Abstract
The level and predictors of medication adherence have not been given attention among populations for whom the cost of treatment is not a concern, especially in the Ghanaian healthcare setting. This study compared generalised linear models to determine the factors associated with adherence to treatment among hypertensive patients with a workplace policy that offers cost-free access in Ghana. We conducted a cross-sectional descriptive study to investigate the level of adherence and factors that influence practice among hypertensive patients with cost-free access to treatment. A total of 254 respondents were conveniently sampled and administered a questionnaire. The predictors of treatment adherence were assessed using Bivariate and generalised linear models (logistic and complementary log–log regression models). The complementary log–log regression model outperformed the logistic regression model in fitting the relationships in the data, by reporting lower AIC and BIC, and higher Nagelkerke and Cox and Snell pseudo-R-squared values. Most (88.9%) respondents scored low adherence, and only 11.1% adhered to treatment. The factors that affect adherence to hypertension medication included age (ARR = 8.58, 95%CI 1.23–83.28, p-value = 0.031), sex (ARR = 0.19, 95%CI 0.03–0.91, p-value = 0.042), location (ARR = 10.78, 95%CI 2.01–79.76, p-value = 0.007), busy schedule (ARR = 0.03, 95%CI 0.00–0.30, p-value = 0.016), time spent (ARR = 0.05, 95%CI 0.00–0.28, p-value = 0.002) and accessibility (ARR = 17.89, 95%CI 4.03–112.38, p-value < 0.001). The high level of non-adherence to treatment among patients, even when the cost of treatment is not a barrier, could contribute to overall treatment outcomes. Despite cost-free access, adherence remains low, highlighting the need for further investigation into non-cost barriers affecting treatment compliance. In addition, there is a need to design and implement measures to improve adherence among patients with hypertension in the country.
Keywords: Non-adherence, Hypertension, Medication, Complementary log–log model, Generalised linear models
Subject terms: Health services, Public health
Introduction
Hypertension has become an important public health problem, especially in sub-Saharan Africa, because of its increasing prevalence and impact on patients and communities1–3. The prevalence of hypertension doubled from 331 million women and 317 million men in 1990 to 626 million women and 652 million men in 20194. Similarly, the prevalence of adult hypertension in Ghana is on the ascendency, ranging from 19 to 48%5. This phenomenon is a cause for concern due to the devastating impact of hypertension on individuals and society2. Hypertension significantly increases the risks of heart attack, stroke, kidney failure, and blindness, and has remained the leading cause of untimely deaths globally1,6. The burden of hypertension in Ghana is projected to increase due to ageing, rapid urbanisation, and unhealthy lifestyles5. However, a significant number of hypertensive clients in Ghana are unaware of their condition7. The current situation could hamper Ghana’s attainment of Sustainable Development Goal 3 target of reducing by one-third premature deaths from non-communicable diseases by the year 20308. Globally, less than 20% of hypertensive patients have their condition under control, with even lower control rates in LMICs1. Less than 20% of people with hypertension in Ghana have their blood pressures well controlled5,7. Non-adherence to treatment is known to contribute to poor blood pressure control9. Adherence refers to the extent to which patients comply with recommendations for prescribed treatments10. A failure of patients to comply with recommended treatment regimens for the management of the condition could thus be described as non-adherence.
Treatment adherence is complex, and several factors may be associated with it. The factors contributing to medication adherence in patients include disease conditions (including the level of blood pressure), medication, healthcare providers, access and availability of medical services, the patient’s perceived effects of the disease, and socioeconomic factors11–14. Also, satisfactory patient-provider relationships, accessible healthcare facilities, presence of specialist clinics, use of technology, affordable cost, trust, fear of complications, and health insurance may influence patients’ adherence to treatment12–14. On the contrary, younger-aged and single patients, unemployment, lack of awareness about the disease, low perception of illness, and unreliable medication and clinic schedules are negatively associated with adherence12,15. The non-affordability of healthcare costs or otherwise has been found to either facilitate or impede adherence to treatment among Ghanaian patients16,17. Highlighting the factors that influence medication adherence among populations that do not have to worry about the cost of treatment is imperative to highlight other factors associated with medication adherence. However, there is a dearth of information on non-adherence to treatment in situations where the patient’s employer bears the cost of treatment. The purpose of this study was to assess the level of treatment adherence and the factors associated with the practice among hypertensive patients with a workplace healthcare policy that grants them cost-free access to treatment. Due to the low probability of success for the dependent variable, generalized linear models, such as logistic and complementary log–log regression models, were compared in fitting the models for predicting treatment adherence.
Materials and methods
Design
This was a facility-based cross-sectional descriptive study that used quantitative methods to assess the level of adherence and factors that influence adherence to treatment among hypertensive patients with cost-free access to treatment. The cross-sectional descriptive design allowed the researchers to collect data at one point in time, and no follow-up was required18.
Setting and population
The study was conducted at the Volta River Authority (VRA) hospital at Akosombo, in the Eastern Region of Ghana. The hospital provides both inpatient and outpatient services as well as specialist clinics like the hypertension clinic. The facility serves staff, dependents of the company, as well as the general public. Eligibility criteria for the study were VRA staff or dependents of staff who are clinically diagnosed as hypertensive for at least six months and on antihypertensive medication. In addition, the cost of the antihypertensive medications was to be borne by the organisation. All participants who met the inclusion criteria were approached to participate in the study. All eligible participants consented to participate in the study.
Sampling
The sample size was calculated using Yamane’s19 formula for determining the sample size of finite populations: n = N/(1 + N(e)2). Where n = sample size, N = population size (450), and e = level of precision (5%). A 10% non-response rate yielded a minimum sample size of 234. However, 253 respondents took part in the study.
A convenience sampling technique was applied in recruiting hypertensive patients for this study. Patients who met the inclusion criteria were recruited based on their availability at the time of data collection18. A total of 253 completed questionnaires were retrieved from 260 questionnaires that were administered, giving a response rate of 97.3%.
Data collection and management
A self-administered pretested questionnaire was used to collect data from November to December 2019. The questionnaire was pretested within the same facility in September 2019. Consequently, participants who completed the pretest were excluded from the final data collection. The questionnaire was divided into three sections. The first section covered the respondents’ socio-demographic characteristics. This includes the residence of patients (rural and urban); Age (20–29, 30–39, 40–49, 50–59 and 60 and above); Sex (female and male); marital status (married, single and divorced/widowed); highest level of education (tertiary and below tertiary); religions (Christianity and non-Christian); duration of diagnosis of hypertension (less than 1 year, 1–5 years and > 5 years) and the number of hypertensive medications participant is on (1, 2 and more than 2). The second section of the questionnaire collected information on respondents’ level of adherence to hypertensive medication using adapted scales20. The items include: I sometimes forget to take my medication; over the past weeks, I did not take my medication on some days due to other reasons than forgetfulness; sometimes when I feel worse, I stopped taking the medication without informing my doctor; sometimes I forget to take my medication with me when I leave home; I took all my medications yesterday; when I feel better, sometimes I stop taking my medication; sticking to my medication daily is very inconveniencing to me; and I often have difficulty remembering to take my medication.
Participants who responded accurately to more than 6 out of 8 items (> 75%) adhered to treatment, while those who responded accurately to at most 6 items did not adhere to medication. The final section assessed the factors that predicted medication adherence, adapted from previous studies21–24.
Two research assistants helped in the data collection process. Participants were given ample time to complete the questionnaire and return it to the hypertension clinic or the research assistants. All completed questionnaires for each day were collected by research assistants and submitted to the principal investigator (the corresponding author) for vetting for completeness and stored safely under lock and key.
Statistical analysis
Data were coded, entered, and analysed using SPSS (version 22) and the R (R version 4.5.0 -2025-04-11 ucrt; https://mirror.metanet.ch/cran/bin/windows/base/R-4.5.0rc-win.exe ) statistical package. SPSS was used for the descriptive and bivariate analyses, while R was used for modelling the relationship between the variables. Descriptive statistics were used to summarise the respondents’ characteristics. The adapted adherence scale reported a Cronbach’s alpha of 0.611. For the factors influencing medication adherence, in addition to the descriptive statistics, a logistic regression model (LRM) and complementary log–log model (CLLM) were used. LRM performs well when it is applied to modelling dependent variables that are dichotomous, such as the presence or absence of the event in the dependent variable, with similar probabilities. However, when the probability of success is very low or high for the dependent variable, the CLLM is usually useful. In this study, the probability of adherence was low (11.1%); hence, the LRM was compared with the CLLM in fitting the relationship between adherence and predictors.
Both models belong to the generalized linear model family, although the logistic regression model is robust in modelling dichotomous outcomes, the existence of an asymmetric probability distribution prohibits its performance. Other possible models that could be used include Bayesian logistic regression, generalized additive, and mixed-effects logistic regression. However, exact information on priors regarding the topic from similar populations was unavailable, so the Bayesian logistic regression model was not used. In addition, the linearity of the relationships was assumed, and there were no nested or clustering variables; therefore, the study could not use generalised additive or mixed-effects logistic regression models.
The adjusted odds ratios (AOR) and risk ratios (ARR) were respectively calculated for LRM and CLLM. The models were assessed using the Akaike information criterion (AIC), Bayesian information criterion (BIC), and pseudo r-squared, such as Nagelkerke, Cox and Snell. The smaller the values of AIC and BIC, the better the model, while the larger the values of pseudo r-squared, the better the model fits the data.
Ethical considerations
The study was granted ethical approval by the Research Ethics Committee of the University of Health and Allied Sciences (protocol identification number: UHAS-REC A.1[50]19–20). The study was conducted following the guidelines of the institutional ethics review committee and the Declaration of Helsinki. Written permission was sought from the management of the facility before data collection. Written informed consent was obtained from each respondent using a Participant’s Informed Consent Form, which explained the objectives of the study. Also, participants who reported having low adherence were referred to the health facility (VRA hospital) for counselling. After two days, the principal investigator (first author) followed up to ensure that the participants visited the health facility and had initiated treatment. Participation in this study was voluntary, and respondents were not under any obligation to respond to questions or participate in the study. Opting out did not affect the treatment received at the clinic. Also, the participants were offered no compensation for participating in the study.
Results
Table 1 shows the demographic characteristics of the respondents. A total of 253 completed questionnaires were retrieved from respondents. The majority (n = 157; 62.1%), had attained tertiary education (n = 211; 83.4%), Christians (n = 147; 58.1%). Again, the majority of respondents who had been diagnosed with hypertension within the last five years (n = 198; 78.2%) had Polypharmacy—patients’ multiple drug use (n = 175; 69.2%). Most (88.9%) respondents scored low adherence, and only 11.1% adhered to treatment (Fig. 1).
Table 1.
Demographic factors associated with treatment adherence.
| Characteristics | Level of Adherence | Total | p-value | COR (95% CI) | |
|---|---|---|---|---|---|
| Adherence | No adherence | ||||
| Age of respondent | |||||
| 20 to 29 | 11(17.5) | 52(82.5) | 63(24.9) | 0.369 | 1.75(0.51,5.94) |
| 30 to 39 | 2(7.1) | 26(92.9) | 28(11.1) | 0.692 | 0.64(0.11,3.74) |
| 40 to 49 | 4(6.3) | 59(93.7) | 63(24.9) | 0.463 | 0.56(0.13,2.38) |
| 50 to 59 | 7(11.3) | 55(88.7) | 62(24.5) | 1.000 | 1.05(0.29,3.86) |
| 60 and above | 4(10.8) | 33(89.2) | 37(14.6) | ||
| Sex | |||||
| Female | 15(15.6) | 81(84.4) | 96(37.9) | 0.071 | 2.05(0.93,4.52) |
| Male | 13(8.3) | 144(91.7) | 157(62.1) | ||
| Marital status | |||||
| Married | 12(10.7) | 100(89.3) | 112(44.3) | ||
| Single | 14(14.1) | 85(85.9) | 99(39.1) | 0.450 | 1.37(0.6,3.13) |
| Divorced/widowed | 2(4.8) | 40(95.2) | 42(16.6) | 0.353 | 0.42(0.09,1.95) |
| Residence | |||||
| Rural | 7(7.1) | 91(92.9) | 98(38.7) | 0.114 | 0.49(0.2,1.2) |
| Urban | 21(13.5) | 134(86.5) | 155(61.3) | ||
| Religion | |||||
| Christianity | 18(12.2) | 129(87.8) | 147(58.1) | 0.482 | 1.34(0.59,3.03) |
| Non-Christian | 10(9.4) | 96(90.6) | 106(41.9) | ||
| Duration of diagnosis | |||||
| Less than 1 year | 12(12.6) | 83(87.4) | 95(37.5) | 0.306 | 1.84(0.56,6.02) |
| 1 to 5 years | 12(11.7) | 91(88.3) | 103(40.7) | 0.385 | 1.68(0.52,5.49) |
| > 5 years | 4(7.3) | 51(92.7) | 55(21.7) | ||
| Number of hypertension medications | |||||
| One | 11(14.1) | 67(85.9) | 78(30.8) | 0.597 | 1.31(0.48,3.61) |
| Two | 10(8.9) | 102(91.1) | 112(44.3) | 0.640 | 0.78(0.28,2.17) |
| More than two | 7(11.1) | 56(88.9) | 63(24.9) | ||
| Education | |||||
| Tertiary | 28(13.3) | 183(86.7) | 211(83.4) | 0.012* | |
| Below tertiary | 0(0) | 42(100) | 42(16.6) | ||
Significant at P < 0.05, *odds ratios with confidence intervals were not computed for cells with zero (0) counts.
Fig. 1.
Level of adherence to medication.
Factors associated with treatment adherence
Aside from the level of education (p-value = 0.012), which was associated with treatment adherence, other demographics did not show any significant associations with adherence to treatment in the bivariate analysis (Table 1).
Non-demographic factors associated with adherence to treatment in the bivariate analysis include getting embarrassed with taking the medication (p-value = 0.004), busy schedule (p-value < 0.001), accessibility (p-value = 0.025), much time spent at the health facility (p-value < 0.001), feeling dizziness (p-value = 0.017), medication side effects (p-value < 0.001), feeling of light-headedness (p-value = 0.012) and erectile dysfunction (p-value = 0.006) (Table 2).
Table 2.
Non-demographic factors associated with adherence to treatment.
| Variables | Adherence | Total | p-value | COR (95% CI) | |
|---|---|---|---|---|---|
| Adherence | Non-adherence | ||||
| Embarrassed taking medication | |||||
| Agree | 3(3.3) | 87(96.7) | 90(35.6) | 0.004 | 0.19(0.06,0.65) |
| Disagree | 25(15.3) | 138(84.7) | 163(64.4) | ||
| Medication decreases sexual interest | |||||
| Agree | 10(8.6) | 106(91.4) | 116(45.8) | 0.254 | 0.62(0.28,1.41) |
| Disagree | 18(13.1) | 119(86.9) | 137(54.2) | ||
| Busy schedule | |||||
| Agree | 1(1.1) | 93(98.9) | 94(37.2) | < 0.001 | 0.05(0.01,0.39) |
| Disagree | 27(17) | 132(83) | 159(62.8) | ||
| Cultural beliefs | |||||
| Agree | 1(3.1) | 31(96.9) | 32(12.6) | 0.146 | 0.23(0.03,1.77) |
| Disagree | 27(12.2) | 194(87.8) | 221(87.4) | ||
| Distance to the hospital | |||||
| Agree | 8(12.7) | 55(87.3) | 63(24.9) | 0.634 | 1.24(0.52,2.96) |
| Disagree | 20(10.5) | 170(89.5) | 190(75.1) | ||
| Short supply of medications | |||||
| Agree | 12(9.7) | 112(90.3) | 124(49) | 0.490 | 0.76(0.34,1.67) |
| Disagree | 16(12.4) | 113(87.6) | 129(51) | ||
| The attitude of the hospital staff discourages me | |||||
| Agree | 6(6.9) | 81(93.1) | 87(34.4) | 0.126 | 0.49(0.19,1.25) |
| Disagree | 22(13.3) | 144(86.7) | 166(65.6) | ||
| Accessibility | |||||
| Agree | 23(14.5) | 136(85.5) | 159(62.8) | 0.025 | 3.01(1.1,8.21) |
| Disagree | 5(5.3) | 89(94.7) | 94(37.2) | ||
| Spend time at the health facility | |||||
| Agree | 8(5.5) | 138(94.5) | 146(57.7) | < 0.001 | 0.25(0.11,0.6) |
| Disagree | 20(18.7) | 87(81.3) | 107(42.3) | ||
| Headache | |||||
| Yes | 9(7.8) | 107(92.2) | 116(45.8) | 0.123 | 0.52(0.23,1.2) |
| No | 19(13.9) | 118(86.1) | 137(54.2) | ||
| Feeling nervous | |||||
| Yes | 4(6.9) | 54(93.1) | 58(22.9) | 0.249 | 0.53(0.18,1.59) |
| No | 24(12.3) | 171(87.7) | 195(77.1) | ||
| Drowsiness | |||||
| Yes | 6(14) | 37(86) | 43(17) | 0.592 | 1.39(0.53,3.65) |
| No | 22(10.5) | 188(89.5) | 210(83) | ||
| Nausea | |||||
| Yes | 2(11.8) | 15(88.2) | 17(6.7) | 1.000 | 1.08(0.23,4.98) |
| No | 26(11) | 210(89) | 236(93.3) | ||
| Dizziness | |||||
| Yes | 12(19.4) | 50(80.6) | 62(24.5) | 0.017 | 2.63(1.17,5.91) |
| No | 16(8.4) | 175(91.6) | 191(75.5) | ||
| Side effects | |||||
| Agree | 4(3) | 129(97) | 133(52.6) | < 0.001 | 0.12(0.04,0.37) |
| Disagree | 24(20) | 96(80) | 120(47.4) | ||
| Ineffective medication | |||||
| Agree | 5(7.7) | 60(92.3) | 65(25.7) | 0.314 | 0.6(0.22,1.64) |
| Disagree | 23(12.2) | 165(87.8) | 188(74.3) | ||
| Lack of knowledge of medication usage | |||||
| Agree | 4(8) | 46(92) | 50(19.8) | 0.440 | 0.65(0.21,1.96) |
| Disagree | 24(11.8) | 179(88.2) | 203(80.2) | ||
| Poor prescription instruction by health workers | |||||
| Agree | 4(8) | 46(92) | 50(19.8) | 0.440 | 0.65(0.21,1.96) |
| Disagree | 24(11.8) | 179(88.2) | 203(80.2) | ||
| Use of other treatment options such as herbal preparations | |||||
| Agree | 7(7.5) | 86(92.5) | 93(36.8) | 0.171 | 0.54(0.22,1.32) |
| Disagree | 21(13.1) | 139(86.9) | 160(63.2) | ||
| Light-headedness | |||||
| Yes | 0(0) | 42(100) | 42(16.6) | 0.012* | |
| No | 28(13.3) | 183(86.7) | 211(83.4) | ||
| Erectile dysfunction | |||||
| Yes | 0(0) | 49(100) | 49(19.4) | 0.006* | |
| No | 28(13.7) | 176(86.3) | 204(80.6) | ||
*odds ratios with confidence intervals were not computed for cells with zero (0) counts.
Comparative analysis of logistic regression and complementary log–log models for predictors of medication adherence
The model results show that the two models were all statistically significant (p-value < 0.001) with no collinearity (GVIF < 2.2), however, CLLM did better than the LRM (Table 3). The CLLM reported AIC and BIC values of 126.1 and 203.8, respectively, while LRM reported AIC and BIC values of 131.2 and 209.0, respectively. The Nagelkerke and Cox and Snell R-squared values for the CLLM were 0.62 and 0.31, respectively, which were higher than the values reported by LRM (Nagelkerke = 0.59 and Cox and Snell = 0.30). Therefore, the results of the CLLM were reported.
Table 3.
Comparative analysis of logistic regression and complementary log–log models for predictors of medication adherence.
| Predictors | Complementary log–log model | Logistic regression model | ||||
|---|---|---|---|---|---|---|
| Risk Ratios | 95% CI | p-value | Odds Ratios | 95% CI | p-value | |
| (Intercept) | 0.07 | 0.01–0.76 | 0.027 | 0.07 | 0.00–1.02 | 0.056 |
| Age (Years) | ||||||
| 20 to 29 | 8.58 | 1.23–83.28 | 0.031 | 12.36 | 1.36–149.12 | 0.032 |
| 30 to 39 | 2.25 | 0.16–36.60 | 0.544 | 3.10 | 0.18–63.15 | 0.434 |
| 40 to 49 | 0.65 | 0.06–7.78 | 0.720 | 0.70 | 0.04–11.67 | 0.804 |
| 50 to 59 | 0.4 | 0.04–2.89 | 0.354 | 0.42 | 0.03–4.34 | 0.476 |
| Sex [Male] | 0.19 | 0.03–0.91 | 0.042 | 0.18 | 0.03–0.98 | 0.059 |
| Location [Urban] | 10.78 | 2.01–79.76 | 0.007 | 12.19 | 1.93–109.22 | 0.014 |
| Occupations | ||||||
| Accountant | 0.11 | 0.01–1.30 | 0.077 | 0.10 | 0.01–1.58 | 0.109 |
| Engineer | 0.24 | 0.04–1.11 | 0.069 | 0.19 | 0.03–1.09 | 0.072 |
| Others | 0.1 | 0.00–1.21 | 0.110 | 0.09 | 0.00–1.28 | 0.107 |
| Embarrassed [Agree] | 0.49 | 0.07–2.44 | 0.415 | 0.45 | 0.05–2.73 | 0.414 |
| Busy schedule [Agree] | 0.03 | 0.00–0.30 | 0.016 | 0.02 | 0.00–0.26 | 0.014 |
| Cultural beliefs [Agree] | 0.24 | 0.01–2.03 | 0.282 | 0.24 | 0.01–2.68 | 0.316 |
| Attitude hospital [Agree] | 0.38 | 0.06–1.91 | 0.231 | 0.44 | 0.06–2.90 | 0.4 |
| Accessibility [Agree] | 17.89 | 4.03–112.38 | < 0.001 | 23.94 | 4.49–173.25 | 0.001 |
| Time spent [Agree] | 0.05 | 0.00–0.28 | 0.002 | 0.04 | 0.00–0.30 | 0.004 |
| Headache [Yes] | 1.29 | 0.30–5.78 | 0.732 | 1.27 | 0.26–6.48 | 0.766 |
| Dizziness [Yes] | 2.34 | 0.48–12.45 | 0.283 | 2.30 | 0.40–14.82 | 0.357 |
| Side effects [Agree] | 0.43 | 0.06–2.65 | 0.367 | 0.33 | 0.04–2.48 | 0.289 |
| Ineffective medication [Agree] | 0.8 | 0.12–4.29 | 0.796 | 0.92 | 0.11–6.50 | 0.936 |
| Treatment options [Agree] | 1.85 | 0.36–9.50 | 0.435 | 1.59 | 0.25–9.80 | 0.613 |
| Model assessment | ||
|---|---|---|
| Measures of fit | Measures of fit | |
| Chi-square statistic (p-value) | 93.96 (p-value < 0.001) | 88.80(p-value < 0.001) |
| AIC | 126.10 | 131.20 |
| BIC | 203.80 | 209.00 |
| Cox and Snell | 0.31 | 0.30 |
| Nagelkerke | 0.62 | 0.59 |
| GVIF(1/(2*Df)) | Less than 2.2 for all predictors | Less than 2.2 for all predictors |
The factors that affect adherence to hypertension medication include age (ARR = 8.58, 95%CI 1.23–83.28, p-value = 0.031), Sex (ARR = 0.19, 95%CI 0.03–0.91, p-value = 0.042), location (ARR = 10.78, 95%CI 2.01–79.76, p-value = 0.007), busy schedule (ARR = 0.03, 95%CI 0.00–0.30, p-value = 0.016), time spent (ARR = 0.05, 95%CI 0.00–0.28, p-value = 0.002) and accessibility (ARR = 17.89, 95%CI 4.03–112.38, p-value < 0.001). Males were less likely to adhere to treatments as compared to females (Table 3). Also, patients who have busy schedules and those who spend more time at their facilities are less likely to adhere to treatments. However, those who have easy access to their health facilities are more likely to adhere to treatment.
Discussion
The study assessed the level of adherence to treatments and factors that influence the behaviour of a population who do not have to bear the cost. We found an adherence level of 11.1%, indicating an extremely low level of adherence to treatment among hypertension patients who do not have to bear the cost of therapy. The extremely low level of adherence is a cause of concern, particularly when cost is not a barrier to accessing treatment. Poor adherence is a major contributor to the increased prevalence of complications of hypertension, including stroke, kidney failure, cardiovascular disorders, and premature deaths1,2. Again, adherence could also contribute to the medical bills borne by the organisation, since it is more costly to treat complications, thus increasing the overhead cost of medical care yet reducing productivity. The extremely high level of non-adherence could also increase the overhead cost of healthcare borne by employers because of lost man-hours and increased cost of treatment of complications arising from non-adherence to treatment. The current findings are at variance with some previous studies in Ghana that found high adherence levels among hypertensive patients25,26. These variations could be attributable to the differences in the study settings. One of the studies was conducted among two hospital-based populations25, while the other was among hypertension patients in the specialist clinic26. It is important to note that while clients in hospital settings may present a positive outlook to medication adherence, community-based or workplace adherence may remain considerably low, warranting the need to institute targeted interventions among this cadre. This is because hypertension patients within the health care institution may have higher health awareness and time compared to those at the workplace or within the communities. A similar study in Japan found a high level of adherence to treatment among elderly hypertensive patients who obtained medication at Veterans Affairs Medical clinics27. Previous studies in sub-Saharan Africa28–30 and in Saudi Arabia31 made congruent findings, even though the level of adherence in their study was relatively higher than in the current study. The difference could be explained by the method used to calculate medication adherence. In the previous studies, unlike the current medium adherence was added to high adherence and considered high adherence. Besides, the study by Sarkodie et al.25 was conducted in relatively rural districts in Ghana, where other communal factors could have contributed to promoting treatment adherence. In addition, the current study involved patients who were still in active employment and could therefore be preoccupied with work, unlike other studies that involved retired elderly adults in care homes with no obstacles to medication refills.
The bivariate analysis found that being embarrassed by taking hypertension medication, having a busy schedule, accessibility to treatment, spending too much time at the health facility, feeling dizzy, having unbearable side effects of hypertension medication, feeling of light-headedness, and having erectile dysfunction significantly influence patients’ hypertension medication adherence. The risk of hypertension side effects (like erectile dysfunction, dizziness, feeling of light-headedness) has to be addressed comprehensively with patients and mitigating measures identified in collaboration with the patient and instituted. Community-level factors like being embarrassed for taking hypertension medication must be addressed through collaborative community education, warranting a change in community-level attitudes. Healthcare providers must discuss with patients to identify the experiences associated with hypertension medication intake so that they can prescribe alternatives with minimal side effects. Also, system-wide challenges like access to health facilities and utilization of health care services, including medication access, may have to be addressed by health care providers. this may include diversifying strategies like task shifting and team-based care provision to ensure that experienced nurses at primary health care facilities are empowered to titrate hypertension medication under the supervision of a medical service provider. Health education regarding hypertension among patients must address the immediate concerns related to treatment side effects. Healthcare providers must also institute prompt and innovative strategies to reduce the waiting time at the outpatient departments to promote easy and accessible refills to limit the rate of default. Also, innovations such as scheduling patient visits for different times of the day could ensure that hypertensive patients are attended to promptly whenever they visit the health facility.
A logistic regression and complementary log–log models were compared in modelling the predictors of medication adherence. When the presence and absence of the event in the dependent variable have similar probabilities, the logistic regression model performs well. The complementary log–log model outperformed the logistic regression model in this study, and this is due to the non-similarity of the probabilities of success (adherence) and failure (non-adherence). The prevalence of adherence was 11.1% while the prevalence of nonadherence was 88.9%. In the multivariable analysis, males were less likely to adhere to treatments as compared to females. This finding agreed with previous studies, which also found that females are more likely to adhere compared with males32,33. However, this finding contradicts that of Noreen et al.34 and Holmes et al.35, which reported higher adherence in males than females. With increasing discrepancies in the study findings of which sex is engaged in better medication adherence, a meta-analysis using a systematic review will be warranted. This will be critical to tailored healthcare interventions aimed at addressing the challenges associated with medication adherence, especially in low-resource settings.
In addition, patients aged 20 to 29 years and those who live in urban areas are more likely to adhere to medication as compared to those aged at least 60 years and those who live in rural areas, respectively. Urban residents and younger people may have more access to health information because of the proliferation of media, the increased presence of healthcare providers, access to health facilities, and the possible observation of persons experiencing the complications of hypertension. This supports the findings of previous studies that non-adherence increases with an increase in age, as well as urban areas reported higher adherence than those in rural areas25,36,37. Also, patients who have busy schedules and those who spend more time at their facilities are less likely to adhere to treatments. A busy schedule and more time at a facility were shown as factors of non-adherence in previous studies16,34. However, those who have easy access to their health facilities are more likely to adhere to treatment. Previous studies have found distance and delays at health facilities to be responsible for non-adherence among patients in similar African countries15,30. High adherence to medication is related to the low or no cost associated with medication for the patients, and the convenience of medication refills, as well as patients receiving patient-centered care at medical homes where they had sufficient access to healthcare providers23,27. It is therefore imperative that healthcare interventions identify these specific barriers to medication adherence and institute measures to promote high adherence. This is particularly important that in the current study, the cost of accessing and utilising medication was paid by the employer instead of the patient, yet we recorded a very low level of medication adherence, especially in the relatively young population.
Strengths and limitations
This study is one of the first to assess the level of adherence to treatment and the factors associated with the practice among a population of hypertensive patients who do not have to bear the cost of treatment. However, some limitations ought to be acknowledged in interpreting the findings. Firstly, the study used a convenience sampling strategy; it did not include patients who attended the hypertension clinic as private clients (as a specific category) and thus had to bear their own cost of treatment in the same setting. The inclusion of these clients could have facilitated a comparative analysis of the level of adherence and the factors that influence adherence, either with the inclusion or exclusion of healthcare costs. However, the study was unique in that it provided insight into a special population of patients that most studies in Ghana have failed to address. The use of the convenience sampling method could also influence the study outcome because of the likelihood that respondents could disproportionately include those who have high health awareness. However, believe that this could have been minimised by the higher sample size used. Secondly, the data collection could have been influenced by recall and social desirability bias. This was, however, mitigated by ensuring data collectors received appropriate training before the commencement of the study. Thirdly, this study was conducted in a single center within a limited geographical zone. Future studies may consider expanding the sample size and including more facilities in resource-limited settings.
Conclusion
The logistic regression model performs well when the probabilities of presence and absence of the event are similar or balanced; however, an imbalance of the probabilities affects its predictive performance. Hence, the complementary log–log model outperformed the logistic regression model due to the large difference between the prevalence of non-adherence (88.9%) and adherence (11.1%).
Treatment adherence is crucial to ensuring the long-term control and management of hypertension. The study discovered an extremely high level of non-adherence to treatment among hypertensive patients in Ghana, even though the cost of treatment was not a challenge. Despite cost-free access, adherence remains low, highlighting the need for further investigation into non-cost barriers affecting treatment compliance. There is thus a need to think about the current measures that aim to improve the treatment outcomes of patients with hypertension. Strategies to improve adherence and treatment outcomes should consider factors that influence adherence in addition to the cost of treatment. There is a critical need to develop health interventions targeting VRA workers by addressing challenges associated with medication side effects, access, time wasted at the health facility, and ignorance regarding the correct use of medication to improve adherence. In addition, authorities of the VRA can incorporate a structured hypertension monitoring program in two weekly work activities to promote medication adherence, ensure efficient control, and promote client outcomes. Given the findings of the study, we recommend the intensification of education focused on mitigating the negative effects of non-adherence to treatment. We also recommend that future studies consider the effect of non-adherence on the treatment outcomes of patients by including blood pressure measurements in the variables to be explored.
Acknowledgements
The authors are grateful to the study participants for agreeing to be part of the study and for providing useful data that generated these findings.
Author contributions
FKN conceptualised the study and supervised the data collection. FKN, KDK and DA provided ideas for the study design and discussion. DA performed the data analysis. FKN, DA, LSA, VB, JG, VD, KDK, and RQ drafted the manuscript. FKN and DA critically reviewed and approved the manuscript content.
Funding
There was no funding for this study.
Declarations
Competing interests
No, I declare that the authors have no competing interests as defined by BMC, or other interests that might be perceived to influence the results and/or discussion reported in this paper.
Ethical approval and consent to participate
Ethics approval for the study was granted by the Research Ethics Committee of the University of Health and Allied Sciences (reference number: UHAS-REC A.1[50]19-20). The study was conducted following the guidelines of the institutional ethics review committee and the Declaration of Helsinki. Before participation in the study, participants signed the informed consent form. Participants granted permission for their data to be included in any publication. All the authors approved the publication of this manuscript.
Consent for publication
Not applicable.
Data availability
The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Kennedy Diema Konlan, Email: dkkonlan@uhas.edu.gh.
David Adedia, Email: dadedia@uhas.edu.gh.
References
- 1.Mills, K. T., Stefanescu, A. & He, J. The global epidemiology of hypertension Katherine. Physiol. Behav.176(1), 139–148 (2020). [Google Scholar]
- 2.World Health Organization. A global brief on hypertension | A global brief on Hypertension [Internet]. 2013. Available from: www.who.int
- 3.Gaye, B. et al. Prevalence of severe hypertension in a Sub-Saharan African community. Int. J. Cardiol. Hypertens.1, 2 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Zhou, B. et al. Worldwide trends in hypertension prevalence and progress in treatment and control from 1990 to 2019: A pooled analysis of 1201 population-representative studies with 104 million participants. The Lancet398(10304), 957–980 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Ghana Health Service. Ghana Health Service 2016 Annual Report. Accra, Ghana; 2017.
- 6.World Health Organization. Global NCD Target Reduce High Blood Pressure Background [Internet]. 2016. Available from: http://apps.who.int/iris/bitstream/10665/79059/1/WHO_DCO_WHD_2013.2_eng.pdf
- 7.Calys-Tagoe, B., Nuertey, B. D., Tetteh, J. & Yawson, A. E. Individual awareness and treatment effectiveness of hypertension among older adults in Ghana: Evidence from the world health organization study of global ageing and adult health wave 2. Pan Afr. Med. J.1(37), 1–16 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.World Health Organization. World Health Statistics 2018: Monitoring Health for the SDGs: Sustainable Development Goals 86 (WHO, 2018). [Google Scholar]
- 9.World Health Organization. Adherence to long-term therapies: evidence for action. World Health Organization. p. 194 (2003).
- 10.Hugtenburg, J. G., Timmers, L., Elders, P. J., Vervloet, M. & van Dijk, L. Definitions, variants, and causes of nonadherence with medication: A challenge for tailored interventions. Patient pref. Adher. 675–682 (2013). [DOI] [PMC free article] [PubMed]
- 11.Peh, K. Q. E. et al. An adaptable framework for factors contributing to medication adherence: results from a systematic review of 102 conceptual frameworks. J. General Int. Med.36(9), 2784–2795 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Gutierrez, M. M., Sakulbumrungsil, R. Factors associated with medication adherence of hypertensive patients in the Philippines: A systematic review. Vol. 27, Clinical Hypertension. BioMed Central Ltd (2021). [DOI] [PMC free article] [PubMed]
- 13.Dougherty, E. C. et al. Examining medication adherence and preferences for a lifestyle intervention among Black and Latinx adults with hypertension: a feasibility study. Pilot Feasibil. Stud.7(1), 209 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Shiraly, R., Khani Jeihooni, A. & Bakhshizadeh Shirazi, R. Perception of risk of hypertension related complications and adherence to antihypertensive drugs: a primary healthcare based cross-sectional study. BMC Prim. Care23(1), 303 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Odusola, A. O. et al. Perceptions of inhibitors and facilitators for adhering to hypertension treatment among insured patients in rural Nigeria: A qualitative study. BMC Health Serv. Res.14(1), 624 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Atinga, R. A., Yarney, L. & Gavu, N. M. Factors influencing long-term medication non-adherence among diabetes and hypertensive patients in Ghana: A qualitative investigation. PLoS ONE13(3), e0193995 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Buabeng KO, Plange-Rhule J. Unaffordable drug prices: the major cause of non-compliance with hypertension medication in Ghana [Internet]. Vol. 7, J Pharm Pharmaceut Sci (www.ualberta.ca/~csps). 2004. Available from: http://www.twnside.org.sg/ [PubMed]
- 18.Gray, J., Grove, S. K. & Sutherland, S. Burns and Grove’s the practice of nursing research: Appraisal, synthesis, and generation of evidence. Elsevier. Vol. 8 1–1192 (2017).
- 19.Yamane, T. Statistics: An Introductory Analysis [Internet]. Harper & Row. (A Harper international edition). Available from: https://books.google.com.gh/books?id=Wr7rAAAAMAAJ (1967).
- 20.Nassar, R. I., Basheti, I. A. & Saini, B. Exploring validated self-reported instruments to assess adherence to medications used: A review comparing existing instruments. Patient Prefer. Adherence24, 503–513 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Johnson, A. B. et al. Medication adherence rating of patients with hypertension in a tertiary care hospital. World J. f Pharm. Pharm. Sci.5(2), 694–710 (2016). [Google Scholar]
- 22.Svarstad, B. L., Chewning, B. A., Sleath, B. L. & Claesson, C. The brief medication questionnaire: A tool for screening patient adherence and barriers to adherence. Patient Educ. Couns.37(2), 113–124 (1999). [DOI] [PubMed] [Google Scholar]
- 23.Mekonnen, H. S., Gebrie, M. H., Eyasu, K. H. & Gelagay, A. A. Drug adherence for antihypertensive medications and its determinants among adult hypertensive patients attending in chronic clinics of referral hospitals in Northwest Ethiopia. BMC Pharmacol. Toxicol.18(1), 27 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Laar, A. K. et al. Health system challenges to hypertension and related non-communicable diseases prevention and treatment: Perspectives from Ghanaian stakeholders. BMC Health Serv. Res.19(1), 693 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Sarkodie, E., Afriyie, D. K., Hutton-Nyameaye, A. & Amponsah, S. K. Adherence to drug therapy among hypertensive patients attending two district hospitals in Ghana. Afr. Health Sci.20(3), 1355–1367 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Yaw Adu-Boakye, B. et al. Adherence to antihypertensive medication in a specialist led-hypertension clinic in Sub-Saharan Africa adherence to antihypertensive medication in a specialist led-hypertension clinic in Sub-Saharan Africa adherence to antihypertensive medications (2021).
- 27.Zhao, M. et al. Validity and reliability of a short self-efficacy instrument for hypertension treatment adherence among adults with uncontrolled hypertension. Patient Educ. Couns.104(7), 1781–1788 (2021). [DOI] [PubMed] [Google Scholar]
- 28.Kretchy, I., Owusu-Daaku, F. & Danquah, S. Spiritual and religious beliefs: Do they matter in the medication adherence behaviour of hypertensive patients?. BioPsychoSoc. Med.7(1), 15 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Boima, V. et al. Factors associated with medication nonadherence among hypertensives in Ghana and Nigeria. Int. J. Hypertens.2015(1), 205716 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Okello, S., Nasasira, B., Muiru, A. N. W. & Muyingo, A. Validity and reliability of a self-reported measure of antihypertensive medication adherence in Uganda. PLoS ONE11(7), e0158499 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Algabbani, F. M. & Algabbani, A. M. Treatment adherence among patients with hypertension: Findings from a cross-sectional study. Clin. Hypertens.26(1), 18 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Lefort, M. et al. Sex differences in adherence to antihypertensive treatment in patients aged above 55: The French league against hypertension survey (FLAHS). J. Clin. Hypertens.20(10), 1496–1503 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.To, T. et al. Global asthma prevalence in adults: Findings from the cross-sectional world health survey. BMC Publ. Health12(1), 204 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Noreen, N. et al. Determinants of adherence to antihypertension medications among patients at a tertiary care hospital in Islamabad, Pakistan, 2019. Prev. Chronic Dis.20, E42 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Holmes, J. & Heaney, L. G. Measuring Adherence to Therapy in Airways Disease Vol. 17 (European Respiratory Society, 2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Kretchy, I. A., Koduah, A., Ohene-Agyei, T., Boima, V. & Appiah, B. The association between diabetes-related distress and medication adherence in adult patients with type 2 diabetes mellitus: A cross-sectional study. J. Diabet. Res.2020(1), 4760624 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Tola Gemeda, A. et al. Adherence to Antihypertensive Medications and Associated Factors Among Hypertensive Patients in Ethiopia: Systematic Review and Meta-Analysis Vol. 8 (SAGE Publications Ltd, 2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.