Year: 2026
Type: article
Study design: Observational Study
Abstract: A latent Gaussian model (LGM) that follows a Gaussian Markov random field was developed for pregnancy risk data. Inference was done through INLA. Sensitivity analysis was done to adjust for underweight bias in BMI. Positive maternal Hepatitis B status (HBsAg) was the strongest risk factor, increasing the odds of high-risk pregnancy more than 300-fold (AOR = 313.69, 95% CI: 43.72–39,858.39). Elevated systolic blood pressure independently increased the odds by 4.3% per unit increase (AOR = 1.043, 95% CI: 1.025–1.061). In contrast, a negative blood or urine sugar test reduced the odds by 54.6% (AOR = 0.454, 95% CI: 0.285–0.723), while each unit increase in BMI within the normal-to-overweight range reduced the odds by 31% (AOR = 0.690, 95% CI: 0.634–0.743). LGM was comparable to classical logistic model (AUC ≈ 0.734; accuracy 70.54% vs. 70.44%). Prior sensitivity analysis demonstrated the robustness of LGM to prior specifications.
Authors: Ekong, A. H., Olawoore, S. A.
Institution: Federal Medical Centre
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7214757367