Searching Strategy
Initially, databases were searched to check for the presence of similar systematic review to avoid duplication using website https://www2.le.ac.uk/library/find/databases/p/Prospero. Searching of primary articles was conducted from PubMed, Hinari, Science Direct, and Cochrane library databases. Furthermore, grey literature was retrieved from Google and Google scholar. Furthermore, the reference lists of published articles were searched to recognize other relevant articles that did not shown in databases. During the search process, to decrease the number of unrelated studies, the search was restricted to only ‘human studies’, ‘women’, and ‘English language’ in the advanced search. The search for primary articles was started on June 26, 2020, and end on August 3, 2020. For searching purposes, we used “Undernutrition OR Underweight AND lactating mothers AND Ethiopia” for objective one and “Determinants OR factors OR predictors AND lactating mothers AND Ethiopia for the second objective as keywords. Both published and unpublished articles that fulfill the eligibility criteria were included in this systematic review and meta-analysis. During writing this review and meta-analysis we used the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines [21]. Articles were downloaded to Endnote version X7 to maintain and manage citations, facilitate the review process, and to check duplication of articles.
Eligibility criteria
Eligibility assessment was executed independently by BG and JN in an unblinded identical manner based on the stated inclusion and exclusion criteria. We solved disagreements by consensus and discussion with the two remaining authors.
Inclusion criteria
All observational studies (cross-sectional, case-control, and cohort studies) conducted in Ethiopia among lactating mothers and published in English were included. Moreover, articles reporting the prevalence of undernutrition (BMI <18.5kg/m2) and associated factors were included. Both published and unpublished full articles were considered. Both institutional and community-based studies were encompassed.
Exclusion criteria
Studies conducted among both lactating and pregnant mothers were excluded.
Outcome measures
This systematic review and meta-analysis have two objectives. The first was to estimate the pooled prevalence of undernutrition among lactating mothers in Ethiopia and it was calculated by dividing the number of lactating mothers with this problem to the total number of lactating mothers included in the study and multiplied by 100. All articles included in this review and meta-analysis used BMI score (<18.5kg/m2)to assess undernutrition among lactating mothers. The second objective was to determine the pooled effects of factors on undernutrition among lactating mothers in Ethiopia. In this systematic review and meta-analysis variables identified as a factor in two and above studies (articles) were considered to include. To express the pooled effects we used odds ratio (OR) that was calculated from the 2x2 table.
Quality assessment and data extraction
Newcastle Ottawa Scale adapted for cross-sectional studies was used to assess the quality of the included studies [22]. BG and JN have appraised the studies independently using the above tool. The tool has the following parameters sampling strategy, inclusion/exclusion criteria, sample size, cut-offs, and reference for the assessment of lactating women undernutrition status, criteria to identify undernutrition, and covariates included in statistical models. The tool comprised 10 criteria for rating different quality elements. After quality assessment studies with high quality (scored 6 and above out of 10) were included for analysis. During the quality assessment, any divergences were solved through discussion, by taking the average result of the two appraisers and by giving the decision for the remaining two authors.
All the necessary data were extracted using a standardized Microsoft Excel data extraction format by two authors (BG and JN) separately. We used two data extraction formats, one for each objective. For the prevalence of undernutrition the data extraction format comprised author name, publication year, region the study conducted, study design, sample size, response rate, outcome measurement tool, study quality score, and prevalence of undernutrition. We also used two by two tables to extract data for objective two (factors for Undernutrition). Any incongruities during the data extraction period between the two authors (BG and JN) were resolved through discussion, twofold checking the varying data together, and third author invitation.
Publication bias and heterogeneity
Publication bias was assessed by both methods, funnel plots that are the subjective method used to test for asymmetry [23], and Egger’s statistical test [24]. To declare the statistical significance of publication bias we used a p-value< 0.05. After a comprehensive examination of the included studies, heterogeneity of the studies was assessed by I2 test statistics. I2 statistics described the total variation across studies and declared as low, moderate, and high heterogeneity if it is < 50, 50–75%, and > 75% respectively [25].
Statistical method and analysis
We extracted important data from each study using a Microsoft excel spreadsheets and the data were exported to STATA software version 14 for analysis. The standard error of prevalence for each original article was calculated using the binomial distribution formula. The effect size of the meta-analysis was the prevalence of Undernutrition and OR of the associated factors. We used a random-effect model for analysis [26]. To check the source of heterogeneity we conducted a leave-one study-out sensitivity analysis and subgroup analysis [27–29]. The effect of the selected associated factors on the outcome variable was examined using separate groups of meta-analysis. To describe the features of the included articles and to display the finding of this review and meta-analysis we used texts, tables, forest plots, and OR and 95% confidence intervals (CI).