Definition of Population and Sample

The National CLAS Physician Survey sample included 2,400 physicians. NAMCS and the National CLAS Physician Survey share the same sample frame and eligibility criteria. The basic sampling unit for the National CLAS Physician Survey was the physician. The sampling frame included nonfederally employed physicians classified by the American Medical Association or American Osteopathic Association as “office-based, patient care” and physicians classified as “hospital-employed” by the American Medical Association. Office-based, hospital-employed physicians work in practices that are owned by a hospital or hospital system. The physician universe excluded physicians from the specialties of anesthesiology, pathology, and radiology, and those who were federally employed or older than age 85 at the time of the survey.

Similar to NAMCS, a respondent was determined to be eligible (or in scope) if they provided direct care for patients in an office-based, nonhospital setting (Appendix I, questions 3 and 4) and confirmed via a checkbox that they were the physician to whom the questionnaire was addressed (Appendix I, after question 42). Nonhospital settings that were in scope for the National CLAS Physician Survey included single or group practices, freestanding clinics or urgent care centers, community health centers, mental health centers, nonfederal government clinics, family planning clinics, health maintenance organizations or other prepaid practices, and faculty practice plans.

A respondent was determined to be ineligible (or out of scope) if they did not provide direct care, was no longer in practice, worked only in ineligible setting(s), was deceased, or if their current contact information could not be located by the end of data collection. A physician was classified as “unknown” eligibility if, after all contact attempts, the physician was located but their eligibility status could not be determined.

Among eligible respondents, those who were defined as having a complete survey were those who provided a valid answer (a nonmissing response) for survey questions 2, 6, 7, 11–15, 18, and 35, and at least one valid answer for questions 30–33 (Appendix I). Among eligible respondents, those who were defined as having a partial survey were those who provided at least one missing response for survey questions 6, 7, 11–15, 18, or 35, but answered questions 1, 2, and 5 and at least one other question. Among eligible respondents, those who were defined as having a refused survey did not answer any questions on the survey except questions 1, 2, and 5.

Data Collection Procedures

The survey used mixed-mode data collection that included a self-administered web questionnaire (n = 48, 12.1%), self-administered mail paper questionnaire (n = 346, 87.1%), and computer-assisted telephone interview (n = 3, 0.8%) (15). The questionnaire included instructions that it should only be completed by the physician to whom it was addressed. The physician had to respond because the questionnaire asked about the physician’s training, knowledge, provision of services, and organizational environment. The paper version of the questionnaire used in the mail mode of data collection can be found in Appendix I. An electronic version of this questionnaire is available from the NCHS National Health Care Surveys website: https://archive.cdc.gov/#/details?url=https://www.cdc.gov/nchs/data/ahcd/2016_NAMCS_CLAS_Sample_Card.pdf.

Fielding of the National CLAS Physician Survey occurred from August to December 2016. The first contact with the sample physician was through an introductory letter from the NCHS director. This introductory letter invited physicians to participate in the survey via the web, informed them of the voluntary nature of the survey, and provided log-in instructions for the web version of the questionnaire. Three weeks after the introductory letter, all sampled physicians who had not responded to the web-based questionnaire received the first mailing. This mailing included a modified introductory letter, a paper questionnaire, and a self-addressed return envelope for the paper questionnaire. Then, about 41/2 weeks after the introductory letter, all sampled physicians received a postcard that thanked them for their participation or reminded them to participate.

Seven weeks after the introductory letter, the physicians who had not responded to the web-based questionnaire and had not returned the paper questionnaire received a second mailing. This mailing included a modified introductory letter, a paper questionnaire, and a self-addressed return envelope for the paper questionnaire. Eleven weeks after the introductory letter, physicians who had not yet responded received a third mailing. The third mailing included a modified introductory letter, a paper questionnaire, and a self-addressed return envelope. Finally, 14 weeks after the introductory letter, all remaining nonresponding physicians were contacted via telephone in a final attempt to obtain survey data. If the physician agreed to participate, the questionnaire was administered via computer-assisted telephone interview. If the physician declined to participate, the interviewer documented the physician as a refusal.

Physicians were asked several eligibility questions to assure that they met National CLAS Physician Survey eligibility criteria. Table D shows the final disposition of sampled physicians from the 2016 National CLAS Physician Survey (15). Of the 2,400 physicians, 697 did not meet all the criteria and were considered ineligible or out of scope (Table D, final disposition codes 3, 4 and 9) for the survey (15). The most frequent reasons for 358 physicians being ineligible or out of scope were that the physician was no longer in practice, did not see ambulatory patients, was retired, or was not working in an office-based setting (final disposition codes 3 and 9) (15). Another 339 physicians could not be located via a verified address or a verified telephone number so were considered ineligible or out of scope (final disposition code 4) (15). Eligibility status for 1,115 physicians was not determined; this was due to lack of resources (final disposition codes 5 and 7) (15).

Table D

Sample size and unweighted and weighted percentages for final disposition codes of sampled physicians: National CLAS Physician Survey, 2016

Of the 588 in-scope or eligible physicians (Table D, final disposition codes 1, 2, 6, and 8), 397 participated in the survey by completing one or more subject-matter item(s) on the questionnaire (final disposition codes 1 and 6) (15). Of the 397 physicians for whom questionnaires were received, 340 participated fully or adequately (completed questions required to satisfy the definition of a complete survey; final disposition code 1), and 57 participated partially (completed some questions but not enough to be considered complete; final disposition code 6) (15).

The overall unweighted response rate was 31.0% (33.8% weighted), based on the number of full responders. Full responders were eligible respondents who either fully or partially completed the survey based on nonmissing responses to predetermined items. The overall unweighted participation rate was 36.1% (39.3% weighted), based on the total of full and partial responders. The weighted response rate used the same computation procedures as used for the unweighted response rate, except that estimated physician counts replaced the unweighted sample counts in those computations. The physician-level response rate for the 2016 National CLAS Physician Survey sample was 33.8% (weighted). This was lower than the 46.0% weighted physician-level response rate for the core 2016 NAMCS (14).

Some factors that may have contributed to the lower response rate in the 2016 National CLAS Physician Survey compared with the 2016 NAMCS include differences in the fielding period, differing modes of data collection, and higher percentages of physicians with unknown eligibility. The 2016 NAMCS physician-level response rate was based on successful collection of data in the physician induction interview. The National CLAS Physician Survey fielding period was slightly delayed. As a result, the data collection schedule was compressed. The data collection ended during the winter holiday season when many physicians are unavailable. This negatively impacted participation rates. The 2016 NAMCS physician induction interview was conducted as a computer-assisted telephone interview or computer-assisted in-person interview. The 2016 National CLAS Physician Survey was conducted via self-administered web questionnaire, self-administered mail paper questionnaire, or computer-assisted telephone interview. Lastly, unlike the CLAS survey, eligibility was able to be assessed for all physicians sampled in the 2016 NAMCS physician induction interview.