IMPLEMENT-ED

Emergency Departments (EDs) are typically a high-workload, dynamic and, often unpredictable clinical environment. In such an environment it is difficult to provide reliably accessible, efficient and safe patient care, as reflected in staff fatigue, prolonged ED length of stay (LOS), delayed ED transfers to wards and inefficient use of staff and resources throughout the system, all leading to crowding. 

Risk stratification tools can help in these situations by identifying sick or vulnerable patients. However, timely risk stratification can be difficult; especially during crowded times or at night there is a risk of omissions or misinterpreting an assessment. Risk stratification is even more useful when it is automated, as it allows immediate operational decisions to be made. A dashboard with automated real-time risk stratification combined with information on healthcare processes would address these matters. The introduction of a machine learning algorithm in such a dashboard would bring the advantage that it always performs and never fatigues at crowded times or in night shifts. 

While healthcare professionals recognize the value of new information technologies, such as automated risk coding dashboards and machine learning algorithms at the workplace, few studies have evaluated the effects on healthcare professionals and relevant clinical outcomes in the ED.

Objective 1: To assess the effect of implementation of a dashboard, with (near) real-time data feedback of risk stratification tools and care processes, on relevant clinical outcomes (in-hospital mortality, ED-LOS, admission rate) and discrepancy between predicted and observed admissions in a tertiary ED in the Netherlands.

Objective 2: To assess the effect of addition of a ML algorithm for hospitalization in this dashboard, on relevant clinical outcomes (in-hospital mortality, ED-LOS, admission rate) and discrepancy between predicted and observed admissions) in a tertiary ED in the Netherlands.

A ‘before-after’ design study of the effects of the implementation of a dashboard with real-time risk stratification and process of care information and of the effects of the introduction of a machine learning algorithm for hospitalization supported by an educational program, with a before-after-after design, performed in the emergency department of a tertiary care centre in (Leiden, LUMC) the Netherlands. Data is collected during six months before implementation (‘before’), during six months after implantation of the dashboard with an educational program (‘after I’), and during six months after implementation of an ML algorithm (‘after II’), using data of the Netherlands Emergency department Evaluation Database (NEED). 

Introduction of a dashboard with real-time information on care processes and in that dashboard a machine-learning based hospitalization prediction tool.  (Note: information provision and standard care, not a linked intervention related to care itself).

Benefits: The adoption of a dashboard at the workplace with automated real-time information about care processes and risks of patients attending the ED, could improve patient outcomes i.e. by helping healthcare personnel prioritizing the sickest and frailest patients and ensure early hospitalization of these groups to an appropriate level of care. In addition, the use of an ML algorithm concerning hospitalization can enable early initiating of the admission process, by automatically updating physicians' knowledge about a patients’ likelihood of admission and thus shortening ED-LOS. 

Burden: Not applicable.

Risks: Not applicable.