This single-center, retrospective cohort study was conducted at Wakayama Medical University Hospital between April 1 and September 30, 2019. The study population comprised consecutive patients who presented to the ED of a tertiary-care academic medical center, regardless of the reason for their visit, and were initially evaluated by the EP team. Patients without a CT scan and those who experienced cardiopulmonary arrest upon arrival were excluded. This study was approved by the Research Ethics Committee of Wakayama Medical University. The requirement for informed consent was waived due to the retrospective nature of the study (approval number: 1782).

Our institution has a tertiary teaching hospital with an advanced emergency center. The EP team initially manages all patients who present urgently to the ED, regardless of the reason for their visit and severity or cause (endogenous or exogenous). The team consists of board-certified EPs, senior residents (SRs), and junior residents (JRs), with the team size varying based on shift schedules. Non-EP specialists provide initial care for acute exacerbations of routine outpatients, perinatal patients, or transfer patients with stable vital signs who have been definitively diagnosed at other hospitals. EPs, SRs, and JRs rotate on two shifts (day shift = 9:01–17:00 and night shift = 17:01–9:00) to ensure continuous 24/7/365 ED coverage. During the day shift, the initial response team includes 2 or 3 EPs, 2 or 3 SRs, and 2 JRs. On the night shift, the team is reduced to a single EP. The ED staffing levels remain the same on weekends and holidays. EPs and SRs collaborate to order imaging studies, including contrast-enhanced CT (CECT). During the study period, CT scans were performed using a General Electric Healthcare 80-slice CT scanner installed in the ED. Consent for CT scans was obtained after explaining the benefits of enhanced diagnostic efficiency and the risks of radiological exposure to the patient or their family. During the study period, no real-time interpretation by radiologists was available. EPs and SRs jointly made decisions based on clinical findings and their own CT interpretations, determining the initial management and ED disposition plan.

The primary outcome was the occurrence of an IECM, reflecting the clinical impact of diagnostic errors. The secondary outcome was the total number of IEs made by EPs (IE), representing the radiological aspect.

During the 6-month study period, clinical and radiological data were collected to closely reflect the real-world conditions. Initial clinical data included age, sex, past medical history, underlying chronic disease, transport type (ambulance or emergency helicopter), time of ED visit (day shift = 9:01–17:00; night shift = 17:01–9:00), ED visit on a weekday or weekend/holiday, transfer from a nursing home or other hospitals, initial blood pressure, initial consciousness level, triage assessment using the Japan Triage and Acuity Scale (JTAS), ED crowding, ED length of stay, magnetic resonance imaging, and hospitalization. The Charlson Comorbidity Index was calculated based on the pre-existing conditions provided during patient interviews [14]. Initial systolic blood pressure below 90 mmHg just upon ED arrival was defined as “initial hypotension.” Impairment of consciousness was graded using the Glasgow Coma Scale score upon ED arrival. The JTAS, developed in 2012 based on the Canadian Triage and Acuity Scale (CTAS), is used in most advanced emergency centers in Japan. It includes five levels: level 1 (resuscitation), level 2 (emergent), level 3 (urgent), level 4 (less urgent), and level 5 (non-urgent). A study on JTAS validity found that levels 1 and 2 were associated with significantly more intensive care unit admissions than lower triage levels [15]. Therefore, JTAS levels 1 and 2 were combined as “emergency triage level” for this study. ED crowding was defined as the occurrence of three or more new ED visits within 1 h of each patient’s arrival. Although four major scales were used to assess ED crowding, the National Emergency Department Overcrowding Study is less effective in hospitals with less frequent crowding [16]. As this study was conducted in a non-urban emergency center, the standard ED crowding scale was not used; instead, local criteria based on EP’s subjective assessments of “ED busy (not overcrowded)” were applied. A “long ED stay” was defined as a length of stay exceeding 180 min.

CT images were interpreted by three board-certified radiologists (Na.S., No.S., and T.Y., with 10, 5, and 15 years of experience in emergency diagnostic imaging, respectively), following the standard institutional protocol that required review within 12 h of acquisition, often sooner and including real-time reading for critical cases. The initial CT interpretations and clinical management decisions made by EPs, documented in the electronic medical record, were subsequently reviewed to identify the IEs and IECM. This workflow—involving initial interpretation by the EP team followed by radiologist review within 12 h—represented the standard procedure for ED CT scans at our institution during the study period and had been established 2 years ago when three emergency radiologists were appointed. The standard protocol for trauma whole-body CT included scanning at least three sites (e.g., head, chest, and abdomen) [4]. Accordingly, a “multisite CT scan” was defined as the simultaneous imaging of three or more regions: head, face, neck, chest, abdominopelvic, spine, extremities, and vessels. Additional data on CECT and magnetic resonance imaging were also collected. IEs were categorized as perceptual errors (failure to recognize findings) or cognitive errors (misinterpretation of recognized findings). ED-affiliated radiologists reviewed all CT images to identify underreading as perceptual errors and faulty reasoning as cognitive errors, based on the 12 types of diagnostic imaging errors described in the previous literature [17]. Underreading referred to missed findings, whereas faulty reasoning denoted errors of overreading and misinterpretation, where findings were detected but attributed to incorrect causes [17]. When an IE was identified, discussions were held to determine whether the attending EP would need to change the initial clinical management (e.g., additional medical treatment or informed consent). IECM, the primary outcome, was defined as an IE that necessitated a change in the initial clinical management plan (e.g., additional treatment, specialist consultation, or altered patient disposition). IE, the secondary outcome, was defined based on the framework described by Kim and Mansfield [17]. It encompasses two main categories relevant to acute care: 1) perceptual errors (underreading), where acute findings were missed, and 2) cognitive errors (faulty reasoning), where acute findings were detected but misinterpreted regarding their cause or significance. Patients with multiple IEs were counted once per case. Incidental findings unrelated to acute management, such as fatty liver, asymptomatic inguinal hernia, aortic disease at unrelated sites, or incidentalomas not requiring urgent treatment as oncology emergencies, were excluded from the outcomes. All collected data were entered into a dataset and analyzed statistically.

Continuous variables were expressed as medians with interquartile ranges (IQRs), whereas categorical variables were expressed as numbers with percentages. A multivariate analysis was performed using binomial logistic regression to determine the independent predictors of IE by calculating the adjusted odds ratios (ORs) and 95% confidence intervals (CIs).

Based on previous studies and considering clinical plausibility, seven variables were included in the multivariate logistic regression model. EPs interpreted CT images in environments subject to various clinical disturbances. Therefore, multivariate analysis was performed under the hypothesis that clinical factors affect EP misinterpretations, in contrast to radiologists who interpret images in settings isolated from direct clinical activity. Four variables were included as clinical factors associated with reduced absolute or relative staffing: night-time ED visits, ED busy status, emergency triage level (JTAS 1 and 2), and prolonged ED stay [11, 18, 19]. Three variables were included as radiological factors that increase the complexity of image interpretation: multisite CT scan, CECT, and abdominopelvic CT [11, 13, 20, 21]. Final models were selected using stepwise backward elimination based on Akaike’s Information Criterion. A P value of less than 0.05 was considered significant. All statistical analyses were conducted using the JMP software (version 14.1; SAS Institute, Cary, NC, USA).