Association between adherence to Antiretroviral Therapy and place of residence among adult HIV infected patients in Ethiopia: A systematic review and meta-analysis

Background According to the World Health Organization, optimal adherence to Antiretroviral Therapy (ART) improves quality of life. Patients who use ART have varying characteristics in terms of where they live. The effect of place of residence on ART adherence is unclear in Ethiopia. Therefore, the aim of this systematic review and meta-analysis was to estimate the pooled association between place of residence and adherence to ART. Methods Articles were retrieved from PubMed, Scopus, African Journals Online (AJOL), Journal Storage (JSTOR), and Web of Science. The data was extracted using Microsoft Excel 2016 spreadsheet. Review Manager 5.3 and STATA version 14 were used for the analysis. The Cochrane Q statistic was used to assess between-study heterogeneity. I2 was used to quantify between-study heterogeneity. A weighted inverse variance random-effects model was used to calculate the pooled odds ratio with 95% confidence interval. Results Seven studies were included in this systematic review and meta-analysis. The Begg’s test (Z = 0.15, P = 0.881) and Egger’s test (t = 0.14, P = 0.894) revealed no evidence of publication bias. Urban residence was associated with an increased likelihood of good adherence (OR 2.07, 95%CI 1.22–3.51). Conclusions The study recommends that policy-makers should enact policies that increase access to ART services in a rural area in order to improve adherence. It is recommended that implementation studies be conducted in order to identify practical and affordable interventions.

The following terms with MeSH (Medical Subject Headings) and Boolean operators were used to search PubMed: The findings of this systematic review and meta-analysis have been reported in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guideline [21] (S1 Checklist). Articles were excluded if they were either review articles or studies that did not report the desired outcome.

Study selection
The author exported the retrieved studies to EndNote X7, which was then used to eliminate duplicate studies. The author determined the eligibility of the candidate studies. The abstract title and content were used to screen the articles. The screened articles were then subjected to a full article review. The inclusion and exclusion criteria were used to screen the articles.

Quality appraisal
The quality of the studies included in the systematic review and meta-analysis was assessed using the Joanna Briggs Institute (JBI) quality appraisal checklist. The appraisal checklist included the following parameters: 1) inclusion criteria, 2) description of study subject and setting, 3) valid and reliable measurement of exposure, 4) objective and standard criteria used, 5) identification of confounder, 6) strategies to handle confounder, 7) outcome measurement, and 8) appropriate statistical analysis. If the quality assessment indicator score was 50% or higher, the study was considered low risk. All of the articles included in this systematic review and meta-analysis were found to be low risk.

Data extraction
A standardized data extraction format was carefully designed in a Microsoft Excel 2016 spreadsheet that was used to extract data from the articles. The first author's name, publication year, study period, study setting, region, study design, sample size, sampling method, adherence definition, measures, and adherence level in both urban and rural areas were all extracted.

Study measures
This study estimated the pooled association between place of residence and adherence. The residence is a dichotomous variable that can be classified as urban or rural. Adherence is another dichotomous variable that can be classified as good or bad. Table 1 shows the studies' definition of good adherence.

Statistical analysis
Pooled analysis was done using Review Manager 5.3(Version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). Publication bias was assessed using STATA 14 software (stataCorp LP, 4905 Lakeway Drive, College Station, TX 77845, USA). The number of adherent and non-adherent HIV-infected patients in urban and rural areas is entered to calculate the odds ratios (OR). The Cochrane Q statistic was used to determine whether there was significant between-study heterogeneity. I 2 was used to quantify betweenstudy heterogeneity, with values of 0%, 25%, 50%, and 75% representing no, low, medium, and increased heterogeneity, respectively. Because of the observed heterogeneity between the studies, a weighted inverse variance random-effects model was used to calculate the pooled odds ratio with a 95% confidence interval.
The publication bias was checked using Begg's rank correlation test and Egger's linear regression test. Publication bias is not the only source of asymmetry in funnel plots. In addition, the visual inspection of funnel plots is subjective and should be supplemented with additional analysis. Therefore, the funnel plot was used to examine small-study effects. Begg's rank correlation test and Egger's linear regression test were used by the author because the tests had stronger discriminatory power in detecting publication bias than Macaskill's method. Furthermore, Deeks' method was not used because it is originally designed for meta-analysis of diagnostic tests.
A p-value of 0.05 was used in this systematic review and meta-analysis to determine the significance of the small study effect. The absence of publication bias was demonstrated by Begg's rank correlation test (Z = 0.15, P = 0.881) and Egger's linear regression test (t = 0.14, P = 0.894).

Results
Initially a total of 2558 articles were retrieved: 1699 from PubMed, 430 from AJOL, 212 from SCOPUS, 165 from JSTOR, and 165 from Web of Science databases. After duplicates were removed, 1077 remained. Two hundred articles were screened based on their abstracts. Of these, 51 articles were assessed using the eligibility criteria. Finally, 7 articles met the eligibility criteria and were included in the quantitative synthesis (Fig 1).

Association between place of residence and adherence
The pooled association between ART adherence and residence was estimated using seven studies. The OR was used to estimate the association. The OR ranges from 0.38(95%CI 0.20-0.71) to 6.21(95%CI 3.28-11.7). The result showed a high level of heterogeneity (Heterogeneity chi-square 45.54, df. 6, I 2 87%, P<0.001). Therefore, the pooled association between ART adherence and residence was determined using a random-effects model. Increased likelihood of good ART adherence was identified among patients living in urban areas. Accordingly, HIV-infected patients living in urban areas were 2.07 times more likely to have good ART adherence as compared to HIV-infected patients living in rural areas (OR 2.07, 95%CI 1. .51) at p = 0.007 (Fig 2).

Publication bias
The funnel plot revealed a symmetrical distribution (Fig 3). The Egger's linear regression test and Begg's rank correlation test were used to objectively identify publication bias. Egger's linear regression test was not statistically significant (t = 0.14, P = 0.894). Furthermore, Begg's rank correlation test was not statistically significant (z = 0.15, P = 0.881) ( Table 2). Therefore, Begg's rank correlation test and Egger's linear regression test revealed no evidence of publication bias.

Discussion
Seven studies were included in this systematic review and meta-analysis to summarize the effect of place of residence on the level of ART adherence among adult HIV-infected patients living in Ethiopia. The findings were compiled from studies conducted in various regions of Ethiopia by the author. Adherence to ART is related to the quality of life of HIV-infected patients [29]. Patients in developing countries are expected to overcome the challenges of the health care system in order to achieve a high quality of life. This problem is exacerbated by a lack of access to ART services in nearby health institutions [8,9,16].
The finding showed that patients living in urban areas are 2.07 times more likely to have a good adherence as compared to those living in rural areas. This finding is supported by a finding from a study conducted in Kenya, which analyzed the results of the second Kenya AIDS indicator survey [30]. It is also supported by a systematic review and meta-analysis, which reported that patients living in urban areas are more likely to have a good adherence as compared to those living in rural areas [31]. This finding is supported by another systematic review conducted in South Africa on equity in utilization of ART [32]. On the other hand, the finding is contrary to a report from a systematic review conducted in Sub-Saharan African countries, which showed that urban residence favors non-adherence [16]. The poor ART adherence  among patients living in rural areas could be attributed to a variety of factors. People in Ethiopia's rural areas have a low socioeconomic status. Likewise, HIV-infected patients living in rural areas have a low income. This might lead to poor adherence. In addition, poor adherence is linked to a lack of access to ART services [33]. HIV-infected patients in urban areas who live close to the health facility use of the service frequently. In Ethiopia, most health facilities are not located in rural areas, and patients are expected to travel to urban areas to receive ART services. Therefore, they may miss an appointment and exhibit poor adherence. The presence of a support system is also related to adherence [34]. Patients in rural areas may have a deficient support system. Asymmetrical funnel plots may indicate publication bias or exaggeration of treatment effects in small, low quality studies. Therefore, the author used Begg's rank correlation test and Egger's linear regression test to check publication bias. The results of the statistical tests showed there was no publication bias.
One of the studies included in this systematic review and meta-analysis used a consecutive sampling method to select study participants and didn't take into account the sampling bias introduced by this method [26]. Studies that collected data using self-reporting did not consider self-reporting bias, social desirability bias, and recall bias [23][24][25]27]. It is possible that it will have an effect on the reported data from the studies. In addition, one of the studies conducted at Jimma University Teaching Hospital selected study participants using convenience sampling, which is a type of non-probability sampling [28]. It is possible that it will have an impact on the study's findings because it is influenced by sampling error and selection bias.
Each article was subjected to a quality assessment before being included in the systematic review and meta-analysis in order to improve the validity of the results. Adherence was measured using a standard method in the articles involved in this study. In addition, publication bias was checked because it affects the validity of the results. In selecting the appropriate database, the author also consulted an expert and a librarian.
The study recommends that policymakers focus on incorporating ART services into rural health centers and health posts. A more stringent engagement strategy should be developed to encourage the involvement of sectors working to combat HIV/AIDS. The health institutions that provide ART services should have a mechanism in place to monitor patient adherence. The facilities could collaborate with health extension workers in the rural area to monitor the patient. Adherence could be improved through a close monitoring and follow-up. Health facilities can also launch a special program that serves as a platform to thoroughly discuss the patient-perceived barriers. The patient should also communicate openly with the health professionals who provide ART services, which can make it easier for health professionals and patients to overcome barriers.
This systematic review and meta-analysis is not free from limitations. These included the author's inability to determine the temporal relation due to the study designs used in the articles. Furthermore, the studies were not conducted across the country. The lack of multiple studies in various regions of the country makes it difficult to conduct a subgroup analysis. Therefore, the factors contributing to heterogeneity were not identified. Since the studies included in this systematic review and meta-analysis had small sample sizes, the findings should be interpreted with caution.

Conclusions
The results of this systematic review and meta-analysis showed a significant association between place of residence and adherence to ART. Accordingly, living in urban areas was significantly associated with an increased likelihood of good ART adherence. Living in rural areas, on the other hand, reduces the likelihood of good adherence. Therefore, all-inclusive and rigorous discussions should be organized at the national level on mechanisms to improve adherence in HIV-infected patients living in rural areas. Furthermore, national-level studies should be conducted to identify the reasons for the effect of the place of residence on ART adherence. Implementation researches that could identify practical and affordable interventions are recommended.