Olga Tennison Autism Research Centre, La Trobe University, Melbourne, Victoria, Australia.
Department Epidemiology & Biostatistics, UC San Francisco, San Francisco, California, USA.
Autism Res. 2024 Nov;17(11):2386-2404. doi: 10.1002/aur.3233. Epub 2024 Sep 20.
Autistic children frequently have one or more co-occurring psychological, behavioral, or medical conditions. We examined relationships between child behaviors, sleep, adaptive behavior, autistic traits, mental health conditions, and health in autistic children using network analysis. Network analysis is hypothesis generating and can inform our understanding of relationships between multiple conditions and behaviors, directing the development of transdiagnostic treatments for co-occurring conditions. Participants were two child cohorts from the Autism Treatment Network registry: ages 2-5 years (n = 2372) and 6-17 years (n = 1553). Least absolute-shrinkage and selection operator (LASSO) regularized partial correlation network analysis was performed in the 2-5 years cohort (35 items) and the 6-17 years cohort (36 items). The Spinglass algorithm determined communities within each network. Two-step expected influence (EI2) determined the importance of network variables. The most influential network items were sleep difficulties (2 items) and aggressive behaviors for young children and aggressive behaviors, social problems, and anxious/depressed behavior for older children. Five communities were found for younger children and seven for older children. Of the top three most important bridge variables, night-waking/parasomnias and anxious/depressed behavior were in both age-groups, and somatic complaints and sleep initiation/duration were in younger and older cohorts respectively. Despite cohort differences, sleep disturbances were prominent in all networks, indicating they are a transdiagnostic feature across many clinical conditions, and thus a target for intervention and monitoring. Aggressive behavior was influential in the partial correlation networks, indicating a potential red flag for clinical monitoring. Other items of strong network importance may also be intervention targets or screening flags.
自闭症儿童常伴有一种或多种共患的心理、行为或医学疾病。我们使用网络分析研究了自闭症儿童的行为、睡眠、适应行为、自闭症特征、心理健康状况和健康之间的关系。网络分析是一种产生假设的方法，可以帮助我们了解多种疾病和行为之间的关系，为共患疾病的跨诊断治疗提供指导。参与者来自自闭症治疗网络登记处的两个儿童队列：2-5 岁（n=2372）和 6-17 岁（n=1553）。在 2-5 岁队列（35 项）和 6-17 岁队列（36 项）中进行了最小绝对收缩和选择算子（LASSO）正则化偏相关网络分析。Spinglass 算法确定了每个网络中的社区。两步预期影响（EI2）确定了网络变量的重要性。最具影响力的网络项目是幼儿的睡眠困难（2 项）和攻击性行为，以及大龄儿童的攻击性行为、社交问题和焦虑/抑郁行为。发现了 5 个幼儿社区和 7 个大龄儿童社区。在前三个最重要的桥接变量中，夜间醒来/睡眠障碍和焦虑/抑郁行为在两个年龄组中都存在，而躯体抱怨和睡眠开始/持续时间分别在年幼和年长组中存在。尽管存在队列差异，但睡眠障碍在所有网络中都很突出，这表明它们是许多临床疾病的跨诊断特征，因此是干预和监测的目标。攻击性行为在偏相关网络中具有影响力，表明这是临床监测的潜在危险信号。其他具有较强网络重要性的项目也可能是干预目标或筛查标志。
