A randomized, controlled open-label trial was conducted in the PED at Lille University Hospital between March 15th and April 23rd, 2021 (NCT05994196), following CONSORT guidelines (Supplementary Material 1). Each day was randomized for the use of Optimum® (i.e., the interventional group) vs. the PED’s standard patient management dashboard (the control group). We assumed that randomization would distribute equally between the two groups patients who were or were not time-consuming for the medical staff. As only 23% of visits took place between midnight and 10 am, with no impact on patient flow, children admitted between 10 am and midnight were included in the analysis. Patients who left without being seen by medical staff and those subsequently admitted to a short-stay unit were excluded. The study protocol was registered with the French National Data Protection Commission (Commission Nationale de l’Informatique et des Libertés, Paris, France; registration number: DEC21-056). In line with the French legislation on analyses of anonymized data from clinical practice, approval by an institutional review board was not required. The patients and their parents were shown a study information sheet at the PED reception desk and were free to object to their child’s participation.

The trial was carried out in our tertiary care center, which receives almost 30,000 PED visits annually, mainly Caucasians. One third of children (0–16 years of age) had an underlying condition and 1/3 needed orthopedic or surgical management. The mean waiting time before being seen by a physician was 1h36 and the mean LOS was 3h25 in 2022 for those without further admission. From 2019 to 2021, 0.9 to 1.6% of patients left the ED without being seen. Physicians working in the PED were pediatricians and the triage at admission was performed by a nurse.

The primary goal of triage is to ensure that patients receive timely, appropriate care, maximizing positive outcomes and minimizing potential harm. Primary triage is a process to determine the urgency of further care at the time of patient arrival [10]. Several triage systems are used, including the Canadian triage scale in our PED. Secondary triage is a new concept for the ED, designed to enable patients to be organized, monitored and assessed effectively [11], even after the initial triage. The concept is developed in this trial with a new tool.

The standard patient management dashboard provides a list of patients in order of arrival with, next to the name, the length of time the patient has been on the unit, the level of urgency (color-coded triage), any tests or medical advices prescribed to be carried out or completed, and the initials of the staff in charge of the patient.

Optimum® was a proof-of-concept software developed in 2015 [12]. Starting from a database of 75,000 visits, 100 different reasons for PED visits were retrospectively defined, and the LOS in the PED was determined for each. The five variables with a statistically significant influence on the LOS were the reason for PED visit, the number of patients present in the ED simultaneously, the prescription of imaging, the prescription of blood tests, and the prescription of treatment [13]. The Optimum®’s interface is shown in Fig. 1. The data in the PED’s standard patient management dashboard required by Optimum® were transferred to Optimum® within five minutes in 83% of cases and within 10 min in 94% of cases [12].

The Optimum software’s interface and annotations on its use. The colored status bar next to the patient’s name corresponds to the LOS in the PED. The status bar’s color depends on the patient’s LOS, relative to that of patients admitted for the same reasons. The bar is green when the patient’s LOS is below the 50th percentile but changes to yellow when the LOS is between the 50th and 75th percentiles, to red when the LOS is between the 75th and 95th percentiles, and lastly to dark red (overcrowding) when LOS is greater than the 95th percentile. Once the PED is overcrowded, the priority is discharging patients rather than seeing new ones

Optimum®’s purpose is to remove the mental load of prioritization from the PED staff. The software first prioritizes the triage of new patients by the ED staff and the first evaluation of a new patient by the medical team. Optimum® then prioritizes blood sample collections and care (for nurses) and review of imaging results, blood test results or an evaluation by a specialist (for physicians). Lastly, Optimum® prioritizes the final step in patient management by a senior physician, when appropriate. It means that a character on a grey background (see Fig. 1, column 2, last thumbnail) appears when all the other actions have been completed, meaning that a decision by the senior physician is awaited: either a blood test, or a new treatment, or to validate a discharge.

The study’s primary endpoint was the LOS for each patient. We hypothesized that there would be a 15-minute difference in the LOS between the intervention and control groups. This would correspond to one less patient in the PED at a given time, when considering 28,500 PED visits per year (78 per day) and a median LOS of 190 min. To assess a median 15-minute difference in LOS between the two groups with an α-risk of 0.05 and a power (1-ß risk) of 0.8, we calculated that a total of 1542 patients had to be included (i.e., 771 per group). The secondary endpoints were the number of patients present at the same time in the PED per day and per period of the day, the time intervals between each stage in patient management from nurse triage to discharge, and the PED staff’s level of satisfaction.

During the two weeks prior to the start of the study, all the PED staff members were trained in use of Optimum®. This secondary triage tool was set up in the PED at this time, so that all the staff members could familiarize themselves with the tool and put any questions to the investigators.

The days were randomized to Optimum® vs. the standard dashboard using the “random” formula in Excel® (Microsoft Corporation, Redmond, WA, USA). Firstly, the study dates were entered into an Excel® spreadsheet and marked with a “0” or a “1” at random. The dates with a “0” were assigned to Optimum® (the control patient management dashboard was turned off) and those with a “1” were assigned to the control patient management dashboard (Optimum® was turned off). Weekdays and weekend/public holiday days were randomized separately.

The principal investigator (TL) was dedicated full-time to this research throughout the recruitment period. He was present in the medical office and had direct access to what was happening to each patient at every moment of the study. He was not involved in management of the study participants. He simply recorded prospectively the time intervals for each patient’s stay in the PED: time of arrival at the PED, evaluation by the triage nurse, the first medical evaluation (by a medical student or a junior physician), the first evaluation by a senior physician, the evaluation by a specialist physician (if applicable), the results of imaging and lab tests (if prescribed), the final medical decision, and discharge.

In addition to the time intervals between the various phases of patient management at the PED, the other variables recorded were age, sex, reason for PED visit, triage level (according to a simplified three-level version of the Pediatric Canadian Triage and Acuity Scale [PaedCTAS] [14]: level 1–2 of the PaedCTAS as level 1 or high-priority level, level 3–4 of the PaedCTAS as level 2 or moderate-priority level, and level 5 of the PaedCTAS as level 3 or low-priority level of the simplified version), and mode of discharge. Five categories of reasons for PED visit were chosen a priori: fever, a respiratory disorder, a digestive tract disorder, trauma, and other reasons.

At the end of the study, the PED staff involved in the study filled out the standardized System Usability Scale (SUS) questionnaire as a guide to the perceived utility of Optimum® and the level of user satisfaction. According to the literature, the SUS score is considered to be very poor if it is less than 51, poor if between 51 and 68, average if 68, good if between 68 and 80.3, and excellent if greater than 80.3 [15,16,17].

Statistical analyses were performed in Lille University Hospital’s biostatistics unit, using SAS software (version 9.4, SAS Institute Inc., Cary, NC, USA). Firstly, the patients’ characteristics were described. Categorical variables were expressed as the frequency (percentage), and continuous variables were expressed as means and standard deviations (SD) in case of normal distribution or medians with interquartile range [IQR] otherwise. The normality of the data distributions was checked graphically and using the Shapiro-Wilk test. Intergroup comparisons of the total LOS and the various time intervals during patient management times were performed by using an analysis of covariance (ANCOVA) (on log-transformed values; or on rank-transformed values for time interval between the prescription of a consultation with a specialist physician and the consultation itself, and time interval final between evaluation by a senior physician and the end of care) adjusted on priority group at triage (low-priority vs. moderate- and high-priority). Standardized differences and their 95% confidence intervals were calculated as effect sizes; absolute values of 0.2, 0.5 and 0.8 are interpreted as small, moderate and large effect size. Heterogeneity of associations between the various tile intervals and the use of Optimum® according to patient’s priority tirage was tested by adding an interaction term to the model. Curves of mean number of patients present simultaneously in the PED throughout the day were compared between the two groups using a functional analysis of variance and plotted on a graph.

For SUS score, a Spearman’s correlation coefficient was used to evaluate the relation between the age of the PED staff and the SUS score.