Real world evidence for altered communication patterns in individuals with autism spectrum disorder
Methods
Study cohort and data collection
We recruited 44 adult individuals with a clinical diagnosis of ASD (the inclusion criteria comprised an IQ of above 70) aside to a control group of 54 individuals with typical neurodevelopment (TD). All participants with ASD received a diagnostic assessment in accordance with the national autism guidelines (Arbeitsgemeinschaft der Wissenschaftlichen Medizinischen Fachgesellschaften (AWMF), 2015), and a formal diagnosis of autism was provided by experienced clinical psychiatrists. Patients were recruited from the “Clinic of Disorders of Social Interaction” at the Department of General Psychiatry 2 at the LVR-Klinikum Düsseldorf.
Passive data collection was conducted using the “JTrack Social” application, which operates passively without necessitating active participation from the participants24. This smartphone application continuously collected data over a period of four months, capturing – among other information – the application usage (including timestamps, duration, and names of the applications used). All data was automatically transferred to a secure server at the Research Centre Jülich. Participants were automatically logged out of the study when reaching the study duration. All smartphone-based data was pseudonymized at the source. All potential participants were provided with comprehensive information about the study and informed written consent was obtained from each participant to ensure ethical compliance. All smartphone data was encrypted during transmission using the Hypertext Transfer Protocol Secure (HTTPS). Access to stored data was restricted to authorized personnel, and strict adherence to data security protocols was maintained.
The study was approved by the Ethics Committee of the Medical Faculty of the Heinrich Heine University Düsseldorf, Germany.
Data Preprocessing and Feature Extraction
Out of 98 recruited participants, all iOS users (n = 17) were excluded from further evaluation of communication preferences as iOS technically does not provide information about specific app names making it impossible to estimate the actual communication mode. Furthermore, to ensure sufficient observation duration only users who contributed data for more than half of the study duration (minimum 60 days) were considered (i.e. leaving the study too early or manually stopping data recording) resulting in further n = 21 drop-outs. These drop-outs were proportionally equally distributed across ASD and TD groups resulting in n = 60 (27 ASD, 33 TD) participants included in further evaluation (Table 1). For these users, the relevant communication applications were categorized based on their primary communication usage into verbal, written, or mixed communication based on app descriptions available in the Android Play Store. Total communication was computed as a sum of these categories. The application usage for each communication app was calculated by summing the daily foreground time, representing the duration users interacted with the app on their screens. For instance, a phone call was classified as verbal communication, while text messaging or email apps were defined as written communication. Dual-functionality apps like “WhatsApp” fell under mixed communication Supplementary Table 1. As app recording times provided by the operating system are not always precise resulting in potential overestimation of phone usage, apps with more than 6 h per application per day were considered as technical outliers and excluded from the analysis. Subsequently, daily average usage for each category within ASD and TD groups was independently computed.
Statistical analysis
Statistical analyses were performed in Jamovi (version 2.3, https://www.jamovi.org.). General Linear Models with mixed effects were used to test for differences in communication preferences for each of the categories (written, verbal, mixed, and total) between the ASD and TD groups:
The dependent variables were log-transformed to approximate a normal distribution before fitting the models. All models were adjusted for age, sex, IQ (fixed effects), and subject (random effect). Considering the exploratory nature of this study, the statistical significance threshold was set at p < 0.05 uncorrected. Cohen’s d effect sizes were used to estimate the magnitude of the observed differences across all observations. In addition, we computed the sign tests to test for consistency of the group preferences for the observed significant group communication categories comparing the number of days where mean group communication time was higher in ASD as compared to TD. Additionally, the analysis was expanded to investigate correlations between AQ scores and the duration of app usage across different communication modalities: mixed, total, verbal, and written, using the Pearson correlation method.
Responses