Accuracy of clinical risk factor-based models as a screening test for detecting gestational diabetes mellitus in a low-resource setting.

Current World Health Organization guidelines recommend fasting 2-hour tests for all pregnant women, a strategy that is burdensome for patients as well as time and labor-intensive for health systems. There have been suggestions for the use of clinical risk factors-based models as alternatives. These have not been widely tested especially in low-resource countries. We aimed to determine the prevalence of GDM and the accuracy of clinical risk factor-based models as screening tools for detecting GDM. This was a prospective cohort study of consenting 400 pregnant women receiving antenatal care at a tertiary health facility in Lagos, Nigeria. All the study subjects were assessed for the risk of GDM using three different clinical risk-based models. They also had universal screening for GDM at 24-28 weeks gestational age using the gold standard 2-hour 75g Oral Glucose Tolerance Test (OGTT). Statistical analysis was done using the statistical package for social science version 24. The Receiver Operating Curve (ROC) was used to determine the accuracy of the risk factor-based models. The mean age of the subjects was 31.0 {+/-} 5.3 years. A total of 76 subjects met the IADPSG/WHO 2013 criteria giving a prevalence of 19.0%. According to the clinical risk scores by Naylor et al, Caliskan et al and Phaloprakarn et al, 340 (85%), 269 (67.3%) and 375 (93.8%) participants respectively had a risk score positive for GDM. If the study participants were selectively screened based on these models, between 71.1-96.1% of the women with GDM would have been identified and 6.3-32.8% of the women would not have performed the diagnostic test. The models had areas under the ROC that ranges between 51.6-52.9%. Conclusions: The prevalence of GDM is high and the clinical risk factor-based prediction models tested in this study could be used to stratify low-risk women out of diagnostic tests.