Evaluating clinical practice variation using a knowledge-based temporal sequence alignment framework
Abstract/Contents
- Abstract
- Entrusted with providing high quality and cost-effective care across the continuum of primary care to quaternary care medicine, health care institutions are turning to electronic medical records to keep pace with the information demands of medicine. The new patient care data collected within electronic medical records provides the computational foundation to build the rapid learning health care system, in which the delivery of health care within an entire institution improves dynamically by transforming the data into knowledge about which clinical practices are most effective. A crucial component to the rapid learning health care system is an understanding of clinical practice variations in medicine. Individual variations in care reflect decision choices of the treating clinician(s). Taken across an entire population, practice variations offer valuable insight on the behaviors and beliefs of an institution. Devising strategies and policies to improve the quality and efficiency of health care would not be possible without the knowledge that studying practice variations provide. Yet, existing methods for measuring clinical practice variations are not designed to handle temporal complexity. They focus on a small set of practices, of limited duration, and with limited scope. With the data that electronic medical records can provide, we have an opportunity to evaluate temporal complexity in medicine by studying patterns of care and entire treatment histories for a population of patients. In this thesis, I present a method, the T3S, for measuring the temporal sequence similarity between two patterns of care. The T3S advances research in temporal data mining by providing methodology that allows for the measurement of complex temporal features in clinical care. Specifically, the T3S measures the similarity of patterns in terms of the temporal ordering, duration, and overlap of its constituent treatments. I implement the T3S in three novel tools that allow population-level clinical practice variations to be studied from electronic medical records. To begin with, I use the T3S with expert derived domain knowledge to match medication treatment data from the medical record to chemotherapy plans so that patterns of care can be abstracted from granular medical data. This automated method for medical record abstraction of treatment information is a crucial first step before clinical data can be analyzed. Next, I use the T3S to find similar patterns of care from an electronic medical record to recommendations from a clinical practice guideline. The evaluation of individual patterns of care against evidence-based guidelines is an important task of health services related outcomes research. Finally, I incorporate the T3S into a new method for discovering patterns of care from a population of treatment histories. I show how this method can be used to summarize the clinical practice patterns within a population cohort and even discover anomalous practice patterns that may be of interest to clinicians and health services researchers. I evaluate each of these methods for its ability to provide clinically meaningful results from the available treatment data. Taken together, the T3S and the methods in which it is implemented offer a novel framework from which temporal complexities in the practice of medicine can be meaningfully explored. Finding and discovering similar patterns of care offers substantial potential in quality of care, outcomes, and comparative effectiveness research. As medicine marches to the digital age of data, measuring temporal similarity will assume a critical role in the development of new informatics methods to address the challenges of population science.
Description
| Type of resource | text |
|---|---|
| Form | electronic; electronic resource; remote |
| Extent | 1 online resource. |
| Publication date | 2012 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Associated with | Lee, Wei-Nchih |
|---|---|
| Associated with | Stanford University, Department of Biomedical Informatics |
| Primary advisor | Das, Amar K. (Amar Kumar) |
| Primary advisor | Musen, Mark A |
| Thesis advisor | Das, Amar K. (Amar Kumar) |
| Thesis advisor | Musen, Mark A |
| Thesis advisor | Goldstein, Mary |
| Advisor | Goldstein, Mary |
Subjects
| Genre | Theses |
|---|
Bibliographic information
| Statement of responsibility | Wei-Nchih Lee. |
|---|---|
| Note | Submitted to the Department of Biomedical Informatics. |
| Thesis | Ph.D. Stanford University 2012 |
| Location | https://purl.stanford.edu/qs342dv7909 |
Access conditions
- Copyright
- © 2012 by Wei-Nchih Lee
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...