Association between pregnancy intention and late initiation of antenatal care among pregnant women in Ethiopia: a systematic review and meta-analysis

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist was used to report the review [24]. Literature was downloaded to Endnote (version X7.2) to maintain and manage citations and facilitate the review process [25].

At the beginning, we checked for the presence of the existing systematic review and meta-analysis on a similar topic using the trial registries and the Cochrane library to avoid duplication. A systematic search of electronic databases such as MEDLINE, Scopes, PubMed, CINAHL, PopLine, MedNar, Cochrane library, the JBI Library, the Web of Science, and Google Scholar was done. All studies were searched using the key terms “antenatal care, late initiation, delayed initiation, early booking of ANC, pregnancy intention, unintended pregnancy, determinant factors, Ethiopia.” Addis Ababa University Digital Library and African digital library were searched to find unpublished papers [26‐29]. The search was conducted from September 15, 2019, to October 30, 2019, among studies conducted until October 30, 2019. All authors participated in the search of the literature.

The pre-defined search terms were used to allow us a comprehensive search of important studies included in our review. All fields within records and Medical Subject Headings (MeSH terms) were used to help enlarge the search in advanced PubMed search. The following search strategies were adapted for the various databases using the two important Boolean operators and search engines with initial search terms “delayed initiation” OR “late initiation” OR “early initiation” AND “associated factors” OR “determinant factors” OR “predictors” AND “Antenatal care” [MeSH Terms] AND “pregnant women.”

The primary outcome variable was the association between pregnancy intention and late initiation of ANC. Pregnancy intention was measured by asking a woman’s desire about her pregnancy at the time she was pregnant (intended/unintended pregnancy). Accordingly, intention to become pregnant was categorized as intended and unintended pregnancy. Intended pregnancy is when the mother indicated that she wanted to become pregnant on their last pregnancy. Unintended pregnancy is classified into mistimed and unwanted. Both show a pregnancy that had not been wanted at the time conception occurred. Mistimed pregnancies are those pregnancies that were wanted by the woman at some time, but which occurred sooner than they were wanted, and unwanted pregnancies are all pregnancies that occurred when the woman did not want to have any more pregnancies at all [30, 31]. For this review, delay or late initiation of the first ANC visit is an initiation of the first ANC visit in public health facilities by skilled health personnel after 12 weeks of gestation in some studies and after 16 weeks of gestation in others (Table 1).

The Joanna Briggs Institute Meta-Analysis of Statistics Assessment and Review Instrument (JBI-MAStARI) was used for quality appraisal [25] (S1 Table). The quality of all the included studies was assessed by three authors (TT, GF, and FK). Any disagreement between the three authors was resolved by involving the other two authors (GF and ET) to independently assess the methodological quality to reach a consensus. Data were extracted using a standardized data extraction checklist on Microsoft Excel. Initially, duplicated literature was excluded by using Endnotes reference management. Then, studies were excluded by their titles. Next, studies that were found to be irrelevant to our study were screened by their abstracts. Full-text articles or reports were assessed for the remaining literature. Based on the preset inclusion and exclusion criteria, eligibility of the studies was evaluated. For the outcome variable (association between pregnancy intention and late initiation of ANC), data were extracted in a format of two by two tables, and then the log OR was calculated based on the findings of the original studies. The checklist for data extraction contains the title, author name, year of publication, region (the area where the study was conducted), study design, sample size, response rate, and the number of participants with the outcome (Table 1). When articles did not have adequate data, the corresponding authors of the research articles were contacted through their email.

Cochrane Q test (chi-squared statistic) and I² test statistic on forest plot were used to check heterogeneity among the included studies. Cochrane’s Q statistical heterogeneity test is considered as statistically significant at p ≤ 0.05. I² values of 0, 25, 50, and 75% were considered as no, low, moderate, and high degrees of heterogeneity, respectively [32]. Subgroup analysis was conducted based on the definition of ANC given by different studies. Funnel plot asymmetry was used to check publication bias. In addition, Egger’s weighted regression test was used to check publication bias [33]. A p value of less than 0.05 was used to declare the statistical significance of publication bias.

The necessary information was extracted from each original study by using a format prepared in a Microsoft Excel spreadsheet. Then the data were exported to STATA for Windows version 14 and used to calculate the pooled effect size with 95% confidence intervals of pregnancy intention on late initiation of ANC using the DerSimonian and Laird random effects meta-analysis (random effects model). The logarithm and standard error of the odds ratio (OR) for each included study were generated using “generate” command in STATA. Meta-regression was conducted to identify the source of heterogeneity, and statistically significant results were declared in the presence of heterogeneity.

A systematic search of electronic databases and library catalogs identified a total of 677 published articles and three unpublished studies. Of the total published and unpublished identified studies, 97 duplicates papers were removed and 553 records were removed by reviewing titles and abstracts. The abstracts and full text of the remaining 30 studies were assessed and screened for eligibility criteria based on the outcome variable (late initiation of ANC) and 16 studies were excluded because they failed to meet the eligibility criteria (poor quality based on the selection criteria, their regression tables did not include intention of pregnancy as an independent variable).

Finally, 14 articles which scored 7 and above on the JBI quality appraisal eligibility criteria were included in the systematic review and meta-analysis. We used the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram to present the systematic review overview (Fig. 1).

Of the total included studies, 13 studies were cross-sectional [12, 29, 34‐44] and only one study was a case-control study design [45]. They had an overall of sample size 6390 pregnant women in Ethiopia, and the study sample sizes ranged from 256 to 997 participants. The studies were conducted from 2013 to 2019 in various regions of the country. Of the fourteen studies included in the final analysis, six studies were conducted in the SNNP region [34‐36, 38, 42, 43], four were conducted in the Amhara region [12, 37, 39, 44], and three were conducted in Addis Ababa [29, 41, 45]. The remaining study was conducted in Tigrai region [40].

The findings of original studies showed varied and inconclusive association between the pregnancy intention and delayed initiation of ANC. The result was a significantly positive and negative association in some studies and insignificant in other studies. Of those studies that showed a significantly positive association, the strongest association was observed in the study conducted in the Tigrai region [40] with an odds ratio of 6.77 (CI 1.98–23.15). The smallest odds ratio for the effect of pregnancy intention on late initiation of ANC (0.74) was reported in the study conducted in the Amhara region [39] (Table 1).

In this meta-analysis, we found significant heterogeneity across studies (I² = 73.5%, p = 0.00), which is an indicator to use random effects model to estimate the pooled effect of pregnancy intention on the delayed initiation of ANC reported by the fourteen studies with inverse variance.

This meta-analysis found that pregnancy intention had statistically significant effects on delayed initiation of ANC. The finding showed that there is an increased odds of delayed antenatal care among women with unintended pregnancies as compared to women with intended pregnancies (OR 2.16, 95% CI 1.62, 2.88) (Fig. 2).

To identify the possible sources of heterogeneity, different factors associated with the heterogeneity such as sample size and year of publication were computed using meta-regression models, but all of these variables were not found to be statistically significant. The presence of publication bias was assessed using a funnel plot and Egger’s test at a 5% level of significance (Fig. 3). Although visual examination of the funnel plot shows it to be asymmetric, Egger’s test was insignificant for the presence of publication bias (p = 0.28) (Table 2).

Sensitivity analysis of the studies was done to test the effect of a single study on the pooled result of the remaining studies using random effect model. We found no strong evidence for influence of individual study on remaining studies (Fig. 4).

Subgroup analysis was conducted based on the definition of late initiation of ANC given by different studies. Accordingly, the odds of delayed ANC initiation among women with unintended pregnancy was 2.10 times higher as compared to women with intended pregnancy with studies used 12 weeks definition of delayed initiation of ANC (OR 2.10, 95% CI 1.23, 3.24). The odds of delayed ANC initiation women with unintended pregnancy was 2.33 times higher as compared to women with intended pregnancy with studies used 16 weeks definition of delayed initiation of ANC (OR 2.33, 95% CI 1.63, 3.32) (Fig. 5).