EHR-Based Machine Learning Pipeline for Refractory Epilepsy Prediction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods fail to accurately predict epilepsy refractoriness in patients due to the lack of seizure frequency data in available sources and incompatibility with electronic medical record systems, making it difficult to implement predictive models for identifying potential refractory patients.

Innovation Solution

A machine learning pipeline is developed to predict epilepsy refractoriness using electronic health records data, constructing patient cohorts, selecting predictive features, and training models to classify patients as refractory or non-refractory, with a computer platform that interfaces with EMR systems to generate predictions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If seizure frequency data is used to predict refractoriness, then prediction accuracy is improved, but data availability deteriorates due to lack of seizure monitoring in available sources

Engineering Contradiction:
Improveprediction accuracyVSAvoiddata availability
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent uses Electronic Health Records (EHR) as an intermediary data source to predict refractoriness. Instead of directly measuring seizure frequency which is unavailable, the system uses EHR data (demographics, comorbidities, treatment history) as a mediator to infer refractory status. This intermediary approach allows prediction without direct access to seizure monitoring data.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent creates a predictive model that copies the refractoriness prediction function from traditional seizure-frequency-based methods to EHR-based methods. The model replicates the ability to identify refractory patients using alternative data sources (EHR) that are already available in clinical settings, eliminating the need for separate seizure monitoring infrastructure.

Inventive Principle:
Principle #26Copying

2Reliability

If a predictive model is developed using specialized seizure data, then prediction reliability is improved, but system compatibility deteriorates due to incompatibility with EMR systems

Engineering Contradiction:
Improveprediction reliabilityVSAvoidsystem compatibility
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent makes the predictive model universal by designing it to work with standard EHR systems that are already widely deployed in healthcare. The model uses common EHR data elements (demographics, comorbidities, treatment history) rather than specialized seizure data, allowing it to be integrated into existing EMR infrastructure without requiring separate specialized systems. This multi-functionality enables the model to operate across different healthcare settings using standard technology.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Reliability

If early identification of refractory patients is achieved, then patient management quality is improved, but implementation difficulty increases due to lack of implementable predictive tools

Engineering Contradiction:
Improvepatient management qualityVSAvoidimplementation difficulty
Core Design Contradiction:
ReliabilityVSEase of manufacture

Solution Approach 1:

The patent enables EHR systems to automatically perform refractoriness prediction using their existing data infrastructure. The model is integrated into the EHR workflow, allowing the system to self-identify refractory patients without requiring additional manual data collection or separate monitoring systems. The EHR system serves itself by utilizing its own stored data to generate predictions, eliminating the need for external specialized tools.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3571700B1Method and system for predicting refractory epilepsy status
Publication Date: 2023.08.02 UCB BIOPHARMA SPRL
  • EP3571700B1 patent drawingFigure 1
  • EP3571700B1 patent drawingFigure 2
  • EP3571700B1 patent drawingFigure 3a~3e

AI summary

A method of building a machine learning pipeline for predicting refractoriness of epilepsy patients is provided. The method includes providing electronic health records data; constructing a patient cohort from the electronic health records data by selecting patients based on failure of at least one anti-epilepsy drug; constructing a set features found in or derived from the electronic health records data; electronically processing the patient cohort to identify a subset of the features that are predictive for refractoriness for inclusion in a predictive model configured for classifying patients as refractory or non- refractory; and training the predictive computerized model to classify the patients having at least one anti-epilepsy drug failure based on likelihood of becoming refractory.