Depressive symptoms and their determinants in patients who are on antiretroviral therapy in the case of a low-income country, Ethiopia: a systematic review and meta-analysis

Background The presence of depression in people living with HIV/AIDS could lead to non-adherence to antiretroviral medications. It also leads to further comorbid and opportunistic illness and then lowering the patient's quality of life. The objective of this study was therefore to determine the pooled prevalence of depression and its related factors in HIV patients. Methods Relevant articles in PubMed, Scopus, and EMBASE were investigated. The Meta-XL version 5.3 was used to extract data and STATA-11 Meta-prop packages with the Random effect model was used to quantify depression and its related factors. Sensitivity and subgroup analysis were performed to explore sources of heterogeneity. The Cochran’s Q-statistic and the Higgs I2 test were also done. Besides, the Eggers test and symmetry in the funnel plot were used to detect the presence/absence of publication bias. Result In this meta-analysis, we included 21 articles that assessed 10,090 participants. The average prevalence of depression among people with HIV/AIDS was 35.8% (95% CI 28.29, 43.25). The average estimated prevalence of depressive symptoms was 59.4% in the Oromia region and 29.25% in southern Ethiopia. Besides, the average prevalence of depression was 45.6% and 26.2% as measured with Beck’s depression inventory and Hospital anxiety and depression scale respectively. Moreover, the prevalence of depression was 47.7% in studies that used a sample size ≤ of 400 participants and 28.5% in studies that used a sample size of > 400 participants. The pooled adjusted odds ratio (AOR) of perceived HIV stigma was 3.75 (95% CI 2.34, 5.16) and that of poor social support was 6.22 (95% CI 2.96, 9.47). Moreover, the average odds ratio of poor medication adherence, opportunistic infection, and advanced stages of AIDS were 3.03 (95% CI 1.00, 5.05), 5.5 (95% CI 1.97, 10.03), and 5.43 (95% CI 1.60, 9.28) respectively. Conclusion The pooled prevalence of depression among individuals living with HIV/AIDS was high. Factors such as perceived HIV stigma, poor social support, opportunistic infection, advanced AIDS stage, and poor medication adherence were related to it. Routine screening and management of depression and its related factors should be given due consideration.


Background
Depression is a psychiatric disorder with clinical presentations of depressed mood, loss of pleasure, reduced motivation, and energy, guilty feelings or low selfesteem, troubled sleep or appetite, suicidal thoughts, and concentration difficulties. Currently, more than 350 million people are living with depression [1]. Globally, it is the fourth leading cause of disability and is the second leading cause of disability with its lifetime prevalence in the general population estimated to be approximately 3 to 17% [2]. Research showed that depression is three times as high as more common in HIV-positive individuals and its lifetime prevalence in this population was estimated to be between 22 and 45% [3].
A systematic review and meta-analysis study by Uthman 2014 [4] found that the prevalence of depression in HIV-positive people in low-, middle-, and high-income countries range from 12.8% to 78%. Another systematic review study done in sub-Saharan Africa on depression among HIV patients incorporated 30 studies and a total of 10,000 participants from 10 countries. The reported pooled estimated occurrence of depression in this study was 31.2% [5]. Similarly, a study in sub-Saharan Africa reported the prevalence of depression range from 9 to 32% [6]. Furthermore, a meta-analysis of depression on east African HIV patients by Ayano et al. [7] revealed that the pooled prevalence of depression in HIV/AIDs patients was 38%. Other individual studies from different countries reported that depression in HIV positive individuals was 28.1% in France [8], 40% in India [9], 37.6% in South Africa [10], 47% in Uganda [11], 56.7% in Nigeria [12], 28% in Kenya, Tanzania and Namibia together [13].
The impacts of depression in people living with the human immune deficiency virus include poor quality of life [34], poor social conditions, poor compliance with prescribed treatment, poor therapeutic outcomes, and risky behaviors [35][36][37]. Furthermore, it predisposes to additional medical and psychiatric problems [38], unemployment, and disability [39][40][41].
Evidence showed that Sub-Saharan Africa countries in general and Ethiopia, in particular, are among the most vulnerable populations to be affected by the high prevalence of HIV AIDS.
Despite this, the mental health aspect of the population living with HIV in general and depression, in particular, were not addressed well. Even, those studies done are single studies that might not be as strong to be generalizable. Therefore, this systematic review and meta-analysis study was aimed to have aggregate empirical evidence on (1) The prevalence of depression in people living with HIV AIDS in Ethiopia and (2) The associated factors for the co-occurrence of depression in people with HIV AIDS and (3) To articulate a recommendation for policymakers, future researchers and clinicians in line with its findings.

Methods
This Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [42] had been utilized as a reference in conducting this study. We practically performed the search strategy for this review in the following two steps. The first step was the exploration of different databases (PubMed, Scopus, and EMBASE) to retrieve scientific evidence regarding depressive symptoms and its related factors in HIV/AIDS patients. The search strategy in the PubMed database was performed using the following key terms and words Epidemiology OR prevalence OR magnitude OR incidence AND depressive symptoms OR depression OR depressive disorder OR depressive AND HIV OR human immunodeficiency virus OR AIDS OR PLWHA OR ART AND factor OR risk OR risk factor OR determinant AND Ethiopia. Besides, we searched EMBASE and Scopus databases following the specific guideline of each database. The next step was a manual search for the reference list of included studies. During the search process, we did not put restrictions on the year of publication of the articles. For the determinants of depression in individuals who live with HIV/ADIS, narrative description, as well as the magnitude of pooled adjusted odds ratio, were employed.

Eligibility criteria
An article was eligible for inclusion in the analysis if it fulfills the following criteria's: (1) The initial criteria was that the study must assess depression in adults HIV patients, (2) the study design should be either, cross-sectional, cohort, or case-control design, (3) the outcome investigated should be depression, (4) studies should assess the associated factors for depression and (5) the study must be conducted in Ethiopia. Previous reviews, studies on non-human subjects, editorials, and articles published in non-English language were excluded. Initially, MN and AB individually screened articles stored in an endnote reference manager based on title and abstract. In the next step, these authors fully read the content of the articles that passed in the initial step, and independently decided on the articles that had to be included for final meta-analysis. Any differences between these authors concerning inclusion/ exclusion criteria were solved by consensus and discussion with a third author (MT).

Methods for data extraction and quality assessment
The previously listed authors (MN and AB) extracted the relevant information independently from the included articles using a standardized data extraction template developed by all of the authors at the beginning. The parameters included in the data extraction template and summary of articles were the author's name, publication year, study setting, study population, sample size, study design, and the assessment instrument for depression in HIV patients. All eligible articles included in the final analysis were summarized in the form of a table. PRISMA guideline [42] has been used as a standard reference during the extraction of data from all of the included studies. We also used the modified Newcastle-Ottawa Scale (NOS) [43] during the quality assessment of studies included in the final analysis. The dimensions of the NOS scale includes representativeness of sample used and sample size, comparability between participants, statistical quality, and ascertainment of cases.

Data synthesis and analysis
In this meta-analysis, the pooled prevalence of depressive symptoms and its associated factors with their 95% CIs were computed using a random-effects model [44]. The Meta-XL version 5.3 [45] and the STATA-11 Meta-prop package [46] were applied during the analysis.
Heterogeneity between the included studies was assessed with the Cochran Q-statistics and the Higgs I 2 test [47]. The I 2 value of zero defines the absence of heterogeneity and I 2 values of 25, 50, and 75% signify little, moderate, and high heterogeneity respectively [47]. Since the study had substantial heterogeneity, we further conducted a sensitivity analysis to know whether the result was influenced by a single study. Furthermore, a subgroup analysis was done by the region, measurement instrument used, and sample size. An eyeball funnel plot test [48] and Eggers regression test were also employed to identify publication bias. During this study, any statistical analysis with a P-value < 0.05 was interpreted as statistically significant.

Identification of studies
Based on the predefined search strategies, a total of 4958 literature were identified. Additionally, 7 articles were identified by a manual search for the reference lists of other articles. This makes the overall search result to be 4965 articles. Of these 35 were duplicates and therefore removed. After further screening, only 56 articles were eligible for full-text revision. Finally, only twenty-one eligible articles that fulfilled our pre-specified inclusion criteria were incorporated in the analysis (Fig. 1).

Quality of included studies
In general, the overall quality assessment score of 21 included studies based on the Newcastle Ottawa quality assessment scale ranges from 6 to 10. Amongst all studies, twenty were found to have good methodological quality and the remaining 1 was having moderate quality. However, none of the studies was found to have poor quality.

The pooled prevalence of depression among HIV/AIDS patients in Ethiopia
Twenty-one studies [14-18, 20-26, 28-33, 49] in Ethiopia were pooled in the final meta-analysis to determine the average magnitude of depression among HIV positive individuals. The reported prevalence of depression in HIV AIDS patients among studies included in the analysis varies from as low as 7.3% in the Amhara region [20] to as high as 76.7%% in the Oromia region [29]. The pooled prevalence of depression among people with HIV AIDS in Ethiopia using the random effect model was 35.8% (95% CI 28.29, 43.25). This average estimate was influenced by a considerable heterogeneity (I 2 = 99%, P-value < 0.001). Detailed information is presented in the forest plot of the pooled prevalence of depression among HIV patients in Ethiopia below (Fig. 2).

Sensitivity analysis
We performed one study leave out at a time sensitivity analysis and its result showed that the pooled estimated prevalence of depression obtained when each of the included studies was left out from the analysis at a time was within the 95% confidence limit of the pooled estimate of depression when all studies were pooled together. This suggested that the overall average prevalence was not influenced by a single particular study.

Publication bias
Egger's publication bias plot test was done to detect publication bias. However, there was no publication bias in this meta-analysis since the P-value of eggers publication bias plot test is insignificant (P-value = 0.86). This can also be strengthened with a visual inspection from a funnel plot for a Logit event rate of prevalence of depression in HIV AIDS patients against its standard error (Fig. 3).
The pooled adjusted odds ratio (AOR) of perceived HIV stigma among the indicated studies was 3.75 (95% CI 2.34, 5.16). This implied that HIV positive individuals who have HIV related perceived stigma were 3.8 times more likely to develop depression than those who have not perceived HIV stigma (Fig. 4). Six studies also reported poor social support as an associated factor for depression in HIV patients and the pooled AOR was found to be 6.22 (95% CI 2.96, 9.47) which means that individuals with poor social support were 6.2 times at increased risk of developing than with good social support (Fig. 5). Moreover, the average odds ratio of poor medication adherence, presence of opportunistic infection, and advanced stages of AIDS were 3.03 (95% CI 1.00, 5.05), 5.5 (95% CI 1.97, 10.03), and 5.43 (95% CI 1.60, 9.28) respectively. Therefore HIV positive individuals with poor medication adherence, who have an opportunistic infection and who are in advanced stages of AIDS were 3, 5.5, and 5.4 times at a higher rate of developing depression as compared to those with good medication adherence, who have no opportunistic infection and in early stages of AIDS respectively.

Discussion
The present meta-analysis study recommends the necessity of efforts to advance screening and management of depressive symptoms and its associated factors. Furthermore, researchers who are interested in assessing depression in HIV patients should use depression measures that exactly indicate the severity levels of depression which is best informative of the risk of the patient and figurative for further management. Moreover, it is imperative to have an improved understanding of depression services that could be integrated into the present ART therapy in Ethiopia.
To the knowledge of investigators, this meta-analysis study is the first of its kind to quantitatively assess both the prevalence as well as associated factors of depressive symptoms in HIV/AIDS patients in the context of Ethiopia. The main objective of the study was therefore to supplement epidemiological evidence concerning the magnitude of depressive symptoms and its related factors in HIV/AIDS patients. Consequently, the information obtained from this study on pooled magnitude and related factors for depressive symptoms would serve as an important baseline to a variety of stakeholders working in the area.
A total of twenty-one studies that assessed depressive symptoms and the related factors in Ethiopian HIV/AIDS patients were included in the analysis. The average estimated magnitude of depressive symptoms in this study was 35.8%. This was much higher than the pooled estimated prevalence of depressive symptoms in the general population in Ethiopia (9.1% to 11%) [52,53]. This signifies the depression had a massive impact on the health of people with HIV/AIDS. This result of this meta-analysis is consistent with the result of a meta-analysis study by Ayano et.al. [7] in which the pooled magnitude of depressive symptoms in East African HIV/AIDS patients was 38%. However, a study on an average prevalence of depressive symptoms among Sub-Saharan HIV/AIDS patients as studied by Bernard et al. [6] obtained a relatively lower result (19%) when compared to the present study.
In this study, the pooled estimated prevalence of depressive symptoms varies based on the region of the country in which the studies were conducted, the sample size studied, the year of publication of the study, and the type of assessment tool used. In the current analysis, the average estimated prevalence of depressive symptoms in southwestern Ethiopia (Oromia region) (59.4%) was considerably greater than the pooled estimated prevalence in southern Ethiopia (29.3%). Regionspecific cultural and socio-economic factors might be responsible for such variation. Moreover, factors related to the difference in ART facilities might also be responsible.
As anticipated, studies that utilized a larger sample size (> 400 samples) resulted in a lower pooled estimated depressive symptom (28.5%) than the pooled estimated depression prevalence of studies that used a smaller sample size (≤ 400);47.7%. This could be due to the minimal possibility of a standard error in larger sample size studies so that providing a more precise result that avoids overestimation of using a small sample.
There was also a substantial variety of pooled estimated depressive symptoms across the measurement instrument used. The average estimated prevalence of depression in HIV/AIDS patients as measured with PHQ-9, HADS, CES-D, BDI, and others (HAM-D&Kessler-6) was 37.6%, 26.2%, 32.4%, 45.6%, and 29.4% respectively. The smaller number of studies integrated into some of the assessment tool categories might minimize the estimate precision and responsible for the difference.
Concerning the associated factors for the occurrence of depressive symptoms in HIV/AIDS patients, a narrative description showed that perceived stigma, opportunistic infection, adverse drug reaction, poor social support, comorbid chronic illness, being female, stage 4 AIDS, living alone, unemployed, low income, negative life events, and non-disclosure of HIV status were some of the associated factors which had a strong positive association with the development of depressive symptoms in HIV patients. Moreover, quantitatively this meta-analysis investigated the pooled odds ratio of perceived HIV stigma, poor social support, poor medication adherence, advanced AIDS stage, and presence of opportunistic infection as they were the most frequently reported factors by the included studies. The pooled AOR of perceived HIV stigma among the included studies was 3.75. This implied that the presence of a perceived HIV stigma increases the probability of developing depressive symptoms by 3.8 times higher as compared with developing depressive symptoms in the absence of perceived stigma. This was strengthened by earlier meta-analysis studies [6,7,51]. The frequent psychological distress associated with perceived stigma might be responsible for this [16,19,22,23,26,27].
Besides, the pooled AOR for studies that reported poor social support as a risk factor for depressive symptoms was 6.2. This means that having poor social support systems increasing the risk of developing depressive symptoms 6.2 times higher as comparatively seen with a good support system. This has also been supported by earlier studies and the reason might be HIV/AIDS patients with poor social support avoids disclosing about their problem due to social humiliation towards themselves, which further increases their separation and loneness [16,19,23,24,26,27,31] as well as decreasing help-seeking intention from professionals.
Moreover, the pooled AOR for advanced AIDS stage, medication non-adherence, and opportunistic infections in this analysis were 5.4, 3, and, 5.5 respectively. This suggests that HIV/AIDS patients with Advanced AIDS stage, Medication non-adherence, and opportunistic infections were 5.4, 3 and, 5.5 times at higher risk of developing depressive symptoms as compared to patients with early

Difference between studies
This meta-analysis study on the prevalence of depressive symptoms and associated factors in HIV AIDS patients was potentially having a high degree of heterogeneity from the difference between the included studies. Therefore it was mandatory to further explore the source of such substantial heterogeneity. For this matter, a subgroup analysis had been done. The subgroup analysis result showed that the type of measurement instrument used to screen depression, the regional location at which the study was done, and the sample size utilized were responsible for the difference in the prevalence of depressive symptoms between included studies. Additionally, a single study leaves out at a time sensitivity analysis had also been performed but none of the studies were obtained to be influential on the overall estimate. This meta-analysis study has some limitations. The primary limitation is that the few numbers of studies are included in a subgroup analysis might affect the precision of the estimate and result in either an overestimation or underestimation of pooled depression prevalence. Besides, the use of different study designs might result in the overestimation and underestimation of depression prevalence. Moreover, the presence of substantial heterogeneity in the pooled depression prevalence among HIV patients might greatly affect the validity of the data. Besides, the average estimate was lower in studies that utilized a larger sample size (> 400 sample); (28.5%) than studies that used a smaller sample (≤ 400); 47.7%. Moreover, the average estimated prevalence of depression in HIV/AIDS patients as measured with PHQ-9, HADS, CES-D, BDI, and Others (HAM-D&Kessler-6) was 37.6%, 26.2%, 32.4%, 45.6%, and 29.4% respectively. The high pooled prevalence of perceived HIV stigma, poor social support, poor medication adherence, and the presence of opportunistic infections and advanced stages of AIDS increase the probability of developing depressive symptoms in HIV positive individuals in Ethiopia.

Implications
The major implication of this meta-analysis study is the high occurrence of depression symptoms among HIV AIDS patients and the variety of contributing factors. This recommends the necessity of additional efforts to advance the screening and management of depressive symptoms and their associated factors. Furthermore, researchers who are interested in assessing depression in HIV patients should use depression measures that exactly indicate the severity levels of depression from mild depression to major depressive disorder which is best informative of the risk of the patient and figurative Author, year of publication  for further management. Moreover, it is imperative to have an improved understanding of depression services that could be integrated into the present ART therapy in Ethiopia (Additional file 1).