Clinical Note Generation Using Historical Note Augmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Clinical notes generated based on doctor-patient conversations often lack important clinical information present in historical patient records, necessitating a system to incorporate this missing information.
Innovation Solution
A machine learning model is trained using historical clinical notes to generate augmented clinical notes by identifying and removing overlapping data from current notes, incorporating missing historical information, and flagging modifications for user review.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If clinical notes are generated solely based on doctor-patient conversations, then the notes can be created quickly and efficiently, but they lack important clinical information that was recorded during previous interactions with the patient
Solution Approach 1:
The system performs preliminary action by proactively retrieving and analyzing historical clinical notes from the EHR before finalizing the current clinical note. This allows the system to identify and incorporate missing historical information in advance, ensuring completeness while maintaining efficient generation based on the current doctor-patient conversation.
Solution Approach 2:
The system introduces an intermediary mechanism (the clinical note augmentation system) that acts as a bridge between the current doctor-patient conversation and historical clinical records. This intermediary automatically retrieves, analyzes, and integrates relevant historical information, resolving the contradiction by enabling efficient current note generation while preventing information loss from historical records.
2Loss of information
If historical clinical notes are manually reviewed and integrated into current clinical notes, then complete clinical information is achieved, but the process becomes time-consuming and complex
Solution Approach 1:
The system implements self-service by automatically retrieving, analyzing, and integrating historical clinical information without requiring manual intervention. The system autonomously identifies relevant historical data, compares it with the current note, and performs the integration, thereby achieving complete clinical information while keeping the process simple and efficient.
Solution Approach 2:
The system replaces the manual mechanical process of reviewing and integrating historical notes with an automated computational system. This substitution eliminates the need for manual EHR searching and note synthesis, reducing process complexity while ensuring comprehensive information integration through algorithmic analysis and comparison.
3Loss of information
If all historical clinical notes are included in the current clinical note, then information completeness is maximized, but redundancy and duplication of information occur
Solution Approach 1:
The system applies the extraction principle by selectively removing redundant and duplicate information from historical clinical notes before integration. It identifies overlapping content between historical and current notes, extracts only the unique and relevant historical information, and incorporates it into the final clinical note, thereby maximizing information completeness while eliminating redundancy.
Solution Approach 2:
The system employs partial action by selectively integrating only the necessary portion of historical clinical information rather than including all historical data. It performs a targeted comparison to identify specific gaps in the current note and fills only those gaps with relevant historical information, avoiding excessive inclusion of redundant content.
Data Source
AI summary
A system includes a non-transitory storage having stored thereon instructions that when executed by a processor cause the processor to train a machine learning model to generate clinical notes by repeatedly: receiving a plurality of clinical notes associated with a patient; selecting a final clinical note; selecting a predecessor clinical note chronologically preceding the final clinical note; determining overlapping data between the final clinical note and the predecessor clinical note; performing an action on the identified overlapping data to generate a synthesized input clinical note; providing the synthesized input clinical note and the predecessor clinical note to the machine learning model; receiving a predicted final clinical note from the machine learning model; and updating the machine learning model based on the predicted final clinical note and the final clinical note.