Ximing Xu

Background: Pediatric care resources are limited and unevenly distributed in many health systems. Caregivers increasingly turn to large language models (LLMs) for children’s health information, but real-world evidence on LLM use and response quality in general pediatric care is limited. Objective: This study aimed to characterize real-world use, consultation needs, and response quality of Xiaoyi, an LLM-based pediatric health assistant, and to evaluate the impact of expert-guided optimization on response quality. Methods: This study collected all consultation sessions submitted between April 7 and December 31, 2025, to Xiaoyi, a pediatric health assistant co-developed by Children’s Hospital of Chongqing Medical University and Zuoyi Tech. Temporal patterns of consultation activity were characterized; all sessions were annotated for pediatric subspecialty, reason for consultation, and presenting symptoms through LLM-assisted labeling with pediatrician validation. A multidisciplinary panel of 40 pediatricians evaluated Xiaoyi’s response quality over 2 rounds using predefined criteria established by expert consensus: an initial evaluation informed system optimization (prompt revision, knowledge-base and tool expansion, and validation rules) and a follow-up evaluation assessed its impact. In each round, 1000 sessions randomly sampled from the corresponding pre- or postoptimization period (2000 in total) were rated at the session level (overall quality and clinical appropriateness) and the turn level (6 predefined deficiency categories). Categorical ratings were compared using the chi-square test, and turn-level deficiency rates were compared using generalized estimating equations to account for within-session correlation. Results: Caregivers of 8266 children (median age 4.2, IQR 1.3-8 years; 4705/8266, 56.9% male) from 31 of 34 provincial-level divisions in China initiated 16,074 sessions, of which 7046 (43.8%) occurred outside routine clinic hours (08:00-17:30). Queries spanned 31 subspecialties, most frequently child health care (2025/16,074, 12.6%), pulmonology (1903/16,074, 11.8%), and gastroenterology (1833/16,074, 11.4%). The leading reasons for consultation were symptom consultation (7845/16,074, 48.8%), medical report interpretation (2531/16,074, 15.7%), and triage guidance (2515/16,074, 15.6%); among symptom consultation sessions, fever (2719/16,074, 34.7%) and cough (2349/16,074, 29.9%) were most common. After expert-guided optimization, the number of sessions rated good (n=1000) in overall quality (the response fully resolving the caregiver’s question with actionable information) increased from 652 (65.2%) to 743 (74.3%), and the number rated clinically appropriate increased from 614 (61.4%) to 874 (87.4%, both P<.001). The turn-level deficiency rate declined from 19.4% (790/4073 turns) to 6.9% (236/3419 turns; odds ratio 0.30, 95% CI 0.24-0.37; P<.001), with suboptimal clinical recommendations (not optimally tailored to the child’s age or clinical context) remaining the most frequent category. Conclusions: This real-world study suggests that an LLM-based health consultation assistant may complement existing pediatric care, with expert-guided optimization substantially improving response quality. However, the persistence of deficiencies highlights the need for continued monitoring and refinement to ensure clinical reliability.

BACKGROUND Pediatric care resources are limited and unevenly distributed in many health systems. Caregivers increasingly turn to large language models (LLMs) for children’s health information, but real-world evidence on LLM use and response quality in general pediatric care is limited. OBJECTIVE This study aimed to characterize real-world use, consultation needs, and response quality of Xiaoyi, an LLM-based pediatric health assistant, and to evaluate the impact of expert-guided optimization on response quality. METHODS This study collected all consultation sessions submitted between April 7 and December 31, 2025, to Xiaoyi, a pediatric health assistant co-developed by Children’s Hospital of Chongqing Medical University and Zuoyi Tech. Temporal patterns of consultation activity were characterized; all sessions were annotated for pediatric subspecialty, reason for consultation, and presenting symptoms through LLM-assisted labeling with pediatrician validation. A multidisciplinary panel of 40 pediatricians evaluated Xiaoyi’s response quality over 2 rounds using predefined criteria established by expert consensus: an initial evaluation informed system optimization (prompt revision, knowledge-base and tool expansion, and validation rules) and a follow-up evaluation assessed its impact. In each round, 1000 sessions randomly sampled from the corresponding pre- or postoptimization period (2000 in total) were rated at the session level (overall quality and clinical appropriateness) and the turn level (6 predefined deficiency categories). Categorical ratings were compared using the chi-square test, and turn-level deficiency rates were compared using generalized estimating equations to account for within-session correlation. RESULTS Caregivers of 8266 children (median age 4.2, IQR 1.3-8 years; 4705/8266, 56.9% male) from 31 of 34 provincial-level divisions in China initiated 16,074 sessions, of which 7046 (43.8%) occurred outside routine clinic hours (08:00-17:30). Queries spanned 31 subspecialties, most frequently child health care (2025/16,074, 12.6%), pulmonology (1903/16,074, 11.8%), and gastroenterology (1833/16,074, 11.4%). The leading reasons for consultation were symptom consultation (7845/16,074, 48.8%), medical report interpretation (2531/16,074, 15.7%), and triage guidance (2515/16,074, 15.6%); among symptom consultation sessions, fever (2719/16,074, 34.7%) and cough (2349/16,074, 29.9%) were most common. After expert-guided optimization, the number of sessions rated good (n=1000) in overall quality (the response fully resolving the caregiver’s question with actionable information) increased from 652 (65.2%) to 743 (74.3%), and the number rated clinically appropriate increased from 614 (61.4%) to 874 (87.4%, both P<.001). The turn-level deficiency rate declined from 19.4% (790/4073 turns) to 6.9% (236/3419 turns; odds ratio 0.30, 95% CI 0.24-0.37; P<.001), with suboptimal clinical recommendations (not optimally tailored to the child’s age or clinical context) remaining the most frequent category. CONCLUSIONS This real-world study suggests that an LLM-based health consultation assistant may complement existing pediatric care, with expert-guided optimization substantially improving response quality. However, the persistence of deficiencies highlights the need for continued monitoring and refinement to ensure clinical reliability. CLINICALTRIAL

BACKGROUND House dust mite (HDM) sensitization commonly begins in early life and contributes to persistent allergic airway inflammation and asthma chronicity. Primary prevention via early-life environmental control is a key pathway to reduce HDM sensitization and asthma risk. OBJECTIVE To characterize child caregivers’ knowledge, attitudes, and practices (KAP) regarding pediatric HDM control using a hybrid literature/expert-driven and social media-driven approach, and examine associations between KAP levels, child age and caregiver social media activity. METHODS This cross-sectional study comprised two interconnected components: (1) mining of content published between August 2023 and July 2025 from five major Chinese social media platforms, analyzed via Latent Dirichlet Allocation (LDA); and (2) a social media-enhanced web-based KAP survey administered in November 2025 to child caregivers in Chongqing, a warm-humid region where HDMs dominate indoor allergens, with participants recruited via local child health facilities. In total, 132,341 social media documents and 2,275 caregivers of children 3 years) and caregiver social media activity (active vs. inactive). RESULTS LDA topic modeling identified five distinct topic clusters in the social media content. Commercial, emotional, and misleading content collectively dominated the information landscape, accounting for 83.3% of included documents, with commercial content often systematically conflating the concepts of “disinfection” and “mite elimination”. Only 16.7% was classified as health educational content focusing on HDM allergy prevention. The average KAP levels of K1, K2, A1, A2, and P domains were 62.9%, 84.7%, 57.0%, 37.8%, and 25.8%, respectively. Social media emerged as the primary knowledge source (80.7%), with methodological knowledge gaps (47.5%) being the top implementation barrier. Caregivers of children ≤3 years had significantly lower self-rated knowledge (23.5% vs. 28.3%, P=.01), stronger endorsement of recommended methods, but also greater information overload (OR 1.39, 95% CI 1.15-1.67, P<.001) and decision difficulties (OR 1.23, 95% CI 1.01-1.52, P<.001). Socially active caregivers showed better performance across multiple items in five domains, but also increased non-recommended practices (ultraviolet irradiation: OR 1.85, 95% CI 1.35-2.53, P<.001) and misconception acceptance (allergy impact exaggeration: OR 1.39, 95% CI 1.04-1.87, P=.03). CONCLUSIONS Complex and suboptimal KAP levels exist, particularly among caregivers of young children (≤3 years). Social media activity associates with both enhanced implementation of control practices and elevated misconception endorsement. These findings reveal critical educational gaps and the necessity of social media intervention. CLINICALTRIAL Not applicable.

In January 2025, China advanced the recommended first pertussis vaccine dose from 3 to 2 months of age. We evaluated the real-world effectiveness and safety of 2-month initiation, which was already practised via pentavalent vaccine schedules prior to the policy change, and assessed the impact of dosing timeliness on pertussis risk. This retrospective birth cohort study linked immunisation, disease surveillance, and adverse event databases in Chongqing, China. Infants born between September 1, 2023, and March 1, 2024, who completed an unmixed three-dose primary series of pertussis-containing combination vaccine were followed until September 1, 2024. We compared 2-month with 3-month initiation for pertussis incidence and adverse events following immunisation (AEFI); infants initiating at other ages (9.3%) or infected before dose 1 were excluded from effectiveness analyses. Among 76,844 eligible infants, 2-month initiation reduced the pre-vaccination infection window: pertussis before dose 1 occurred in 3 (0.02%) versus 38 (0.07%) in the 3-month group. Of 69,612 infants in effectiveness analyses (13,815 and 55,797 in the 2-month and 3-month groups, respectively), post-primary series incidence was similar between groups (6.99 versus 6.80 per 1000 person-years). Timely inter-dose intervals were associated with lower pertussis risk regardless of initiation age (3.99 versus 10.66 per 1000 person-years for on-time versus non-adherent schedules), regardless of initiation age. AEFI were rare (0.34%) and comparable between groups, with no serious vaccine-related events. Two-month initiation provides equivalent protection and safety while narrowing the vulnerable pre-vaccination window. Timely inter-dose intervals are as critical as early initiation, a finding with global relevance. These data support China’s 2025 policy change. The online version contains supplementary material available at 10.1186/s12889-026-27239-x.

Sleep disturbances significantly impact children with autism spectrum disorder (ASD), yet their heterogeneous manifestations remain poorly understood. This multicenter prospective cohort study employed latent profile analysis to identify distinct sleep phenotypes among 631 children with ASD (aged 3–6 years) and 768 typically developing (TD) controls across three Chinese cities representing Northern, Central, and Western regions. Analysis of Children's Sleep Habits Questionnaire data revealed three distinct sleep phenotypes based on optimal model fit determined by Bayesian information criterion. Compared to TD children who showed generally better sleep patterns with lower sleep onset delay and fewer disturbances overall, the ASD groups exhibited distinctive profiles: Cluster 1 (9.2%) exhibited severe disturbances across multiple domains (sleep anxiety, parasomnias, night wakings and sleep‐disordered breathing) and demonstrated the most severe autism symptoms; Cluster 2 (36.0%) presented a mixed profile with comparable bedtime resistance, sleep duration, and daytime sleepiness to TD children but elevated sleep‐disordered breathing; and Cluster 3 (54.8%) showed reduced sleep‐disordered breathing but elevated night waking and bedtime resistance. One‐year follow‐up data indicated that Cluster 3, characterized by mild sleep‐disordered breathing, showed significant improvements in core symptoms particularly in social cognition, communication, and motivation domains, whereas Clusters 1 and 2 demonstrated modest changes. These findings suggest that early identification of sleep phenotypes may predict treatment response and inform personalized intervention strategies. Our results underscore the importance of incorporating comprehensive sleep assessment and management into ASD care protocols.