Statistical methods for predicting individualized intervention effects with clustered and longitudinal data

Identifying individuals likely to respond to a specific intervention is a critical challenge in medical research. In the age of big data, where vast amounts of information are accessible, the potential for personalized interventions based on individual characteristics has become increasingly feasible. However, advancement comes with significant challenges. The sheer volume of data often leads to datasets with numerous factors which might influence outcomes. Moreover, the data may exhibit clustering or repeated measurements, with potentially informative cluster sizes, adding complexity to the analysis. For instance, researchers are interested in understanding why some pregnancies are more vulnerable to maternal immune activation (MIA), which impacts brain and behavioral development in offspring and increases the risk of autism, schizophrenia and other neurodevelopmental disorders. However, this task is complex due to the multitude of biomarkers and clustered or repeatedly measured outcomes over time and brain regions. Current statistical tools available are inadequate for handling the complexities of such data, thus impeding the progress of precision medicine. To address this significant gap, this proposal underscores the urgent need for innovative statistical methodologies that can adeptly handle the complexity of clustered and longitudinal datasets with numerous covariates, thereby advancing the field of precision medicine. By developing novel methods building on our preliminary statistical framework, integrating machine learning techniques, rigorously evaluating these methods through simulation studies that mimic real data, and applying these methodologies to real-world longitudinal and clustered datasets, we aim to make significant contributions to this field. Our preliminary simulation results and real-world examples demonstrate both the scientific merit and computational feasibility of these methods. We will apply these newly developed statistical tools to existing datasets as a proof-of-concept to uncover factors that predict susceptibility and resilience to MIA regarding the brain and behavior development outcomes in offspring. The innovative statistical methods developed hold significant promise for identifying biomarkers that elucidate the link between environmental exposure during human pregnancy and brain mechanisms associated with neurodevelopmental disorders. This advancement will assist in identifying high- risk pregnancies and tailoring interventions for offspring at risk due to MIA exposure. Furthermore, these innovative statistical approaches can be adapted to various interventions and a wide range of medical conditions. We will provide free, user-friendly programs and software to enable research communities to apply these methods easily. Consequently, this project presents a unique opportunity to tackle a complex issue in precision medicine and leverage existing datasets for groundbreaking insights. Project Number: 1R21MH140288-01A1 | Fiscal Year: 2026 | NIH Institute/Center: National Institute of Mental Health (NIMH) | Principal Investigator: Shuai Chen | Institution: UNIVERSITY OF CALIFORNIA AT DAVIS, DAVIS, CA | Award Amount: $442,750 | Activity Code: R21 | Study Section: Analytics and Statistics for Population Research Panel A Study Section[ASPA] View on NIH RePORTER: https://reporter.nih.gov/project-details/11304344