ML-Based Patient Risk Prioritization for Non-Emergent Care
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The prioritization of non-emergent medical procedures and visits is challenging due to resource constraints, such as shortages of skilled human resources and equipment, and existing methods fail to accurately assess patient risk variations, leading to inadequate care distribution during pandemics or routine operations.
Innovation Solution
A machine-learning based system that models patient risk trajectories over time to predict the impact of delayed non-emergent care, distinguishing patients who require immediate attention from those who can wait, using a comprehensive analysis of tens of thousands of patient attributes and observational data to generate personalized and accurate prioritization recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If non-emergent care is shut down entirely during pandemics to minimize exposure, then patient exposure risk is reduced, but patient health outcomes deteriorate due to delayed necessary care
Solution Approach 1:
The patent segments the homogeneous group of non-emergent patients into risk strata based on machine-learning predictions. High-risk patients are identified and prioritized for continued care, while low-risk patients are safely deferred. This segmentation resolves the contradiction by allowing differentiated care strategies rather than blanket shutdowns.
Solution Approach 2:
The system performs preliminary risk assessment and prioritization before resource allocation decisions are made. By predicting which patients are most likely to deteriorate if care is delayed, the system enables proactive identification of patients who should receive continued attention, preventing adverse outcomes before they occur.
2Ease of operation
If non-emergent care is prioritized for all patients during resource constraints, then equitable care distribution is achieved, but resource depletion occurs and essential emergency care is compromised
Solution Approach 1:
The patent applies local quality by assigning different priority levels to different patient groups based on their individual risk characteristics. Instead of uniform treatment, high-risk patients receive preferential access to limited resources while low-risk patients are deferred, optimizing resource allocation to where it is most needed.
Solution Approach 2:
The system changes the parameter of care priority from a static first-come-first-served metric to a dynamic risk-based metric. Machine-learning models continuously assess patient risk parameters and adjust prioritization accordingly, enabling flexible resource allocation that responds to changing patient needs and resource availability.
3Device complexity
If traditional first-come-first-served prioritization is used, then operational simplicity is maintained, but patient risk variations are not accounted for leading to inadequate care distribution
Solution Approach 1:
The patent replaces the mechanical first-come-first-served scheduling system with an intelligent machine-learning-based prioritization system. This substitution enables sophisticated risk assessment using electronic health record data, accurately identifying patients at highest risk of deterioration while automating the complexity through established AI methodologies.
Solution Approach 2:
The system enables self-service by automatically analyzing patient data and generating prioritization recommendations without requiring manual clinical assessment for each patient. The machine-learning model processes electronic health records and independently determines priority levels, reducing burden on healthcare providers while improving assessment consistency.
Data Source
AI summary
Embodiments of the present disclosure include systems and methods for generating a medical recommendation. Methods according to the present disclosure include receiving patient data associated with a first patient and receiving second patient data associated with a second patient. The methods further include inputting the first patient data and the second patient data into a trained machine-learning model to determine a first set of one or more risk values for the first patient and a second set of one or more risk values for the second patient. The methods further include comparing the first set of one or more risk values and the second set of one or more risk values to determine a priority for distributing care to the first patient and the second patient. In accordance with the determination that the first patient has priority, the system can generate a medical recommendation.