Exploratory analysis of machine learning approaches for surveillance of Zika‐associated birth defects

Abstract

In 2016, Centers for Disease Control and Prevention (CDC) established surveillance of pregnant women with Zika virus infection and their infants in the U.S. states, territories, and freely associated states. To identify cases of Zikaassociated birth defects, subject matter experts review data reported from medical records of completed pregnancies to identify findings that meet surveillance case criteria (manual review). The volume of reported data increased over the course of the Zika virus outbreak in the Americas, challenging the resources of the surveillance system to conduct manual review. Machine learning was explored as a possible method for predicting case status. Ensemble models (using machine learning algorithms including support vector machines, logistic regression, random forests, k-nearest neighbors, gradient boosted trees, and decision trees) were developed and trained using data collected from January 2016-October 2017. Models were developed separately, on data from the U.S. states, non-Puerto Rico territories, and freely associated states (referred to as the U.S. Zika Pregnancy and Infant Registry [USZPIR]) and data from Puerto Rico (referred to as the Zika Active Pregnancy Surveillance System [ZAPSS]) due to differences in data collection and storage methods. The machine learning models demonstrated high sensitivity for identifying cases while potentially reducing volume of data for manual review (USZPIR: 96% sensitivity, 25% reduction in review volume; ZAPSS: 97% sensitivity, 50% reduction in review volume). Machine learning models show potential for identifying cases of Zika-associated birth defects and for reducing volume of data for manual review, a potential benefit in other public health emergency response settings.

AIGDR: An Artificial Intelligence Framework for Disability Research – A Multiple Sclerosis Case

“…The predictors of visual-motor integration in children with intellectual disability have also been identi ed using AI (35). Other research areas in disability include prediction of hospital-associated disability (36), prediction of on-road driving ability in healthy older people (37), prediction of swallowing-related quality of life of the elderly living in a local community (38), determination of whether a patient has any geriatric syndromes (39), identi cation and characterization of cognitive subtypes within the atrial septal defect population (40), identi cation of cases of Zika-associated birth defects (41), and classi cation of samples of speech produced by children with developmental disorders versus typically developing children (42,43).…”

“…For example, much of the existing research (13,21,23,29,31,37,38) has used support vector machines (SVMs), a supervised ML method. Linear regression (LR), which is also a form of supervised learning, is another of the most widely used methods (13,14,23,33,34,35,36,41). Other supervised learning algorithms that have been used include random forest (RF) (13,14,32,36,40,41,43) and decision tree (DT) (13,21,41) algorithms.…”

“…Linear regression (LR), which is also a form of supervised learning, is another of the most widely used methods (13,14,23,33,34,35,36,41). Other supervised learning algorithms that have been used include random forest (RF) (13,14,32,36,40,41,43) and decision tree (DT) (13,21,41) algorithms. Other work has implemented k-nearest neighbors (KNN) (13, 30, and 41), adaptive boosting (13), extreme learning machines (ELMs) (30), gradient-boosted trees (41), linear discriminant analysis (LDA) (42), and natural language processing (NLP) (39,44).…”

Feature based analysis of endometriosis using machine learning

“…Decision tree (C5.0) and C4.5 was used for evaluation. C4.5 algorithm outperforms well in 9.33 seconds and 94.15% of accuracy [7].…”

Leveraging automated approaches to categorize birth defects from abstracted birth hospitalization data

“…SET‐NET contains tens of thousands of birth outcomes for pregnant people exposed to COVID‐19, hepatitis C, or syphilis, which would require a large amount of time from analysts to prepare the data for review and from clinicians to manually review each outcome to synthesize and categorize individual birth defect findings for dissemination. Machine learning algorithms have previously been shown to accurately predict manual review by clinicians of the classification of Zika‐associated birth defects and autism cases in surveillance data, and automated approaches have the potential to improve the timeliness of those data to inform clinical and public health action (Lee et al, 2019; Lusk et al, 2020).…”