We conducted a prospective study at Maharaj Nakorn Chiang Mai Hospital which was university hospital from Dec 24, 2019, to Jan 31, 2020. The facility is a 1,500-bed university hospital, level I trauma, and tertiary center with an ED census of about 30,000 patients per year. The ED has 16 treatment bays and monitors. The medical ED physicians include emergency medicine residents, interns, and attending emergency physicians.
The research assistants were well briefed on the study protocol and prospectively collected the data. Data included the number of physicians, the number of patients at each triage level, patients with a plan of disposition (patients waiting for admission and referral), ED code activations including of ST-Elevation Myocardial Infarction (STEMI), acute ischemic stroke, sepsis, and severe Traumatic Brain Injury (TBI), adverse events of the code activations, in-hospital cardiac arrest, and overcrowding perception of ED attending physicians. The data were collected every period, which is two hours. Therefore, we collected twelve periods per day. This is a similar method to an original EDWIN study [19]. Adverse events of the code activations include the delay of more than 90 min of wire crossing in cases of STEMI, a delay of more than 60 min of thrombolytic treatment in acute ischemic stroke, the delay of more than 60 min in the administration of antibiotics in sepsis, and the delay time of more than 60 min in surgery for severe traumatic brain injury.
All participants, attending emergency physicians or the chief emergency residents were asked to assess how overcrowded the ED was at that particular time using the six-point Likert Score: (1) not busy at all, not crowded, (2) steady, easily keeping up, (3) average: working hard, (4) more crowded and busy than desirable, (5) very busy, need external resources (doctors, nurses, ventilators or other equipment), and (6) extremely busy, hospital overcrowding code activation. All physicians were orientated, and eight different ED scenarios tested the agreement of the busy ED before participating. Interrater reliability of overcrowded ED among the physicians showed Fleiss’ Kappa of 0.582 (95% CI 0.493–0.671), p < 0.001 [20].
All variables were recorded by research assistants who were blinded to all physicians at that moment. “The triage system is based on Canadian Triage and Acuity Scale (CTAS). The triage score is reversed when calculating the EDWIN score, for example, the Score of 5 is for the most severe case (Level I– CTAS), and the Score of 1 is for the least (Level V– CTAS).” All identifiers were removed from the information as it was obtained.
The analysis team analyzed the overcrowding scores; EDWIN, occupancy rate, Work Score and modified EDWIN. The studied overcrowding scores were defined as follows:
$$ {EDWIN}^{20}= \sum {n}_{i}{t}_{i}Na(BT-BA)$$
(1)
number of patients in ED in triage category; ti is the triage category in CTAS (ordinal scale 1–5, 5 being most acute); Na is the number of attending physicians on duty at a given time; BT is the number of treatment bays, available in ED; and BA is the number of admitted patients who held in ED. For BA, the authors included the patients who were planned to be referred to another hospital and generated a new score, modified EDWIN.
The Emergency Department occupancy rate or EDOR is calculated by dividing the number of patients currently in the ED by the total number of available treatment beds in the ED. The formula is as follows:
$$\begin{aligned} {ED\,OR}^{12} = &Total\,Equation\,Number\, of \,patients / \\&Total\, Equation\,Number \,of \,licensed \,treatment \,bays\end{aligned}$$
(2)
In this study, the maximum number of monitored beds in the ED, which is 15, is used to calculate the EDOR score. For the Work Score, it is similar to EDWIN score in taking into account the triage level, effective ED size, and a number of providers. It divides the CTAS by the number of nurses rather than physicians. Work Score was calculated by using the following formula:
$$ \begin{aligned}{Work Score}^{21}=& (3.23 \times {P}_{wait}/ {B}_{T}) + (0.097 \times \\& \sum {n}_{i}{t}_{i}/{N}_{n}) + (10.92 \times {B}_{A}/{B}_{T})\end{aligned}$$
(3)
where Pwait is the number of patients in waiting room; BT is the number of treatment bays; ni is the number of patients in ED in triage category; ti is the triage category in CTAS (ordinal scale 1–5, 5 being most acute); Nn is the number of nurses on duty at a given time; and BA is the number of admitted patients who held in ED. Thus, the primary outcome was the correlation between the ED crowding scores and the Likert Score of overcrowding reflected by the emergency physician. The secondary outcome was to identify which Score has the highest sensitivity and specificity and which relates to the code activations’ adverse events.
Statistical analysis
The experimental size calculation was conducted based on prior statistical evidence [12]. We used the simple regression method for the primary outcome. With a power of 90% and an alpha value of 0.05, the total estimated sample size was 437. We decided to include 459 samples (5% estimation) in the data analysis in case of missing data.
The data collected for this study were organized and entered into a Microsoft Excel 2010 spreadsheet. Statistical analysis was performed using SPSS (Statistical Package for Social Sciences) version 22 (SPSS Inc., Chicago, USA). Descriptive statistics, such as median and interquartile ranges (IQR), were used for non-normally distributed variables.
Differences between the means of two groups were compared using either independent t-tests or Wilcoxon rank sum tests, depending on the data distribution. The Wilcoxon rank sum test is a non-parametric statistical hypothesis test employed to ascertain whether there exists a significant difference between the distributions of two independent samples. This test is specifically designed for situations where the assumptions of parametric tests, such as the t-test, are not satisfied, especially when dealing with ordinal or non-normally distributed data. While ANOVA was used for comparing the differences of means among more than two groups in the case of continuous data.
Categorical variables were presented as frequency and percentage and compared using a chi-square test. The correlation between crowding scores and physician perception was analyzed using the Spearman correlation coefficient.
To identify which Score shows the highest correlation with adverse events, the researchers assumed that a Likert score of 5 and 6 represents an out-of-control busyness in ED and the need for external resources as a reference standard to diagnose ED overcrowding in this study. The diagnostic incidence of adverse events was calculated using the area under the receiver operating characteristic curve (AUC). The cutoff value to predict adverse events was identified to indicate the highest sensitivity and specificity. A value of p < 0.05 was considered statistically significant. We presented the association of each calculated overcrowding score and the incidence of the adverse events and compared these using Chi-square tests.
Add Comment