Machine-Learning prediction of comorbid substance use disorders in ADHD youth using Swedish registry data.
使用瑞典注册数据对 ADHD 青年共病物质使用障碍的机器学习预测。
- 作者列表："Zhang-James Y","Chen Q","Kuja-Halkola R","Lichtenstein P","Larsson H","Faraone SV
BACKGROUND:Children with attention-deficit/hyperactivity disorder (ADHD) have a high risk for substance use disorders (SUDs). Early identification of at-risk youth would help allocate scarce resources for prevention programs. METHODS:Psychiatric and somatic diagnoses, family history of these disorders, measures of socioeconomic distress, and information about birth complications were obtained from the national registers in Sweden for 19,787 children with ADHD born between 1989 and 1993. We trained (a) a cross-sectional random forest (RF) model using data available by age 17 to predict SUD diagnosis between ages 18 and 19; and (b) a longitudinal recurrent neural network (RNN) model with the Long Short-Term Memory (LSTM) architecture to predict new diagnoses at each age. RESULTS:The area under the receiver operating characteristic curve (AUC) was 0.73(95%CI 0.70-0.76) for the random forest model (RF). Removing prior diagnosis from the predictors, the RF model was still able to achieve significant AUCs when predicting all SUD diagnoses (0.69, 95%CI 0.66-0.72) or new diagnoses (0.67, 95%CI: 0.64, 0.71) during age 18-19. For the model predicting new diagnoses, model calibration was good with a low Brier score of 0.086. Longitudinal LSTM model was able to predict later SUD risks at as early as 2 years age, 10 years before the earliest diagnosis. The average AUC from longitudinal models predicting new diagnoses 1, 2, 5 and 10 years in the future was 0.63. CONCLUSIONS:Population registry data can be used to predict at-risk comorbid SUDs in individuals with ADHD. Such predictions can be made many years prior to age of the onset, and their SUD risks can be monitored using longitudinal models over years during child development. Nevertheless, more work is needed to create prediction models based on electronic health records or linked population registers that are sufficiently accurate for use in the clinic.
背景: 注意缺陷/多动障碍 (ADHD) 儿童发生物质使用障碍 (SUDs) 的风险较高。早期识别高危青年将有助于为预防计划分配稀缺资源。 方法: 精神和躯体诊断，这些疾病的家族史，社会经济困扰的措施,关于出生并发症的信息从瑞典 19,787 和 1989年出生的 1993 例 ADHD 儿童的国家登记处获得。我们训练了 (a) 一个横断面随机森林 (RF) 模型，使用 17 岁之前可用的数据来预测 18 岁至 19 岁之间的 SUD 诊断; 和 (b) 具有长短期记忆 (LSTM) 架构的纵向递归神经网络 (RNN) 模型，用于预测每个年龄的新诊断。 结果: 随机森林模型 (RF) 的受试者工作特征曲线下面积 (AUC) 为 0.73(95% CI 0.70-0.76)。从预测因子中去除先前的诊断，RF 模型在预测所有 SUD 诊断 (0.69，95% CI 0.66-0.72) 或新诊断 (0.67, 95% CI: 0.64，0.71) 在 18-19 岁期间。对于预测新诊断的模型，模型校准良好，Brier 评分较低，为 0.086。纵向 LSTM 模型能够在最早诊断前 10 年预测 2 岁以后的 SUD 风险。预测未来 1 、 2 、 5 和 10 年新诊断的纵向模型的平均 AUC 为 0.63。 结论: 人群登记数据可用于预测 ADHD 个体的高危共病 SUDs。这样的预测可以在发病年龄前许多年进行，并且可以在儿童发育期间使用多年的纵向模型监测其 SUD 风险。尽管如此，需要做更多的工作来创建基于电子健康记录或链接人口登记的预测模型，这些模型足够准确，可用于临床。
METHODS:Abstract Background Attention-deficit/hyperactivity disorder (ADHD) is a psychosocially impairing and cost-intensive mental disorder, with first symptoms occurring in early childhood. It can usually be diagnosed reliably at preschool age. Early detection of children with ADHD symptoms and an early, age-appropriate treatment are needed in order to reduce symptoms, prevent secondary problems and enable a better school start. Despite existing ADHD treatment research and guideline recommendations for the treatment of ADHD in preschool children, there is still a need to optimise individualised treatment strategies in order to improve outcomes. Therefore, the ESCApreschool study (Evidence-Based, Stepped Care of ADHD in Preschool Children aged 3 years and 0 months to 6 years and 11 months of age (3;0 to 6;11 years) addresses the treatment of 3–6-year-old preschool children with elevated ADHD symptoms within a large multicentre trial. The study aims to investigate the efficacy of an individualised stepwise-intensifying treatment programme. Methods The target sample size of ESCApreschool is 200 children (boys and girls) aged 3;0 to 6;11 years with an ADHD diagnosis according to Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) or a diagnosis of oppositional defiant disorder (ODD) plus additional substantial ADHD symptoms. The first step of the adaptive, stepped care design used in ESCApreschool consists of a telephone-assisted self-help (TASH) intervention for parents. Participants are randomised to either the TASH group or a waiting control group. The treatment in step 2 depends on the outcome of step 1: TASH responders without significant residual ADHD/ODD symptoms receive booster sessions of TASH. Partial or non-responders of step 1 are randomised again to either parent management and preschool teacher training or treatment as usual. Discussion The ESCApreschool trial aims to improve knowledge about individualised treatment strategies for preschool children with ADHD following an adaptive stepped care approach, and to provide a scientific basis for individualised medicine for preschool children with ADHD in routine clinical care. Trial registration The trial was registered at the German Clinical Trials Register (DRKS) as a Current Controlled Trial under DRKS00008971 on 1 October 2015. This manuscript is based on protocol version 3 (14 October 2016).
METHODS:Prefrontal volume reductions commonly are demonstrated in ADHD, but the literature examining prefrontal volume in reading disorders (RD) is scant despite their also having executive functioning (EF) deficits. Furthermore, only a few anatomical studies have examined the frontal lobes in comorbid RD/ADHD, though they have EF deficits similar to RD and ADHD. Hence, we examined frontal gyri volume in children with RD, ADHD, RD/ADHD and controls, as well as their relationship to EF for gyri found to differ between groups. We found right inferior frontal (RIF) volume was smaller in ADHD, and smaller volume was related to worse behavioral regulation. Left superior frontal (LSF) volume was larger in RD than ADHD, and its size was negatively related to basic reading ability. Left middle frontal (LMF) volume was largest in RD/ADHD overall. Further, its volume was not related to basic reading nor behavioral regulation but was related to worse attentional control, suggesting some specificity in its EF relationship. When examining hypotheses on the etiology of RD/ADHD, RD/ADHD was commensurate with ADHD in RIF volume and both RD and ADHD in LSF volume (being midway between the groups), consistent with the common etiology hypothesis. Nevertheless, they also had an additional gyrus affected: LMF, consistent with the cognitive subtype hypothesis in its specificity to RD/ADHD. The few other frontal aMRI studies on RD/ADHD supported both hypotheses as well. Given this, future research should continue to focus on frontal morphology in its endeavors to find neurobiological contributors to the comorbidity between RD and ADHD.
METHODS:BACKGROUND:Mechanistic endophenotypes can inform process models of psychopathology and aid interpretation of genetic risk factors. Smaller total brain and subcortical volumes are associated with attention-deficit hyperactivity disorder (ADHD) and provide clues to its development. This study evaluates whether common genetic risk for ADHD is associated with total brain volume (TBV) and hypothesized subcortical structures in children. METHODS:Children 7-15 years old were recruited for a case-control study (N = 312, N = 199 ADHD). Children were assessed with a multi-informant, best-estimate diagnostic procedure and motion-corrected MRI measured brain volumes. Polygenic scores were computed based on discovery data from the Psychiatric Genomics Consortium (N = 19 099 ADHD, N = 34 194 controls) and the ENIGMA + CHARGE consortium (N = 26 577). RESULTS:ADHD was associated with smaller TBV, and altered volumes of caudate, cerebellum, putamen, and thalamus after adjustment for TBV; however, effects were larger and statistically reliable only in boys. TBV was associated with an ADHD polygenic score [β = -0.147 (-0.27 to -0.03)], and mediated a small proportion of the effect of polygenic risk on ADHD diagnosis (average ACME = 0.0087, p = 0.012). This finding was stronger in boys (average ACME = 0.019, p = 0.008). In addition, we confirm genetic variation associated with whole brain volume, via an intracranial volume polygenic score. CONCLUSION:Common genetic risk for ADHD is not expressed primarily as developmental alterations in subcortical brain volumes, but appears to alter brain development in other ways, as evidenced by TBV differences. This is among the first demonstrations of this effect using molecular genetic data. Potential sex differences in these effects warrant further examination.