孤独症儿童睡眠障碍的影响因素及预测模型

Influencing factors and predictive model of sleep disorders in children with autism

摘要:

目的 探讨孤独症儿童睡眠障碍的影响因素，以此构建列线图预测模型并进行验证。方法 回顾性分析2023年1月—2023年12月宁波大学附属妇女儿童医院收治的132例孤独症儿童的临床资料，根据8∶2定律随机分为训练集106例和验证集26例。依据患儿是否存在睡眠障碍分为睡眠障碍组和对照组。统计孤独症患儿的睡眠障碍发生情况，比较训练集睡眠障碍组和对照组的基本资料，分析孤独症儿童睡眠障碍的影响因素，构建列线图模型并验证，预测孤独症患儿发生睡眠障碍的效能。结果 训练集106例患儿中28例（26.42%）患儿发生睡眠障碍，验证集26例患儿中6例（23.08%）患儿发生睡眠障碍。睡眠障碍组睡前电子屏幕时间长于对照组（P <0.05），新生儿窒息史率、母孕期抑郁率均高于对照组（P <0.05），父母关系和谐率低于对照组（P <0.05）。训练集睡眠障碍组与对照组患儿性别、年龄、照顾者、午间睡眠时间、户外活动时间、电子屏幕时间及父母年龄、父母性格、居住环境、3岁前与父母分离时间、分娩方式比较，差异均无统计学意义（P >0.05）。多因素一般Logistic回归分析结果显示：睡前电子屏幕时间长［O^R =1.078（95% CI：1.003，1.158）］、存在新生儿窒息史［O^R =4.867（95% CI：1.400，16.919）］、父母关系差［O^R =6.818（95% CI：1.575，29.509）］、母孕期抑郁［O^R =2.632（95% CI：1.036，6.686）］均是孤独症儿童睡眠障碍的危险因素（P <0.05）。经Bootstrap法验证，模型C-index指数为0.859（95% CI：0.769，0.948），Calibration曲线结果显示，校正曲线与理想曲线趋近，经Hosmer-Lemeshow检验，提示拟合度较好（P >0.05）。训练集模型预测孤独症儿童发生睡眠障碍的曲线下面积为0.826，敏感性为85.70%（95% CI：0.775，0.934），特异性为73.10%（95% CI：0.753，0.814）；验证集模型预测孤独症儿童发生睡眠障碍的曲线下面积为0.788，敏感性为71.40%（95% CI：0.634，0.813），特异性为79.50%（95% CI：0.716，0.894）。结论 睡前电子屏幕时间长、存在新生儿窒息史及母孕期抑郁、父母关系差的孤独症患儿更易发生睡眠障碍，基于此构建的列线图预测模型对孤独症患儿睡眠障碍的发生风险具有较好的预测效能。

Abstract:

Objective To analyze the factors affecting sleep disorders in children with autism, and to construct and validate a nomogram-based prediction model.Methods The clinical data of 132 children with autism admitted to our hospital from January 2019 to December 2019 were retrospectively analyzed. They were randomly assigned into a training set (n = 106) and a validation set (n = 26) at a ratio of 8:2. Based on whether the children had sleep disorders, they were divided into the sleep disorder group and the control group. The incidence of sleep disorders in children with autism was calculated. The characteristics of the sleep disorder group and the control group in the training set were compared. The influencing factors of sleep disorders in children with autism were analyzed, and a nomogram model was constructed and validated. The predictive efficacy of the nomogram for sleep disorders in children with autism was evaluated.Results Sleep disorders occurred in 28 (26.42%) of 106 children in the training set and 6 (23.08%) of 26 children in the verification set. The screen time before bedtime in the sleep disorder group was longer than that in the control group (P < 0.05). The rates of neonatal asphyxia and maternal depression during pregnancy were higher in the sleep disorder group than those in the control group (P < 0.05). The rate of harmonious parental relationships in the sleep disorder group was lower than that in the control group (P < 0.05). There were no statistically significant differences between the sleep disorder group and the control group in terms of children’s sex, age, caregiver, duration of midday naps, outdoor activity time, screen time, as well as parental age, personality, living environment, duration of separation from parents before age three, or mode of delivery (P > 0.05). Multivariable Logistic regression analysis showed that long screen time before bedtime [O^R = 1.078 (95% CI: 1.003, 1.158) ], history of neonatal asphyxia [O^R = 4.867 (95% CI: 1.400, 16.919) ], maternal depression during pregnancy [O^R = 6.818 (95% CI: 1.575, 29.509) ] and poor parental relationships [O^R = 2.632 (95% CI: 1.036, 6.686) ] were all risk factors for sleep disorders in autistic children (P < 0.05). The C-index of the model as validated by the Bootstrap method was 0.859 (95% CI: 0.769, 0.948). The calibration curve closely aligned with the ideal curve, and the Hosmer-Lemeshow test indicated good model fit (P > 0.05). The area under the curve of the model based on the training set for predicting sleep disorders in children with autism was 0.826, with a sensitivity of 85.70% (95% CI: 0.775, 0.934) and a specificity of 73.10% (95% CI: 0.753, 0.814). The area under the curve of the model based on the validation set was 0.788, with a sensitivity of 71.40% (95% CI: 0.634, 0.813) and a specificity of 79.50% (95% CI: 0.716, 0.894).Conclusions Autistic children with prolonged screen time before bedtime, a history of neonatal asphyxia, maternal depression during pregnancy, and poor parental relationships are more likely to develop sleep disorders. The nomogram prediction model constructed based on these factors demonstrates good predictive performance for assessing the risk of sleep disorders in children with autism.