PHI-Segmented Translation Workflow for Discharge Instructions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Patients who do not speak English face difficulties understanding hospital discharge instructions due to the lack of translated content, and transmitting protected health information (PHI) across jurisdictions is restricted, complicating accurate and secure translation processes.
Innovation Solution
A system that parses and separates burdened content (PHI) from unburdened content, translates the unburdened content in a different jurisdiction, and recombines it with the burdened content for secure, near-real-time delivery, utilizing translation memory and AI for accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If PHI is transmitted to external computing systems for translation, then translation accuracy and speed are improved, but security and compliance with jurisdictional restrictions deteriorate
Solution Approach 1:
The system segments content into burdened content (containing PHI) and unburdened content (without PHI). Only unburdened content is transmitted to external computing systems for translation, while burdened content remains locally encrypted. This segmentation resolves the contradiction by enabling fast external translation of non-sensitive content while maintaining security compliance through local handling of sensitive content.
Solution Approach 2:
The system extracts PHI from content before transmission to external systems. By removing burdened content and retaining only unburdened content for external translation, the system achieves both fast translation processing and security compliance, as external systems receive no PHI that would violate jurisdictional restrictions.
2Reliability
If encrypted content is transmitted for translation, then security is improved, but translation accuracy and speed deteriorate
Solution Approach 1:
The system extracts and removes burdened content (PHI) before transmission to external systems. This extraction enables unburdened content to be transmitted in an unencrypted or lightly encrypted format to external translation systems, achieving both security compliance and fast translation processing without the overhead of encrypting/decrypting the entire content.
Solution Approach 2:
The system segments content into encrypted burdened content kept locally and unburdened content transmitted for translation. This segmentation allows the translation process to proceed rapidly on unburdened content while encrypted burdened content remains securely stored locally, eliminating the need to decrypt and re-encrypt large portions of content during translation.
3Measurement precision
If all content is translated through external systems, then translation accuracy is improved, but network dependency and processing time increase
Solution Approach 1:
The system segments content into burdened and unburdened portions, transmitting only unburdened content to external translation systems. This reduces network transmission time and external processing time while maintaining translation accuracy for the unburdened portion, as external systems specialize in high-quality translation processing.
Solution Approach 2:
The system performs preliminary segmentation and identification of burdened content before transmission. By pre-processing content to remove PHI and identify what needs external translation, the system minimizes network dependency and processing time while ensuring accurate translation of the remaining unburdened content through specialized external systems.
Data Source
AI summary
Embodiments of a system and method as disclosed herein translate content such as discharge instructions received in English, scan, parse, remove and store burdened content such as protected health information (PHI) from the discharge instructions, apply a translation memory to the unburdened content, optionally employ artificial intelligence (AI) translation of text not addressed by the translation memory, optionally employ post editing (PE) or other post-processing, import the translated text back into the translation memory for improved future processing, and reassemble and transmit the final translation product including burdened and unburdened content together, with translation of the unburdened content. In various embodiments, burdened content is stored in a first jurisdiction and unburdened content is translated and/or otherwise processed in a second jurisdiction different from the first jurisdiction.