Perceptions of Patients With Primary Nonadherence to Statin Medications

Derjung M. Tarn, MD, PhD; Maureen Barrientos, BA; Mark J. Pletcher, MD, MPH; Keith Cox, MA; Jon Turner, PhD; Alicia Fernandez, MD; Janice B. Schwartz, MD


J Am Board Fam Med. 2021;34(1):123-131. 

In This Article


Participant Identification and Recruitment

Participants were recruited from 1) lists of patients with primary nonadherence at the University of California, Los Angeles (UCLA), identified by querying electronic health records linked to Surescripts medication fill data;[20] 2) placing internet advertisements on in 22 United States metropolitan areas over a 6-month period; and 3) a large, Internet-based CVD cohort (the Health eHeart Study). Advertisements contained a link to a study information sheet and to an eligibility screening questionnaire; the study team contacted potentially eligible participants.

Eligibility criteria were: aged 18 years and older, received a new statin prescription within 2 years before contact, and did not start taking the prescription. We oversampled for minority patients. The UCLA Institutional Review Board approved the study protocol, and served as the Institutional Review Board of record for the University of California, San Francisco (UCSF). No written informed consent was required, but participants gave verbal consent before the focus groups.

Data Collection

Focus groups were conducted between February and July 2018 by 2 project coordinators with psychology backgrounds who were experienced interviewers. We chose interviewers without medical backgrounds to lead the focus groups so that patients would not feel inhibited sharing their honest opinions about their medical care. To minimize variation in technique, both interviewers were thoroughly oriented to the research problem, participated in developing the interview guide and potential probes, and engaged in debriefing sessions after each focus group. A physician-investigator with expertise conducting focus group interviews concerning patient medication use[21–24] was present for all but one focus group discussion, and worked with the interviewers to probe participant perspectives. To avoid influencing patient responses, this investigator was introduced only by first name, and patients were informed only that the investigator was a research team member. Focus groups lasted a mean of 80.2 (SD, 11.1) minutes and were audio recorded and transcribed verbatim.

Interviewers used a focus group discussion guide to lead conversations and probed participants as needed for detailed answers (Table 1). Interviewers met with 2 of the investigators (JBS [a cardiologist] and DMT [a family physician]) after each focus group to reflect on interview findings, put them into context of earlier focus group discussions, and to modify the focus group guide as needed to prompt greater depth of responses.[25,26] All participants were surveyed about their demographics; history of heart disease/heart attack, stroke, or diabetes; and number of prescription medications taken. Participants received a $60 gift card for participating.

Focus Group Analyses

Four investigators with different backgrounds (family physician with expertise in physician-patient communication, medical sociologist, cardiologist, and project coordinator with psychology background) formed the coding team. On completion of data collection, they independently reviewed a subset of 2 focus group transcripts, using inductive content analysis,[27,28] existing literature,[29] and clinical expertise to generate themes. The coders used open coding to identify comments in focus group discussions that related to patient decisions about not starting a prescribed statin. Themes were classified as "major themes" if they emerged in every focus group discussion, and as "minor themes" if they were raised in only a subset of discussions. Subthemes (specific themes within major and minor themes) also were identified. The coding team engaged in discussions about the themes, resolved disagreements via consensus, and generated a codebook describing the themes. After theme generation, 3 coders performed focused coding using the codebook, with at least 2 coders analyzing each transcript. Coding discrepancies were resolved through discussion. ATLAS.ti 8.0 (Scientific Software Development, GmbH, Berlin, Germany) was used for coding. Theoretical saturation, when no new themes can be generated from the data, was reached after 8 transcripts. This was assessed by using ATLAS.ti to track the codes applied to each transcript.