Explainable Disease Severity Prediction from Real-World Clinical Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing disease severity scores are inconsistently calculated in real-world settings, limiting their availability and hindering the understanding of disease impact in routine clinical practice.

Innovation Solution

A computer-implemented method using a generalized additive model, specifically an explainable boosting machine, processes input data to predict disease severity scores based on physiological measurements, enabling the generation of reliable disease severity scores for inflammatory conditions like rheumatoid arthritis, atopic dermatitis, and ulcerative colitis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If disease severity scores are calculated using standard methods (DAS-28, EASI, Mayo score), then the scores are reliable and valid for research, but they are not consistently calculated by physicians in real-world settings, leading to limited availability

Engineering Contradiction:
Improvedisease severity score reliabilityVSAvoiddisease severity information availability
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

The patent creates a computational copy of the disease severity scoring system that can be applied to real-world data. Instead of relying on physicians to manually calculate scores like DAS-28, EASI, or Mayo scores, the system uses a generalized additive model that replicates the scoring logic and applies it automatically to electronic health record data, thereby preserving reliability while improving availability.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent replaces the mechanical/manual calculation process with an automated computational model. The generalized additive model processes electronic health record data automatically, substituting the manual physician assessment mechanism with an algorithmic system that can process data at scale without human intervention.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Loss of information

If generalized additive model is used to predict disease severity, then disease severity scores can be generated for all patients in real-world data, but the model complexity increases

Engineering Contradiction:
Improvedisease severity information availabilityVSAvoidpredictive model complexity
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent develops a universal predictive model framework that can be applied across multiple disease types and settings. The generalized additive model is designed to handle various input feature types (demographic, clinical, laboratory data) and can be adapted to different diseases by changing the input features and target variable, thereby reducing the need for disease-specific models.

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

Solution Approach 2:

The patent uses parameter estimation techniques to simplify the model. The generalized additive model estimates parameters from training data and uses these parameters to make predictions on new data. This parameter-based approach reduces model complexity by replacing complex non-linear relationships with learned parameter transformations.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4593035A1Predicting disease severity
Publication Date: 2025.07.30 SANOFI SA(FR)
  • EP4593035A1 patent drawingFigure 1
  • EP4593035A1 patent drawingFigure 2~3
  • EP4593035A1 patent drawingFigure 4~5

AI summary

A computer-implemented method of predicting disease severity in a subject comprises: receiving, at a generalized additive model, input data corresponding to a set of one or more features for the subject, and generating, as an output of the generalized additive model, a predicted disease severity score for the subject, wherein generating the predicted disease severity score for the subject comprises processing the received input data using the generalized additive model.