Researchers: Anat Ratnovsky¹, Shai Rozenes¹, Pinchas Halpern²

Background: A panel of 20 experts from six collaborating medical centers selected 47 key performance indicators (KPIs) to improve the quality of care and patients' satisfaction.

Objectives: Examine the availability of the 47 KPIs in the six EDs, perform exploratory data analysis (EDA) on a dataset, design an interactive dashboard to display the KPIs and to develop a model to predict patient's admission probability.

Method: Meetings were held with members of the panel to determine measurable KPIs. Dataset of 172,695 records from 2016 was retrospectively collected from one ED. The relationships within the dataset was analyzed using EDA followed by confirmatory data analysis. Machine learning algorithms were used for the admission predictive model. A dashboard system, displaying the KPIs, was designed.

Findings: Five EDs can collect 85% of KPIs, 2.72% of all patients left against medical advice, the length of stay (LOS) was not affected by the week day. A linear increase was found between LOS and patients' age. The dashboard system can be employed. The developed predictive model yield ~79% prediction rate.

Conclusions: The dataset illustrated that overall ED 1 is allocating its resources properly to cope with ED overcrowding as can be seen from the lower rate of patients leaving against medical advice and insignificant difference in the mean LOS between weekdays.

Recommendations: The ability of most EDs to measure 85% of the KPIs enables standardization and a design of a unified quality indicators system to increase the effectiveness of EDs in Israel. Implementation of the designed dashboard and the predictive model should be employed to fill information gaps in the ED, support decision making and resource allocation for the improvement of KPIs.

Research end date: 10/2018