Skip to main content
SMFM 2025 Pregnancy Meeting Main banner
Final Program


Program Book
Icon Legend
Presentation Icons
Additional registration fee required
Additional registration fee required
CME Credit Offered
CME Credit Offered
Abstract Award
Abstract Award
Recording
Recording
Poster Icons
Additional registration fee required
Additional registration fee required
Faculty have requested this content not be shared on social media
Faculty have requested this content not be shared on social media
CME Credit Offered
CME Credit Offered
Abstract Award
Abstract Award
Back

Poster Session 1

(109) Leveraging Machine Learning to Identify Predictors of Passive and Active Influenza Immunity in Infants

Thursday, January 30, 2025
10:30 AM – 12:30 PM  
Objective: The optimal conditions for inducing durable infant immunity to influenza virus are unclear. This study aims to identify predictors of infant serum influenza antibodies using unsupervised machine learning. 

Study Design: We enrolled infants for serial infant capillary blood collection. Samples collected at 3-5 months of age (prior to pediatric influenza vaccine recommendation) were used to study passive immunity from maternal antibody transfer. Samples collected at 6-12 months (age when pediatric vaccines recommended) were used to evaluate infant immunity after vaccination or natural infection. Serum IgG antibodies specific to influenza A and B virus were quantified using an electrochemiluminescence assay. Thirteen covariates, including maternal and infant health and demographic data, vaccine and infection history were obtained by self-report. Continuous variables were imputed using a random forest imputer and normalized with the Yeo-Johnson method. Machine-learning models (Random Forest, boosted tree, RIDGE, LASSO linear regression) were compared using mean squared error values, and predictors were identified with the best-performing model based on the mean decrease in impurity. 

Results: Twenty-three infant serum samples at 3-5 months and 57 serum samples from 6-12 months were analyzed. The Random Forest model had the lowest mean squared error. Top p</span>redictors for passive immunity included: infant age, days since the last maternal influenza vaccination/infection, maternal age, and infant weight. Predictors for later infant immunity included: days since the last infant influenza vaccination/infection, total number of infant vaccinations/infections, infant age, infant weight, and days since the last parent vaccination/infection. 

Conclusion: Unsupervised machine-learning techniques can be utilized to identify clinical factors for early and late infant immunity. Future studies should integrate more longitudinal samples to develop a more comprehensive understanding of flu immunity and evaluate for maternal antibody inhibition. 
Back to Top