Prediction of patient evolution in terms of Clinical Risk Groups form routinely collected data using machine learning

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Predictive Analytics

Background:

• Chronic diseases significantly impact patient quality of life and global healthcare expenditures.
• Proactive patient and healthcare system management is crucial for mitigating the effects of chronic conditions.
• Predictive modeling offers a pathway to enhance proactive healthcare strategies.

Purpose of the Study:

• To develop a supervised machine learning model for predicting the healthcare status of chronic patients.
• To utilize Clinical Risk Groups (CRGs) as a measure of disease chronicity.
• To assess the predictive power of diagnosis and prescription data for patient health status.

Main Methods:

• Employed supervised learning techniques to construct the predictive model.
• Utilized patient diagnosis and prescription data as input features.
• Explored various data encoding strategies suitable for machine learning applications.
• Applied the model to the entire healthcare system population, irrespective of specific diseases.

Main Results:

• The developed model demonstrated accurate predictions using only diagnosis and prescription data.
• Prescription information exhibited a higher predictive value compared to diagnosis data within the study's dataset.
• The findings indicate the feasibility of building robust predictive models from routinely collected healthcare data.

Conclusions:

• Routinely available clinical data, particularly prescription information, can effectively predict chronic patient healthcare status.
• Machine learning models can be instrumental in proactive healthcare management for chronic conditions.
• This approach supports a population-wide strategy for managing chronic disease, enhancing healthcare system efficiency.