Table 1 Selected studies (see additional file 4 for more detail)  

[11]

To reduce cognitive load on clinicians by predicting the risk for admission

[15]

To reduce mortality by predicting the risk for (severe) sepsis in the ED

[16]

To help physicians by predicting the need for hospitalization

[17]

To help streamline crowded EDs by developing an AI tool that could remove the need for an expert emergency medicine physician during triage

[18]

To enhance ED triage systems by predicting mortality risk and risk for cardiac arrest

[19]

To prevent overcrowding of EDs by predicting future ED visits

[20]

To reduce ED morbidity and mortality by predicting the disposition of asthma and COPD exacerbation after triage

[21]

To increase physician satisfaction and reduce physician burnout by improving the efficiency and quality of structured data

[22]

To reduce/prevent overcrowding of EDs and improve patient care by predicting the need for hospitalization

[23]

To reduce ED morbidity and mortality costs by predicting risk for sepsis at triage and by implementing protocolized care

[24]

To reduce the length of stay (LOS) in ED by predicting clinical ordering at triage

[25]

To reduce/prevent overcrowding of EDs by predicting the risk for cardiac arrest in ED

[26]

To reduce ED morbidity and mortality and overcrowding of EDs by predicting triage levels for patients with suspected cardiovascular disease (CVD)

[27]

To cope with the increasing demand for clinical care in EDs by predicting septic shock at triage

[28]

To alleviate overburdened EDs and increase patients’ throughput by identifying patients’ need for a head CT scan at triage

[29]

To alleviate overburdened EDs by improving patient categorization by predicting ED mortality

[30]

To improve patients’ throughput in EDs by identifying severe thorax injury

[31]

To reduce overcrowding of EDs by predicting patient waiting times

[32]

To reduce overcrowding of EDs by developing an e-triage system

[33]

To improve patient outcomes and reduce adverse effects by identifying patients at risk for acute kidney failure

[34]

To prevent adverse outcomes by predicting/identifying the geriatric need for hospitalization

[35]

To improve patient outcomes by identifying scaphoid fractures

[36]

To improve patient outcomes by predicting patient waiting times

[37]

To cope with overcrowding of EDs through predicting critical care and hospitalization outcomes at triage

[38]

To improve patient outcomes by linking prehospital records to hospital records

[39]

To safely reduce hospital admissions by predicting risk for 30-day adverse severe events

[40]

To improve patient outcomes and enhance physician ability by identifying ECG outcomes

[41]

To increase patient throughput in crowded EDs by predicting patient disposition during triage

[42]

To reduce diagnostic errors (and costs & overutilization of resources) by predicting/identifying urinary tract infections (UTIs) early

[43]

To improve healthcare delivery by predicting future hospital demand

[44]

To improve healthcare provider wellbeing and preserve patient safety by predicting clinician workload

[45]

To cope with overcrowding of EDs by predicting adverse clinical outcomes at tirage

[46]

To improve patient outcomes by identifying septic shock at an early stage

[47]

To reduce diagnostic errors and excess costs by predicting and identifying severe cardiac events