Implementation of machine learning algorithms to create diabetic patient re-admission profiles
Mohamed Alloghani1,2, Ahmed Aljaaf3,4, Abir Hussain3
1The Artificial Intelligence Department-, Dubai, UAE. phdmn2015@gmail.com.
BMC Medical Informatics and Decision Making
|December 14, 2019
Summary
Machine learning identified factors predicting diabetes readmission. Women, Caucasians, and patients with less rigorous treatment, especially those discharged without improvement, are at higher risk for readmission.
Area of Science:
- Artificial Intelligence
- Machine Learning
- Healthcare Informatics
Background:
- Machine learning (ML) is a subset of Artificial Intelligence (AI) focused on algorithm development, enabling computer learning.
- ML is increasingly vital across various sectors, including healthcare, education, and business.
Purpose of the Study:
- To apply machine learning techniques to a diabetes dataset.
- To identify patterns and factors associated with patient readmission in diabetes care.
Main Methods:
- Utilized multiple classification algorithms: Linear Discriminant Analysis, Random Forest, k-Nearest Neighbor, Naïve Bayes, J48, and Support Vector Machine.
- Analyzed a dataset of 100,000 cases, with 78,363 identified as diabetic.
Main Results:
- Over 47% of diabetic patients in the study were readmitted.
- Key predictors for readmission included being female, Caucasian, an outpatient, or receiving less intensive procedures/medication.
- Patients discharged without improvement, particularly those with positive HbA1c tests but without insulin administration, showed higher readmission rates.
Conclusions:
- Diabetic patients, especially women and Caucasians, with less rigorous assessments and treatments are prone to readmission.
- Discharge without improvement and inadequate insulin administration are critical factors for readmission in diabetes patients.
