A systematic review and meta-analysis of epidemiology of depression in people living with HIV in east Africa

Background Depression is the most prevalent psychiatric disorder among people living with HIV (PLWHIV) and is associated with poor quality of life, additional comorbidities, disability, unemployment, poorer therapeutic outcomes and risky behaviors. The present systematic review and meta-analysis aims to systematically summarize empirical evidence and to formulate recommendations for future research. Methods We searched PubMed, EMBASE, SCOPUS, and relevant literature for possible studies to include. A qualitative and quantitative analysis was undertaken for this systematic review. Subgroup and sensitivity analysis were performed. Cochran's Q- and the I2 test were used to assess heterogeneity. The presence of publication bias was evaluated by using Egger's test and visual inspection of the symmetry in funnel plots. Results Of 283 titles initially identified, 81 abstracts were eligible for review. Of these, 46 articles qualified for full text review and 19 were retained. In our meta-analysis the pooled prevalence of depression in PLWHIV was 38% (95% CI 29.30-47.54). The pooled prevalence estimates of depression was 49.79% in Ethiopia and 30.88% in Uganda. In addition, the prevalence of depression was 12.40% and 46% as measured by diagnostic and screening instrument respectively. Our qualitative synthesis showed that factors such as having opportunistic infection, perceived stigma, negative life event, WHO clinical staging of disease, hospitalization in the past one month, stressful life events, food insecurity, self-efficacy, missed frequency of clinic visit, frequency of follow-up, older age, low income, urban residence and being government employee were strongly and significantly associated with depression in PLWHIV in east Africa. Conclusion The pooled prevalence estimates of prevalence of depression in PLWHIV was 38%. The prevalence estimates of depression in PLWHIV in Ethiopia was significantly higher than Uganda. In addition the prevalence of depression was significantly higher in studies conducted by screening than diagnostic instrument. Routine screening and integrated management of depression into the existing HIV care services is warranted. Validation and use of standard instrument to assess depression in PLWHIV is needed. Moreover, longitudinal and community based studies focusing on incidence and determinates of depression in PLWHIV are recommended. Electronic supplementary material The online version of this article (10.1186/s12888-018-1835-3) contains supplementary material, which is available to authorized users.

Evidence shows that east Africa is the region most affected by HIV in the world next to south Africa [32]. However, the there are no systematic review and meta-analysis studies conducted on the epidemiology of depression in PLWHIV in east Africa. Therefore, the main objective of this systematic review and meta-analysis is to summarize the available empirical evidence in east Africa on: [1] the prevalence of depression in people living with HIV (PLWHIV) and [2] the determinants of depressionin PLWHIV and to formulate recommendations for future clinical practice and research.

Methods/design
We conducted extensive search of literature as indicated in the guideline of reporting systematic review and meta-analysis (PRISMA) [32]. Three databases such as PubMed, Embase, and Scopus were consulted for our literature search. We conducted our search in Pubmed using the following terms and keywords: "Epidemiology OR prevalence OR magnitude OR incidence AND depression OR depressive symptoms OR depressive disorder OR depressive AND HIV OR human immunodeficiency virus OR AIDS AND factor OR risk OR risk factor OR determinant AND Ethiopia OR Eritrea OR Kenya OR Uganda OR Tanzania OR Sudan OR Djibouti OR Somalia OR Rwanda OR east Africa". For the other two databases (Embase and Scopus) we employed specific-subjects headings as advised for each databases. In addition, in order to identify other relevant articles we manually searched the reference lists of eligible articles.

Eligibility criteria
Two reviewers (GA and MA) evaluated the relevant articles using their title and abstracts prior to retrieval of full-text articles. The retrieved full-text articles were further screened according to pre-specified inclusion and exclusion criteria. We resolved disagreements by discussing with a third reviewer (MS).

Inclusion criteria
Design type -Cross sectional and other observational studies Study subject-people living with HIV Articles published in English language Studies which reported magnitude of depression in PLWHIV Articles that assessed risk factors of depression in PLWHIV Studies done in east Africa

Exclusion criteria
We excluded commentaries, editorials, letters, reviews, and interventional studies Duplicate studies were also excluded Methods for data extraction and quality assessment We used standardized data extraction form to extract data from identified studies. The following information were extracted for each included study: The name of the first author, publication date, study design, associated factors, sample size study setting, tools used for assessing outcome, confounders adjusted for, risk estimate (OR) and their 95% confidence interval. Data extraction from source documents was done independently by two investigators. Disagreements were resolved by consensus.
The quality of included studies was evaluated by using a modified version of the Newcastle-Ottawa Scale (NOS) [33]. Sample representativeness and size, comparability between participants, ascertainment of depressive symptoms, and statistical quality were the domains NOS scale uses to assess the quality of each studies. Actual agreement and agreement beyond chance (unweighted Kappa) were used to evaluate the agreement between the two reviewers. We considered the values 0 as poor agreement, 0.01-0.20 as slight agreement, 0.21-0.40 as fair agreement, 0.41-0.60 = moderate agreement, 0.61-0.80 as substantial agreement, and 0.81-1.00 as almost perfect agreement [34].

Data synthesis and analysis
Comprehensive meta-analysis software version3 was used for meta-analysis and forest plots that showed combined estimates with 95%CI. The overall pooled prevalence was estimated by random effect meta-analysis [35]. Heterogeneity was evaluated using Q statistic and the I 2 statistics [35]. The magnitude of statistical heterogeneity between studies was assessed using I 2 statistic and values of 25,50 and 75% were considered to represent low, medium and high, respectively [36]. For the data identified as heterogeneous, a random-effects model was used during analysis. When statistical pooling is not possible, non-pooled data was presented in table form. Meta-regression was performed to explore the potential source of heterogeneity.
The instrument used to assess depression and the country where the studies were conducted were used to determine the possible source of heterogeneity between the studies. In addition, we carried out a leave-one-out sensitivity analysis to evaluate the key studies that exert major impact on between-study heterogeneity. Publication bias was assessed by funnel plot and Egger's regression tests

Included
Full text included in qualitative and quantitative analysis (n=19)

Identification of studies
We identified 280 articles in our database search. Our manual search resulted additional six articles. Of these we excluded 205 articles during the review of duplicate and titles as they did not met the inclusion criteria ( Fig. 1). In our review of abstracts and keywords we excluded another 46 studies and 19 articles with full-text that met the inclusion criteria were included in our final analysis.

Quality of included studies
We used a modified version of the Newcastle-Ottawa scale (NOS) scale to evaluate the quality of included studies. The methodological quality was good for all nineteen studies. The risk of selection, measurement and non-response bias was low for all studies. The agreement between reviewers regarding the level of bias was moderate or almost perfect for all studies (Kappa statistic range 0.50-1 (Additional file 1: Table S1).
Subgroup analyses of the prevalence of depression in people living with HIV (PLWHIV) Subgroup analysis were performed by type of instrument used to assess depression and the country where the studies were conducted as possible source of heterogeneity between the studies as well as to explore the prevalence in each country. For our subgroup analysis based on the country where the studies were conducted, we identified representative data only for Uganda and Ethiopia. For the rest of east African countries (Kenya, Tanzania, Rwanda and Sudan), there is limited data for subgroup analysis.
Of the included studies, ten reported the prevalence of depression in PLWHIV in Uganda [1-4, 6-9, 11, 19, 20]. The pooled prevalence estimates of depression in PLWHIV in Ugandawas30.88% (95% CI 20.84-43.13) (See Table 2).  Table 2). In our analysis the prevalence of depression in PLWHIV in Ethiopia is significantly higher than the prevalent in Uganda (P<.00001). For the rest of east African countries, there is limited data for subgroup analysis. We found only one study conducted in each of the other four east African countries such as Kenya [18], Tanzania [15], Rwanda [14] and Sudan [40].
Finally, we performed subgroup analysis based on the setting where the studies were conducted. The pooled prevalence of depression in people living with HIV in east Africa for institution based studies was 36.86% (95% CI 27.87-46.86) (see Table 2).

Publication bias
The funnel plot and Egger's regression tests (B=-9.17, SE=6.16, P=0.155) provided no evidence of substantial publication bias for the prevalence of depression in PLWHIV in east Africa (Fig. 3).

Sensitivity analysis
For the purpose of further investigating potential source of heterogeneity in the analysis of the prevalence of depression in PLWHIV, we performed leave-one-out sensitivity analysis. Our sensitivity analysis showed that our findings were strong and not dependent on a single study. Our pooled estimated prevalence varied between 35.86(27.53-44.25) and 41.39(32.33-51.07) after deletion of a single study (see Table 3).

Determinants of depression in people living with HIV
Of 19 included studies, 11 studies that reported factors associated with depression in peoples living with HIV in east Africa were included in the qualitative analysis [2,3,5,7,9,10,16,17,20,26,38] (See Table 4). Association between WHO clinical staging of disease and depression was observed in three studies [2,10,17].

Discussion
To our knowledge, this is the first systematic review and meta-analysis of depression and determinants in PLWHIV in east Africa. The main aim of the study was identification of epidemiological information on the overall prevalence and common determinants of depression in PLWHIV in the area. It is also believed that the results of this review provide greater precision than the results of the studies individually considered. Nineteen studies that analyzed the prevalence of depression and associated factors in PLWHIV in east Africa were included.
The pooled prevalence estimate of depression in this meta-analysis was found a to be 38% in PLWHIV. The magnitude of depression in PLWHIV differed by the setting and the country where the studies were conducted as well as the type of instrument used to measure depression. When measured by DSM or ICD the estimated pooled prevalence of major depressive disorder appeared to be 12.40 % , but the pooled prevalence of depression was 46% as measured by screening instruments. The magnitude of major depressive disorder in PLWHIV in our meta-analysis is much higher compared with the 4% found in the general population [42]. Our estimated pooled prevalence of depression (38%) was higher than the results of systematic review and meta-analysis studies in sub-Saharan Africa which reported the pooled prevalence estimates of depression was 19% in people living HIV with sub-Saharan Africa [22]. These difference might be due to the socioeconomic and cultural differences of the countries.
In our study the pooled prevalence estimates of depression in PLWHIV in Ethiopia (49.79%) was significantly higher than the pooled prevalence estimates of depression in Uganda(30.88%). The difference might be due to all studies done in Ethiopia used screening instrument to determine depression and based on the pooled prevalence estimate, the prevalence of depression measured by screening instrument(46%) was higher than the prevalence of depression measured by diagnostic instrument (12.40%).
As expected, the magnitude of depression showed a considerable variations depending on the measurement used to determine depression in PLWHIV. The pooled prevalence of depression in PLWHIV was apparently higher in studies conducted using screening instrument than diagnostic instrument. The estimated pooled prevalence of depression in PLWHIV was found to be 66.14% , 51.30%, 12.40%, 32.15% and 30.50% as measured by CES-D, HADS, DSM or ICD, PHQ-9, and HSC-D respectively. The fact that diagnostic instrument give more weights towards the high specificity as compared to screening instrument which gives more emphasis towards high sensitivity might be the main reason for the observed difference. The other possible explanation might be diagnostic instrument uses strict criteria as compared to screening instrument as screening instrument aimed to identify potential indicators for disease or potential disease rather than establishing the presence or absence of disease which is the main purpose of diagnostic instrument. Finally, the possible interpretation for the variation in magnitude of depression among the different screening instrument may be due to difference in the sensitivity and specificity the tools used to screen depression in PLWHIV. The findings support the view that validation and use of standard instrument for screening as well as diagnosis of depression in PLWHIV.
Moreover, in our subgroup analysis of prevalence in PLWHIV by study setting show the pooled prevalence of depression in PLWHIV for the studies conducted institution-based setting was 36.86%. This results are lower than unpooled results of depression in community setting which reported magnitude of depression 59%. (see Tables 1  and 2). The fact that the community based study included in this review used self report of depression and all institution based studies used standard diagnostic or screening instrument to assess depression might be the possible reason for the observed difference in magnitude of depression in clinical and community based settings. In addition, the community based study is only one study so the observed result might be highly biased as it is self-report. The study suggest replication of studies in community samples to strengthen our finding as well as understanding of the precise estimates of depression in PLWHIV.
Concerning associated factors, Our qualitative synthesis showed that factors such as having opportunistic infection, perceived stigma, negative life event, WHO clinical staging of disease, hospitalization in the past one month, stressful life events, food insecurity, self-efficacy, missed frequency of clinic visit, frequency of follow-up, older age, low income, urban residence and being government employee were strongly and significantly associated with depression in people living with HIV in east Africa.

Difference between studies
The variation between the studies included in our systematic review and meta-analysis led to high level of heterogeneity in our analysis. The possible contributing factors for the variance in prevalence rates of depression in PLWHIV in east Africa includes the instrument used to measure depression, the study setting and populations. We conducted leave-one-out sensitivity analysis for the purpose of further investigating potential source of heterogeneity in the analysis of the prevalence of depression in PLWHIV. Our   sensitivity analysis showed that our findings were strong and not dependent on a single study.

Strength and limitations
The strength of the study includes the use of predefined search strategy in order to reduce reviewers bias as well as performing data extraction and quality assessment by two independent reviewers to minimize the possible reviewer bias. The other strength is performing sensitivity and subgroup analysis based on study setting, instrument used, and country of study to identify the small study effect and the risk of heterogeneity. Finally, evaluating the quality of the studies included in analysis and based on the results from the assessment of the study quality the methodological quality was generally good. Nevertheless, the study has some limitations: first, small number of studies were included in our subgroup analysis which reduce the precision of the estimate; second, apparent heterogeneity was identified among the studies; due to inconsistent adjustment and inclusion of factors determining depression in PLWHIV we done only qualitative analysis for associated factors of depression.

Conclusion
Our study found that the pooled prevalence estimate of depression in PLWHIV was 38% (95% CI 29.30-47.54). In our study the pooled prevalence estimates of depression in PLWHIV in Ethiopia was significantly higher than the pooled prevalence estimates of depression in Uganda. In addition the prevalence of depression was significantly higher in studies conducted by using screening as compared to diagnostic instrument. The study also resulted in a significant variation in prevalence of depression based on the instrument used. These findings have several implications for clinical practice and research. Based on this finding routine screening and integrated management of mental health condition into the existing HIV care services is warranted in east Africa. Validation and use of standard instrument to assess depression in PLWHIV is needed. In addition longitudinal and community based studies focusing on incidence and determinates of depression in PLWHIV are recommended. Attention need to be given for those people who have opportunistic infection, older age, advanced HIV stage, experienced stressful life events,who have perceived stigma and those who are on ART to prevent risk of non-adherence and treatment resistance and further to improve the quality PLWHIV.

Additional files
Additional file 1: Table S1. Summary of agreed level of bias and level of agreement on the methodological qualities of included studies in meta-analysis based on sampling, outcome, response rate and method of analysis. (DOCX 15 kb)