Enabling Medical Experts to navigate clinical text for cohort identification (meTAKES)

Enabling Medical Experts to navigate clinical text

for cohort identification (meTAKES)Stephen Wu, Mayo Clinic

SHARPn Summit 2012June 12, 2012

Outline

• Motivation• Methods (current)

• System architecture• Data retrieval• Search• Cohort management

• Conclusion & Future Work

Motivation

• Clinical NLP out-of-the-box• Comprehensive knowledge• Customize? Collaborate!

• Diverse requirements• Physician/Researcher tasks

• Enroll patients in study• Define retrospective cohort• Case abstraction

Somali patients (unique terms)

Drug-induced liver injury (rel’ns)

Pediatric asthma (temporal)

“Medical expert”-driven NLP

• Use case-specific• Streamlined• On-the-fly NLP Diverse requirements

• Interactive interface• Delivery mechanism• Available data vs. expert knowledge

• Use case-agnostic• Comprehensive• Pre-computed NLP Known requirements

text semantics expert criteriasource

Client Server

• Web interface (GWT)

GUI

Lucene

ranked records

EHRdata pool parameters

records

NLP (MedTagger)

query parser

query

cohort manipulation

records

cohort mgmt

Data retrieval

• Parameters (current)• Patient ID• Date

• Sources (current)• Enterprise Data Trust (EDT)

@ Mayo Clinic• Text files on server

Search

• Parameters (current)• Term lists• Logic• Expansion

• Techniques (current)• Dictionary (Lucene)• NLP results

(e.g., negation)

Cohort Management

• Parameters:• Cohort name/tag• Selecting patients

• Export• Iterative refinement

Conclusion and Future Work

• NLP / search• Text characteristics• Semantic search• Relationships

• HCI / cohort management• Learning• Collaboration

• Interoperability• Structured data• API

• Mayo delivery: DDQB Clinical Notes

Search Tool

Evaluation framework

THANK YOU.

https://sites.google.com/site/stephentzeinnwu

[email protected]

meTAKES team: Stephen Wu Ravikumar K.E. Hongfang Liu

Special thanks to: Siddhartha Jonnalagadda James Masanz Vinod Kaggal Sean Murphy Tom Suther Erik Voldal Carlos Garcia Melissa Gregg

This work was supported in part by the SHARPn (Strategic Health IT Advanced Research Projects) Area 4: Secondary Use of EHR Data Cooperative Agreement from the HHS Office of the National Coordinator, Washington, DC. DHHS 90TR000201.