Predictive Model Integrating EHR and Patient-Reported Outcomes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning algorithms for predicting mortality in cancer patients rely heavily on structured electronic health record (EHR) data, which provides limited insight into patient symptoms and functional status, resulting in suboptimal performance with true positive rates under 50%, and the role of patient-reported outcomes (PROs) in risk stratification remains unexplored.
Innovation Solution
Developing a predictive model that integrates both EHR data and patient-reported outcome (PRO) data using a two-phase methodology, including a preliminary LASSO model and logistic regression with an offset term, to improve the accuracy of mortality risk prediction for cancer patients.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning algorithms rely heavily on structured EHR data for mortality prediction, then the prediction system can be implemented with available data infrastructure, but the predictive performance remains suboptimal with true positive rates under 50%
Solution Approach 1:
The patent combines structured EHR data with unstructured PRO data from patient interviews into a unified predictive model. This merging of data sources integrates objective clinical measurements with subjective patient-reported symptoms and functional status, resolving the contradiction by maintaining implementation feasibility while significantly improving predictive performance through complementary information sources.
Solution Approach 2:
The predictive model functions as a composite system that processes two distinct data types: structured EHR data and unstructured PRO data. Each data source contributes unique information properties, with EHR providing objective clinical metrics and PROs providing subjective patient experience data. This composite approach enables the system to overcome the limitations of either data source alone, achieving superior predictive accuracy while maintaining practical implementability.
2Measurement precision
If patient-reported outcome data is collected and integrated into the predictive model, then the accuracy of mortality risk prediction improves, but the complexity of data collection and processing increases
Solution Approach 1:
The patent introduces natural language processing as an intermediary layer that automatically transforms unstructured PRO data from patient interviews into structured features suitable for predictive modeling. This intermediary component handles the complexity of unstructured data processing, allowing the system to leverage rich patient-reported information while shielding the rest of the system from the computational complexity of natural language analysis.
Solution Approach 2:
The system replaces manual data collection and processing methods with automated electronic data capture and computational processing. Instead of relying on manual chart review or structured questionnaires that require extensive processing, the system uses electronic interview transcripts and automated natural language processing to extract relevant features, significantly reducing the operational complexity while improving measurement precision.
Data Source
AI summary
Methods, systems, and computer readable media for predicting patient outcomes. In some examples, a method includes training, using at least one processor, a predictive model by fitting a first model using patient outcome data for a number of individuals and electronic health record data for individuals. Training the predictive model includes fitting a second model using patient reported outcome data for a subset of the plurality of individuals. The method includes supplying patient data for a patient to the predictive model and using the predictive model to predict at least one patient outcome for the patient.