| To Register SMDM Homepage | |
|
|
|
||||
Purpose. Little work has examined whether questionnaire formatting or sender affects physicians' likelihood to respond to surveys. Furthermore, the extent to which physician surveys are filled out completely and carefully has received little study. We conducted a factorial manipulation of printing format and sender, and examined their effects on physician survey response and completion.
Methods. We asked members of the Canadian Association of Emergency Physicians (CAEP) to complete a 4-page mail survey. The survey was carried out according to standard survey methodology (Dillman, 2000). The 2 x 2 factorial design varied Print Format (single vs. double sided printing) and Sender (Known [IS, well known to this audience] vs. Unknown [JB, unknown to the audience]) as the manipulated variables. Outcomes included response rate and survey completion, i.e. the number of items skipped without any discernible reason. The sample frame included 400 active CAEP members; those listed as non-MD's, retired, and not currently residing in Canada were excluded.
Results. Of our initial sample of 400 names, 376 were eligible potential respondents with correct addresses. Overall response rate was 261/376, or 69.4%. Tests for response bias showed no effect of gender on likelihood to respond χ2(1)=2.38, p = 0.12, but a marginal effect of location χ2(1) =3.48, p = 0.06; physicians who came from Ontario were marginally more likely to respond than those coming from other provinces. A 2 x 2 factorial ANOVA with Print Format and Sender predicting response rate showed no significant effect of Print Format (65.8% for double-sided vs. 73.0% for single sided; F(1,372) = 2.39, p = 0.12), no significant effect of Sender (72.5% Known vs. 66.3% Unknown; F(1,372)=1.76, p=0.19), and no interaction F(1,372) = 1.35, p = 0.25. Chi-square tests showed that Print Format did not predict the likelihood of leaving more fields blank χ2(1) = 0.70, p = 0.40, but that fewer fields were left blank when the Sender was Known χ2(1 )= 5.30, p = 0.02.
Conclusions. Print Format and Sender did not affect response rate significantly, although effect sizes of 7.2% and 6.2% might prove important (and statistically significant) with larger sample sizes. The extent to which a questionnaire is fully completed can be a useful variable in survey methodology research.
See more of Poster Session - CEA: Methods and Applications; Health Services Research
See more of The 26th Annual Meeting of the Society for Medical Decision Making (October 17-20, 2004)