Stata 12.0 statistical

All statistical analyses were performed by Meta-DiSc and STATA 12.0 statistical software [22] (link). All accuracy data from each study (true positives, false positives, true negatives and false negatives) were extracted to obtain pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), positive predicted value, negative predicted value, diagnostic odds ratio (DOR) and their 95% confidence interval [95% CI], simultaneously, generate the summary receiver operator characteristic (SROC) curve and calculate the area under the curve (AUC). The sensitivity, specificity, positive and negative predicted value, diagnostic odds ratio of miR-92a were presented as forest plots. Moreover, the heterogeneity between the studies caused by threshold effect was quantified using Spearman correlation analysis. The Non-threshold effect was assessed by the Cochran-Q method and the test of inconsistency index (I²), and a low p-value (≤0.05) and high I² value (≥50%) suggest presence of heterogeneity by caused Non-threshold effect. If the Non-threshold effect existed, meta-regression would be used to find out the sources. For publication bias, all eligible studies were assessed by Begg’s test and Egger’s test using STATA 12.0 statistical software. The P value with less than 0.05 shows a result of statistical significance.

We calculated age-and gender-adjusted prevalence rates for each study variable according to the place of residence and estimated perceived discrimination prevalence by sex, age, ethno racial self-classification and educational attainment using Poisson regression. To verify differences among categories, we used the proportion test, considering a level of significance of 0.05. Use of Poisson models was justified because our outcomes were relatively common (prevalence > 10%), especially the main one (any discrimination) [25 (link)]. To verify the univariate and multivariate associations between “any discrimination” and the independent variables, we fit Poisson regression to estimate the prevalence ratios (PR) and their respective 95% confidence interval (95% CI). The multivariate model was tested including all independent variables. To examine whether the place of residence had a different effect on race, we include an interaction term. Those variables with a p-value > 0.05 were removed. From the multivariate model, we plotted the predicted probabilities of reporting any discrimination.
Data analysis was performed using Stata 12.0 statistical program [26 ]. Through the svy command, we took into account the sample design, individual weights and aggregation.

Descriptive analysis was used for the characteristics of the study population. To study how independent variables were interrelated, we used the chi-squared test and Spearman correlation. A multinomial logistic regression analysis was used to study associations of the dependent variable (with three categories comparing ‘1-9 days sick leave’ and ‘10 and more days sick leave’ with no sick leave) with socio-demographic, type of work and lifestyle-related factors. The odds ratio (OR) was estimated as measure of association with corresponding 95% confidence intervals (95% CI). In order to study the influence of educational level, type of work, stress, lifestyle factors, and risk for CHD on the association between gender and sick leave, these factors were added separately to the basic statistical model describing the association between gender and sick leave. All analyses were carried out with the STATA 12.0 statistical software package.