Crowell David H, Kulp Thomas D, Kapuniai Linda E, Hunt Carl E, Brooks Lee J, Weese-Mayer Debra E, Silvestri Jean, Ward Sally Davidson, Corwin Michael, Tinsley Larry, Peucker Mark
Department of Pediatrics, John A Burns School of Medicine, University of Hawaii and Kapi'olani Medical Center for Women and Children, Honolulu, HI 96826, USA.
J Clin Neurophysiol. 2002 Oct;19(5):469-83. doi: 10.1097/00004691-200210000-00010.
Infant arousal scoring based on the Atlas Task Force definition of transient EEG arousal was evaluated to determine (1). whether transient arousals can be identified and assessed reliably in infants and (2). whether arousal and no-arousal epochs scored previously by trained raters can be validated reliably by independent sleep experts. Phase I for inter- and intrarater reliability scoring was based on two datasets of sleep epochs selected randomly from nocturnal polysomnograms of healthy full-term, preterm, idiopathic apparent life-threatening event cases, and siblings of Sudden Infant Death Syndrome infants of 35 to 64 weeks postconceptional age. After training, test set 1 reliability was assessed and discrepancies identified. After retraining, test set 2 was scored by the same raters to determine interrater reliability. Later, three raters from the trained group rescored test set 2 to assess inter- and intrarater reliabilities. Interrater and intrarater reliability kappa's, with 95% confidence intervals, ranged from substantial to almost perfect levels of agreement. Interrater reliabilities for spontaneous arousals were initially moderate and then substantial. During the validation phase, 315 previously scored epochs were presented to four sleep experts to rate as containing arousal or no-arousal events. Interrater expert agreements were diverse and considered as noninterpretable. Concordance in sleep experts' agreements, based on identification of the previously sampled arousal and no-arousal epochs, was used as a secondary evaluative technique. Results showed agreement by two or more experts on 86% of the Collaborative Home Infant Monitoring Evaluation Study arousal scored events. Conversely, only 1% of the Collaborative Home Infant Monitoring Evaluation Study-scored no-arousal epochs were rated as an arousal. In summary, this study presents an empirically tested model with procedures and criteria for attaining improved reliability in transient EEG arousal assessments in infants using the modified Atlas Task Force standards. With training based on specific criteria, substantial inter- and intrarater agreement in identifying infant arousals was demonstrated. Corroborative validation results were too disparate for meaningful interpretation. Alternate evaluation based on concordance agreements supports reliance on infant EEG criteria for assessment. Results mandate additional confirmatory validation studies with specific training on infant EEG arousal assessment criteria.
基于阿特拉斯特别工作组对短暂脑电图觉醒的定义对婴儿觉醒评分进行评估，以确定：（1）能否在婴儿中可靠地识别和评估短暂觉醒；（2）训练有素的评分者之前评分的觉醒和无觉醒时段能否得到独立睡眠专家的可靠验证。评分者间和评分者内可靠性评分的第一阶段基于从孕龄35至64周的健康足月儿、早产儿、特发性明显危及生命事件病例以及婴儿猝死综合征婴儿的兄弟姐妹的夜间多导睡眠图中随机选择的两个睡眠时段数据集。训练后，评估测试集1的可靠性并识别差异。重新训练后，由相同的评分者对测试集2进行评分以确定评分者间的可靠性。之后，来自训练组的三名评分者对测试集2重新评分以评估评分者间和评分者内的可靠性。评分者间和评分者内可靠性kappa值及其95%置信区间，一致性水平从实质性到几乎完美。自发觉醒的评分者间可靠性最初中等，之后达到实质性。在验证阶段，将315个之前评分的时段呈现给四位睡眠专家，以评定是否包含觉醒或无觉醒事件。评分者间专家的一致性各不相同，被认为无法解释。基于对之前采样的觉醒和无觉醒时段的识别，睡眠专家一致性的协调性被用作次要评估技术。结果显示，在协作家庭婴儿监测评估研究中，86%的觉醒评分事件得到了两名或更多专家的认同。相反，协作家庭婴儿监测评估研究中评分的无觉醒时段只有1%被评定为觉醒。总之，本研究提出了一个经过实证检验的模型，该模型包含使用修改后的阿特拉斯特别工作组标准在婴儿短暂脑电图觉醒评估中提高可靠性的程序和标准。通过基于特定标准的训练，在识别婴儿觉醒方面证明了评分者间和评分者内的高度一致性。确证性验证结果差异太大，无法进行有意义的解释。基于协调性协议的替代评估支持依靠婴儿脑电图标准进行评估。结果要求进行额外的确证性验证研究，并针对婴儿脑电图觉醒评估标准进行特定培训。
