Risk Prediction Models Using Iterative Population Segregation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current prognostic models for predicting pregnancy disorders suffer from suboptimal performance, particularly in achieving high positive predictive value (PPV) and sensitivity (Sn), especially for low-prevalence diseases, and traditional AUROC analysis is inadequate for assessing their clinical utility.
Innovation Solution
An iterative population segregation methodology involving at least two segregation steps, using rule-in and rule-out probability-defined models to enrich sub-populations, enhancing PPV and NPV, thereby improving the accuracy of risk prediction models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional AUROC optimization is used to develop prognostic models, then the model maximizes area under ROC curve, but the model does not necessarily achieve high positive predictive value (PPV) or sensitivity (Sn) for low-prevalence diseases
Solution Approach 1:
The patent segments the population into multiple sub-populations through iterative segregation steps. Each step divides the current population into high-risk and low-risk sub-groups based on predictive markers, allowing the model to achieve high PPV and Sn in the final high-risk sub-population while maintaining overall model reliability
Solution Approach 2:
The patent transitions from two-dimensional AUROC optimization to a multi-dimensional approach involving iterative segregation steps that simultaneously optimize for PPV, Sn, and NPV. This dimensional expansion allows the model to achieve superior performance metrics that cannot be obtained through AUROC alone
2Measurement precision
If a single prognostic model is used to identify high-risk population, then the model provides a unified classification, but it cannot simultaneously achieve high PPV and high sensitivity for low-prevalence conditions
Solution Approach 1:
The patent employs iterative segregation that divides the population into multiple sub-populations through sequential steps. Each segregation step uses different criteria (PPV-defined or NPV-defined models) to create enriched sub-groups, enabling high detection rates in the final high-risk population while managing complexity through structured division
Solution Approach 2:
The patent uses a dynamic, iterative model structure where the classification criteria change across different segregation steps. The model adapts its parameters and thresholds at each step to optimize for different objectives (PPV in early steps, Sn in later steps), allowing simultaneous achievement of high detection rates and controlled complexity
3Adaptability or versatility
If prevalence-independent statistics (AUROC, Sn, Sp) are used to assess prognostic tests, then the statistics remain valid across different populations, but they fail to capture the clinical usefulness in specific clinical contexts with different prevalences
Solution Approach 1:
The patent changes the assessment parameters from prevalence-independent (AUROC) to prevalence-dependent (PPV, NPV, Sn, Sp) statistics. By adjusting the criteria and thresholds at each segregation step according to the specific clinical context and population prevalence, the model achieves accurate measurement of clinical usefulness while maintaining adaptability across different settings
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system and method of generating a model M that can detect or predict the predisposition to an outcome, for example a person's predisposition to developing a health condition such as preeclampsia, with a positive or/and negative predictive value better or equal to a predefined predictive value target, where the method factors in the impact of prevalence on test performance. The system and method employ an iterative population segregation methodology, involving at least two population segregations steps in which each step, independently, employs a probability-defined model. The first segregation step is selected from one of a rule-in probability defined model and a rule-out probability defined model, and the second segregation step is selected from the other of a rule-in probability defined model and a rule-out probability defined model. The present invention overcomes the technical limitations of existing predictive models and provides a solution to achieve a superior predictive model delivering an accurate risk prediction result, and in a less computationally intensive way. By isolating and segmenting particular population subsets according to the invention a much more robust and accurate way of detecting, or predict risk of, an outcome is achieved.