Skewed logit model for analyzing correlated infant morbidity data

Area of Science:

• Epidemiology
• Biostatistics
• Maternal and Child Health

Background:

• Infant morbidity serves as a global health care indicator.
• Bacterial vaginosis (BV) is strongly associated with infant morbidity.
• Imbalanced data in morbidity severity prediction necessitates advanced modeling.

Purpose of the Study:

• To compare standard logit and skewed logit models for analyzing longitudinal imbalanced binary data.
• To investigate the longitudinal effect of maternal BV on infant morbidity in HIV-positive mothers.
• To highlight the utility of skewed logit models in epidemiological research.

Main Methods:

• Utilized a Kenyan dataset of infants born to HIV-positive mothers screened for BV.
• Derived a morbidity incidence score based on reported illnesses.
• Employed Generalized Estimating Equations to fit standard binary logit and skewed logit models, adjusting for covariates.

Main Results:

• Accounting for skewness in imbalanced binary data confirmed expected associations.
• Maternal BV was significantly associated with increased infant morbidity over time.
• The skewed logit model provided insights consistent with existing literature.

Conclusions:

• Maternal BV status is positively associated with infant morbidity incidence.
• Early intervention is crucial for HIV-infected pregnant women with BV.
• Skewed logit modeling offers a valuable approach for analyzing imbalanced longitudinal health data.