Katholieke Universiteit Leuven, Faculty of Medicine, School of Dentistry, Forensic Odontology Department, Kapucijnenvoer 7, Leuven B-3000, Belgium.
Forensic Sci Int. 2010 Sep 10;201(1-3):106-11. doi: 10.1016/j.forsciint.2010.04.040. Epub 2010 May 18.
Unquestionable forensic age investigations are based on statistical models constructed on a sample containing subjects of identical origin as the examined individual. In cases where corresponding models are unavailable, the established report has to describe the possible effects of this unrelated information on the predicted age outcome. The aim of this study is to collect country specific databases of third molar development and to verify how the related dental age estimations are influenced if we were to use dental developmental information only from Belgium or from all collected countries together. Data containing third molar developmental stages scored following Gleiser and Hunt (modified by Köhler) were collected from 9 country specific populations (Belgium, China, Japan, Korea, Poland, Thailand, Turkey, Saudi-Arabia and South-India). Age predictions were obtained from a training dataset and validated on a test dataset. Bayes rule using the repeated third molar scores is applied to get age predictions and prediction intervals. Three age predictions were compared for males and females separately. For the first prediction, the training dataset contains only Belgian subjects. For the second prediction, the training dataset for each country consists only of subjects of the country itself. For the final prediction, subjects from all countries are pooled into one common training dataset. Besides the (absolute) difference between the chronological age and the predicted age, specific interest lies in the juvenile-adult distinction. In the age range from 16 to 22 years 6982 subjects (3189 male and 3793 female) were analyzed. Using information on third molar development from Belgium compared to information from the country specific databases hardly increased the mean absolute differences (MAD) and mean squared errors (MSE): the MAD and MSE increased on average with 0.5 and 2.5 months with maximal increases of, respectively 1.6 and 7.3 months. Using information from all countries pooled compared to country specific information provided even on average negligible increases (0.05 and 0.2 months for MAD and MSE, respectively). For the juvenile-adult discrimination, using information from all countries instead of country specific information yielded comparable performances. Using Belgium instead of country specific information increased the percentage of correctly identified juveniles, but decreased the percentage of correctly identified adults. The adult-juvenile discrimination based on information used from Belgium provides judicially the best applied reference.
毫无疑问，法医年龄调查是基于在与被检查个体具有相同来源的样本上构建的统计模型进行的。在没有相应模型的情况下，所建立的报告必须描述这种不相关信息对预测年龄结果的可能影响。本研究的目的是收集特定国家的第三磨牙发育数据库，并验证如果我们仅使用来自比利时或所有收集国家的牙齿发育信息，相关的牙齿年龄估计会受到怎样的影响。从 9 个特定国家的人群（比利时、中国、日本、韩国、波兰、泰国、土耳其、沙特阿拉伯和印度南部）中收集了包含 Gleiser 和 Hunt（经 Köhler 修改）评分的第三磨牙发育阶段的数据。年龄预测是从训练数据集获得的，并在测试数据集上进行验证。使用重复的第三磨牙评分的贝叶斯规则来获得年龄预测和预测区间。分别对男性和女性进行了三种年龄预测的比较。对于第一种预测，训练数据集仅包含比利时受试者。对于第二种预测，每个国家的训练数据集仅由该国家的受试者组成。对于最终预测，将所有国家的受试者汇集到一个共同的训练数据集中。除了与预测年龄的（绝对）差异外，人们还特别关注青少年和成年人的区别。在 16 至 22 岁的年龄范围内，分析了 6982 名受试者（男性 3189 名，女性 3793 名）。与特定国家的数据库相比，使用来自比利时的第三磨牙发育信息几乎没有增加平均绝对差异（MAD）和平均平方误差（MSE）：MAD 和 MSE 平均增加 0.5 至 2.5 个月，最大增加分别为 1.6 和 7.3 个月。与特定国家的信息相比，使用所有国家汇总的信息甚至平均增加可以忽略不计（MAD 和 MSE 分别为 0.05 和 0.2 个月）。对于青少年和成年人的鉴别，使用所有国家的信息而不是特定国家的信息产生了可比的结果。使用比利时而不是特定国家的信息增加了正确识别青少年的百分比，但降低了正确识别成年人的百分比。基于来自比利时的信息进行的成年人-青少年鉴别提供了司法上最佳的应用参考。
