Healthcare Data Analytics and Patient Health – A Way Forward to Improve Decision-making and Accountability in Healthcare System

Abstract

Artificial Intelligence (AI) has been a major factor to transform healthcare delivery and decision-making in healthcare. Computer vision, machine learning, and deep learning are some of the AI techniques widely used in workflows of healthcare institutions to promote risk assessment, diagnosis, and care planning. This study aims to investigate the use of AI in healthcare sector to predict readmission within 30 days among diabetic patients and how it ensures accountability in making decisions. This study is also based on analyzing secondary data collected from previous studies and credible online sources. In addition, secondary data was also collected from a healthcare dataset “Diabetes 130 US hospitals for years 1999-2008” which is publicly available and consists of 101,766 encounters. The 30-day readmission was predicted using Random Forest and Logistic Regression models and performance of AI models was evaluated with precision, accuracy, F1-score, recall, and PR-AUC and ROC-AUC. It is found that AI can deliver vital information related for making clinical decisions to predict readmission of patients at risk. Hence, it is important to assess fairness, auditability, and explainability to determine predictive performance. This study recommends healthcare units to use AI to ensure accountability. Instead of replacing human decision-makers, AI should be used as a helping hand to provide transparent details that will help in decision-making. In addition, AI shows better performance in some tasks, such as, prediction and image-based uses. There is still need to wait until AI matures enough to be reliable in complex environments.