Predictive Analytics for Prescription Drug Plans Using Synthetic Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for forecasting the profitability of prescription drug plans rely on actuarial techniques that aggregate historical data, failing to consider individual plan member factors such as enrollment and dis-enrollment, which leads to inaccurate predictions.
Innovation Solution
The system uses virtual or simulated data to model potential changes in prescription drug plans, including enrollment and dis-enrollment of members, to provide more accurate predictions of future events and plan performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If aggregate historical data methods are used, then calculation simplicity is improved, but prediction accuracy deteriorates
Solution Approach 1:
The patent segments the aggregate historical data into individual plan member level data. Instead of treating all members as a homogeneous group, the system divides the population into discrete individuals and models each member's enrollment, dis-enrollment, and drug utilization behaviors separately. This segmentation allows the prediction system to capture individual variations and dynamics that are lost in aggregate calculations, thereby improving prediction accuracy while maintaining computational feasibility through efficient data structures and algorithms.
2Measurement precision
If individual plan member factors are considered, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing key metrics and patterns from historical individual member data before the actual prediction process. The system pre-processes enrollment histories, drug utilization patterns, and demographic characteristics into structured formats that can be quickly queried during predictions. This preliminary preparation reduces the computational burden during runtime, allowing the system to consider detailed individual factors without proportionally increasing computational complexity during the prediction phase.
Solution Approach 2:
The system changes parameters by transforming detailed individual member data into aggregated statistical measures that capture essential dynamics without requiring processing of every individual record during predictions. The patent uses parameters such as enrollment rates, dis-enrollment rates, and drug utilization metrics derived from individual-level analysis, allowing the system to maintain prediction accuracy while reducing computational complexity through parameter aggregation and optimization.
3Measurement precision
If virtual or simulated data is used, then forecasting accuracy is improved, but data processing requirements increase
Solution Approach 1:
The patent uses copying by creating virtual or simulated representations of plan members and their behaviors based on historical patterns. Instead of processing additional real-world data, the system generates synthetic copies of enrollment scenarios, dis-enrollment events, and drug utilization patterns that mirror real-world dynamics. These virtual data copies allow the system to test multiple forecasting scenarios and improve accuracy without proportionally increasing the quantity of actual data that must be collected, stored, and processed.
Data Source
AI summary
A system includes a processor and memory providing instructions for execution by the processor. The instructions include, responsive to a request from a user device indicating a group type and a time frame, selecting a set of users based on the group type. The instructions include retrieving the user data and historical data from a database and identifying a set of new users during the time frame. The instructions include adding the set of new users to the set of users and generating synthetic historical data for the set of users. The instructions include, based on the synthetic historical data for each user, computing a likelihood of discontinuation of a drugs and computing a likelihood of initiation of a drug. The instructions include estimating, for the group type, an aggregated output based on the likelihoods of initiation and discontinuation. The instructions include displaying the aggregated output on the user device.