Wellbeing Metric Computation via Unstructured Data Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methodologies for determining wellbeing indices require user input and are time-consuming, unable to utilize existing data, and are specific to particular questionnaires, making them inefficient and limited in their ability to predict answers and adapt to missing information.
Innovation Solution
A computer-implemented method that extracts information from unstructured data to categorize questions and predict answers using machine-learning neural networks, allowing for the computation of wellbeing metrics without user input and adapting to various questionnaires by extracting relevant data through natural language processing and machine learning techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If user input is required for determining wellbeing indices, then measurement precision can be maintained, but time consumption increases and productivity decreases
Solution Approach 1:
The system performs preliminary actions by automatically extracting and structuring relevant information from unstructured patient data before the questionnaire is even presented. This pre-processing of data includes identifying key health metrics, social determinants, and clinical indicators that will be needed for wellbeing assessment, thereby reducing the actual time users need to spend on questionnaire completion while maintaining measurement precision through comprehensive data collection
Solution Approach 2:
The system enables self-service by automatically generating wellbeing indices without requiring manual user input for data collection. The machine learning models autonomously extract information from electronic health records, perform sentiment analysis on patient notes, and compute wellbeing metrics independently, eliminating the time-consuming manual questionnaire completion process while preserving accurate measurements through automated information retrieval
2Measurement precision
If traditional questionnaire methods are used, then specific metric accuracy is maintained, but adaptability to different questionnaires and missing data handling deteriorates
Solution Approach 1:
The system achieves universality by developing a multi-functional machine learning framework that can handle multiple types of questionnaires and wellbeing metrics simultaneously. The same core architecture processes diverse data sources including electronic health records, social determinants of health, and various standardized questionnaire formats, enabling the system to adapt to different metric requirements while maintaining precision through specialized processing for each metric type
Solution Approach 2:
The system applies parameter changes by dynamically adjusting the weighting and prioritization of different data sources based on the specific wellbeing metric being calculated. For example, when calculating mental health metrics, the system increases the weight of sentiment analysis from clinical notes, while for physical health metrics, it prioritizes objective clinical measurements. This flexible parameter adjustment enables adaptation to different questionnaire requirements while maintaining measurement accuracy for each specific metric
3Measurement precision
If manual data extraction from unstructured data is performed, then information accuracy is maintained, but processing time and complexity increase
Solution Approach 1:
The system replaces manual mechanical data extraction processes with automated machine learning-based information extraction. Natural language processing models automatically parse unstructured clinical notes, social history data, and patient narratives to extract relevant wellbeing indicators, eliminating the need for manual review while maintaining high accuracy through trained algorithms that have learned to identify and extract pertinent information from diverse text sources
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer-implemented method comprising extracting information from unstructured data relating to a target patient; categorizing each of a plurality of questions as a numerical question requiring a numerical answer or a categorical question requiring a categorical answer, based on a set of question keywords and corresponding question keyword categorizations; for each question, when the extracted information includes patient information corresponding to an answer to the question, selecting that patient information as the answer, and when the extracted information does not include patient information corresponding to the answer, predicting the answer; and computing, using the selected answers and using rules, a health metric.