目的 探究串联质谱（tandem mass spectrometry, MS/MS）联合人工智能（artificial intelligence, AI）技术对新生儿甲基丙二酸血症（methylmalonic acidemia, MMA）的诊断准确率，并分析影响诊断准确性的因素。 方法 召回秦皇岛市妇幼保健院2019年1月—2024年12月间MMA初筛结果阳性的新生儿246例进行复查。以基因诊断为金标准，对所有患儿进行MS/MS诊断、AI诊断和联合诊断（MS/MS联合AI），评估不同诊断方法的灵敏度、特异度等诊断效能指标，同时分析其与基因诊断的一致性，并采用多因素logistic回归分析诊断准确性的影响因素。 结果 联合诊断的灵敏度为92.3%，特异度为97.0%，准确率为96.7%，与基因诊断的一致性高（Kappa=0.733）。联合诊断、MS/MS诊断和AI诊断的曲线下面积分别为0.914（95%CI：0.867~0.935）、0.759（95%CI：0.635~0.816）和0.669（95%CI：0.584~0.776），联合诊断的曲线下面积大于MS/MS诊断、AI诊断（P<0.05）。甲硫氨酸、苯丙氨酸、鸟氨酸、甘氨酸、血氨、乳酸水平低，以及电生理检查结果异常均是影响诊断准确性的因素（P<0.05）。 结论 MS/MS联合AI诊断对新生儿MMA的诊断准确率较高；甲硫氨酸、苯丙氨酸、鸟氨酸、甘氨酸、血氨、乳酸水平低，以及电生理检查结果异常可影响MMA诊断的准确性。

Objective To evaluate the diagnostic accuracy of tandem mass spectrometry (MS/MS) combined with artificial intelligence (AI) for neonatal methylmalonic acidemia (MMA), and to identify factors associated with diagnostic accuracy. Methods A total of 246 neonates with positive initial MMA screening at Qinhuangdao Maternal and Child Health Hospital from January 2019 to December 2024 were recalled for confirmatory testing. Using genetic diagnosis as the reference standard, all cases underwent MS/MS diagnosis, AI diagnosis, and combined diagnosis (MS/MS plus AI). Sensitivity, specificity, accuracy, and other diagnostic performance indices were calculated; agreement with genetic diagnosis was assessed; and multivariable logistic regression was performed to identify factors associated with diagnostic accuracy. Results The combined diagnosis yielded a sensitivity of 92.3%, a specificity of 97.0%, and an accuracy of 96.7%, with high agreement with genetic diagnosis (Kappa=0.733). The areas under the receiver operating characteristic curve for combined, MS/MS, and AI diagnoses were 0.914(95%CI: 0.867-0.935), 0.759(95%CI: 0.635-0.816), and 0.669(95%CI: 0.584-0.776), respectively; the area under the curve for the combined diagnosis was significantly higher than that for either MS/MS or AI alone (P<0.05). Lower levels of methionine, phenylalanine, ornithine, glycine, blood ammonia, and lactic acid, as well as abnormal electrophysiological findings, were independently associated with diagnostic accuracy (P<0.05). Conclusions MS/MS combined with AI shows high diagnostic accuracy for neonatal MMA. Lower levels of methionine, phenylalanine, ornithine, glycine, blood ammonia, and lactic acid and abnormal electrophysiological findings may affect the diagnostic accuracy of MMA.