Monday, October 19, 2015: 4:45 PM
Grand Ballroom A (Hyatt Regency St. Louis at the Arch)

Aaron M. Scherer, PhD1, Holly O. Witteman, PhD2, Angela Fagerlin, PhD3, Predrag Klasnja, PhD1, Beth A. Tarini, MS, MD1, Nicole L. Exe, MPH1, Knoll Larkin, MPH1 and Brian J. Zikmund-Fisher, PhD1, (1)University of Michigan, Ann Arbor, MI, (2)Université Laval, Quebec City, QC, Canada, (3)VA Ann Arbor Healthcare System & University of Michigan, Ann Arbor, MI
Purpose: With the advent of online patient portals to electronic health record systems, patients are increasingly receiving their laboratory test results directly. Other research from our lab highlights that presenting test results with a visual line graph reduces confusion about the results compared to the tables typically used on online portals. However, it is unclear what specific line graph features might improve understanding of test results.

Methods: Participants (N=4,833) completed an online survey experiment in which they were presented with a hypothetical scenario where they received hemoglobin A1c test results from a blood draw done between clinical visits for management of Type 2 diabetes. Test results were randomly presented in one of three line graph formats: (1) solid, single-color that only showed the standard range, (2) color gradient utilizing a stoplight color theme (green to red) to indicate risk categories or (3) solid-color blocks that indicated discrete risk categories utilizing the stoplight color theme. The A1c test value was also randomized, being within the standard range (5.4%) or one of three higher levels (6.4%, 7.1%, or 8.4%). Our primary outcome measures were 1) how good or bad the participant thought their test result was for their health, with “don’t know” as a response option and 2) graph preferences. Individual difference measures included subjective and objective numeracy, and graphical literacy.

Results: Controlling for numeracy and graphical literacy, both A1c test value and graph format had significant effects on risk perceptions (p’s<.05). More importantly, we observed a significant A1c test value by format interaction (p<.001): Participants were more sensitive to changes in A1c values when they viewed the stoplight-colored block design with discrete risk categories than when they saw either of the other line graph formats. Respondents also significantly preferred the stoplight-colored block design (p’s<.03). The A1c test value was not a significant predictor of “don’t know” responses (p=.88), but the stoplight-colored gradient design did result in higher rates of “don’t know” responses than either the solid single-color and stoplight-colored block designs (p=.01).

Conclusion: Presenting laboratory test results using a line graph that utilizes stoplight-colored blocks that clearly demarcate discrete risk categories decreases confusion, increases sensitivity to differences in test values, and is more preferred to line graphs that utilize a solid, single-color or a stoplight-colored gradient.