not measured according to a validated crowding score because existing crowding scores are not validated in the Dutch ED setting. However, the number of patients present in the ED was measured in real time and is a major factor related to overcrowding [23].

Secondly, our results may not be applicable for all hospitals because tertiary care centres usually treat more complex patients. As a result, the impact of consultations on ED LOS could be overestimated in urban hospitals, because of a higher consultation rate in tertiary care centres [24]. Thirdly, although our sampling method should not have introduced selection bias [17], it is theoretically possible that during night hours, laboratory and radiology services are more short-staffed, probably delaying diagnostic testing results. In addition, staff members are less available for supervision. Therefore, the impact of consultations and diagnostic testing on ED LOS could have been underestimated.

Finally, although in our hospital the number of patients per month is fairly stable throughout the year, there could be some impact of season. However, given the low impact of number of patients present in the ED on ED LOS (Fig. 3), it is not likely that variation in number of ED presentation with season does affect the independent determinants of prolonged ED LOS. It is possible that the type of patients varies with season but we adjust for that in the multi-variable regression analysis.