The dataset shows the results of individuals who were tested for pregnancy, diabetes, and hypertension. However, individuals without a test could potentially have been misclassified as ‘negative’. To quantitatively assess for this kind of systematic error, a sensitivity analysis is recommended [20 (link)]. Using a misclassification spreadsheet [21 ], results from the regression analysis were explored for exposure bias. The misclassification spreadsheet provides adjusted bias data based on observed data on pregnant women stratified by dengue severity. As suggested for best practices [22 (link)], pairs of sensitivity and specificity were explored to study exposure misclassification.

Spatial analysis

The sums of severe and non-severe dengue cases were calculated by pregnancy status. The spatial distribution of severe and non-severe dengue was visualized and compared across pregnancy status using ArcGIS. To measure spatial autocorrelation, Moran’s Index (I) was calculated for both pregnant and non-pregnant women with severe dengue. Moran’s I for the attributes pregnant (0.010, p = 0.13) and non-pregnant women (0.003, p = 0.54) were both not statistically significant, and the spatial distributions of these two attributes were random.
Free full text: Click here