Selection and identification of studies
In the search process, we identified 52 published studies. Of those retrieved, 22 duplicate studies were removed and 8 studies were excluded after reviewing of their titles, abstracts and design based on the settled inclusion criteria. The remaining 22 studies were screened for full text for eligibility and whether they report outcome of interest, of which 7 studies were excluded due to lack of outcome of interest and 4 studies were excluded because they failed to meet eligibility criteria. After quality assessment with JBI eligibility criteria, 11 studies that scored seven and above were included in the final meta-analysis.
(Figure 1).
Characteristics of included studies
All included studies were cross sectional studies that directly or indirectly estimated the prevalence of food insecurity and gender disparity among adult people living with HIV in Ethiopia. A total of 11 studies were included in the review with a total of sample of 4031 individual living with HIV. It was found that the sample size of each study was almost similar. The range of sample size was between 305 and 452 participants in the studies conducted in Butajira of SNNPR Region[6] and Dembia of Amhara Region respectively [14]. The studies were conducted from 2012 to 2017 in different regions of the country.
Of the total 11 studies, two studies [10, 14] were conducted in Amhara Region, five studies [6-9, 16] in SNNPR Region, three studies [4, 5, 11] in Oromia Region, and one study [15] conducted in Tigray Region. Six studies[5, 7, 9-11, 16] were conducted directly to assess food insecurity among adults with HIV while the remaining five studies[4, 6, 8, 14, 15] were conducted to assess under nutrition, but indirectly assessed the prevalence of food insecurity as a variable (Table 1).
Prevalence of food insecurity among adult people living with HIV and receiving antiretroviral therapy
Our review of 11 included studies showed that the lowest prevalence of food insecurity among adult people living with HIV, receiving antiretroviral therapy was 18.36% in the study conducted at Dembia District health facilities [14] of Amhara Region, with the highest prevalence of 92.82% in the study conducted at Fitche Hospital [4] of Oromia Region (Table1). The analysis indicated significant heterogeneity across studies (I2=99.5%, p <0.001) which suggest that the use of a fixed effect model might lead to unreliable estimates. Statistically, fixed-effects model is not usually recommended, since it assumes that all the heterogeneity can be explained by the covariates and such assumptions could further lead to excessive type I errors, when there is residual, or unexplained, heterogeneity.
Therefore, we used a random effects model to estimate the pooled prevalence of food insecurity among adult people living with HIV, receiving antiretroviral therapy reported by the 11 studies. The overall pooled prevalence of food insecurity among adult people living with HIV, receiving antiretroviral therapy in Ethiopia was 63.44 % (95% CI (46.33% - 80.54%)).
(Figure 2).
To identify the source of heterogeneity, we further assessed the heterogeneity using different statistical techniques. We performed meta-regression model using publication year and sample size as covariates by specifying the method for estimating the between-study variance. The analysis indicated that none of the two variables were statistically significant for explaining the presence of heterogeneity (Table 2).
We assessed the presence of Publication bias using funnel plot and Egger and Begg statistical tests at 5% significant level. There was no statistical evidence of publication bias. The funnel plots were symmetry, the Begg and Egger tests were not statistically significant with p-value = 0.213 and p-value = 0.599 respectively.
(Figure 3).
We also performed sensitivity analysis and the result of sensitivity analysis suggested that no strong evidence for the influence of single study the overall prevalence of food insecurity among adult people living with HIV receiving antiretroviral therapy.
(Figure 4).
Gender disparity pooled prevalence of food insecurity
From the 11 studies included in the overall meta-analysis, only 4 studies reported gender disparity as outcome variable of interest. Therefore, a total of 4 studies with 1436 participants were included to assess gender disparity in prevalence of food insecurity among adult people living with HIV[7, 9, 11, 16]. The finding of meta-analysis revealed significant gender disparities in the prevalence of food insecurity among adult people living with HIV. The findings showed a higher pooled prevalence of food insecurity among female (68.85%, 95% CI: 57.88%, 79.81%), I2=91.5%, p<0.001) compared to males (32.24%, 95% CI: (20.19%, 42.12%), I2=91.5%, p<0.001).
We further assessed the heterogeneity using different statistical techniques to identify the source of heterogeneity. We performed meta-regression using the same variable with general prevalence (publication year and sample size) as covariates by specifying the method for estimating the between-study variance. The analysis indicated that none of the two variables were statistically significant for explaining the presence of heterogeneity (Table 3).
We assessed the presence of Publication bias using funnel plot and Egger and Begg statistical tests at 5% significant level. There was no statistical evidence of publication bias. The Begg and Egger tests were not statistically significant with p-value = 0.734 and p-value = 0.480 respectively. We also performed sensitivity analysis and the result of sensitivity analysis suggested that no strong evidence for the influence of single study on the gender disparity prevalence of food insecurity among adult people living with HIV receiving antiretroviral therapy.