Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form


Mowery, Danielle (2014) DEVELOPING A CLINICAL LINGUISTIC FRAMEWORK FOR PROBLEM LIST GENERATION FROM CLINICAL TEXT. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

Primary Text

Download (852kB) | Preview


Regulatory institutions such as the Institute of Medicine and Joint Commission endorse problem lists as an effective method to facilitate transitions of care for patients. In practice, the problem list is a common model for documenting a care provider's medical reasoning with respect to a problem and its status during patient care. Although natural language processing (NLP) systems have been developed to support problem list generation, encoding many information layers - morphological, syntactic, semantic, discourse, and pragmatic - can
prove computationally expensive. The contribution of each information layer for accurate problem list generation has not been formally assessed. We would expect a problem list generator that relies on natural language processing would improve its performance with the addition of rich semantic features

We hypothesize that problem list generation can be approached as a two-step classification problem - problem mention status (Aim One) and patient problem status (Aim Two) classification. In Aim One, we will automatically classify the status of each problem mention using semantic features about problems described in the clinical narrative. In Aim Two, we will classify active patient problems from individual problem mentions and their statuses.

We believe our proposal is significant in two ways. First, our experiments will develop and evaluate semantic features, some commonly modeled and others not in the clinical text. The annotations we use will be made openly available to other NLP researchers to encourage future research on this task and other related problems including foundational NLP algorithms (assertion classification and coreference resolution) and applied clinical applications (patient timeline and record visualization). Second, by generating and evaluating existing
NLP systems, we are building an open-source problem list generator and demonstrating the performance for problem list generation using these features.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Mowery, Danielledlm31@pitt.eduDLM31
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairChapman,
Committee MemberSchleyer, Titusschleyer@regenstrief.orgTITUS
Committee MemberVisweswaran, Shyamshv3@pitt.eduSHV3
Committee MemberWiebe, Janycewiebe@cs.pitt.eduJMW106
Committee MemberMeystre,
Date: 8 August 2014
Date Type: Publication
Defense Date: 30 May 2014
Approval Date: 8 August 2014
Submission Date: 8 August 2014
Access Restriction: 1 year -- Restrict access to University of Pittsburgh for a period of 1 year.
Number of Pages: 121
Institution: University of Pittsburgh
Schools and Programs: School of Medicine > Biomedical Informatics
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: natural language processing, semantics, clinical texts, problem lists, informatics, medicine
Date Deposited: 08 Aug 2014 17:08
Last Modified: 19 Dec 2016 14:42


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item