Hospital Flowsheet Row Matching With Speech-to-Text and RAG
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manually determining relevant rows in hospital-specific flowsheets for patient data documentation is time-consuming and inefficient, leading to increased cognitive burden, delayed accessibility, and potential inaccuracies.
Innovation Solution
A system utilizing generative artificial intelligence (AI) with retrieval augmented generation (RAG) to automatically populate relevant rows in flowsheets by processing speech-to-text transcripts, segmenting meaningful information, and applying schema and example similarity-based searches to match key-value pairs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If nurses manually determine relevant rows and insert data in flowsheets, then data entry can be completed with existing systems, but nursing efficiency is reduced and cognitive burden increases
Solution Approach 1:
The system enables self-service by automatically populating flowsheet rows without requiring manual nurse intervention. The speech-to-text system captures patient information, and the AI model automatically matches and inserts data into relevant flowsheet rows, allowing the system to serve itself rather than requiring continuous manual input from nursing staff.
Solution Approach 2:
The patent replaces the mechanical manual data entry process with an automated speech-to-text and AI-based system. Instead of nurses manually typing or selecting data points, the system uses speech recognition and machine learning models to automatically capture and populate patient information, substituting human manual labor with automated technological processes.
2Loss of time
If nurses manually complete flowsheets, then data can be entered into the EHR system, but time consumption increases and data accessibility is delayed
Solution Approach 1:
The system performs preliminary action by capturing speech-to-text transcripts and pre-processing patient information before the actual flowsheet population is needed. The AI model pre-identifies relevant rows and prepares data mappings in advance, so that when data needs to be entered, the work is already substantially completed, reducing both time consumption and delays in data accessibility.
3Reliability
If manual data entry is used, then nurses can control the data entry process, but accuracy is reduced due to potential human errors
Solution Approach 1:
The patent replaces the mechanical process of manual data entry with automated speech-to-text recognition and AI-based data matching. This substitution eliminates human errors associated with manual typing and selection, while the AI model's systematic approach to matching speech transcripts with flowsheet rows ensures consistent and accurate data population without requiring continuous manual oversight.
Data Source
AI summary
A transcript including a plurality of portions is processed. A reduced schema set from a structured document is generated by identifying a subset of schema rows from the structured document using the transcript. A reduced example set is generated by identifying a subset of examples from the structured document using the transcript. The reduced schema set and the reduced example set are composed into a reduced composite schema set. An instance of information is extracted from the transcript by processing a prompt including the reduced composite schema set with a generative artificial intelligence (AI) model using retrieval augmented generation (RAG). A schema row of the structured document is populated with the instance of information from the transcript.