Heterogeneity In Longitudinal Trajectories of Children with Autism

Background:  Although autism is typically diagnosed in early childhood, it is a lifelong disorder from which only a minority recover.  Currently, little is known about the long-term symptom trajectories of children diagnosed with autism.  A few studies have followed the long-term outcomes of persons with autism over the life course, but these studies, in general, have utilized small or non-representative samples or infrequent follow-ups.  Further, the few extant studies tend to analyze average or typical outcomes with little attention to the heterogeneity of outcomes. 

Objectives:  In this paper, we identify and describe the common patterns of symptom expression over time among children with autism from ages two to fourteen. These trajectories are induced from the data, using an innovative statistical strategy, rather than imposed based on diagnostic or other categories. We establish their frequencies in a large population of children with autism and then explore the correlates – including birth cohort, age of diagnosis, maternal characteristics, and risk factors – for group membership.

Methods:  We begin with a population-level data set of the birth certificate records of all children with autism born in California in 1992-2000 and enrolled with the Department of Developmental Services (DDS).  We extract the annual evaluation items from the CDER and construct longitudinal trajectories for social interaction and communication indices for 4,376 children from diagnosis until they are up to 14 years old.  Using group-based latent trajectory modeling, we identify and describe sub-groups within the data of similar developmental trajectories along both dimensions.  We then estimate a series of binomial and multinomial logistic regression models on the trajectory assignments uncovered by the trajectory models.

Results:  The final models split the social and communication trajectories into six typical patterns. Most children develop substantially over time, but the extent of this development is heterogeneous; some children improve much more than others.  The period of most rapid development occurs before age 6, and several of the trajectories subsequently flatten.  A group of “Bloomers”, children who develop especially rapidly, was identified.  These children begin with low scores, comparable to the lowest-functioning trajectories, and quickly improve, ending with scores comparable to the two highest-functioning trajectories. Regression results show that the Bloomer group is relatively socioeconomically advantaged with white mothers of average or better education, and they are unlikely to be born on Medi-Cal.  They are also geographically concentrated in such a way that is not entirely explained by regional center catchment areas.

Conclusions:  There is significant heterogeneity in longitudinal autism outcomes, suggesting phenotypic and etiological heterogeneity. Children born on Medi-Cal and to non-white mothers or to mothers with less education are less likely to experience rapid, upward mobility on their symptom trajectories. This suggests that even in an environment like California, where children with autism diagnoses are provided with state-funded developmental services, there is substantial heterogeneity in outcomes associated with parental resources.  Without additional targeted resources, some children with autism may never realize their full potentials.