This study uses a mixed-methods design. A randomized intervention (RCT) will be a mailed letter sent to approximately 1310 patients, disclosing their physicians' financial relationships with industry before patients meet with their physicians. There are 4 experimental conditions, systematically varying the wording of the COI disclosures, and one control condition. These patients will be briefly surveyed after they see their physicians, using standardized measures of trust and other questions. The goal is to achieve a 65% response rate (final n=850).
The sample size was determined based on a 0.3 unit change in Mayer Trust Scale adapted to refer to "physician" rather than "management" between the subgroups using the POWER procedure in SAS software version 9.3. In its original presentation, the Mayer Trust Scale lists a standard deviation for the 4-item trust score as 0.66 - 0.68. A sample size of 850 patients (170 in each of the 5 groups) would provide 82% power to detect a 0.3 unit difference in Mayer Trust Scale after Bonferroni correction for the 10 possible comparisons, assuming a standard deviation of 0.7 and that the significance level of 0.005 (0.05 / 10 comparisons) would be used in these analyses. In addition, this sample size would allow for multivariable regression models sufficient parameters to be fit following the suggested ratio of 10 patients for each predictor in the model.
Additionally, to supplement patient-derived RCT survey results, approximately 25 Cleveland Clinic physicians with industry relationships over $20,000 will be interviewed to gather their perspectives on COI disclosure. These interviews will occur before and after the RCT, and will collect qualitative and quantitative data. Qualitative data will be analyzed using content analysis, an iterative process of data immersion and data coding, whereby narrative data is categorized into discrete codes based on thematic content and then used to compute descriptive statistics.