Cardiovascular complications in a diabetes prediction model using machine learning: a systematic review

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Cardiovascular disease (CVD) is a significant risk for type 2 diabetes (T2DM) patients.
• Accurate prediction models are crucial for early diagnosis and prognosis.
• Machine learning (ML) offers advanced tools for developing robust prediction models.

Approach:

• A systematic literature search was conducted on Scopus and Web of Science (WoS).
• Risk of bias (ROB) was assessed using the Prediction model Risk of Bias Assessment Tool (PROBAST).
• Reporting standards were evaluated against the Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) standard.

Key Points:

• Neural networks demonstrated high performance (76.6% precision, 88.06% sensitivity, 0.91 AUC) for CVD risk prediction in T2DM.
• The overall applicability of included studies was low, with some studies having high or unknown ROB.
• Adherence to TRIPOD standards averaged 53.75%, indicating room for improvement in reporting.

Conclusions:

• Neural networks are a reliable ML algorithm for CVD risk prediction in T2DM patients.
• Future models require adherence to PROBAST and TRIPOD for reduced bias and enhanced clinical applicability.
• Incorporating lipid peroxidation markers could further improve prediction model performance.