Building Early Warning Scores With Statistical-Based Structured Methods on Patient Clinical Data

In order to facilitate the healing process and prevent mortality, hospital medical staff must provide appropriate and timely medical treatment. One method by which hospital personnel can reduce mortality rates is to accurately access the condition of patients upon their arrival at the hospital, thereby enabling the provision of appropriate medical treatment. Consequently, the utilization of a structured and statistically-based methodology will result in the automatic generation of EWS tables, which will assist in the identification of potential health hazards within the clinical data of patients. The dataset that was used in this study comprise of approximately 1000 records of patient vital signs, that were sourced from a combination of public dataset from Kaggle and real clinical data from MIMIC III. those data then put through comprehensive data preprocessing to ensure the data that will be used do not have any missing null values. The aforementioned data will then be utilized to generate an automated EWS Table through the application of statistically-based structured techniques on patient clinical data. This technique will employ the percentile method to assist in the identification of potential health risks. This technique will facilitate the identification and comparison of various risk variables in each patient. The efficacy of the generated Early Warning Score (EWS) table was evaluated by comparing it to an existing EWS table developed by the Royal College of Physicians, known as NEWS. The results of this comparison demonstrated an accuracy rate of seventy-one percent.