Using Publicly Available Electronic Healthcare Records MIMIC-III for Process Mining with Cardiovascular Patients Case Study

Abstract

Researchers in process mining in healthcare find difficulties in accessing healthcare datasets from hospital for several reasons. The Medical Information Mart for Intensive Care III: MIMIC-III provides an opportunity to solve this challenge by making an unidentifiable electronic healthcare record publicly available. It contains information for more than forty thousand patients whom are over ten years old from different clinical departments at an American hospital. The events occurred to patients stored with timestamps data; this making MIMIC-III usable for process mining purposes. However, since it was not built with process mining requirements in consideration, there was a need to investigate MIMIC-III and document how to use it to mine processes existing within its tables. There were previous researchers who investigated and used MIMIC-III for process mining successfully, but had no detailed documentation provided.This project aimed to deeply investigate MIMIC-III and identify how to use it for process mining, then providing ready made scripts, SQL queries, for future researchers. The investigation work was a success, and it was documented thoroughly in this project, along with provided SQL queries. The project deliverables were supported with a case study included to show how process mining researchers utilized and provided guiding methodology and supporting SQL queries in action. The experiment and case study were evaluated deeply using various methods of evaluation. This project can be seen as a contribution to process mining academic communities, as it can be useful in teaching and learning for beginners in process mining before they set off to work on real client data.