Advisory Committee Chair
James J Cimino
Advisory Committee Members
James H Willig
John Osborne
Seung Park
Michael Mugavero
Eddy Yang
Document Type
Dissertation
Date of Award
2019
Degree Name by School
Doctor of Philosophy (PhD) Heersink School of Medicine
Abstract
Since their introduction in the 1960s, electronic medical systems have brought with them tremendous opportunities and difficult challenges. In order to address patient safety, Clinical Decision Support (CDS) was added to the system, many times in the form of "pop-up" alerts. However, traditional alert typically do not incorporate enough patient-specific context resulting in inaccurate warnings. In response, clinicians override many of them. However, the high exposure to false positive alerts results in alert fatigue, a desensitization to future ones. This issue decreases patient safety by causing clinicians to ignore legitimate alerts. Despite some shortcomings, the data in modern Electronic Health Record (EHR)s might be able to provide additional context to alerts in order to decrease false positives. Therefore, the primary hypothesis investigated by this dissertation is that clinicians override alerts using an understanding of patient-context that traditional alerts do not use; however, the large amount of information stored in the EHR and other external systems can be used to understand the patterns of medical alert overrides. To this end, several studies were performed focusing first on the data contents itself, its interpretability, and finally its use for predicting and understanding medical overrides. The first study addressed the data in the EHR for its utility. This project surveyed the literature in order to categorize the EHR data that clinical researchers find inadequate for their studies or completely missing. The results showed that, while there are improvements to be made, the information is still valuable for secondary use. A second study looked at methods of improving the interpretability of the data for clinical researchers through use of infobuttons. These tools are effective at meeting the information needs of clinicians during patient care and show promise for doing the same for researchers. Finally, this dissertation used a metric of patient health stored in the EHR in order to predict the likelihood of a clinician overriding an alert on that patient. While the results indicated mild success, the current data in the EHR is insufficient to perform the task effectively. Taken together, this dissertation suggests that while EHR data has utility for secondary use, there are several shortcomings to be overcome in order to be used effectively for both research improved patient safety through CDS.
Recommended Citation
Kennell, Timothy, "The Electronic Health Record and the Clinical Informatics Researcher: A Journey to Predicting False Positive Alerts with Patient Characteristics" (2019). All ETDs from UAB. 2128.
https://digitalcommons.library.uab.edu/etd-collection/2128