Prediction of disease comorbidity using explainable artificial intelligence and machine learning techniques: A...

Area of Science:

• Biomedical informatics
• Artificial Intelligence in Healthcare
• Computational Medicine

Background:

• Disease comorbidity presents significant challenges in healthcare, impacting patient quality of life and increasing costs.
• AI-driven prediction of comorbidities offers a pathway to improved precision medicine and holistic patient care.

Approach:

• A systematic literature review following the PRISMA framework was conducted across Ovid Medline, Web of Science, and PubMed.
• The review identified and summarized machine learning (ML) methods for comorbidity prediction, evaluating model interpretability and explainability.

Key Points:

• 61 ML models from 22 studies were reviewed, with 33 achieving high accuracy (80-95%) and AUC (0.80-0.89).
• 72% of the reviewed studies exhibited high or unclear risk of bias.
• Limited scope of comorbidities (1-34) and data (phenotypic, genetic) restricted the discovery of novel comorbidities.

Conclusions:

• This review highlights the diverse application of ML in predicting disease comorbidities.
• Advancements in explainable AI for comorbidity prediction hold potential for identifying previously unrecognized patient risks and unmet health needs.