| To Register SMDM Homepage | |
|
|
|
||||
Purpose: Reports of randomized clinical trials often use survival curves to summarize clinical outcomes over time and graphically demonstrate evidence of treatment effectiveness. Survival curves can also be used in patient communications to display how health risks accumulate over time. However, for survival graphs to be an effective communication method, people must recognize how long a time period is shown and correct their interpretations accordingly. But, do people actually think about the time axis labels when they view these graphs?
Methods: In a randomized survey experiment, administered online, we tested whether people viewing survival curves appropriately adjust their risk perceptions according to the time period shown. Internet users (N=864) were recruited from a demographically balanced U.S. panel. Participants read about a hypothetical disease and then viewed one of four survival graphs that displayed mortality risks with and without treatment. They were also informed that all risks were constant over time. Survival graphs showed either a visually large or visually small difference between treatments and were labeled to represent either 5 year or 15 year risk statistics. Participants then provided ratings of disease seriousness, as well as treatment effectiveness for each possible treatment.
Results: Variations in ratings corresponded more with the visual similarity of the graphics than with changes in the statistical risk exhibited, with participants perceiving greater disease seriousness and differences in treatment effectiveness in longer term (15 year) graphs. For example, when comparing two possible treatments, survey participants perceived a 1% difference in annual mortality risk to be much more significant when displayed over 15 years than when it was displayed for only 5 years (p<0.001). Conversely, perceptions of the disease were statistically indistinguishable in a pair of visually similar graphs even when the 5 year graph showed an untreated mortality risk almost three times larger (e.g., 20.9% vs. 7.5%) than the 15 year graph (p>0.14).
Conclusions: When people interpret survival curves, they do not pay adequate attention to the time frame represented in the curves and respond, instead, to the visual image presented in the graphic. This cognitive bias will impact the use of survival graphs as a patient communication tool and raises concerns about clinicians’ abilities to accurately interpret published research results.
See more of Oral Concurrent Session B - Methods in Judgment and Decision Making
See more of The 26th Annual Meeting of the Society for Medical Decision Making (October 17-20, 2004)