Untitled

Abstract

We conducted a meta-analysis of data available from studies on determinants of intermittent preventive treatment (IPTp) and insecticide treated net (ITN) use in pregnancy. Odds ratios (ORs) were used to represent effect size and data were extracted from studies in the following hierarchy: raw data (numbers) were used in preference to unadjusted ORs and 95% CI; unadjusted ORs were used where raw numbers were not provided; adjusted ORs were used if unadjusted ORs or raw counts were not provided. Several studies reported a cluster design, but it was not always clear if the design was taken into account when the results were presented. For this reason the use of a cluster design as provided by the studies is indicated in the meta-analysis forest plots and notes, in addition to the use of adjusted (by multivariate analysis) ORs. If studies contributed data for "1+ doses" and "2+ doses" of IPTp, the data for "1+ doses" was removed to avoid overrepresentation of that study population in the analysis. A random effects model was used to provide a summary OR estimate of effect using the approach by DerSimonian and Laird. The random effects model recognises that the studies are a sample of all potential studies and incorporates a between study component to the estimate of variability. We present the ORs with 95% confidence intervals (CI), the I 2 and 95% CI to describe the percentage of the variability in the summary OR that is attributable to between study heterogeneity and the P-value for the Z-