Typology of Temporal Patterns: Identifying Subgroups of Individuals with ASD

Background: Technological advances in data acquisition and analytics have increased the feasibility and usefulness of employing idiographic methods in autism research. While there are many advantages to employing an idiographic approach, a major criticism has been lack of generalizability from single subject research to the larger population of interest. This is an especially important issue in Autism Spectrum Disorder (ASD) given the high heterogeneity observed in the population, and fact that important individual differences in data can be obscured by group-level averages.

Objectives: Develop a novel analytic method that combines time series analysis and dynamic cluster analysis (Typology of Temporal Patterns; TTP) to identify subgroups of individuals who share similar longitudinal trajectories, helping address the issue of generalizability in idiographic research in autism research. 

Methods: Apply TTP to the assessment of cardiovascular arousal to environmental stressors in individuals with ASD. Data analyses were performed on heart rate (HR) data collected telemetrically from 43 severely affected (M IQ = 31) individuals with confirmed diagnoses of ASD (M = 14.55 yrs, SD = 4.24) during rest and several psychological and physical challenges established in previous studies by Goodwin and colleagues (Goodwin et al., 2004; 2006; Groden et al., 2005). Interrupted time series analysis was performed for each participant to examine individual-level HR patterns across rest and challenge conditions. High variability observed across interrupted time series results demonstrated the presence of subgroups of individuals with similar HR patterns. Accordingly, dynamic cluster analysis was conducted on HR time series data from the 43 participants. 

Results: While HR response patterns were generally elevated for all participants (80bpm <), the first cluster analysis revealed a three-cluster solution (Low, Middle, and High) largely dominated by differences in HR level (i.e., mean). Importantly, if only the total sample level average was considered, results suggest that all participants are in the Middle group (denoted by the black line in Figure 1, top panel). A second cluster analysis, focused on shape and scatter of HR patterns, revealed two subgroups (Autonomic Stabiles and Autonomic Labiles) that differed in their patterns of HR reactivity to stressors and HR recovery during rest conditions. Thus, TTP yielded a combined six-cluster solution defined by level and shape (Figure 1, bottom panel). Following these clustering results, a series of ANOVAs produced statistically significant differences between identified subgroups, even after controlling for age, sex, IQ, verbal ability, medication status, and motor movements. 

Conclusions: Our findings provide support for the utility of TTP to evaluate idiographic data at both individual and subgroup levels, and suggest that cardiovascular reactivity is a useful biological marker for identifying meaningful individual differences in the heterogeneous population of ASD. Furthermore, this is the first report of clearly defined ASD subgroups relating to cardiac stability/lability. This variable may represent an important dimension for assessing ANS as an indicator of hypo- or hyper-reactivity to environmental conditions (i.e., not just level and variability, but shape) – a new feature in the DSM-5 symptom description of ASD.