We conducted an internet-based cross-sectional survey of U.S. PCPs. The survey development and implementation process was described previously (Bleich et al., 2012
). Between February–March 2011, we surveyed 500 PCPs from The Epocrates Honors panel — an opt-in panel of 145,000 U.S. physicians. A total of 2010 invitations were sent to a random sample of PCP panel members, who received a $25 incentive to participate; 58 invitations were returned as undeliverable. The response rate, calculated as completed interviews over the total of working emails sent an invitation was 25.6%. The Johns Hopkins Bloomberg School of Public Health IRB approved this study.
We assessed physician beliefs about the causes of obesity with the question, “How important is each of the following possible causes of obesity for your patients?” For each cause (over consumption of food, restaurant/fast food eating, consumption of sugar-sweetened beverages (SSB), genetics/family history, and metabolic defect), physicians indicated whether it was very important, somewhat important, not very important, or not at all important. We dichotomized variables for each cause where 1 was “very important” and 0 was otherwise, which we selected based on cut-points in the data.
We assessed nutritional counseling habits using the question, “How frequently do you provide each of the following types of nutritional counseling to your obese patients?” For each nutritional recommendation (reading nutritional labels to determine calorie/nutrition content, avoiding high calorie ingredients when cooking, avoiding high calorie menu items when eating outside the home, reducing consumption of SSB, and reducing portion size), physicians indicated whether they provided it very frequently, somewhat frequently, not very frequently, or not at all frequently. Based on cut-points in the data, we dichotomized variables for each type of nutritional counseling where 1 was “very/somewhat frequently” and 0 was “not very/not at all frequently.”
All analyses were conducted with STATA 11.0 (College Station, TX). We used multivariate logistic regression to assess the association between physician beliefs about obesity’s causes and frequency of corresponding nutritional recommendations, adjusting for age, race, gender, body weight category, years since completing medical school (<20 years, ≥20 years), specialty (family practice, internal medicine), region (northeast, north central, south, west), and practice site (inpatient, outpatient, both inpatient and outpatient). We selected these covariates based on prior literature, and were included regardless of statistical significance (Bleich et al., 2012
). We used SVY to account for the complex survey design to address systematic under- or over-representation of the physician sub-populations, systematic non-response along known PCPs’ demographics, and sampling biases due to differences in non-response rates. The survey’s weighted margin of error was +/−5.3%.