Abstract
OBJECTIVE:
The goal of this study was to describe the typical longitudinal developmental trajectories of social and communication functioning in children with autism and to determine the correlates of these trajectories.
METHODS:
Children with autism who were born in California from 1992 through 2001 and enrolled with the California Department of Developmental Services were identified. Subjects with <4 evaluations present in the database were excluded, resulting in a sample of 6975 children aged 2 to 14 years. Score sequences were constructed based on 9 evaluative items for social, communication, and repetitive behavior functioning. Typical trajectories were identified by using group-based latent trajectory modeling, and multinomial logistic regression models were used to determine the odds of classification within each trajectory varied by individual and family-level factors.
RESULTS:
Six typical patterns of social, communication, and repetitive behavior functioning were identified. These trajectories displayed significant heterogeneity in developmental pathways, and children whose symptoms were least severe at first diagnosis tended to improve more rapidly than those severely affected. One group of ∼10% of children experienced rapid gains, moving from severely affected to high functioning. Socioeconomic factors were correlated with trajectory outcomes; children with non-Hispanic, white, well-educated mothers were more likely to be high functioning, and minority children with less-educated mothers or intellectual disabilities were very unlikely to experience rapid gains.
CONCLUSIONS:
Children with autism have heterogeneous developmental pathways. One group of children evidenced remarkable developmental change over time. Understanding what drives these outcomes is thus critical.
KEY WORDS: autistic disorder, longitudinal outcomes, trajectory models
What’s Known on This Subject:
Autism is widely considered a heterogeneous disorder in terms of etiology and phenotype. Although autism is usually a lifelong disorder, little is known about the rate or timing of how children develop regarding their communication and social functioning.
What This Study Adds:
Utilizing annual evaluations for a large population of children with autism, we describe the 6 most common trajectories from diagnosis through age 14 years. Trajectories revealed considerable variation, and high socioeconomic status children were more likely to experience rapid improvement.
Autism is a neurodevelopmental disorder with varying phenotypic expressions typically diagnosed in early childhood and characterized by deficits of communication and social interaction as well as repetitive and stereotyped behaviors. In recent years, autism has become more visible, as both incidence and prevalence have increased. Recent prevalence estimates have reached 1 per 110 children aged 8 years.1
Autism’s phenotypic heterogeneity is likely associated with different etiologies. Recent studies found inherited and de novo deletions and duplications in a wide range of locations on the genome in autism cases compared with controls, suggesting heterogeneous genetic predispositions.2–4 Family aggregation studies similarly suggest that core characteristics of autism have different levels of heritability.5–7 Autism has also been associated with various factors relating to prenatal and perinatal environments, labor and delivery complications, parental characteristics, and socioeconomic status, each possibly pointing to mechanisms that lead to heterogeneous symptom presentation.8–15
A paucity of data on developmental trajectories for children with autism compounds the difficulty in understanding whether different etiologies are associated with differing phenotypic expression. Examining autism’s varied presentations only at first diagnosis neglects the developmental course of the disorder.16 Although most children diagnosed with autism retain the diagnosis as adolescents17,18 and adults,19 what we know about the unfolding dynamics of the disorder is limited. Recent longitudinal studies reveal substantial diversity in trajectories through childhood and adolescence.20,21 However, most extant studies have largely spanned relatively short time frames,22–26 included infrequent follow-ups,18,19,27 used small sample populations,18,22–25,27,28 are often inconsistent in findings of stability or change,26,28 and pay little attention to how characteristics of children and their families might affect these trajectory patterns.
Filling in this knowledge gap with regard to autism symptom trajectories may point to how etiology and/or resources or treatment shape dynamics of the disorder. Researchers have noted the need for more knowledge about developmental trajectories in autism,29 arguing that it would improve intervention and care.16,18 Furthermore, conceptualizing developmental delay in terms of trajectories can highlight different causal mechanisms30 and improve diagnostic systems.31
We observed autism symptom trajectories of children in California from diagnosis until they were up to 14 years old. We used group-based trajectory modeling to identify heterogeneous subsets of symptom trajectories. We describe here the typical patterns of longitudinal development along autism’s 3 main symptom domains, consider their relationships to one another, and then examine the correlates of trajectories: what kinds of children follow each pattern, and what makes them different? Finally, we explored the implications of our findings for autism research and treatment.
Methods
Population
Our data set links birth records from the California birth master files with autism caseload records from the California Department of Developmental Services (DDS). In California, the DDS, through its 21 regional centers, is responsible for coordinating diagnoses, services, and support for persons with developmental disabilities (including autism). The DDS provides services to patients with autistic disorder (Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, code 299.032) but not to those with other disorders on the autism spectrum unless they have another qualifying condition. Although enrollment is voluntary, the vast majority of individuals with autism in California are enrolled, making the DDS the largest administrative source of data on autism diagnoses.33 All autism diagnoses are confirmed by DDS staff, either by on-site diagnostic teams or by professionals with expertise in diagnosing autism.34 Diagnoses were informed by using 1 or more of several diagnostic instruments, including the Autism Diagnostic Interview or the Autism Diagnostic Observation Schedule. We used the records of children who were born between 1992 and 2001 and who received an autism diagnosis by 2006. Of these 16 681 children, we excluded those born outside of California and those with <4 annual records. This left us with a sample population of 6975 individuals from whose records the trajectories are derived. From the autism caseload records, we also extracted information on the presence of an intellectual disability. This diagnosis was made or confirmed by regional center clinicians on the basis of IQ test results.
Demographic information was obtained from the linked birth records. Eighty-five percent of DDS records for children ever diagnosed with autism were successfully linked to birth records. Those not linked were typically born outside of California. Variables extracted from the birth records include: maternal age at birth, education, and race and ethnicity; place of birth; child’s gender and birth weight (<2500 g was considered low birth weight); and whether the birth was paid for by Medi-Cal (California’s Medicaid program). Missing data on covariates reduced the sample to 6968 for some analyses.
Symptom scores were obtained from the Client Development Evaluation Report (CDER) database of the DDS. During the CDER interview, DDS clients’ symptom severity and functioning across several dimensions are recorded. The CDER is intended to determine appropriate services based on level of functioning, not serve as a diagnostic instrument. Nonetheless, the CDER is completed by trained DDS staff and covers the 3 core domains of autism symptoms. It is conducted with someone who interacts closely with the child on a regular basis, typically a parent or caregiver. Although CDERs are missing for some children in some years, most children were evaluated approximately annually.
To create scores for communication and social functioning, we summed the CDER’s 3 items that measure communication and 5 items that measure social interaction, weighting each item equally, and then rescaled the sums such that the minimum, or low-functioning (LF), score was 0 and the maximum possible score was 100. (Details on items, score construction, and the distributions of these scores are described in the Supplemental Information.) Scores between these extremes can be achieved through multiple combinations of behaviors along the facets of communication and social interaction measured by CDER items. In addition, we used the evaluation item describing the presence of repetitive behavior. This item has 5 possible levels, ranging from 1 (repetitive body movements occur without cessation during waking hours) to 5 (no apparent repetitive body movements) (Supplemental Table 4; Distributions of scores on the three symptom dimensions at selected ages can be found in Supplemental Figure 2). Evaluations for children too young or too disabled for repetitive behavior to be observed were coded as missing, reducing the sample for that dimension to 6938 children. Finally, we assembled each individual’s sequence of scores over time for each of the 3 dimensions. Sequence lengths vary, depending on birth cohort, age at diagnosis, time in the DDS caseload, and frequency of evaluation.
This study was approved by the Columbia University institutional review board and the California Committee for the Protection of Human Subjects. A waiver of informed consent was obtained because the risk to confidentiality arising from contacting subjects to obtain consent would far exceed any risk from the study itself.
Modeling
The goal of this analysis was to identify and describe subgroups within the data that have similar developmental trajectories. To achieve this goal, we used group-based latent trajectory modeling implemented with the PROC TRAJ module for SAS (SAS Institute, Inc, Cary, NC).35–37 This modeling strategy uses a polynomial equation to capture the relationship between age and behavior, where the outcome is a latent variable characterizing individual i’s autism symptom level at time t. Symptom outcomes then depend on a cubic function of i’s age at time t and membership in 1 of j groups. The β parameters describe the shape of the trajectories and are permitted to vary across the j groups, allowing each group’s trajectory to be shaped differently. This method reveals the heterogeneity across a population both in age-specific presentation of symptoms and in the way symptoms change over time.
The model parameters describe the functional form of each trajectory as well as the population-level probabilities of the size of each trajectory group, which are estimated by using maximum likelihood methods. After testing the model with different numbers of trajectory groups, 6 group models were selected. Goodness of fit criteria included the Bayesian Information Criterion (BIC), which selects the model with the highest likelihood while penalizing complex models, and conceptual clarity. Posterior probabilities for each individual’s membership to each group were used to assign each child to his or her most likely trajectory.
To describe the composition of each trajectory’s population, we calculated proportions for selected demographic variables and autism risk factors for the population and for each trajectory subgroup and conducted χ2 tests of independence. For each communication and social group, we then calculated the relative risks (RRs) of being in each of the other’s trajectory categories and each repetitive behavior category for children in that group compared with all other children in the sample and conducted 2-tailed significance tests with α = 0.05. To clarify which factors distinguish the trajectories, we recoded the 6 trajectory groups of the communication and social dimensions into 3 groups representing high, low, and rapid improvement, and estimated multinomial logistic regression models predicting the log odds of group membership on the basis of demographic and autism risk factors. Odds ratios (ORs) and their 95% confidence intervals are presented.
Results
Symptom Trajectories
The fitted trajectories for the 6 group models are displayed graphically in Fig 1. (Coefficients can be found in Supplemental Table 5.) Figure 1 A, B, and C displays the communication, social, and repetitive behavior trajectories, respectively. The communication and social graphs have 4 features of note. First, many children experienced substantial development. Particularly on the communication dimension, most of the trajectories increased steadily over time. Second, there was substantial heterogeneity in the extent of this development; some children improved much more than others. This heterogeneity is especially visible in the social dimension; some groups improved whereas others, particularly the “low” group, were relatively flat. With 1 important exception, groups that were high-functioning (HF) at the start improved more rapidly. Third, the most rapid development occurred before age 6 years, and several of the trajectories tended to flatten out after that. Finally, both the communication and social dimensions have a group, here identified as “bloomers,” that improved especially quickly. These children started with low scores, comparable to those on the LF trajectories, and improved at a rapid pace, ending with scores comparable to those on the HF trajectories.
FIGURE 1.
(A) Modeled communication, (B) social, and (C) repetitive behavior symptom trajectories based on CDER scores by age.
In contrast, the repetitive behavior trajectories remained relatively stable, except for 1 that decreased in severity and 1 that increased. Although there was diversity in the amount of repetitive behavior observed, only ∼15% of children seemed to change significantly over time, one-half of these improving and one-half getting worse. In the analyses that follow, we describe the differences in the composition of the trajectory groups and distinguish bloomers from other children with autism.
Composition of Trajectory Groups
Table 1 contains descriptive statistics on the population, as well as according to subgroup membership. In general, children on the highest trajectories tended to have been born in the most recent cohorts. Trajectory groups did not differ according to gender. The LF trajectory groups were less likely to be white, more likely to be born as a Medi-Cal recipient, and to have younger, foreign-born mothers with fewer years of education. The HF children tended to be born to older, more educated, white mothers, and were less likely to be a Medi-Cal recipient. Children with intellectual disabilities were more likely to be on LF trajectories.
TABLE 1.
Descriptive Statistics for Sample of Persons With Autism With at Least 4 Valid Evaluations
| Total Population | Communication Trajectories | Social Trajectories | Repetitive Behavior Trajectories | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Low | Low– Medium | Medium | Bloomers | Medium– High | High | Low | Low– Medium | Medium | Bloomers | Medium– High | High | Usually | Daily | When Stressed | Improving | Declining | Never | ||
| Trajectory Group | |||||||||||||||||||
| n | 6975 | 769 | 1293 | 1758 | 451 | 1827 | 877 | 582 | 1719 | 2107 | 701 | 1378 | 488 | 554 | 1894 | 2015 | 476 | 439 | 1597 |
| % | 100.0 | 11.0 | 18.5 | 25.2 | 6.5 | 26.2 | 12.6 | 8.3 | 24.6 | 30.2 | 10.1 | 19.8 | 7.0 | 7.9 | 27.2 | 28.9 | 6.8 | 6.3 | 22.9 |
| Birth cohorta,b,c | |||||||||||||||||||
| 1992–1994 | 36.7 | 35.5 | 39.4 | 34.1 | 42.8 | 36.7 | 35.8 | 31.6 | 40.2 | 38.8 | 35.8 | 32.1 | 35.9 | 36.5 | 40.7 | 33.7 | 28.4 | 46.7 | 35.5 |
| 1995–1997 | 40.4 | 41.0 | 37.8 | 40.7 | 41.2 | 43.2 | 37.1 | 43.1 | 38.4 | 41.2 | 41.2 | 41.6 | 36.5 | 40.3 | 38.6 | 40.7 | 41.0 | 39.2 | 42.4 |
| 1998+ | 22.9 | 23.5 | 22.7 | 25.1 | 16.0 | 20.1 | 27.1 | 25.3 | 21.4 | 20.0 | 23.0 | 26.3 | 27.7 | 23.3 | 20.6 | 25.5 | 30.7 | 14.1 | 22.1 |
| Male | 82.0 | 78.9 | 80.7 | 83.1 | 83.8 | 81.9 | 83.2 | 84.0 | 82.7 | 81.7 | 82.7 | 80.8 | 80.3 | 82.1 | 82.3 | 82.1 | 81.7 | 83.1 | 81.2 |
| Low birth weight | 7.8 | 9.1 | 8.0 | 8.0 | 7.3 | 6.6 | 8.7 | 6.5 | 8.0 | 7.4 | 6.4 | 8.6 | 10.0 | 7.6 | 7.3 | 7.8 | 8.0 | 8.4 | 8.1 |
| Medi-Cala,b,c | 28.5 | 41.9 | 35.7 | 28.0 | 31.0 | 24.3 | 14.6 | 40.5 | 31.2 | 26.9 | 27.4 | 23.7 | 27.0 | 29.2 | 29.6 | 26.1 | 30.3 | 33.0 | 28.2 |
| Age of autism diagnosis, ya,b,c | |||||||||||||||||||
| ≤3 | 58.3 | 69.6 | 65.4 | 60.5 | 57.4 | 49.6 | 52.2 | 59.1 | 60.8 | 56.7 | 51.6 | 60.4 | 59.0 | 65.0 | 59.7 | 56.6 | 65.1 | 61.5 | 53.6 |
| 4 | 20.0 | 18.9 | 20.2 | 19.7 | 18.0 | 20.7 | 20.8 | 23.4 | 19.8 | 18.1 | 24.4 | 20.0 | 18.0 | 16.4 | 19.1 | 20.3 | 22.7 | 22.1 | 20.5 |
| 5 | 10.6 | 7.3 | 8.6 | 10.4 | 11.1 | 13.7 | 10.1 | 10.7 | 10.5 | 11.7 | 12.3 | 8.2 | 10.5 | 10.6 | 10.2 | 9.7 | 10.1 | 10.9 | 12.2 |
| ≥6 | 11.1 | 4.3 | 5.9 | 9.4 | 13.5 | 15.9 | 16.9 | 6.9 | 8.8 | 13.5 | 11.7 | 11.3 | 12.5 | 7.9 | 11.0 | 13.4 | 2.1 | 5.5 | 13.7 |
| Intellectual disabilitya,b,c | 37.3 | 58.8 | 52.0 | 41.0 | 33.9 | 26.0 | 14.4 | 50.5 | 43.4 | 37.9 | 31.2 | 29.2 | 28.7 | 44.0 | 40.8 | 33.2 | 31.9 | 46.0 | 35.1 |
| Mother foreign-borna,b,c | 39.4 | 53.2 | 49.3 | 40.5 | 46.3 | 32.1 | 21.9 | 58.8 | 45.8 | 39.4 | 35.9 | 29.7 | 25.8 | 40.6 | 35.5 | 37.3 | 37.4 | 43.1 | 45.7 |
| Mother Non-Hispanic, white (n = 6968)a,b,c | 41.8 | 26.9 | 31.0 | 36.1 | 36.8 | 50.7 | 66.1 | 26.3 | 36.4 | 43.3 | 45.2 | 48.3 | 49.2 | 39.4 | 44.9 | 43.8 | 39.9 | 42.6 | 36.8 |
| Maternal educationa,b,c | |||||||||||||||||||
| <High school | 17.1 | 29.8 | 24.2 | 16.7 | 16.6 | 12.8 | 5.9 | 30.8 | 19.6 | 15.3 | 16.5 | 12.7 | 13.5 | 22.0 | 17.1 | 14.9 | 17.6 | 17.5 | 18.1 |
| High school graduate/some college | 53.2 | 54.2 | 53.8 | 53.1 | 61.0 | 53.4 | 47.3 | 47.4 | 53.9 | 53.8 | 52.1 | 52.8 | 58.2 | 50.2 | 54.4 | 53.8 | 57.4 | 51.9 | 51.3 |
| College+ | 29.6 | 16.0 | 22.0 | 30.2 | 22.4 | 33.8 | 46.8 | 21.8 | 26.5 | 30.9 | 31.4 | 34.5 | 28.3 | 27.8 | 28.6 | 31.3 | 25.0 | 30.5 | 30.6 |
| Maternal age, y (n = 6974)a | |||||||||||||||||||
| <20 | 4.5 | 5.9 | 5.4 | 5.2 | 5.5 | 3.5 | 2.1 | 5.3 | 4.5 | 4.3 | 4.4 | 4.0 | 5.7 | 6.9 | 5.0 | 3.9 | 5.3 | 3.2 | 3.9 |
| 20–34 | 72.9 | 74.9 | 75.3 | 72.2 | 72.1 | 72.9 | 69.8 | 74.7 | 72.2 | 75.0 | 71.5 | 70.8 | 72.3 | 72.0 | 72.2 | 72.7 | 74.6 | 74.7 | 73.4 |
| ≥35 | 22.6 | 19.2 | 19.3 | 22.5 | 22.4 | 23.6 | 28.2 | 19.9 | 23.3 | 20.6 | 24.1 | 25.2 | 21.7 | 21.1 | 22.8 | 23.3 | 20.2 | 22.1 | 22.7 |
Data are presented as %, unless otherwise indicated.
χ2 test of independence for communication trajectories significant at P ≤ .05.
χ2 test of independence for social trajectories significant at P ≤ .05.
χ2 test of independence for repetitive behavior trajectories significant at P ≤ .05.
Relationship Between Trajectories
Table 2 describes the relationships between trajectory group memberships. Those on the lowest communication trajectory had ∼6 times the risk of being on the lowest social trajectory and were extremely unlikely to be on HF social trajectories. Communication bloomers were heavily concentrated among the social bloomers (RR: 2.58) as well as among the medium–high and high groups, which, like the bloomers, show significant improvement over time, and social bloomers are heavily concentrated among the communication bloomers (RR: 2.75). Meanwhile, communication and social bloomers were unlikely to be in the other’s LF group. Similarly, those on LF social and communication trajectories were more likely to exhibit frequent repetitive behaviors, whereas those on HF trajectories were more likely to exhibit fewer or none.
TABLE 2.
RRs for Trajectory Group Membership
| A: RRs for Communication Group Members of Assignment to Each Social and Repetitive Behavior Group, Relative to All Others | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Communication trajectory group | Social Trajectory Group | Repetitive Behavior Trajectory Group | ||||||||||
| Low | Low–Medium | Medium | Bloomers | Medium–High | High | Usually | Daily | When Stressed | Improving | Declining | Never | |
| Low | 6.08a | 1.87a | 0.58a | 0.22a | 0.17a | 0.10a | 3.22a | 1.32a | 0.72a | 0.35a | 0.79 | 0.66a |
| Low–medium | 2.21a | 2.00a | 0.93 | 0.41a | 0.40a | 0.17a | 1.66a | 1.39a | 0.81a | 0.48a | 1.12 | 0.76a |
| Medium | 0.59a | 1.22a | 1.36a | 0.50a | 0.92 | 0.50a | 0.76a | 1.16a | 0.97 | 0.80a | 0.85 | 1.06 |
| Bloomers | 0.28a | 0.88a | 1.10a | 2.58a | 0.79a | 0.46a | 0.86 | 0.81a | 1.08 | 1.08 | 1.37a | 1.05 |
| Medium–high | 0.12a | 0.42a | 1.15a | 1.72a | 1.79a | 1.49a | 0.43a | 0.78a | 1.12a | 1.67a | 1.10 | 1.18a |
| High | 0.06a | 0.14a | 0.63a | 1.88a | 1.97a | 5.25a | 0.36a | 0.52a | 1.34a | 1.80a | 0.93 | 1.24a |
| B: RRs for Social Group Members of Assignment to Each Communication and Repetitive Behavior Group, Relative to All Others | ||||||||||||
| Social trajectory group | Communication Trajectory Group | Repetitive Behavior Trajectory Group | ||||||||||
| Low | Low–Medium | Medium | Bloomers | Medium–High | High | Usually | Daily | When Stressed | Improving | Declining | Never | |
| Low | 5.29a | 1.95a | 0.64a | 0.27a | 0.15a | 0.06a | 3.07a | 1.26a | 0.65a | 0.28a | 0.75 | 0.84a |
| Low–medium | 2.23a | 2.17a | 1.22a | 0.86 | 0.43a | 0.12a | 1.98a | 1.43a | 0.82a | 0.66a | 0.95 | 0.66a |
| Medium | 0.52a | 0.92 | 1.40a | 1.14 | 1.16a | 0.58a | 0.71a | 1.05 | 1.14a | 0.72a | 1.09 | 0.96 |
| Bloomers | 0.23a | 0.44a | 0.55a | 2.75a | 1.53a | 1.83a | 0.46a | 0.92 | 1.12 | 1.78a | 0.97 | 0.94 |
| Medium–high | 0.16a | 0.39a | 0.92 | 0.76a | 1.68a | 2.15a | 0.38a | 0.64a | 1.16a | 1.35a | 1.03 | 1.44a |
| High | 0.10a | 0.19a | 0.55a | 0.46a | 1.36a | 4.19a | 0.27a | 0.51a | 0.99 | 2.25a | 1.12 | 1.52a |
RR statistics significant at 2-tailed P < .05.
In general, we found that the 3 dimensions were associated with one another, such that children who were HF on 1 dimension were unlikely to be LF on the others. Yet, there is some independence, so that children may develop at a different pace along these 3 dimensions.
Correlates of “Blooming”
We next investigated the factors that distinguish bloomers, who improve rapidly, from the LF children they initially resemble and the HF children they resemble at the end of the observed trajectory. In Table 3, we present the adjusted ORs for being in 1 of the 2 lowest or 2 highest social and communication trajectories, relative to being a bloomer. The middle trajectory was excluded from this analysis to provide a cleaner comparison.
TABLE 3.
ORs and 95% Confidence Intervals From Multinomial Logistic Regression Results Showing Correlates of LF and HF Trajectories Versus Blooming
| Variable | Communication Trajectories (n = 4828) | Social Trajectories (n = 4497) | ||
|---|---|---|---|---|
| LF | HF | LF | HF | |
| Male | 0.86 (0.65–1.14) | 0.98 (0.75–1.30) | 1.04 (0.82–1.32) | 0.85 (0.67–1.08) |
| Medi-Cal recipient | 1.19 (0.92–1.56) | 0.90 (0.69–1.17) | 1.07 (0.85–1.34) | 0.92 (0.72–1.16) |
| Age of diagnosis, y | ||||
| ≤3 | 1.00 | 1.00 | 1.00 | 1.00 |
| 4 | 1.00 (0.75–1.32) | 1.49 (1.13–1.97)a | 0.73 (0.59–0.91)a | 0.75 (0.60–0.93)a |
| 5 | 0.62 (0.43–0.88)a | 1.49 (1.07–2.09)a | 0.73 (0.55–0.96)a | 0.67 (0.50–0.90)a |
| ≥6 | 0.34 (0.24–0.48)a | 1.45 (1.06–1.98)a | 0.62 (0.46–0.83)a | 0.89 (0.67–1.18) |
| Intellectual disability | 2.15 (1.71–2.7)a | 0.60 (0.48–0.76)a | 1.55 (1.28–1.88)a | 0.75 (0.61–0.92)a |
| Mother foreign-born | 0.99 (0.77–1.28) | 0.61 (0.48–0.79)a | 1.51 (1.23–1.87)a | 0.74 (0.59–0.92)a |
| Mother race/ethnicity | ||||
| Non-Hispanic, white | 0.69 (0.50–0.93)a | 1.86 (1.38–2.51)a | 0.70 (0.55–0.90)a | 1.03 (0.80–1.34) |
| Hispanic | 0.73 (0.55–0.98)a | 1.38 (1.03–1.85)a | 0.68 (0.53–0.87)a | 1.11 (0.85–1.45) |
| All other races | 1.00 | 1.00 | 1.00 | 1.00 |
| Maternal education | ||||
| <High school | 1.82 (1.30–2.54)a | 1.01 (0.72–1.42) | 1.26 (0.95–1.66) | 0.86 (0.64–1.17) |
| High school or some college | 1.00 | 1.00 | 1.00 | 1.00 |
| College graduate | 1.06 (0.80–1.40) | 1.93 (1.48–2.51)a | 0.81 (0.66–1.01) | 0.97 (0.78–1.20) |
ORs significant at P ≤ .05 (2-tailed).
Beginning with the communication trajectories, we found that children diagnosed later were more likely to be on HF trajectories, and were less likely to be LF, than bloomers. Children with intellectual disabilities were more than twice as likely to be LF than bloomers (OR: 2.15) and very unlikely to be HF. Children with foreign-born mothers were less likely to be HF than bloomers, but there was no significant difference for the LF group. LF children were less likely to be white (OR: 0.69), whereas HF children were more likely to be white (OR: 1.86). However, among nonwhites, Hispanic children had greater odds of being HF and lower odds of being LF. The LF children were likely to have less-educated mothers, and the highest-educated mothers were more likely to have HF children. Finally, there was no significant difference according to gender or being a recipient of Medi-Cal.
The pattern for social bloomers was similar. There was no difference according to gender or being a Medi-Cal recipient, and maternal education did not affect social trajectories. Both HF and LF groups were less likely than bloomers to be diagnosed after age 3 years. Having an intellectual disability was strongly associated with LF trajectories, as was having a foreign-born mother. Finally, children of white mothers were less likely to be LF than bloomers, as were children of Hispanic mothers (relative to nonwhites).
Discussion
There is significant heterogeneity in the developmental trajectories of these study children with autism. Some children improved rapidly, whereas the trajectories of others were both slower and less likely to reveal significant improvement. This is especially true of the social and communication dimensions; most children showed little change in repetitive behaviors over the observed period. Furthermore, these dimensions show some association with one another yet are also independent; 1 child may show different patterns of change on these 3 dimensions, further contributing to the heterogeneity of autism. The pattern of overrepresentation of HF trajectories among more recent cohorts may be evidence of a broadening of diagnostic criteria over time, as more mild cases are identified relative to previous years.38 Improved awareness and screening in recent years may also have contributed to this result.
Our findings are consistent with some previous research showing heterogeneous outcomes, as well as change in autism symptoms over time.16,18,27,28 Some researchers have found a subgroup of children with autism who improve substantially and may even lose their diagnosis over time.16,21,39 However, this research differs from previous work by providing a more detailed and nuanced understanding of the pace and timing of symptom change across a broad swath of childhood and early adolescence for a large population in a diverse state.
Children who begin with HF scores tend to improve more rapidly over time. One significant exception is the group we call bloomers; these children start out with scores placing them among the most severely affected but improve so substantially that they enter adolescence with scores comparable to the HF children.
Bloomers differ from other children with respect to intellectual disability and socioeconomic characteristics. Among young children with severe autism, those most likely to “bloom” are those without intellectual disability and those with more educated, nonminority mothers. Although we are unable to identify the specific mechanisms through which socioeconomic status affects trajectory outcomes, the intervening variables likely include home and neighborhood environments, quality and intensity of treatment, quality of education, the efficacy with which parents are able to advocate for their children with institutions providing services, and many other factors in various permutations. If this heterogeneity in outcomes is associated with parental and community resources, then equal access to early intervention and treatment resources for less-advantaged children is vital. Although some trajectories may be associated with different etiologic drivers, if etiology alone were driving outcomes, we would be less likely to observe the strong socioeconomic effects unless socioeconomic status was associated with exposure to some biological risk factor for a particular autism subtype.
Linking the DDS data to California’s birth master files required that our data sets contain information only on those children who were both born in California and enrolled with the DDS, possibly biasing the results. There are several limitations associated with use of the DDS data. First, the symptom severity items were collected by the DDS for the purpose of resource allocation, not diagnostic evaluation. Second, the fact that the DDS only provides services to those with autistic disorder but not to those with other disorders on the autism spectrum may result in overdiagnosis of autism. Third, we used the first entry into the DDS as a proxy for age of diagnosis, which may lag first diagnosis in the community by several months or more. In addition, longer follow-up through adolescence and into adulthood would have been preferable, but data were available on outcomes only into 2006. Moreover, regional centers operate independently and may be applying evaluative items according to different criteria. However, results were robust to fixed-effects models adjusting for regional center catchment areas (results not shown). Finally, given the socioeconomic differences we find, it is unfortunate that we do not have any information on types and quantities of services used by individual children. This subject topic is a critical avenue for future research.
Conclusions
Six trajectories characterized the typical social, communication, and repetitive behavior development of children with autism in this population. This research differs from other longitudinal work on autism outcomes by using data for a large sample with far more frequent follow-ups over a wide swath of childhood and early adolescence. It is important to observe the developmental pathways children with autism follow over time to understand the pace and timing of changes. More work is needed to discover whether these longitudinal patterns will help us not only to understand the diversity of autism but also to better target interventions and improve treatment.
Supplementary Material
Acknowledgments
We thank Marissa King, Keely Cheslack-Postava, Katherine Keyes, Ka-yuet Liu, Diana Dakhlallah, Kinga Makovi, and Soumya Mazumdar.
Glossary
- BIC
Bayesian Information Criterion
- CDER
Client Development Evaluation Report
- DDS
Department of Developmental Services
- HF
high-functioning
- LF
low-functioning
- OR
odds ratio
- RR
relative risk
Footnotes
Dr Fountain contributed to the conception and design, analyzed and interpreted data, drafted and revised the article, and had final approval; Ms Winter contributed to the interpretation of data, drafted and revised the article, and approved the final version; and Dr Bearman contributed to conception and design, acquisition of data, and interpretation, drafting and revision of the article, and approved the final version.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Funded by the NIH Director’s Pioneer Award program, part of the NIH Roadmap for Medical Research, through grant 1 DP1 OD003635-01. Funded by the National Institutes of Health (NIH).
References
- 1.Autism and Developmental Disabilities Monitoring Network Surveillance Year 2006 Principal Investigators . Prevalence of Autism Spectrum Disorders—Autism and Developmental Disabilities Monitoring Network, United States, 2006. MMWR Surveill Summ. 2009;58(SS10):1–20 [PubMed] [Google Scholar]
- 2.Sebat J, Lakshmi B, Malhotra D, et al. Strong association of de novo copy number mutations with autism. Science. 2007;316(5823):445–449 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Marshall CR, Noor A, Vincent JB, et al. Structural variation of chromosomes in autism spectrum disorder. Am J Hum Genet. 2008;82(2):477–488 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Weiss LA, Shen Y, Korn JM, et al. Autism Consortium . Association between microdeletion and microduplication at 16p11.2 and autism. N Engl J Med. 2008;358(7):667–675 [DOI] [PubMed] [Google Scholar]
- 5.Ronald A, Happé F, Bolton P, et al. Genetic heterogeneity between the three components of the autism spectrum: a twin study. J Am Acad Child Adolesc Psychiatry. 2006;45(6):691–699 [DOI] [PubMed] [Google Scholar]
- 6.Georgiades S, Szatmari P, Zwaigenbaum L, et al. Structure of the autism symptom phenotype: a proposed multidimensional model. J Am Acad Child Adolesc Psychiatry. 2007;46(2):188–196 [DOI] [PubMed] [Google Scholar]
- 7.Freitag CM. The genetics of autistic disorders and its clinical relevance: a review of the literature. Mol Psychiatry. 2007;12(1):2–22 [DOI] [PubMed] [Google Scholar]
- 8.Larsson HJ, Eaton WW, Madsen KM, et al. Risk factors for autism: perinatal factors, parental psychiatric history, and socioeconomic status. Am J Epidemiol. 2005;161(10):916–925, discussion 926–928 [DOI] [PubMed] [Google Scholar]
- 9.Hultman CM, Sparén P, Cnattingius S. Perinatal risk factors for infantile autism. Epidemiology. 2002;13(4):417–423 [DOI] [PubMed] [Google Scholar]
- 10.Cheslack-Postava K, Liu K, Bearman PS. Closely spaced pregnancies are associated with increased odds of autism in California sibling births. Pediatrics. 2011;127(2):246–253 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Croen LA, Grether JK, Selvin S. Descriptive epidemiology of autism in a California population: who is at risk? J Autism Dev Disord. 2002;32(3):217–224 [DOI] [PubMed] [Google Scholar]
- 12.King MD, Fountain C, Dakhlallah D, Bearman PS. Estimated autism risk and older reproductive age. Am J Public Health. 2009;99(9):1673–1679 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Croen LA, Najjar DV, Fireman B, Grether JK. Maternal and paternal age and risk of autism spectrum disorders. Arch Pediatr Adolesc Med. 2007;161(4):334–340 [DOI] [PubMed] [Google Scholar]
- 14.Durkin MS, Maenner MJ, Newschaffer CJ, et al. Advanced parental age and the risk of autism spectrum disorder. Am J Epidemiol. 2008;168(11):1268–1276 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.King MD, Bearman PS. Socioeconomic status and the increased prevalence of autism in California. Am Sociol Rev. 2011;76(2):320–346 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Seltzer MM, Shattuck P, Abbeduto L, Greenberg JS. Trajectory of development in adolescents and adults with autism. Ment Retard Dev Disabil Res Rev. 2004;10(4):234–247 [DOI] [PubMed] [Google Scholar]
- 17.Robinson EB, Munir K, Munafò MR, Hughes M, McCormick MC, Koenen KC. Stability of autistic traits in the general population: further evidence for a continuum of impairment. J Am Acad Child Adolesc Psychiatry. 2011;50(4):376–384 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.McGovern CW, Sigman M. Continuity and change from early childhood to adolescence in autism. J Child Psychol Psychiatry. 2005;46(4):401–408 [DOI] [PubMed] [Google Scholar]
- 19.Billstedt E, Gillberg IC, Gillberg C. Autism after adolescence: population-based 13- to 22-year follow-up study of 120 individuals with autism diagnosed in childhood [published correction appears in J Autism Dev Disord. 2007;37(9):1822]. J Autism Dev Disord. 2005;35(3):351–360 [DOI] [PubMed] [Google Scholar]
- 20.Anderson DK, Maye MP, Lord C. Changes in maladaptive behaviors from midchildhood to young adulthood in autism spectrum disorder. Am J Intellect Dev Disabil. 2011;116(5):381–397 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Baghdadli A, Assouline B, Sonié S, et al. Developmental trajectories of adaptive behaviors from early childhood to adolescence in a cohort of 152 children with autism spectrum disorders. J Autism Dev Disord 2011. Available at: www.ncbi.nlm.nih.gov/pubmed/21928042. Accessed September 20, 2011 [DOI] [PubMed]
- 22.Landa R, Garrett-Mayer E. Development in infants with autism spectrum disorders: a prospective study. J Child Psychol Psychiatry. 2006;47(6):629–638 [DOI] [PubMed] [Google Scholar]
- 23.Werner E, Dawson G, Munson J, Osterling J. Variation in early developmental course in autism and its relation with behavioral outcome at 3-4 years of age. J Autism Dev Disord. 2005;35(3):337–350 [DOI] [PubMed] [Google Scholar]
- 24.Bernabei P, Cerquiglini A, Cortesi F, D’Ardia C. Regression versus no regression in the autistic disorder: developmental trajectories. J Autism Dev Disord. 2007;37(3):580–588 [DOI] [PubMed] [Google Scholar]
- 25.Munson J, Faja S, Meltzoff A, Abbott R, Dawson G. Neurocognitive predictors of social and communicative developmental trajectories in preschoolers with autism spectrum disorders. J Int Neuropsychol Soc. 2008;14(6):956–966 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Darrou C, Pry R, Pernon E, Michelon C, Aussilloux C, Baghdadli A. Outcome of young children with autism: does the amount of intervention influence developmental trajectories? Autism. 2010;14(6):663–677 [DOI] [PubMed] [Google Scholar]
- 27.Howlin P, Goode S, Hutton J, Rutter M. Adult outcome for children with autism. J Child Psychol Psychiatry. 2004;45(2):212–229 [DOI] [PubMed] [Google Scholar]
- 28.Charman T, Taylor E, Drew A, Cockerill H, Brown JA, Baird G. Outcome at 7 years of children diagnosed with autism at age 2: predictive validity of assessments conducted at 2 and 3 years of age and pattern of symptom change over time. J Child Psychol Psychiatry. 2005;46(5):500–513 [DOI] [PubMed] [Google Scholar]
- 29.Szatmari P, Bryson SE, Streiner DL, Wilson F, Archer L, Ryerse C. Two-year outcome of preschool children with autism or Asperger’s syndrome. Am J Psychiatry. 2000;157(12):1980–1987 [DOI] [PubMed] [Google Scholar]
- 30.Thomas MS, Annaz D, Ansari D, Scerif G, Jarrold C, Karmiloff-Smith A. Using developmental trajectories to understand developmental disorders. J Speech Lang Hear Res. 2009;52(2):336–358 [DOI] [PubMed] [Google Scholar]
- 31.Fecteau S, Mottron L, Berthiaume C, Burack JA. Developmental changes of autistic symptoms. Autism. 2003;7(3):255–268 [DOI] [PubMed] [Google Scholar]
- 32.American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders-IV-TR. Washington, DC: American Psychiatric Association; 2000 [Google Scholar]
- 33.Croen LA, Grether JK, Hoogstrate J, Selvin S. The changing prevalence of autism in California. J Autism Dev Disord. 2002;32(3):207–215 [DOI] [PubMed] [Google Scholar]
- 34.Autism spectrum disorders. Best practice guidelines for screening, diagnosis and assessment. 2002. Available at: www.dds.ca.gov/Autism/docs/ASD_Best_Practice2002.pdf. Accessed October 26, 2011
- 35.Nagin D. Group-Based Modeling of Development. Cambridge, MA: Harvard University Press; 2005 [Google Scholar]
- 36.Jones BL, Nagin DS, Roeder K. A SAS procedure based on mixture models for estimating developmental trajectories. Sociol Methods Res. 2001;29(3):374–393 [Google Scholar]
- 37.Jones BL, Nagin DS. Advances in group-based trajectory modeling and an SAS procedure for estimating them. Sociol Methods Res. 2007;35(4):542–571 [Google Scholar]
- 38.King MD, Bearman PS. Diagnostic change and the increased prevalence of autism. Int J Epidemiol. 2009;38(5):1224–1234 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Sutera S, Pandey J, Esser EL, et al. Predictors of optimal outcome in toddlers diagnosed with autism spectrum disorders. J Autism Dev Disord. 2007;37(1):98–107 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.