Using Machine Learning to Improve Personalised Prediction: A Data-Driven Approach to Segment and Stratify Populations...

Area of Science:

• Health Informatics
• Data Science in Healthcare
• Clinical Decision Support

Background:

• Traditional Population Health Management (PHM) often uses subjective methods for patient segmentation and stratification.
• Existing approaches may lack the precision needed for effective, personalized healthcare interventions.
• The need for data-driven, objective methodologies in PHM is increasingly recognized.

Purpose of the Study:

• To develop and evaluate a novel 'cluster-then-predict' methodology for PHM.
• To improve the accuracy and personalization of population segmentation and stratification.
• To enhance the effectiveness of healthcare system decision-making.

Main Methods:

• Combined unsupervised clustering for population segmentation.
• Employed supervised classification, personalized to identified clusters, for stratification.
• Compared the 'cluster-then-predict' approach against an unlinked methodology.

Main Results:

• Observed a significant increase in cluster homogeneity.
• Demonstrated improvements in sensitivity and positive predictive value.
• The personalized, cluster-based approach outperformed unlinked methods.

Conclusions:

• The 'cluster-then-predict' methodology offers a more effective and personalized approach to PHM.
• This data-driven strategy has the potential to significantly improve healthcare system decision-making.
• Personalized stratification enhances the value and impact of health management programs.