Stacking Average Model for Non-Linear Healthcare Cost Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting healthcare costs are inaccurate due to the non-linear nature of health data, which is influenced by demographics, disease interactions, and lifestyle factors, leading to challenges in developing effective computerized models for cost prediction and treatment optimization.
Innovation Solution
A system and method that involves converting electronic health information into a tabular format, using semantic unification and hierarchical condition category mapping, and training a stacking average model with algorithms like Kernel Ridge Regression, Elastic Net, and Gradient Boosting to identify non-linearity, enabling the creation of a predictive data set for prioritizing patient treatment paths.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If simple linear regression models are used to predict healthcare costs, then the model complexity is low, but the prediction accuracy is poor due to non-linear data trends
Solution Approach 1:
The patent transforms the non-linear healthcare cost data through mathematical transformations (log transformation, Box-Cox transformation) to convert it into linear relationships that can be effectively modeled by linear regression algorithms, thereby maintaining model simplicity while improving prediction accuracy
Solution Approach 2:
The patent introduces intermediate variables and features (demographic factors, disease burden indices, lifestyle parameters) as mediators between the input data and cost predictions, allowing linear models to capture complex non-linear relationships through these intermediate representations
2Device complexity
If parsimonious models using only demographics information are used, then the model simplicity is high, but the prediction accuracy is very low
Solution Approach 1:
The patent merges multiple data sources including demographics information, claims data, and lifestyle factors into a unified predictive model, combining the strengths of each data type to achieve high prediction accuracy while maintaining reasonable model simplicity through integrated analysis
3Reliability
If models using claims data are used, then the disease burden is captured, but the demographics information is not taken into account
Solution Approach 1:
The patent creates a universal predictive model that serves multiple functions simultaneously: it captures disease burden from claims data, incorporates demographics information, and accounts for lifestyle factors, making the model adaptable to various prediction scenarios without requiring separate models for each data type
4Measurement precision
If models taking demographics and claims data together are used, then the prediction accuracy is good, but episodic events such as operations or readmissions are not captured
Solution Approach 1:
The patent implements a dynamic modeling approach that adapts to different types of healthcare events: it uses claims data for chronic disease burden, incorporates demographics for baseline risk, and specifically captures episodic events like operations and readmissions through targeted feature engineering and time-dependent analysis, making the model versatile across different event types
Data Source
AI summary
A system and method for training a computerized data model for the algorithmic detection of non-linearity in a data set includes providing two master data sets corresponding to two discrete time periods, respectively, and a third data set for a third discrete time period. The two master data sets are mapped to at least one code model. A stacking average model is trained with the at least two master data sets corresponding to two discrete time periods by using a stacked regression algorithm. A box-cox transformation function is applied to the models to provide a predicted value for the third data set of the third discrete time period. An ensemble is created using the predicted value for the third data set and the first, second, and third models of the trained stacking average model to identify a non-linearity in the third data set.