Public Abstract
First Name:Andrew
Middle Name:Hargrove
Last Name:Hutson
Adviser's First Name:Suzanne
Adviser's Last Name:Boren
Co-Adviser's First Name:
Co-Adviser's Last Name:
Graduation Term:SP 2016
Department:Informatics
Degree:PhD
Title:Measuring the Speed and Efficacy of Clinical Decision-Making When Comparing Two Different Data
Visualizations for Medications
Background: The percentage of patients with polypharmacy needs is increasing among a growing patient
population. As a result, the amount of time health care professionals require to make clinical decisions
based on current and past medications is increasing. Health care professionals need methods for
increasing the speed of clinical decision making without sacrificing the quality of care. The goal of this
study is to demonstrate how modifying the data visualization for patient medication histories will change
decision making speed or efficacy.
Methods: We compared two groups across five randomized blocks. Group 1 responded to questions based
on the control data visualizations derived from an existing electronic health record. Group 2 responded to
questions based on the experimental data visualization based on a medication history developed by a team
led by Dr. Jeffrey Belden. All medical information presented to both groups is identical.
Each block represents a core clinical task associated with leveraging the medication history for a clinical
decision extrapolated from anecdotal scenarios in primary care. Block 1 asks the participant to identify
current prescriptions. Block 2 asks the participant to identify past prescriptions. Block 3 asks the participant
to identify the length of time a patient has been prescribed a specific drug. Block 4 asks the participant to
identify all new prescriptions in a given time interval. Block 5 asks the participant to identify a dosage
change for any prescription in a given time interval.
Each block holds two questions, identical in wording, differing only on the visualization presented to the
participant. The survey is configured to randomly present one question from each block to each participant.
Regardless of the question presented, we additionally track the response time for each block measured as
the last click on the survey page before the “submit” or “next” button is clicked. Participants are shown only
one question per page to increase the relevance of time tracking.
Results: Twenty-three participants enrolled in the study. A total of 112 observations were collected across
five randomized blocks. The average task time for control was 1366.3+/-10.35 and the average response
time for treatment 1773.23+/-10.4; however, the T-value was -1.313, thus the results were not statistically
significant. The average task correctness for control was 30.61% and the average task correctness for
treatment was 66.67% with a p-value of 0.000502.
Conclusions: Task correctness saw a significant increase in the probability for a correct response when
using the treatment visualization versus the control visualization. Additional research is required to
determine the effect of the treatment visualization on task time. The findings may have a significant impact
on how medication histories are presented to care provided through the electronic health record.