Editor's Note: Survey methodology for the professional survey was updated Jan. 5 to provide additional detail into how the survey was done.
The survey, Patient Prejudice: When Credentials Aren’t Enough, was conducted by WebMD/Medscape in collaboration with STAT, a national publication produced by Boston Globe Media that focuses on health, medicine and scientific discovery.
The survey of professionals presents findings from more than 1,000 health care professionals, including doctors, registered nurses, nurse practitioners, and physician assistants. A survey of 1,000 patients asked about their biases toward doctors and other health care professionals.
See professional survey results here.
See patient survey results here.
Professional Survey Methods
Survey invitations were emailed to a random sample of 84,000 Medscape member healthcare providers from July 17 through August 22, 2017. Respondents were invited to participate in an online study of “Healthcare Provider-Patient Interactions.” The sampling frame (84,000 healthcare providers) was established from the Medscape membership by key demographic characteristics. African-American/Black and Hispanic Medscape member physicians were quota sampled to support subgroup analysis; they were then weighted to reflect the proportions of each in the Medscape member population.
Respondents were required to be practicing U.S. physicians, registered nurses, nurse practitioners, or physician assistants. The total sample size was 1,186 healthcare providers, comprising 822 physicians, 100 registered nurses, 160 nurse practitioners, and 104 physician assistants. The survey completion rate was 7% based on the 11,558 people who received and opened the email invitation, excluding the supplemental minority sample. The margin of sampling error for this survey among physicians is +/-3.5% (n=822); among registered nurses, +/- 10.0% (n=100); among nurse practitioners, +/- 8.0% (n=160); and among physician assistants, +/-10.0% (n=104), at the 95% confidence level for a statistic at 50%. Statistical difference from 50% is the most conservative in measuring sample error. Also note that Medscape members who completed the survey were entered into a sweepstakes that awarded 25 random winners a $100 Amazon gift card.
Other factors may also contribute to error in survey estimates, including measurement error, construct and content validity, question wording and question order, etc. The survey designers worked to ensure that the questionnaire was worded clearly, questions were ordered properly, and that appropriate scales and response categories were used. In addition, during the field period, the survey was checked to make certain that respondents were answering all questions logically.
The final data were cleaned to ensure data quality. A comparison of the sample to the sampling frame for validity on three key physician characteristics—age, sex, and specialty—yielded close results that did not require weighting on those characteristics.
The key measure in the study for perceived bias experiences used a five-year timeframe. The wording was “In the past five years, has a patient ever made an offensive remark in your presence about your ‘gender, sexual orientation, race, ethnicity/national origin, medical education from outside the US, religion, age, weight, accent, disability, smoking, political views, other (specify), none of the above.’”
Patient Survey Methods
WebMD’s survey was completed by 1,019 respondents from July 13 to July 17, 2017 using the NORC AmeriSpeak® Panel, which is a probability-based panel designed to be representative of the US household population. The margin of error for a statistic at 50% is +/- 4.06% at the 95% confidence level, for the entire sample of 1,019 respondents. Statistics for subgroups of the sample have larger margins of error, as do statistics greater or less than 50%. Most questions in the survey were completed by a subgroup of 947 (934 weighted) respondents who had visited a healthcare provider in the past 5 years.
Funded and operated by NORC at the University of Chicago, AmeriSpeak is a probability-based panel designed to be representative of the US household population. Randomly selected US households are sampled with a known, non-zero probability of selection from the NORC National Sample Frame, and then contacted by US mail, telephone, and field interviewers (face to face). AmeriSpeak panelists participate in NORC studies or studies conducted by NORC on behalf of governmental agencies, academic researchers, and media and commercial organizations. NORC at the University of Chicago is an independent research institution.