Uncovering distinct motor development trajectories in infants during the first half year of life

Infants undergo significant developmental changes in the first few months of life. While some risk factors increase the risk of developmental disability, such as preterm birth, the developmental trajectories of infants born pre-term (PT) and full-term (FT) present with individual variability. This study aims to investigate whether the utilization of data-driven unsupervised machine learning can identify patterns within groups of infants and categorize infants into specific developmental trajectories. Thirty-four infants, 19 FT and 15 PT, were assessed with the gross and fine motor subscales of the Bayley Scales of Infant and Toddler Development, version III (BSID-III) monthly for 2-5 visits between the ages of 1 and 6 months. Latent class growth analysis (LCGA) models were adopted to identify clusters of motor developmental trajectories during this critical time. Based on statistical significance, the linear, 2-class trend was selected as the best-fitting model for both gross and fine motor trajectories. Within this, LCGA reveals 2 developmental trends with varying beginning scores and developmental rates, including the low-baseline slow-growth (LBSG) subgroup, and the high-baseline fast-growth (HBFG) subgroup, with age (adjusted for prematurity) being equally distributed across both subgroups. Both subgroups, HBFG and LBSG, had a combination of infants born FT and PT (55% FT in HBFG, 56% FT in LBSG), supporting that preterm birth alone may not sufficiently categorize an infants developmental trajectory. The later BSID-III gross motor score showed marginal difference between groups (p = 0.062). Similarly, the fine motor model displayed a mixture of both infants born FT and PT (68% FT in HBFG, 40% FT in LBSG). In this case, the late motor composite BSID score was different between groups (p = 0.04). Our study uses a novel approach of LCGA to elucidate heterogeneous trajectories of motor development for gross and fine motor skills during the first half of life and offers potential for early identification of subgroup membership. Furthermore, these findings underscore the necessity for individualized risk assessments and intervention strategies tailored to individual needs. Ultimately, further validation of these models may provide usefulness in uncovering distinct motor development trajectories in infants. Author NoteThe authors would like to thank the infants and families who participated in these study, and Katy Kelley, Carolina Panceri, Judy Zhou, and Dana Fine for assisting with recruitment and data collection. Thank you to Eisner Health, Childrens Hospital Los Angeles, and Ventura County Medical Center for assistance with recruitment. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The studies secondarily analyzed in this work were supported by grants from the Bill & Melinda Gates Foundation [OPP1119189] (PI: B.A.S.) and the Eunice Kennedy Shriver National Institute of Child Health & Human Development of the National Institutes of Health under Award Number R03HD096137 (PI: Smith). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Study data were collected and managed using REDCap electronic data capture tools hosted at the Southern California Clinical and Translational Science Institute at the University of Southern California NIH grant UL1TR001855 (PI: T. Buchanan). REDCap (Research Electronic Data Capture) is a secure, web-based application designed to support data capture for research studies, providing (i) an intuitive interface for validated data entry; (ii) audit trails for tracking data manipulation and export procedures; (iii) automated export procedures for seamless data downloads to common statistical packages; and (iv) procedures for importing data from external sources.