A decision support system for cost-effective diagnosis.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Clinical Decision Support Systems

Background:

• Optimizing diagnostic processes is crucial for efficient healthcare delivery.
• Balancing speed, cost, and accuracy in disease diagnosis presents a significant challenge.
• Sequential decision-making models offer a framework for dynamic diagnostic pathways.

Purpose of the Study:

• To develop and evaluate a machine learning-based expert system for optimizing sequential diagnostic test selection.
• To enhance diagnostic decision-making by balancing speed, cost, and accuracy.
• To reduce unnecessary testing and associated healthcare expenditures.

Main Methods:

• The proposed algorithm integrates lazy-learning classifiers, confident diagnosis, and locally sequential feature selection (LSFS).
• LSFS identifies optimal test sequences to reach a diagnostic threshold with adequate certainty.
• Speed-based and cost-based objective functions guide test selection.

Main Results:

• Consistent results across four datasets (heart disease, thyroid disease, diabetes, hepatitis) demonstrated significant reductions in tests and costs.
• Average cost savings ranged from 22% to 57%, and average test savings ranged from 24% to 73%.
• Diagnostic accuracies were comparable to or surpassed baseline methods using all available tests.

Conclusions:

• A novel approach dynamically estimates and determines the optimal sequence of diagnostic tests.
• The method maximizes information gain or disease probability based on available patient data.
• This optimizes diagnostic efficiency and resource utilization in clinical settings.