Opioid Education and Prescribing Practices

Sample

The online national HCP survey was conducted with 2000 HCPs representing primary care physicians (PCPs), specialists (SPs), physician assistants (PAs), and nurse practitioners (NPs). PCPs were from family practice, general practice, internal medicine specialties, and SPs from cardiology, dermatology, endocrinology, neurology, obstetrics/gynecology, oncology, ophthalmology, psychiatry, rheumatology, and urology. NPs and PAs included a mix of primary care and specialists. The sample was randomly drawn from WebMD's Medscape subscriber network, stratified by HCP group, and designed to yield target numbers of completed interviews per group. Eligible HCPs practiced in an office-based setting and spent at least 50% of their time providing patient care. Physicians were required to be American Medical Association-validated physicians. NPs and PAs were required to have prescribing authority in their state. Honorariums of $50 and $60 were provided for survey completion. Institutional review boards of each author's institution approved the study.

Measures

Survey measures were developed using a multi-stage process, including an expert review, cognitive interviews, and a pretest. The survey assessed knowledge, attitudes, and practices in multiple domains, including prescription drug promotions, clinical trials, FDA policies, and opioids. Data presented in this article focuses on opioid questions. Key measures included the number of patients prescribed opioids for chronic pain, exposure to opioid educational information, recall of opioid prescription drug promotions, and knowledge of ADFs.

Data Collection

The survey was administered between October and November 2019. Participants received an initial e-mail about the survey with a link to the screener. Eligible HCPs were sent the survey within 24 hours of completing the screener. Response rates were calculated based on the American Association of Public Opinion Research¹¹ standards. Seventy-five percent of the initial e-mails were never opened by the provider, resulting in a low overall response rate, ranging between 3–6% across the 4 HCP groups. However, the extended interview response rates (completed interviews/eligible screeners) were over 90% across all groups.

Nonresponse bias analysis was conducted to compare respondents and nonrespondents on known characteristics for the screener and survey. A high cutoff (P < .20) was used to test for significant differences for response rate bias considerations. At this level, differences were found in primary specialty, geographic region, race, and age groups. These differences were adjusted for by raking, a proportional fitting technique for calibrating sample weights using external data for population estimates. Raking was conducted to national totals from 2 separate data sources, the National Ambulatory Medical Care Survey (NAMCS) restricted data (https://www.cdc.gov/nchs/data/ahcd/Availability_of_NAMCS_and_NHAMCS_Restricted_Data.pdf) and the American Community Survey (ACS) public use microdata sample (https://www.census.gov/programs-surveys/acs/microdata.html). Both data sources provide national estimates on the 4 HCP groups. For the survey, weights were computed for each completed interview to correct unequal selection probabilities, differential response rates, and coverage. The weights were used to ensure distributions by HCP characteristics matched national benchmarks for PCPs, SPs, NPs, and PAs. Weights were developed for gender, age groups, and the 4 U.S. Census Regions based on external totals from national surveys (NAMCS and ACS). In addition, external totals were used to develop weights for specialty and practice sizes for PCPs and SPs.

Analyses

Analyses were conducted using SPSS Version 25 complex samples module and utilized Taylor Series variance computations to account for weighting and sample design correctly. χ² and Benjamini-Hochberg (BH) procedures were used to examine differences across HCP groups. Logistic regression was conducted to examine the likelihood of prescribing opioids to fewer patients at 3 months relative to 12 months ago. Predictors included exposure to opioid educational information during the last 12 months, the number of patients prescribed opioids for chronic pain during the last 3 months, recall of opioid prescription drug promotion, ADF knowledge score, HCP group, provider demographics, and practice type.