Prevalence of depression or depressive symptoms among people living with HIV/AIDS in China: a systematic review and meta-analysis

Background The number of people living with HIV/AIDS (PLHA) in China continues to increase. Depression, a common mental disorder in this population, may confer a higher likelihood of worse health outcomes. An estimate of the prevalence of this disorder among PLHA is required to guide public health policy, but the published results vary widely and lack accuracy in China. The goal of this study was to estimate the pooled prevalence of depression or depressive symptoms among PLHA in China. Methods A systematic literature search of several databases was conducted from inception to June 2017, focusing on studies reporting on depression or depressive symptoms among PLHA in China. The risk of bias of individual studies was assessed using a modified version of the Newcastle-Ottawa scale. The overall prevalence estimates were pooled using random-effects meta-analysis. Differences according to study-level characteristics were examined using stratified meta-analysis and meta-regression. Results Seventy-four observational studies including a total of 20,635 PLHA were included. The pooled prevalence of depression or depressive symptoms was 50.8% (95% CI: 46.0–55.5%) among general PLHA, 43.9% (95% CI: 36.2–51.9%) among HIV-positive men who have sex with men, 85.6% (95% CI: 64.1–95.2%) among HIV-positive former blood/plasma donors, and 51.6% (95% CI: 31.9–70.8%) among other HIV-positive populations. Significant heterogeneity was detected across studies regarding these prevalence estimates. Heterogeneity in the prevalence of depression among the general population of PLHA was partially explained by the geographic location and baseline survey year. Conclusions Because of the significant heterogeneity detected across studies regarding these prevalence estimates of depression or depressive symptoms, the results must be interpreted with caution. Our findings suggest that the estimates of depression or depressive symptoms among PLHA in China are considerable, which highlights the need to integrate screening and providing treatment for mental disorders in the treatment package offered to PLHA, which would ultimately lead to better health outcomes in PLHA. Electronic supplementary material The online version of this article (10.1186/s12888-018-1741-8) contains supplementary material, which is available to authorized users.


Background
Human immunodeficiency virus (HIV) infection remains a significant social issue worldwide. Estimates reported by the World Health Organization (WHO) suggested that 36.7 million people were living with HIV infection and acquired immune deficiency syndrome (AIDS) at the end of 2015, with 2.1 million new infections and 1.1 million deaths due to HIV-related causes. Most people living with HIV/AIDS (PLHA) are in low-income and middle-income countries [1]. As a middle-income country, the number of PLHA continues to increase in China, although the nationwide epidemic situation of HIV/ AIDS remains at a low rate. According to the Chinese Center for Disease Control and Prevention, there were 0.50 million people infected with HIV at the end of 2014 in China [2], and in June 2017, this number increased to 0.66 million, 41.7% of whom were AIDS patients [3]. From 2014 to 2016, more than 10,000 people were infected with HIV every year [2][3][4].
Since the introduction of highly active antiretroviral therapy (ART) in the late 1990s, a large percentage of individuals with HIV-infection have been able to avoid death and live longer in a healthy condition. Nevertheless, due to social stigma, sexual dysfunction, long-term physical discomfort and illness, side effects of antiretroviral therapy, and neurobiological changes [5,6], PLHA are at a higher risk of mental disorders, particularly depression. Evidence suggests that, depression occurs more commonly in HIV-positive individuals, with a prevalence that is two to four times higher compared with comparable HIV-negative individuals or the general population [7][8][9]. Individuals with HIV infection and depression perform more poorly on clinical outcomes [10]. In fact, evidence suggests that depression may reduce antiretroviral therapy adherence and quality of life, weaken the physical function and therapeutic effect [11,12], and confer a higher rate of medical comorbidities [13,14]. Moreover, in several studies, depression has been found to be associated with higher HIV viral loads and lower CD4 counts, even after controlling for the effects of adherence, which predict a worsening disease progression and mortality [11,[15][16][17][18][19][20]. Even depressive symptoms, which do not necessarily meet the entire diagnostic criteria for a depressive disorder, have been identified as a significant factor associated with worse health outcomes among people with HIV infection, including impaired immunological response and mortality [21][22][23][24][25][26][27]. Therefore, screening for depression or depressive symptoms is an overriding concern in identifying significant risk factors for health outcomes among those who are living with HIV/AIDS.
Given the importance of the association between HIV infection and depression, scholars have been committed to the epidemiological study of depression in China.
While revealing a high occurrence of depression or depressive symptoms among PLHA, the results of existing studies have been fragmentary and inconsistent. For example, the prevalence among PLHA in Changsha City was 18.3% [28], whereas among PLHA in Wuhan City, it was 40.4% [29], among PLHA in Shanghai City, it was 60.3% [29], and among PLHA in Kunming City, it was 81.5% [30]. This knowledge gap is an obstacle to policy and practice. For example, the success of a screening program is sensitive to base prevalence. As the inconsistencies are outstanding in the current literature, it would be useful to analyze the data provided in the scientific literature using integrated approaches to establish the extent of depression or depressive symptoms among PLHA and clarify the reasons for the differences.
Therefore, in the present study, the objective was to conduct a systematic review and meta-analysis of studies to determine the prevalence of depression or depressive symptoms among PLHA in China and to explore the possible causes of the inconsistencies in the current estimates.

Search strategy
Two reviewers independently searched the EMBASE, Web of Science, PubMed, Wanfang, China Biology Medicine disc, China National Knowledge Infrastructure, and Weipu databases from inception to June 2017 for articles in English and Chinese, with no restriction on the year of the study.. The following search terms were used: human immunodeficiency virus, acquired immune deficiency syndrome, HIV, AIDS, depression, depressive disorder, depressive symptom, mental disorder, mental health, mood disorder, affective disorder, psychological health, and psychiatric. Search strategy details are shown in Additional file 1. In addition, the reviewers manually searched the reference lists of identified articles to identify any relevant studies missed in the initial search.

Study selection
At the stage of titles and abstracts screening, we purposely broadened the inclusion criteria to obtain any relevant study. First, studies were considered for inclusion if they were published in Chinese or English and reported on depression or depressive symptoms among PLHA. Then, the full texts of all selected studies were reviewed. Articles were included if they 1) were cross-sectional or cohort in design, 2) reported PLHA in China as a primary study population, 3) used a standard instrument to assess for depression or depressive symptoms, and 4) provided information about prevalence estimate of depression or depressive symptoms among PLHA. Conversely, articles were excluded if they 1) were review papers, conference abstracts, case reports, experimental studies, qualitative studies or case-control studies, 2) had incomplete or unclear data, or 3) were duplicate publications. Studies using only the data obtained from the National Health Insurance Research Database (NHIRD) were also excluded because of the possibility of underestimation. When there was more than one study involving the same population of PLHA, only the most recent published or comprehensive one was included. In addition, if the same data were published in both Chinese and English, then the articles published in Chinese were excluded.

Data extraction
Two reviewers independently extracted and evaluated the data for each included article using a self-designed data abstraction form. Disagreements were resolved through discussion or consultation with a third reviewer when consensus could not be achieved. The following data were extracted: the first author, year of publication, duration of data collection, geographic location, study design, sample source, subjects, sample size, average age of participants (mean or median), number and percentage of male participants, screening or diagnostic method, outcome definition (screening instrument cutoff or diagnostic criteria) and reported prevalence estimates of depression or depressive symptoms among PLHA. If a study reported more than one estimate assessed by different measurement tool, the one detected by the more valid measurement tool (i.e., the tool with higher specificity and sensitivity) was extracted. When there were multiple estimates over time in the same sample of a study, the first one was chosen.

Assessment of risk of bias
The risk of bias in the included studies was assessed using a modified version of the Newcastle-Ottawa scale (NOS) which was referred to the version used in the meta-analysis conducted by Rotenstein et al. to estimate the prevalence of depression or depressive symptoms in medical students [31]. The tool contained five items, which determine the risk of bias, including sample representativeness, sample size, response rate, ascertainment of depression, and quality of descriptive statistics reporting (for details, see Additional file 2). The five criteria were assessed as either "1 point" or "0 point". The higher the score, the lower the risk of bias in an individual study. According to Rotenstein et al. [31], a study was rated as having a high risk of bias if less than 3 points were given, and a low risk of bias if 3 or more points were given.

Statistical analysis
All analyses were performed using R version 3.4.1 (R Foundation for Statistical Computing), 'meta' package (version 4.8-4). In the presence of between-study heterogeneity, the pooled prevalence estimates and corresponding 95% confidence intervals (CIs) were calculated using random-effects meta-analyses. Data from studies based on HIV-positive sub-populations with specific characteristics (i.e., men who have sex with men [MSM], pregnant women, tuberculosis [TB] patients, injected drug users [IDUs] and former blood/plasma donors [FBPD]) were analyzed separately when at least six studies were available. As fewer than six studies reported data on HIV-positive pregnant women, HIV-TB co-infected individuals and HIV-positive IDUs, studies on those sub-populations were combined as "other HIV-positive population" to estimate the pooled prevalence. Cochran Q test and the I 2 statistic were used to assess the between-study heterogeneity. The Cochran Q test was used to evaluate whether the variation across studies was compatible with chance, and p < 0.1 was considered to indicate significant heterogeneity. The I 2 statistic was a quantitative indicator used to evaluate the percentage of total variance in prevalence estimates due to statistical heterogeneity rather than chance, or sampling error (I 2 > 75% indicates high heterogeneity, 51-75% indicates substantial heterogeneity, 26-50% indicates moderate heterogeneity, and ≤ 25% indicates low heterogeneity).
Results from included studies were grouped according to pre-specified study-level characteristics, and then they were compared using subgroup meta-analysis (for screening instrument cut-off or diagnostic criteria, geographic location, sample source and total NOS score) or random-effects meta-regression (for baseline survey year, sample size, age and sex). The difference between subgroups was examined using the Cochran Q test (p < 0.05 indicated statistically significant differences). To determine the influence of individual studies on the pooled prevalence estimates, sensitivity analyses were performed by serially repeating the meta-analysis after the exclusion of each included study. If the point estimate of the new pooled prevalence is outside of the 95% confidence interval of the original pooled prevalence, it can be determined that the study which has been excluded to get the new prevalence has an significant effect on the original pooled prevalence. Publication bias was evaluated using Egger's line regression test (p < 0.05 indicated statistically significant differences). Preferred Reporting Items for Systematic Reviews and Meta-analysis guidelines were strictly adhered to wherever appropriate [32].

Identification and characteristics of studies
In total, 54,005 unique citations were identified after an initial search, 53,771 of which were excluded after removing duplicate papers and screening titles and abstracts (Fig. 1). Then, the full text of 234 articles were reviewed, 74 of which [9,[28][29][30] were considered to be eligible and included in the systematic review and meta-analysis.
In the 74 studies, there were a total of 20,635 PLHA. The median number of participants in those studies was 185 (range: 28 to 4103). Sixty-seven studies were conducted in one of the seven areas (twenty-one in East China, seventeen in Central China, ten in South China, nine in Southwest China, six in North China, three in Northeast China and one in Northwest China), six studies were conducted in two or more areas and one study did not report the study site. The papers were published between 2004 and 2017, and more than 70% (54/74 studies) were published between 2011 and 2017. Seventy-one cross-sectional studies (n = 20,154) and three longitudinal studies (n = 481) reported on the prevalence of depression or depressive symptoms, and twenty-three of the seventy-three observational studies focused on specific sub-populations (ten on MSM, seven on FBPD, two on pregnant women, two on HIV-TB co-infected individuals, and two on IDUs). More details are shown in an additional table file (see Additional file 3).

Study quality
Modified NOS score components for all 74 individual studies are shown in Additional file 4 and Additional file 5. Fifty-seven studies (77.0%) had an overall rating of low risk, while the rest were rated as high. One-fifth of the studies scored 1 point on each of these five items. The overall sample representativeness was fair, as more than half of the studies (41, 55.4%) sampled PLHA from HIV-infected individuals databases of the provincial or municipal Center for Disease Control and Prevention or from multiple study sites. Forty-one studies reported responses of at least 70%, and of these more than 90% sampled 100 or more PLHA.
Depression or depressive symptoms among the general PLHA Estimate of overall prevalence of depression or depressive symptoms among the general population of PLHA The prevalence estimates of depression or depressive symptoms among the general PLHA reported by 50 included studies ranged from 18.3 to 86.9%. Meta-analytic pooling of these prevalence estimates yielded a crude summary prevalence of 50.8% (8023/14,824 individuals, 95% CI: 46.0-55.5%), with significant between-study heterogeneity present (I 2 = 96.4%, p < 0.001) (Fig. 2). No evidence of publication bias was detected using the Egger's test (t = − 1.549, p = 0.128). Sensitivity analysis showed that none of the studies had a significant influence on the pooled prevalence estimate (see Additional file 6).
To further characterize the range of prevalence estimates of depression or depressive symptoms, a stratified analysis was conducted, based on the screening instruments and cut-off scores used in these methodologically diverse studies (Table 1). Summary prevalence estimates of depression or depressive symptoms ranged from 18.3% (95% CI: 13.0-24.8%) for Psychological "Comput- Similarly, there were no significant differences in the prevalence estimates of depression or depressive symptom between studies with total NOS score < 3 points and studies with total ≥ 3 points (Q = 2.5, p = 0.117). Data are shown in Table 2.
The results of the random-effects meta-regression showed that the prevalence estimates of depression or depressive symptoms significantly varied with the baseline survey year (slope = − 8.

Heterogeneity within the depression survey instruments
To identify potential sources of heterogeneity independent of assessment method, stratified meta-analysis and univariate meta-regression analysis were conducted within subgroups of studies using the same instruments when at least five studies were available. An additional file shows this process in more detail (see Additional file 7). No significant differences were observed between community-based and hospital-based studies, as well as studies with total NOS score < 3 points and ≥ 3 points, within any instruments. Heterogeneity was partially accounted for by geographic location, as studies conducted in North China yielded lower depression or depressive symptoms prevalence estimates than studies conducted in Central China ( Table S1).
The baseline survey year significantly contributed to the observed notable heterogeneity among the studies using the Zung Self-Rating Depression Scale (Zung SDS), with a cut-off score of 50 or greater, and the 90-item Symptom Checklist (SCL-90), with a cut-off score of 2 or greater. Similarly, age also accounted for between-study heterogeneity within two instruments, Zung SDS score ≥ 50 and CES-D-20 score ≥ 16. Sample size also significantly contributed to the observed notable heterogeneity within three instruments (Zung SDS score ≥ 50, SCL-90 score ≥ 2 and CES-D-20 score ≥ 16), although the results were inconsistent (i.e., two analyses suggested that the prevalence estimate of depression was increasing with sample size, while a third one suggested that it was decreasing). Sex and ART did not significantly contribute to the between-study heterogeneity within any of the four instruments (see Additional file 7 Table S2).

Discussion
In the present systematic review and meta-analysis, we quantified the proportion of depression or depressive symptoms among PLHA using data from seventy-four studies involving 20,635 individuals in seven areas of China. On average, the pooled prevalence estimates were 50.8% for depression or depressive symptoms among the general PLHA. We also quantified these proportion among specific PLHA. As significant heterogeneity was detected across studies for all these prevalence estimates, the results must be interpreted with caution. To the best of our knowledge, this study represents the first time that the epidemic of depression among PLHA in China was exhaustively reviewed. As depression among PLHA is a public health issue, the risk of burden on human resources and the health care systems is considerable. The study could help to estimate the public health burden of depression among PLHA in China and to guide policy, as well as advocacy efforts. Furthermore, the study represents the first step in developing effective interventions to prevent and treat associated sequelae.
Evidence suggested that the prevalence of depression among the general population in China ranged between 1.2 and 6.9% [103,104], significantly lower than the prevalence rate reported in our study, which further confirmed that depression was an outcome conforming to logic among PLHA [105]. However, due to the common symptoms associated with HIV illness, such as pain, fatigue, insomnia, anorexia and cognitive impairment, it is difficult to diagnose depression among PLHA [106][107][108][109][110]. Based on a nationally representative sample, it is demonstrated that depression among PLHA is under-diagnosed in clinical practice in the United States [111]. Although there is no study on this issue in China, we can speculate that depression in the Chinese people with HIV/AIDS is also under-diagnosed in clinical practice because more than three-quarters of non-psychiatric clinicians in China lack adequate knowledge of depression [112], which has proven to contribute to the difficulty in identifying individuals with depression [113]. Moreover, a strong stigma against PLHA lead them to defer seeking health care services or to disclose their own HIV status to the health care workers [114], which is an additional obstacle to early detection and treatment of depression among PLHA. In fact, the serious shortness and uneven distribution of mental health resources are obstacles to directing adequate attention toward those health issues [115,116]. To improve the current situation, the National Heath and Family Planning Commission of the People's Republic of China issued the Nation Mental Health Program (2015-2020) [116] in 2015, in which a series of specific goals aimed at ultimately promoting public mental health have been proposed, including   general improvement of the public cognition of depression and other common mental disorders and the public awareness of forwardly seeking medical advice, as well as obvious improvement in ability of medical workers to identify depression. In our study, depression was found to be associated with the baseline survey time, on the decrease over time, even among some studies using common instruments. Economic development in the past decades may be a possible reason for this decrease in depression prevalence, which has greatly increased the investment of mental health as well as the availability of mental health services [115]. The growing awareness of AIDS-related knowledge among the public [117,118] helps to reduce discrimination against PLHA and, hence, may be conductive to decreasing the prevalence of depression.
Even today, no consensus has been reached on the impact of ART on depression among PLHA in China. Several studies [54,61,68,72,78,119,120] have reported a higher prevalence of depression or depressive symptoms among PLHA who have undergone ART when compared with those who have not, and only two studies have reported a statistical significance [119,120]. Nevertheless, some other studies [53,58,73,121] have found the prevalence to be lower in patients who had received ART than in those who had not, while only one study has reported statistical significance for this opposite result [121]. Due to the lack of data available on depression prevalence estimates among the people using and not using ART, stratified meta-analyses could not be conducted in these two sub-populations. Instead, random-effects meta-regression analysis was used to explore the relationship between ART and depression or depressive symptoms prevalence. As a result, there was no significant association between them.
Given the higher reported prevalence estimates of depression among females in the general population, females were considered to be more vulnerable than males to the onset of depression [122], a finding supported by evidence from studies conducted in PLHA populations. In a observational cross-sectional study conducted in central India recruiting a large sample of 1181 PLHA, Deshmukh et al. have found that a greater percentage of females was screened positive for depressive symptoms when compared with males (59.9% vs 43.7%, p < 0.001) [123]. In another cross-sectional study MSM, men who have sex with men; FBPD, former blood/plasma donors conducted in Nigeria, a significantly higher prevalence of major depressive disorders was reported among females than among males [124]. However, in a current study targeted on newly diagnosed HIV-patients, being female was found to be protective against depression but without significance (OR = 0.48, p = 0.078) [125]. In our study, no significant association between the prevalence estimates of depression or depressive symptoms and gender was determined. In addition, in the general population, age has also been proven to be associated with variations in the prevalence estimates of depression, with younger participants having higher prevalence of current and lifetime depression than participants older than 50 or 55 years [126]. However, the association between age and depression among PLHA remained unclear. The results showed that a younger age was significant associated with the higher prevalence of depression screened by CES-D-20 [127], as well as diagnoses by psychiatrists, according to DSM-IV-TR [128], even after adjusting for confounding factors. However, the study conducted in HIV-infected adults undergoing anti-retroviral treatment demonstrated that participants older than 50 years old had a two times higher risk of depression when compared with participants with between 18 and 30 years old [129]. In addition, the result from the study which used the Depression, Anxiety, Stress subscales, and full Scale (DASS-21) for depression screening showed that no significant effect of age on the rate of depression was found among PHLA [123]. In this meta-analysis, although no significant association was found between age and the pooled prevalence of depression, age was demonstrated to be linked with a higher risk of depression in studies with Zung SDS scores ≥50 or CES-D-20 scores ≥16 as the criteria for screening positive, which might support the positive association between age and depression among PLHA to some extent. However, as there were few studies using those two screening instrument cut-offs as screening criteria, the results must be interpreted with caution. Further studies are needed to clarify the associations between gender/age and the risk of depression among PLHA, which will help to identify individuals in high-risk.
When interpreting the results of this study, note that the data synthesized in this meta-analysis were nearly entirely extract from studies using self-report inventories of depressive symptoms as the survey instruments, which had a wide range of sensitivity and specificity for diagnosing major depressive disorder (Additional file 4). Instruments such as the Psychological "Computerized Tomography" 4.0 Vision (PCT V4.0) have high specificity and sensitivity for diagnosing depression, whereas others instruments, such as the SCL-90, have low specificity and should be regarded as screening tools. Furthermore, evidences suggest that screening tools tend to over-estimate prevalence relative to diagnostic tools, which may lead to an over-estimation of true rates in the meta-analysis with all the included studies relying on screening instruments. Despite the limitations in self-report inventories of depressive symptoms, these inventories are still essential for assessing depression in HIV-positive individuals because they are easier and more cost-effective for use in busy specialty medical clinics and epidemiological surveys than formal diagnostic interviews [130,131], particularly in epidemiological surveys. Because of the high prevalence in China, it is nearly impossible to assess depression through formal interviews between psychiatrists and HIV-positive individuals in epidemiological studies. As an alternative, self-report inventories are the best choice. Nevertheless, for primary care physicians, it is better to remember that the diagnosis of depression should not be based solely on the results of the screening questionnaire [132]. In this meta-analysis, to control the diversity in these inventories, stratified analyses were conducted based on survey instrument and cut-off scores that identified a range of prevalence estimates not presented in the previous review [133].
This study has important limitations. As with other meta-analyses, significant heterogeneity was found in the prevalence estimates in our study, which was incompletely explained by the stratified meta-analyses and meta-regressions analyses. We hypothesize that other variables might affect the heterogeneity, such as poor income adequacy, unemployment, homeless, lower CD4 counts, higher viral loads, the severity of depressive symptoms, duration of HIV/AIDS, poor self-efficacy and lack of social support. However, we were unable to obtain adequate information about these variables. For example, less than 1/3 of the studies reported the average or median counts of CD4 cells among HIV infected populations, and fewer than 10 studies provided employment-specific prevalence estimates of depression. In addition, although an extensive document retrieve was performed in multiple databases, the existence of non-indexed studies in those databases might have led to some relevant studies being ignored. Moreover, although an attempt was made to minimize the possible bias in the process of document retrieving with specific searches in major English-Chinese databases (including master and doctoral theses), there may still be some unidentified papers. Fortunately, as the results of Egger's test results showed, there was no publication bias found in all results because we obtained a certain percentage of data from unpublished papers (fourteen theses [38,56,57,66,70,71,75,82,87,89,91,94,96,99]).

Conclusions
Our findings suggest that the estimates of depression or depressive symptoms among PLHA in China are considerable. Given that the progression of depression are associated with a higher short-term suicide risk and a higher long-term risk of cardiovascular disease and cancer [134,135], the findings in this study highlight the need for screening and treatment for mental disorders to be integrated in the treatment package offered to PLHA, which will ultimately lead to better health outcomes for PLHA [136].