Crohn’s disease (CD) is a serious chronic inflammatory bowel disease (IBD) which was once considered rare in the pediatric population. Recently it has been recognized as one of the most important chronic. diseases that affect children and adolescents, with increasing incidence in youths of varying ages1–4. As child-hood is a time of physical, emotional, and social maturation, onset of CD in childhood can seriously delay growth,jeopardize mental health, and the negative impact may last a lifetime.

In current PCD research, disease activity is commonly evaluated at pre-specified time points post treatment for comparison of treatments and prediction of long-term outcomes9–11. Findings from these studies are based on snapshots of patients’ responses, with limited information on disease activity between the time points. In addition, discrete analyses of disease activity fail to provide insights on functional relationship between disease activity and other health outcomes, which is essential for selection of treatment to induce andmaintain remission while optimizing the patient’s well-being.

Aim 1: To characterize disease trajectory and identify unique patterns to explain heterogeneity of disease. trajectory using EHR data. We will develop and evaluate a framework, named as TArgET-FPCA, based on functional principal component analysis (FPCA), for the identification and characterization of patterns of diseasetrajectories in PCD. The methods will be developed to tackle two unique challenges in studying PCD using EHR. data, including the ability to extract patterns that represent both global and local features for three major phases of treating PCD, i.e., remission induction, short-term maintenance, and long-term maintenance, and the ability to reduce bias in estimation and extraction of patterns induced by the outcome-dependent sampling schemes in EHR data.

Aim 2: To identify PCD subgroups and build dynamic prediction for long-term PCD outcomes by incorporating information in disease trajectory. We will develop a method to identify subgroups based on information contained in disease trajectory and study the effects of group membership and its interaction with treatment on long-term PCD outcomes. We will also development of a time-varying coefficients model to incorporate temporal and historical information contained in patient’s disease trajectory to dynamically update the predicted risk.

Aim 3: Application and Software development. The proposed methods will be applied to PEDSnet datato investigate novel hypotheses on improving long-term outcomes. We will identify unique patterns of disease trajectories and subgroups of PCD patients in PEDSnet PCD population and identify risk factors that are related to long-term outcomes. We will develop software and visualization tools to assist clinical decision-makings.