Risk Prediction System Using Ensemble ML and Social Determinants
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current healthcare models for predicting patient outcomes and interventions are often inaccurate and lack predictive power, especially for Medicaid populations, due to reliance on non-representative data and traditional electronic health records, which fail to account for social determinants of health and individual patient needs.
Innovation Solution
A system that combines external data sources, such as insurance claims and electronic health records, with patient engagement data using machine learning models to generate personalized risk predictions and intervention recommendations, incorporating social determinants and real-time feedback for continuous model improvement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional electronic health records and insurance claims data are used for risk prediction, then data availability is improved, but prediction accuracy deteriorates due to lack of social determinants and individual patient context
Solution Approach 1:
The patent combines multiple data sources including electronic health records, insurance claims, social determinants of health, and patient engagement data into a unified patient profile. This merging of diverse data types enables both comprehensive data availability and improved prediction accuracy by capturing the full spectrum of factors influencing patient outcomes.
Solution Approach 2:
The patent adds new dimensions to traditional risk prediction by incorporating social determinants of health (housing, employment, education) and patient engagement data alongside clinical information. This dimensional expansion transforms the prediction model from a narrow clinical focus to a holistic view of patient health, improving accuracy while maintaining data availability.
2Productivity
If aggregate population models are used, then computational efficiency is improved, but predictive power for individual patients deteriorates
Solution Approach 1:
The patent segments the population into individual patient profiles while maintaining efficient computational processing. By organizing data at the individual level and using targeted machine learning models for each patient, the system achieves both computational efficiency and high predictive power for individual patients, avoiding the need for computationally intensive aggregate modeling.
Solution Approach 2:
The patent applies local quality by customizing risk predictions and interventions for each individual patient based on their unique profile combining clinical, social, and engagement data. This localized approach maintains computational efficiency through targeted processing while significantly improving predictive power for individual patients compared to generic aggregate models.
3Measurement precision
If non-representative data subsets are used, then data quality is improved, but model representativeness deteriorates especially for Medicaid populations
Solution Approach 1:
The patent creates a universal data collection framework that serves multiple functions: it captures clinical data, social determinants, patient engagement information, and insurance claims for all patients including Medicaid populations. This multi-functional approach ensures both high data quality through standardized collection and broad representativeness by including diverse patient populations in the model.
Solution Approach 2:
The patent changes the parameters of data collection to include social determinants of health and patient engagement metrics alongside traditional clinical data. This parameter expansion enables the model to accurately represent Medicaid and underserved populations while maintaining data quality through systematic collection and validation of diverse data types.
4Measurement precision
If comprehensive data collection is implemented, then prediction accuracy is improved, but system complexity increases
Solution Approach 1:
The patent implements a nested structure where patient profiles contain nested data from multiple sources (EHR, claims, social determinants, engagement data) organized in a hierarchical manner. This nesting approach improves prediction accuracy by capturing comprehensive information while managing system complexity through structured data organization and modular architecture.
Solution Approach 2:
The patent introduces an intermediary data integration layer that mediates between multiple data sources and the prediction model. This intermediary layer standardizes and harmonizes data from diverse sources, improving prediction accuracy through comprehensive data collection while reducing system complexity by providing a unified interface for data access and processing.
Data Source
AI summary
Systems and methods for predicting changes in risk based on interventions are presented herein. In an example computer-implemented method, a computing device may receive, from a first source, first information and receive, from a second source, second information. The computing device may generate patient information by linking the first information and the second information using a linkage, corresponding to a patient. The computing device may generate using one or more trained machine learning models, a risk prediction for the patient and a change in risk prediction for the patient corresponding to an intervention. The computing device may output the risk prediction and the change in risk prediction for the patient.