Search strategy
Medical literature on income levels and gestational diabetes mellitus were searched in the Pubmed, Web of Science, Cochrane library, Ovid databases, with search keywords ‘socioeconomic status’, ‘income’, ‘financial statement’ and ‘gestational diabetes mellitus’ from January 3, 2020. As a result, 566, 230, 6, and 2974 related studies were obtained from the above four databases. A total of 280 articles were included in the preliminary selection after filtering. Subsequently, two researchers with related background knowledge (He Qiong and Zhang Mengyuan) independently conducted the second-round literature screening. If there was any objection, a third researcher (Liu Yunfeng) would be asked for their views regarding article inclusion. After reading the titles, abstracts, and full text, 13 articles were finally included, and a total of 1,817,801 patients entered the network meta-analysis study (Figure 1).
Inclusion and exclusion criteria
According to the study purpose of this study, the specific inclusion criteria for the correlation between income levels and incidences of gestational diabetes were: i. A study population of pregnant women, including both primiparous and multiparous women; ii. Economic income levels including at least four levels of low, lower middle, medium, upper middle, and high income, stratified by quartiles, quintiles or numeric. In this study, income is defined as household income or maternal income. To make the income levels comparable between studies, currencies of other countries were converted to U.S. dollars based on the current exchange rates;c. Studies with a clear meaning of income level and diagnostic criteria for gestational diabetes; d. Studies in which the number of enrollment and the number of patients with gestational diabetes has been given, or can be calculated based on the data available in the article.
Exclusion criteria were: i.. If the study population involved the general social population, including both men and women; ii. Women with type 1, type 2 and other types of diabetes before pregnancy; iii. Patients with impaired glucose tolerance or other diseases that do not meet the diagnostic criteria for gestational diabetes; iv. Non-clinical research articles such as reviews, guidelines, or case reports.
According to the above inclusion and exclusion criteria, the quality of the selected documents was strictly controlled, and a total of 13 studies that met the conditions entered the statistical analysis stage. The information and characteristics of the included literature were summarized. The characteristics of each study including the author, year of publication, region, type of study, inclusion and exclusion criteria, diagnostic criteria for gestational diabetes, income level grading, total number of groups, number of gestational diabetes patients, and prevalence were extracted [see Additional file 1].
Literature quality assessment
According to the Cochrane Handbook for Systematic Reviews of Interventions, Revman 15.3 software was used to evaluate the quality of the 13 included clinical studies. Quality is mainly evaluated from seven aspects; random sequence generation, allocation concealment, blindness of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting of research results and other source biases. Quality evaluation is divided into three levels of low, unclear and high. The higher the quality of the article, the more reliable its conclusions. After comprehensive analysis of the 13 included articles, it was concluded that their quality was high, and subsequent statistical analysis could, therefore, be trusted. The results of the evaluation are shown in Figure 2.
Bias Evaluation
For each of the 13 studies included in this article, five different income levels were compared in pairs to evaluate publication bias, as shown in Figure 3. From the basic symmetry of the funnel graph, it was concluded that the possibility of small sample impact and publication bias in this study was low, the quality of included articles was high, and the statistical results data have clinical significance.
Overall inconsistency estimation
The overall inconsistency of the included literature was tested. The statistical result was p = 0.083 (>0.05). Therefore, the overall inconsistency of the 13 studies included in this article was not significant (Figure 4). There were reasons to believe that the consistency of the 13 articles was good, therefore, local inconsistencies were used to further evaluate the quality of the included 13 articles.
Local inconsistency test
Local inconsistency was tested through the node-splitting method. It was found that the p value in each comparison was greater than 0.05. Therefore, there was no significant local inconsistency in the network meta-analysis, and it can be tested by the consistency model (Figure 5).
Consistency evaluation
A total of 13 studies and 5 income levels formed a closed loop, and the closed-loop was tested for consistency (Figure 6). It was found that the closed-loop consistency was good.