Characteristics of the included studies
Studies of 434 were retrieved from PubMed, Scopus, HINARI, Google Scholar, African Journals, WHO’s Global Health Library, and other gray and online repositories accessing articles regarding the prevalence and determinant factors of cervical cancer screening in Ethiopia. After duplicates were expunged, 268 studies remained. Out of the remaining articles, 106 articles were excluded after review of their abstracts and titles. Therefore, 62 full-text articles were assessed and assessed for inclusion criteria, which resulted in the further exclusion of 54 articles primarily due to inaccessibility of full text and outcome of interest were not reported. As a result, 8 studies were met the inclusion criteria to undergo the final systematic review and meta-analysis (Fig. 1).
Study characteristics
Different factors such as poor/fair antiretroviral treatment adherence, knowing baseline CD4 level, disclosure of HIV status to partners, experienced drug side effects, residing in rural areas, and educational level were included in this study. Eight studies with a total of 5,854 study participants were included in this review. All included articles were facility-based study settings. Regarding the study area, four of the studies were conducted in Amhara region, three in Oromia, and one in Tigray region (Table 1).
The Magnitude of loss to follow-up in HIV care
The pooled magnitude of loss to follow-up among pregnant and lactating mothers living with HIV and enrolled in HIV care clinics in Ethiopia is presented with a forest plot (Fig. 2). Therefore, the national estimated magnitude of loss to follow-up in Ethiopia was 14.28% % (95% CI: 10.12–18.44, I2 = 91.9%, P < 0.001).
Publication bias
The funnel plot was assessed for asymmetry distribution of prevalence of lost to follow-up among pregnant and lactating mothers living with HIV and enrolled in HIV care clinics in Ethiopia by visual inspection (Fig. 3). Egger's regression test showed a P value of 0.869 with no evidence of publication bias.
Sensitivity analysis
This systematic review and meta-analysis showed that the point estimate of its omitted analysis lies within the confidence interval of the combined analysis. Therefore, the trim-and-fill analysis was not further computed (Fig. 4).
Subgroup analysis
Subgroup analysis was employed with evidence of heterogeneity. In this study, the Cochrane I2 statistic was 91.9%, P < 0.001, which showed evidence of marked heterogeneity. Therefore, subgroup analysis was done using the study region and sample size. As a result, the prevalence of loss to follow-up among pregnant and lactating mothers living with HIV and enrolled in HIV care clinics was highest in Amhara 14.8%, whereas 14.48% in the study conducted with a sample size > 334(Figs. 5 and 6).
Determinants of lost to follow-up in Ethiopia
Relationship between women's educational level and lost to follow-up
The risk of LTFU among women who had no education was 4.28 times (OR = 4.28, 95% CI: 3.62–5.00) higher than women who had secondary and above education level. In this meta-analysis, the included studies were characterized by the existence of no heterogeneity (I2 = 0.0%, P = 0.826) (Fig. 7).
Relationship between women's residency and lost to follow-up
Likewise, the loss to follow-up among women who were residing in rural areas was 2.66 times higher than women who were residing in urban areas (OR = 2.66, 95% C: 2.34–3.03). The included studies were characterized by the existence of no heterogeneity (I2 = 0.0%, P = 0.503) (Fig. 8).
Relationship between experienced drug side effects and lost to follow-up
The loss to follow-up among women who experienced antiretroviral drug side effects was 7.91 times higher than women who don't experience antiretroviral drug side effect s (OR = 7.91, 95% CI: 2.94–21.33). The included studies were characterized by the existence of high heterogeneity (I2 = 96.4%, P <0.001) (Fig. 9).
Relationship between disclosure of HIV status to partners and lost to follow-up
The risk of loss to follow-up among women who did not disclose their status to their sexual partners was 2.69 times higher than women's disclose their status (OR = 2.69, 95% CI: 2.08–3.48). In this Meta-analysis and systematic review, the included studies were characterized by the existence of high heterogeneity (I2 = 94.8%, P <0.001) (Fig. 10).
Relationship between poor treatment adherences and lost to follow-up
The risk of loss to follow-up among pregnant and lactating women who were on ART and had Poor/fair antiretroviral treatment adherence was 2.69 times higher than women who had good adherence (OR = 3.68; 95% CI: 2.89–4.69). In this Meta-analysis and systematic review, the included studies were characterized by the existence of moderate heterogeneity (I2 = 77.4%, P =0.012) (Fig. 11).
Relationship between knowing baseline CD4 level and lost to follow-up
The risk of loss to follow-up among women who do not know their baseline CD4 level was 3.68 times higher than women who know their status (OR = 3.68; 95% CI: 2.40–5.63). In this Meta-analysis and systematic review, the included studies were characterized by the existence of high heterogeneity (I2 = 97.5%, P <0.001) (Fig. 12).