Latent Constructs Underlying Sensory Subtypes: An Independent Components Analysis

Background:   Sensory features (SF) are atypical responses to sensory stimuli that influence the functioning of children with neurodevelopmental disorders (NDD) including Autism Spectrum Disorder (ASD) and Attention Deficit Hyperactive Disorder (ADHD). Due to the heterogeneous nature of SF, there is a need to define clinically meaningful subtypes, which may differentially respond to intervention. Sensory features are highly relevant to the ASD population due to their recent inclusion in diagnostic criteria. Distinct sensory subtypes have been identified in children with ASD through parent responses on the Short Sensory Profile (SSP) (Lane et al., 2014). Clinical profiles led to the hypothesis that two sensory dimensions underlie SF: 1) sensory reactivity (the intensity of a response to a stimulus), and/or 2) multisensory integration (the ability to process multiple concurrent stimuli). The use of Lane et al.’s sensory subtypes facilitates systematic examination of behavior and symptom profiles associated with each subtype. It is necessary, however, to have a sound understanding of the theoretical basis underlying the subtypes before they can be adopted confidently in clinical settings. As a next step to validate this schema, we performed independent component analysis (ICA) to elucidate the latent constructs underlying sensory subtype classification.

Objectives:  To determine the latent constructs that underpin sensory subtype classification.

Methods:  Two datasets were analyzed in the present study. The first is a sample of 228 children with ASD aged 2-10. The second included 155 children aged 4-10 years with ADHD, Sensory Processing Disorder or ASD and typically developing controls.

Parents of study participants completed the SSP, which measures behaviors associated with responses to everyday, environmental sensory stimuli in children aged 3–10 years. Parents respond to each item using a 5-point ordinal scale, with higher scores indicating more typical performance.

 

Independent component analyses (ICA) were conducted via a multistep process using z-scores from the seven SSP domains and item-level data. First, our hypothesis of a two-component structure was tested on domain z-scores. Next, z-scores were analyzed using model-fitting software to determine if any model had superior model fit than the hypothesized solution. The optimal number of components was determined, balancing interpretability with model fit. Lastly, item-level data were analyzed via ICA to determine which specific items on the SSP contribute most to the underlying latent constructs.

 

Results:   A three-component model best explained the data for both samples, each with major contributions from one of the following domains: 1) taste/smell sensitivity, 2) low energy/weak, and 3) underresponsive/seeks sensation and auditory filtering. Key items from those domains that heavily contributed to the corresponding latent components were identified.

Conclusions:   We propose that components one and two represent specific foci of SF, vis a vis taste/smell sensitivity and low energy/weak. Component three represents the construct of severity. Indeed, the Lane et al.’s sensory subtypes can be mapped onto these components in a conceptually meaningful way. The present study asserts a novel conceptual framework of constructs (focus and severity) that underlie sensory subtypes to guide future research and clinical practice on SF in ASD.