Fall Risk Prediction Tool Using ML for Post-Acute Care
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional technologies for predicting patient falls in post-acute care settings are inaccurate and inefficient, leading to high false positives and alert fatigue, as they were developed for acute care settings and do not effectively utilize relevant data available early in a patient's stay.
Innovation Solution
A computerized decision support tool using machine learning models that incorporate clinical and non-clinical patient data, including demographics, medications, and natural language processing techniques, to predict falls early in a post-acute care patient's encounter, reducing false positives and improving resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional fall prediction technologies (Morse Fall Risk Scale) are used in post-acute care settings, then the system can provide fall risk assessment, but the accuracy is poor with area under curve of 0.5572, leading to high false positives
Solution Approach 1:
The patent changes the parameters used for fall risk assessment by transitioning from conventional acute care-based metrics (Morse Fall Risk Scale) to post-acute care-specific parameters including fall risk assessments, functional mobility assessments, cognitive functional assessments, and medication data tailored to rehabilitation settings. This parameter transformation enables accurate prediction in post-acute care while reducing false positives.
Solution Approach 2:
The system dynamically updates fall risk predictions by continuously incorporating new patient data including real-time fall risk assessments, functional mobility changes, cognitive functional status, and medication adjustments throughout the post-acute care stay. This dynamic approach allows the system to adapt to changing patient conditions and provide accurate predictions at different stages of recovery.
2Loss of time
If conventional technologies are used, then the system can operate with simple structure, but it fails to utilize data available earlier in patient's stay, causing 15.5% of falls to be missed
Solution Approach 1:
The system performs preliminary fall risk assessment early in the patient's post-acute care stay by incorporating initial fall risk assessments, functional mobility assessments, and cognitive functional assessments conducted during admission. This preliminary action identifies high-risk patients before falls occur, enabling early intervention and preventing 15.5% of falls that would otherwise be missed.
Solution Approach 2:
The system maintains continuous monitoring and assessment throughout the patient's stay, continuously updating fall risk predictions by incorporating ongoing fall risk assessments, functional mobility changes, cognitive functional status, and medication adjustments. This continuous action ensures no fall is missed regardless of when it occurs during the extended post-acute care period.
3Ease of operation
If conventional technologies are used, then the system can be easier to operate, but it causes alert fatigue and inefficient use of resources due to high false positives
Solution Approach 1:
By changing the assessment parameters to post-acute care-specific metrics including fall risk assessments, functional mobility assessments, cognitive functional assessments, and medication data, the system reduces false positives while maintaining ease of operation through automated data collection and processing of these standardized parameters.
Solution Approach 2:
The system incorporates feedback mechanisms that continuously monitor patient outcomes and adjust predictions based on actual fall events and non-fall incidents. This feedback loop refines the algorithm to reduce false positives over time, improving resource allocation efficiency while maintaining operational simplicity through automated learning and adaptation.
Data Source
AI summary
Systems, methods, and media are provided for predicting fall risk for a post-acute care patient. Patient data for the post-acute care patient is received. Features from the patient data are extracted. The features include one or more polypharmacy features. Other features may include lab or vital features. Based on the features extracted from the patient data, a prediction of the post-acute care patient suffering a fall within the future is generated using one or more machine learning models. The one or more machine learning models may be trained using one or more of binary features, continuous features, categorical features, free-text features, or a combination thereof. An action is initiated based on the prediction of the post-acute care patient suffering a fall. The action is associated with reducing the risk of the patient fall, such adjusting the lighting in the patient's room, for example.