This was a cross-sectional analytic study conducted from 12 November 2013 to 11 March 2014, a period of four months in the Buea Health District (BHD).

Buea is located at the eastern slope of the base of Mount Cameroon and is the capital of the South West Region. The BHD shares boundaries with Mount Cameroon to the west and north, Mutengene to the south, and Ekona to the east. Its population is estimated at 86,272 [21], and is made up of seven health areas including Buea road, Muea, Molyko, Bova, Buea Town, Bokwaongo and Tole.

Using the following formula: \( \boldsymbol{n}=\frac{\boldsymbol{p}\left(\mathbf{1}-\boldsymbol{p}\right){\boldsymbol{z}}^{\mathbf{2}}}{{\boldsymbol{d}}^{\mathbf{2}}} \)

We estimated a minimum acceptable sample size (n) of 174 participants for this study. The standard normal variate for significance (z) was 1.96 and margin of error (d) set at 5%, and the prevalence (p) of adherence rate was considered to be 13% according to a community-based study conducted by Ekram et al. [22].

Participants were recruited using a two-staged cluster randomised sampling technique. The preliminary stage involved random selection through balloting to select four out of the seven aforementioned health areas. The Tole, Muea, Bokwaongo and Bova health areas were selected. With the permission of the local leaders, the objective of our study was channeled to the community through the local radio stations, focal person for communication, churches and meeting houses. This was followed by an exhaustive invitation of consenting eligible participants in every household. For every household, we inquired about the hypertension status of all adults who were 21 years and older. Participants who attested to have been diagnosed with hypertension and placed on treatment were required to state the name(s) and physically present the drug(s) they were taking for hypertension. This information was then confirmed with the participants’ hospital records. Only one participant was required per household. In case there was more than one person in a household who were eligible for this study, balloting was done to select the final participant to be included.

Consenting participants at least twenty-one years old living with hypertension who were on hypertensive medication(s) for at least one month were recruited for this study.

We excluded pregnant women, individuals who declared being hypertensive but had no proof that they were on or had been prescribed drugs, those who had smoked, or consumed alcohol or other cardiostimulants 30 min prior to data collection, and individuals who could not express themselves in either English or French.

Using a predesigned questionnaire, data was collected in two steps. First, through a one-on-one interview, data on socio-demographic characteristics (age, gender, marital status, religion, educational status, occupational status), clinical parameters (BP, duration of treatment, number of tablets and number of doses per day), level of adherence using the Morisky 8-item validated questionnaire [23], and factors potentially associated with non-adherence (like forgetfulness, absence of symptoms, use of herbal medicine, smell and taste of drugs, lack of funds to purchase drugs, busy schedule, and adverse drug effects) were collected. Second, the BP was measured respecting standardized protocol. For every participant, two BP readings were recorded at least 10 min apart using an automatic machine (Medical Rossmax-Automatic Upper Arm Blood Pressure Monitor, Model AK150f, manufactured by Rossmax International Ltd., Botzstrasse 6,D-07743 Jena, Germany) approximated to the nearest 1 mmHg. The average of both readings was then computed and used for analysis.

A controlled BP was defined as a systolic BP and diastolic BP below 140 mmHg and 90 mmHg, respectively.

Adherence to antihypertensive pharmacotherapy was defined as a Morisky score of ≤2 while non-adherence was defined as a Morisky score above 2.

Occupational status was categorised as low (require no expert training or no technical expertise like farming), medium (a technical expertise is required but no expert training e.g. carpentry, commercial bike riding and commercial taxi, etc.) and high (for professionals and require advanced training e.g. health personnel, teachers, accountants etc.).

Ethical approval was granted by the Faculty of Health Sciences, Institutional Review Board (IRB) of the University of Buea (reference number: 2013/0106/UB/FHS/IRB) prior to conducting this study. In addition, administrative authorization was obtained from the South West Regional Delegation of Public Health before data collection.

Table 1 portrays the socio-demographic and clinical characteristics of our study population. Out of 183 participants, 118 (64.5%) were females and the mean age was 55.9 ± 12.3 years. The ages of the participants ranged from 21 to 90 years. One hundred and twenty-six (68.9%) of the participants were married. One hundred and sixty-three (89.1%) were Christians. Majority (n = 139, 79.9%) of the participants had only up to primary school level of education. Ninety-five (51.9%) were employed.

Out of 183 participants, 122 were non-adherent to their blood pressure lowering medication according to the Morisky medication adherence tool, giving a non-adherence prevalence of 66.7% (95%CI: 59.8–73.6%).

The following candidate variables were used in univariate analysis to test their association with non-adherence: multiple daily doses, multiple antihypertensive drugs, use of herbs, cost of treatment, absence of symptoms, forgetfulness, adverse drug effects, busy schedule. We found that forgetfulness (adjusted odds ratio (aOR) = 7.9; 95% confidence interval (CI): 3–20.8, p <  0.001), lack of finances (aOR = 2.8; CI: 1.1–6.9, p = 0.024), multiple daily doses (aOR =2.5; 1.2–5.6, p = 0.02) and drug side effects (aOR = 7.0 CI: 1.7–33.6, p = 0.007) were independent predictors of non-adherence after controlling for potential confounders in multivariate analysis. Our model, shown in Table 2, explained about 32.7% of the variability in the outcome variable (non-adherence). When adjusted for age, gender, number of antihypertensive medications and duration on medications, participants who were adherent to medication were more likely to have good BP control (OR = 13.82; 95% CI = 5.46–34.95; p <  0.001), Table 3.

Our study had a number of limitations. First, our cross-sectional design precluded assessment of temporality; rather we could only obtain associations. Second, only a single questionnaire was used to assess adherence in this study. The use of a second scale to measure medication adherence, such as the Medication Adherence Rating Scale would potentially have improved on the reliability of our findings. Third, pathologies such as chronic kidney disease, which can influence blood pressure control and indirectly adherence rates, were not excluded. Fourth, all seven health areas of this district could not be sampled due to financial restrictions, which limited a more complete picture on antihypertensive medication non-adherence in this health district. However, the authors believe that with a random selection of four out of the seven health areas gave each health area a fairly equal probability to be included. Fifth, participants on antihypertensive pharmacotherapy for just a year were recruited in this study. This could be a source of bias. Indeed, shorter treatment durations have been associated with higher non-adherence rates. In addition, predictors of medication non-adherence such as a poor patient-provider relationship was not captured in this study [11]. The Morisky medication adherence scale (MMAS) is largely based on self-report, making it liable to recall bias. Additionally, this can potentially lead to misclassification with individuals providing inaccurate responses. Despite these shortcomings, the MMAS has been widely validated in patients with hypertension [23] and other chronic diseases [52] with fairly good performance (sensitivity of 93% and specificity of 53%) in the clinic setting. It is thus the most widely accepted self-report tool for assessing medication adherence, especially to detect medication non-adherence with good blood pressure control data [23].