生成式人工智能与孕妇及产后人群的孕产妇健康结局：一项随机临床试验。
完整摘要 · 中文译文
重要性：生成式人工智能（AI）聊天机器人越来越多地被用于支持孕产妇健康，但随机对照证据有限。 目的：评估使用生成式AI孕产妇健康聊天机器人是否能改善12个月孕产妇健康结局。 设计、设置和参与者：这项两组随机临床试验于2023年10月至2024年12月纳入400名参与者，12个月随访于2025年12月完成。参与者通过线上方式在美国49个州和哥伦比亚特区招募。符合条件的参与者为英语使用者，年龄14岁或以上，处于妊娠期或正在抚养6个月以下婴儿，并自我认同为种族和/或族裔少数群体成员。 干预：参与者按1:1随机分配接受聊天机器人使用权限，或接受由每月儿童书籍组成的主动对照。 主要结局和测量指标：主要结局为12个月时使用广泛性焦虑障碍-7量表和患者健康问卷-9测量的焦虑和抑郁症状。次要结局包括急诊就诊、妊娠相关并发症、产后保健检查、有效避孕措施使用和睡眠时长。 结果：在400名随机参与者中，平均（SD）年龄为32.6（5.1）岁；200名（50%）在入组时正在抚养婴儿；10名（2%）认同为美洲印第安人或阿拉斯加原住民，80名（20%）为亚洲人，188名（47%）为黑人或非裔美国人，57名（14%）为多种族，65名（16%）为西班牙裔或拉丁裔。AI聊天机器人与焦虑评分的差异无关 es（校正均差，0.10；95% CI，-0.77至0.96）或抑郁评分（校正均差，-0.23；95% CI，-1.05至0.59），也与急诊就诊、产后护理、避孕使用或睡眠时长无关。在入组时妊娠的参与者中，聊天机器人在次要12个月终点分析与更多自我报告的妊娠相关并发症相关，但在纵向分析中则不相关。在探索性组内分析中，更高的聊天机器人参与度与更长的睡眠时长和更高的焦虑评分相关。 结论与意义：在这项针对种族和族裔少数群体妊娠及产后个体的随机临床试验中，与主动对照相比，使用生成式人工智能孕产妇健康聊天机器人并未改善12个月时的孕产妇焦虑或抑郁。次要和探索性发现应谨慎解读。 试验注册：ClinicalTrials.gov标识符：NCT06053515。
完整摘要 · 英文原文
IMPORTANCE: Generative artificial intelligence (AI) chatbots are increasingly used to support maternal health, but randomized evidence is limited. OBJECTIVE: To evaluate whether access to a generative AI maternal health chatbot improved 12-month maternal health outcomes. DESIGN, SETTING, AND PARTICIPANTS: This 2-group randomized clinical trial enrolled 400 participants from October 2023 to December 2024, with 12-month follow-up completed in December 2025. Participants were recruited online across 49 US states and the District of Columbia. Eligible participants were English-speaking, aged 14 years or older, pregnant or parenting an infant younger than 6 months, and self-identified as members of racial and/or ethnic minoritized groups. INTERVENTIONS: Participants were randomized 1:1 to receive access to the chatbot or an active control consisting of monthly children's books. MAIN OUTCOMES AND MEASURES: Primary outcomes were anxiety and depression symptoms at 12 months measured using the Generalized Anxiety Disorder-7 and Patient Health Questionnaire-9. Secondary outcomes included emergency department visits, pregnancy-related complications, postpartum care checkups, effective contraception use, and sleep duration. RESULTS: Among 400 randomized participants, the mean (SD) age was 32.6 (5.1) years; 200 (50%) were parenting an infant at enrollment; and 10 (2%) identified as American Indian or Alaska Native, 80 (20%) Asian, 188 (47%) Black or African American, 57 (14%) multiracial, and 65 (16%) Hispanic or Latino. The AI chatbot was not associated with differences in anxiety scores (adjusted mean difference, 0.10; 95% CI, -0.77 to 0.96) or depression scores (adjusted mean difference, -0.23; 95% CI, -1.05 to 0.59), nor with emergency department visits, postpartum care, contraception use, or sleep duration. Among participants pregnant at enrollment, the chatbot was associated with more self-reported pregnancy-related complications in the secondary 12-month end point analysis, but not in longitudinal analyses. Higher chatbot engagement was associated with longer sleep duration and higher anxiety scores in exploratory within-group analyses. CONCLUSIONS AND RELEVANCE: In this randomized clinical trial of pregnant and postpartum individuals from racial and ethnic minoritized groups, access to a generative artificial intelligence maternal health chatbot did not improve maternal anxiety or depression at 12 months compared with an active control. Secondary and exploratory findings should be interpreted cautiously. TRIAL REGISTRATION: ClinicalTrials.gov Identifier: NCT06053515.