Patient Beliefs Have a Greater Impact Than Barriers on Medication Adherence in a Community Health Center

Monica D. Gagnon, MPH; Eve Waltermaurer, PhD; Adam Martin, MPH; Colette Friedenson, MPH; Eric Gayle, MD; Diane L. Hauser, MPA


J Am Board Fam Med. 2017;30(3):331-336. 

In This Article


We developed a 31-question survey to capture participants' concerns about medicines and their perceptions of the necessity of the medicines. The survey incorporated 2 previously used instruments: the complete 8-item Morisky Medication Adherence Scale (MMAS-8),[16] and questions from the Beliefs about Medicine Questionnaire (BMQ). The MMAS-8 is used to measure patient self-reported adherence, or the extent to which the patient does not take medicines as prescribed. The BMQ is used to assess patients' beliefs about taking their medicines.[13,17,18] We limited BMQ items to 7 statements specifically pertaining to beliefs about medicines, and we changed the wording of some questions to make them more applicable to our study population. In addition to these 2 instruments, we asked questions about patient characteristics and questions about barriers to getting their medicine prescriptions filled. Barriers to getting medicines consisted of 7 factors including cost, access to refills, and perceived side effects.

The survey was administered to a convenience sample of patients coming to the health center for regularly scheduled appointments between September 2013 and May 2014. Patients over age 18 years were eligible. Study personnel obtained consent in private examination rooms before beginning the survey. Questions were read aloud by survey administrators, and answers were recorded, without patient identifiers, on iPads using iSURVEY software (Harvest Your Data, Wellington, New Zealand). The survey was available in both English and Spanish. The research protocol was approved by the institutional review board at the Institute for Family Health.


Sociodemographic factors including age, sex, race/ethnicity, education, and income, as well as chronic disease prevalence and number of medicines taken, were examined. Descriptive analyses included general univariate values (numbers and percentages). Corresponding with past research,[19–22] low adherence was defined as an MMAS-8 score <6. Specific positive BMQ measures and reported "barriers" were explored independently based on adherence level (low vs high), using the Mantel-Hanszel χ2 test of proportions.

For the predictive models, affirmative responses to questions regarding "beliefs" about medicine (adapted BMQ) were summed for a low value of 1 and an upper value of 7; these were further dichotomized to produce low beliefs (1–5) and high beliefs (6–7). "Barriers" were dichotomized as no reported barrier (49%) or ≥1 barrier (51%). Low adherence rates were examined by sociodemographics, chronic disease prevalence, and medicine utilization. Log-binomial predictive models were used to test the outcome of low adherence by any external barrier and by higher negative beliefs, adjusting for race, education, income, number of medicines, and disease burden to patients. Log-binomial regression was used as the preferred model when the prevalence of the outcome was not rare.[23] All analyses were conducted using SPSS version 22 (IBM, Chicago, IL).