脑震荡后返校学习AI信息评价
脑震荡后返校学习(RTL)的定义口径、时长与预测因素实证研究
共同点:均以RTL这一核心结局变量为因变量,采用系统综述、回顾性队列与大样本登记数据,界定RTL的定义差异(任何形式返校 vs 无特殊照顾的完全返校)、量化返校天数,并检验性别、初始症状严重度、精神心理病史、恐惧回避信念、学校层级与运动级别等预测因素。研究重点是为“AI给出的返校时间与返校准入门槛是否与真实恢复进程一致”提供循证对照基准。
- Return-to-learn following sport-related concussion: a systematic review.(A. Bishay, O. Jallow, J. Jo, K. Williams, Eunyoung Hong, Jackson H. Allen, S. Zuckerman, D. Terry, 2025, Journal of Neurosurgery: Pediatrics)
- Return-to-learn after sport-related concussion: does school level matter?(J. Jo, K. Williams, Ty Hill, G. Perry, O. Prosak, A. Amedy, T. Anesi, D. Terry, S. Zuckerman, 2023, Journal of neurosurgery. Pediatrics)
- Examining for gender differences in return to learn following sport-related concussion in high school student athletes.(J. Jo, J. Dugan, G. Rigney, K. Williams, P. Berkner, G. Iverson, S. Zuckerman, D. Terry, 2024, Neurosurgical Focus)
- What factors must be considered in ‘return to school’ following concussion and what strategies or accommodations should be followed? A systematic review(L. Purcell, G. Davis, G. Gioia, 2018, British Journal of Sports Medicine)
- Concussion Return to Learn or Work and Return to Play(Suraj A. Achar, William Ward, Rachel Buehler Van Hollebeke, 2020, No journal)
- Fear avoidance and return to work after mild traumatic brain injury(D. Snell, Josh W. Faulkner, J. Williman, N. Silverberg, A. Theadom, L. Surgenor, J. Hackney, R. Siegert, 2023, Brain Injury)
- Factors Associated With Return-to-Learn Duration After Concussion in US Service Academy Cadets and Midshipmen.(Jeremy D. Ross, Michael J. Aderman, M. Roach, Steven R. Malvasi, A. Susmarski, R. Brodeur, G. McGinty, Joel B. Robb, Jonna Jackson, S. Broglio, M. McCrea, T. McAllister, P. Pasquina, K. Cameron, 2026, The journal of head trauma rehabilitation)
- Factors Influencing Time to Return to Learn Among NCAA Student-Athletes Enrolled in the Concussion Assessment, Research, and Education (CARE) Study(A. Memmini, T. Snedden, A. Boltz, Benjamin A. Benson, Eric Margolin, P. Pasquina, T. McAllister, M. McCrea, S. Broglio, A. Hoy, Louise A. Kelly, J. Ortega, J. Miles, Nicholas L. Port, S. Chrisman, D. Langford, J. McDevitt, C. Master, Patrick G. O'Donnell, Christopher J D'Lauro, G. McGinty, K. Cameron, A. Susmarski, Joshua T. Goldman, H. Benjamin, T. Buckley, T. Kaminski, J. Clugston, Luis A. Feigenbaum, J. Eckner, J. Mihalik, A. Kontos, A. Brooks, S. Rowson, L. Lintner, Chris Miles, Jody Harl, Janetta Matesan, M. Menser, A. Rettmann, N. L'heureux, M. McEachern, M. Jarrett, Vibeke Brinck, Bianca Byrne, Melissa M. Baker, Christy Collins, William Felix, Bethany Morath, 2024, Sports Medicine)
RTL与重返运动(RTS)的并行进程、急性症状群分化与校内恢复监测
共同点:不再把返校视为单一时间点,而是把RTL与RTS作为相互牵制、存在先后依赖的两条进程加以刻画;同时用急性期认知症状群、初始症状负荷预测恢复分支,并在中学现场实施系统的症状追踪与阶段性返校评估。研究重点是为AI信息中“分阶段、可监测、需动态调整”的返校建议提供依据,并揭示单纯症状消退不足以代表学业准备度。
- Associations of Concussion Return-to-Sport and Return-to-Learn Strategies with Demographics in Parents of Child Athletes(R. Moran, 2025, Journal of Science and Medicine in Sport)
- 7.30 Acute cognitive symptom cluster predicts return-to-learn duration following a sport-related concussion(A. Tang, P. Davis, Kristen Williams, A. Grusky, K. Hajdu, Brian Q. Hou, A. Yengo-Kahn, S. Zuckerman, D. Terry, 2024, Second Round Abstract Submissions)
- Clinical recovery from concussion–return to school and sport: a systematic review and meta-analysis(M. Putukian, L. Purcell, K. Schneider, A. Black, J. Burma, A. Chandran, A. Boltz, C. Master, J. Register-Mihalik, V. Anderson, G. Davis, P. Frémont, J. Leddy, D. Maddocks, Z. Premji, P. Ronksley, S. Herring, S. Broglio, 2023, British Journal of Sports Medicine)
- Investigating the Student in Returning to Learn After Concussion: A Systematic Review and Meta-Analysis.(Ya-Lian Pei, Amy M. Kemp, Katy H. O'Brien, 2023, Journal of School Health)
- Assessment Of A Middle School Return To Learn Concussion Protocol(Aaron Yaksich, Laura L. Harris, 2026, Journal of Sports Medicine and Allied Health Sciences: Official Journal of the Ohio Athletic Trainers' Association)
症状报告谱、临时性住宿措施使用与学业结局(成绩、缺勤、住宿实效)
共同点:聚焦“学生实际报告了什么症状、获得了什么支持、产生什么学业后果”的闭环,涵盖学生自评RTL护理方案的使用过程、个别化临时住宿措施的类型与落实、脑震荡对成绩与出勤的可量化影响、住宿措施是否真能加速恢复,以及中学后阶段可用支持与资源的系统梳理。研究重点是把学业端结局作为评价AI返校建议“可操作性”和“落地收益”的判据。
- Student concussion symptoms and tailored accommodations during use of a return to learn program in 13 public high schools(Erik B Philipson, Aspen D. Avery, Julian Takagi-Stewart, Qian Qiu, Thomas M. Jinguji, D. Coppel, M. Vavilala, 2024, Pm & R)
- Concussion symptoms and temporary accommodations using a student-centered return to learn care plan(Erik B Philipson, Emma L. Gause, K. Conrick, Scott Erickson, A. Muma, Zhinan Liu, Rajiv C Ayyagari, M. Vavilala, 2021, NeuroRehabilitation (Reading, MA))
- Academic outcomes and accommodations following adolescent sport-related concussion: a pilot study(K. Russell, Erin Selci, S. Chu, Adrian Rozbacher, Michael J. Ellis, 2017, Concussion)
- Academic outcomes following adolescent sport-related concussion or fracture injury: A prospective cohort study(K. Russell, Erin Selci, Brian Black, Karis Cochrane, Michael J. Ellis, 2019, PLoS ONE)
- Do academic accommodations help students recover following sport-related concussion? A retrospective study of 96 youth athletes.(G. Rigney, J. Jo, Carter Burns, K. Williams, D. Terry, S. Zuckerman, 2023, Journal of neurosurgery. Pediatrics)
- The effects of concussive brain injury on learning and memory in rats: a short and mid-term post-concussion investigation(Chang-Mao Guo, Xiao-hu Xu, Guang-Fan Ma, 2003, No journal)
- Rapid accommodations for university students with concussion are feasible: nationwide return-to-learn perspectives of accessibility staff(Zachary W. Bevilacqua, Nicole Mucica, Kaitlyn Bjelko, C. Finn, 2026, Frontiers in Neurology)
- Available supports and resources for postsecondary students with traumatic brain injury: A systematic review of the literature(Emily J. Tarconish, A. Lombardi, Joseph W. Madaus, Ashley V. Taconet, C. Coelho, 2021, Journal of Vocational Rehabilitation)
临床管理路径、症状恢复轨迹异质性与神经认知/眼动评估技术
共同点:提供返校决策的医学与神经科学底座——权威临床实践指南与分级推荐、基于组轨迹建模揭示的症状恢复异质性、认知控制相关脑功能改变,以及传统神经心理学测验的局限与眼动追踪、认知—运动整合指标、可解释机器学习等新型评估手段;并延伸到神经心理测评如何转化为具体课堂住宿建议、门诊多学科主动康复范式。研究重点是为判断AI信息的专业准确性、时效性与“能否落到具体住宿措施”提供知识标尺。
- The management and rehabilitation of post-acute mild traumatic brain injury(B. Eapen, Amy O. Bowles, J. Sall, A. E. Lang, Carrie W. Hoppes, Katharine C. Stout, T. Kretzmer, D. Cifu, 2022, Brain Injury)
- Longitudinal Trajectories of Post-Concussive Symptoms Following Mild Traumatic Brain Injury(Eva S. Keatley, K. Bechtold, K. Psoter, Matthew E. Peters, A. Everett, V. Rao, T. van Meter, Hayley J. Falk, F. Korley, D. Roy, 2023, Brain Injury)
- Neural correlates of cognitive control deficits in pediatric mild traumatic brain injury(H. J. van der Horn, A. Dodd, Tracey V. Wick, C. Robertson-Benta, Jessica R. McQuaid, Anne K. Hittson, J. Ling, V. Zotev, S. Ryman, E. Erhardt, John P. Phillips, Richard A. Campbell, R. Sapién, Andrew R. Mayer, 2023, Human Brain Mapping)
- What traditional neuropsychological assessment got wrong about mild traumatic brain injury. IV: clinical applications and future directions(E. Bigler, Steven Allder, B. Dunkley, Jeff Victoroff, 2025, Brain Injury)
- Integration of eye-tracking systems with sport concussion assessment tool 5th edition for mild TBI and concussion diagnostics in neurotrauma: Building a framework for the artificial intelligence era(A. M. Fiedler, R. Anghinah, Fernando De Nigris Vasconcellos, A. Morell, Timoteo Almeida, Bernardo de Assumpção, J. Graciolli Cordeiro, 2023, Neuroscience Informatics)
- Predicting NHL Performance from Cognitive-Motor Metrics After Concussion: An Explainable AI Approach(M. Moschitto, L. Sergio, Marin Litoiu, Joydeep Mukherjee, S. Mukhopadhyay, 2025, 2025 International Conference on Artificial Intelligence, Computer, Data Sciences and Applications (ACDSA))
- The Role of Neuropsychology in the Evaluation of Concussion.(N. McGrath, J. Eloi, 2019, Seminars in Pediatric Neurology)
- Outpatient Management of Sport-Related Concussion, Return to Learn, Return to Play.(P. Kriz, James P. MacDonald, 2021, Clinics in Sports Medicine)
脑震荡风险识别、预防策略与面向儿童青少年人群的中文实践与教育文献
共同点:从风险识别与健康教育侧切入目标人群——运动性脑震荡风险检查表/评估表编制与三维预防策略、获得性脑损伤儿童的教育安置与学习支持、小儿脑震荡与球类运动脑震荡的中文临床教育类文献。研究重点是为AI信息评价提供“目标读者画像+专业术语底座+本土语境”,并揭示中文科普语料中常见的表述模糊与建议粗放问题。
- 体育类专业大学生运动性脑震荡风险识别及预防策略研究(Hao Yang, 小琴 彭, 都娅 许, 红平 黄, 2026, 学知 Xue Zhi)
- Educating children with acquired brain injury(S. Walker, B. Wicks, 2005, No journal)
- 小儿脑震荡352例分析(Bao Yang, Peng Liu, Hong-Sheng Guo, 1997, No journal)
- 脑震荡早期血清TNF-α和IL-1β的变化(Yuancheng Zhang, Jin-Quan Zhang, An-Ping Wang, Xing-Yi Xiao, 2007, No journal)
- 头球潇洒,但有风险——给青少年足球爱好者提个醒(Ran Zhuo, 2014, No journal)
脑震荡临床治疗技术与共病焦虑抑郁的心理护理干预
共同点:属于临床治疗与护理干预分支,包括病例系列总结的临床特征、以EEG脑生物标志物支持的高压氧治疗疗效评估,以及针对脑震荡后焦虑伴抑郁患者的心理护理对情绪与生活质量的作用。研究重点在于刻画“生理—心理”双重前置条件:返校学习能力不仅取决于症状消退,还受情绪共病与治疗进程制约,AI信息若忽略心理共病维度即构成内容缺陷。
- 对120例脑震荡患者焦虑抑郁综合征的心理干预及效果评价(Gui-Ling Wang, Ming-Tao Xu, Xu-Jun Di, 2013, International Journal of Nursing)
- EEG-based brain biomarker supports hyperbaric oxygen therapy for acute concussions.(D. W. Denham, Menley A. Denham, 2025, Undersea & Hyperbaric Medicine)
- 探究心理护理对脑震荡后焦虑症伴抑郁症患者的临床影响(Zong-Mei Xu, 2025, 国际护理与健康)
州级与校级RTL立法、指南文本分析及健康公平差异
共同点:从政策与制度层面进行文本与内容分析——各州青少年脑震荡立法是否包含面向全体学生(而非仅运动员)的RTL条款、州卫生部门RTL指南的传播与可获得性、校园RTL政策分析、高校RTL团队建设倡议、性侵与运动脑震荡政策的交叉比较,以及社会决定因素(SDoH)在RTL研究中的系统性缺位。研究重点是为AI信息评价补充“制度合规性与公平性”维度:同一份AI建议对不同州、不同学段、不同资源条件学生的可执行性差异显著。
- It is time to rewrite state youth sports concussion laws(J. Howland, H. Hackman, A. Torres, Julia K. Campbell, J. Olshaker, 2021, BMJ Open Sport & Exercise Medicine)
- Return-to-learn after concussion in Washington state public high schools during the COVID-19 pandemic(Shyam J. Deshpande, Aspen D. Avery, Julian Takagi-Stewart, Brianna Mills, Q. Qiu, M. Vavilala, 2023, Concussion)
- An Evaluation of Post-Concussion Return to School Guidelines: A Survey of Massachusetts School Nurses(J. Howland, Julia K. Campbell, Rebekah J Thomas, Steven Smyth, Aynsley Chaneco, A. Torres, J. Olshaker, 2021, Journal of School Nursing)
- Policy Analysis of Return to Learn After Sport and Recreational Related Concussion for Secondary Schools in New England: Relevance to School Nurses and Nursing Practice(Joseph M. Fetta, A. Starkweather, T. V. van Hoof, Robert A. Huggins, D. Casa, Jessica Gill, 2023, Policy, Politics & Nursing Practice)
- Commentary: Establishing the college Return to Learn team for concussion: a practical approach(Zachary Bevilacqua, Jacob I. McPherson, 2023, Frontiers in Public Health)
- Comprehensive policies for victims of sexual assault returning to the campus classroom: Lessons from university sports-related concussion policies(Elizabeth A. Moschella, Cheyenne Quilter, S. Potter, 2021, Journal of American College Health)
- A - 13 Social Determinants of Health and Health Equity in Return to Learn Following Sport-Related Concussion(A. Kissinger-Knox, K. Stephenson, J. Santangelo, K. Adubofour, G. Iverson, N. Cook, 2024, Archives of Clinical Neuropsychology)
RTL协议的实施科学评估与跨专业、跨机构协作机制
共同点:以实施科学与协作网络视角,考察RTL协议在中学与高校落地时的可行性、可接受性、适宜性与组织变革准备度,识别沟通链条断裂、教职员工培训不足、角色边界模糊等障碍,并呈现学业支持人员、教师运动代表、运动防护师之间的协作诉求与社区参与的转化模型。研究重点是为AI信息工具界定“谁在用、在什么流程节点用、需要与谁对接”,避免输出与实际工作流脱节的建议。
- Implementation of Return to Learn Protocols for Student Athletes with Sport and Recreation Related Concussion: An Integrative Review of Perceptions, Challenges and Successes(Joseph M. Fetta, A. Starkweather, Robert A. Huggins, T. V. van Hoof, D. Casa, J. Gill, 2021, Journal of School Nursing)
- Considerations for Implementing the Post-Concussion Collegiate Return-to-Learn Protocol in the National Collegiate Athletic Association Power 5 Conferences(A. Memmini, D. Kinnett-Hopkins, R. Hasson, S. Rifat, S. Broglio, 2023, The journal of head trauma rehabilitation)
- Academic Professionals’ Perspectives on Return to Learn Following Concussion in NCAA Student-Athletes(T. Bowman, M. Lininger, Alicia M. Pike Lacy, J. Register-Mihalik, Steven Myers, 2025, Journal of issues in intercollegiate athletics)
- Using a Community-Informed Translational Model to Prioritize Translational Benefits in Youth Concussion Return-to-Learn Programs(Julian Takagi-Stewart, Aspen D. Avery, Shyam J. Deshpande, Stephanie Andersen, T. Combs, M. Vavilala, Laura C. Prater, 2023, Health Promotion Practice)
教师、学校管理者与家长对脑震荡及住宿措施的认知、角色体验与培训转化
共同点:以问卷调查与质性访谈/主题分析为主要方法,从一线执行者与信息使用者视角揭示RTL落地的真实障碍——对脑震荡与住宿措施的知晓度与误解、角色分工不清、政策知晓度低、沟通与资源不足带来的信心不足与挫败感,以及教师培训项目对知识、态度与实际支持行为的改善效果。研究重点是为“AI信息面向学校人员时应采用的表述方式、颗粒度与可信度提示”提供需求侧证据。
- Facilitating students’ return to school following a concussion: Perspectives of Canadian teachers and school administrators(H. A. Shepherd, I. Shill, J. Caron, N. Reed, K. Yeates, Kathryn J. Schneider, A. Black, C. Emery, 2023, Cogent Education)
- The Experiences of Collegiate Professors in Providing Academic Accommodations to Students Who Have Suffered a Concussion(Nicholas D. Wright, Angela R. Merlo, M. Romero, R. Foster, Brittany A. Clason, 2025, Journal of Rehabilitation Practices and Research)
- Survey About Academic Accommodations Provided to Students with Concussions(Brad Nakamura, R. Carl, Adam J. Potteiger, 2018, Pediatrics)
- Academic Accommodations for Middle and High School Students Following a Concussion: Perspectives of Teachers and School Administrators.(H. A. Shepherd, K. Yeates, N. Reed, J. Caron, Kathryn J. Schneider, C. Emery, A. Black, 2023, Journal of School Health)
- Return-to-Learn Accommodations and Concussion Perceptions Among Indiana High School Principals(Isaac A. Janson, Vincent C. Nittoli, Derek White, K. Tekulve, 2019, The journal of head trauma rehabilitation)
- Return to school after concussion: perceptions of school administrators(M. McCart, G. DeSalvo, Jody Slocumb, P. Reyna, Deanne K. Unruh, Jeff M. Gau, A. Glang, 2026, Preventing School Failure: Alternative Education for Children and Youth)
- Effects of a faculty training program on knowledge and awareness of student concussion and academic performance(C. Finn, 2020, Journal of American College Health)
- Mild Traumatic Brain Injury: Facilitating School Success(K. Hux, Carolyn Hacksley, 1996, Intervention in School and Clinic)
面向患者的LLM输出安全性、准确性与共情沟通的横断面比较评价
共同点:采用标准化患者提问—多模型输出—医师/专家盲评的范式,以Likert量表与安全性、准确性、共情沟通维度,系统比较ChatGPT、Gemini、Copilot、DeepSeek、Claude等模型与常规搜索引擎的应答差异,并检验不同诊断/健康场景下的表现波动与潜在风险。研究重点是可迁移到脑震荡返校信息评价的核心指标集与“流畅但可能不安全、共情不足”的共性缺陷判定。
- Large language model chatbots for patient-facing erectile dysfunction information: a cross-sectional evaluation of safety, accuracy, empathy, information quality and readability.(Da-Lian Liang, Zheng-Kun Zhang, Wen-Li Zhao, Jing-Li Shang, Hai-Yan Li, 2026, Sexual Health)
- Large language models as sources of diabetic retinopathy patient education: a comparative evaluation of safety, health information quality, and readability(Zhi-Qi Chen, Wen Liu, Jing Zhang, Tian Zheng, Ying Li, Qing Bao, Ping-Ting Ji, Hong-Mei Yao, Ying He, 2026, Frontiers in Public Health)
- Examining the Performance of Large Language Models in Health Information Quality Evaluation Tasks: An AI Post-Alignment Perspective(Shenghui Zhang, Jun Zhang, Peng Li, Min Zuo, Liqiang Feng, 2026, Information)
- Artificial intelligence large language model ChatGPT: is it a trustworthy and reliable source of information for sarcoma patients?(Marisa Valentini, J. Szkandera, M. Smolle, S. Scheipl, A. Leithner, D. Andreou, 2024, Frontiers in Public Health)
- Safety, accuracy, empathic communication, information quality, and readability of five large language model interfaces answering public questions about interstitial cystitis/bladder pain syndrome(Jiang-Tao Zhu, Zhen-Hua Zhao, Song Li, Wen-Bin Li, Jie Guo, Jia-Ning Li, Yu-Jie Ma, 2026, Frontiers in Public Health)
- Evaluating Large Language Models for Diagnostic Accuracy and Health Information Quality in Oral Mucosal Diseases(Melisa Iacob, Ayham Qawas, R. Balasubramaniam, A. Frydrych, O. Kujan, 2026, Journal of Personalized Medicine)
以DISCERN/EQIP/JAMA/GQS与可读性指数为核心的信息质量评价及专科化、非英语模型证据
共同点:以标准化信息质量工具与可读性指数为核心测量框架,横向比较不同大模型在信息完整性、来源透明度、可读性、共情与信任度上的表现,并扩展到中文语料、月经健康等专科教育场景以及微调/基座模型差异、患者与公众信任感知,检验微调数据质量对表现的放大或削弱作用。研究重点是为脑震荡返校信息建立“工具化、可复算”的评分规程与阈值经验。
- A cross-sectional evaluation of large language model chatbot interfaces for patient-facing herpes zoster information: safety, information quality, and readability(Da-Lian Liang, Cong Mai, Zheng-Kun Zhang, Jie-Yun Lin, X. Wu, Hai-Ying Lu, Jing-Li Shang, Hua-Jun Wang, Hai-Yan Li, 2026, Frontiers in Public Health)
- Information quality, readability, and empathy of AI-generated public mental health information: a comparative evaluation of eight large language models(Qi-Long Wang, Si-Heng Ma, Jian Sun, Xin Qi, Jian-Peng Zhao, Dong Ma, Xiao-Fei Liu, Qiang An, Dong-Rong Zhao, Songkai He, 2026, Frontiers in Public Health)
- Evaluating Chinese large language models for HIV/AIDS health information: a multidimensional comparative study of response quality(Jin-Hao Gu, Jian-Wei Liu, Yun-Rong Liu, Ping-Ping Wang, De-Hua Hu, Qi-Zhen Zhu, 2026, Frontiers in Public Health)
- Menstrual Health Education Using a Specialized Large Language Model in India: Development and Evaluation Study of MenstLLaMA(P. Adhikary, I. Motiyani, G. Oke, M. Joshi, K. Pathak, Salam Michael Singh, Tanmoy Chakraborty, 2025, Journal of Medical Internet Research)
- Assessing the Quality of Mental Health Support in LLM Responses through Multi-Attribute Human Evaluation(Abeer Badawi, Md Tahmid Rahman Laskar, Elahe Rahimi, Sheri Grach, Lindsay Bertrand, L. Danok, F. Rudzicz, J. Huang, Elham Dolatabadi, 2026, arXiv.org)
- Trust, Safety, and Accuracy: Assessing LLMs for Routine Maternity Advice(V. Divya, A. Bhanusree, Rimjhim, K. Rao, 2026, arXiv.org)
- Evaluating Fine-Tuned and Base Language Models in Maternal and Vaccination Healthcare for African Settings(A. Ajibade, Oluwaseun Odunsi, Iyinoluwa Animasaun, Chioma Nwakanma-Akanno, Oluwasegun Oguntuase, Oluwafunke Akinbuwa, Abiodun Adereni, 2026, arXiv.org)
- Artificial intelligence for automating the establishment of an Arabic benchmark dataset for enhancing health information quality assessment(Yousef Baqraf, P. Keikhosrokiani, Y. Cheah, Amany Alomarji, Fatima Baqraf, 2025, Discover Artificial Intelligence)
个性化生成与多语言、语域敏感的读者适配评价
共同点:关注“同一信息对不同读者是否讲得对”,通过读者画像驱动的个性化简明语言摘要实验,以及英语/法语/阿拉伯语多语域(专业、通俗、含糊、情绪化、信息不足)对照评价,揭示理解力提升与偏见强化、幻觉及安全失败之间的权衡。研究重点直接对应返校场景的核心难点:学生、家长、教师与校医的信息需求与语言水平差异极大,评价指标必须包含“适读性而非仅正确性”。
- ReLay: Personalized LLM-Generated Plain-Language Summaries for Better Understanding, but at What Cost?(Joey Chan, Yi-Kun Han, Jingyuan Chen, Samuel Fang, Lauren D. Gryboski, Alexandria Lee, Sheel Tanna, Qing-Qing Zhu, Zhi-Yong Lu, Lucy Lu Wang, Yue Guo, 2026, arXiv.org)
- HerHealthEval: Evaluating Multilingual and Register-Sensitive Understanding of Women's Health Communication(Hassan Saeed Hassan Albattra, Mazen Bahgat, Rahatara Ferdousi, Hana Essam Sayed Ahmed Amrya, Mariam Mousa, 2026, arXiv.org)
自动评价方法学(LLM-as-a-Judge)、输出可复现性与用户采纳信任
共同点:从评价工具与可信度治理角度切入——以LLM-as-a-Judge对齐专家评分实现可扩展自动评价、以重复推断刻画小参数开源模型输出的可复现性缺口,并考察患者与医务人员对LLM参与信息获取与共同决策的态度与信任差异,以及患者社群如何接受、转译与消解AI输出的确定性。研究重点是为构建可持续、可审计、被使用者接受的AI信息评价机制提供方法学与治理路径。
- Automating Evaluation of AI Text Generation in Healthcare with a Large Language Model (LLM)-as-a-Judge(E. Croxford, Yan-Jun Gao, E. First, Nicholas Pellegrino, Miranda Schnier, J. Caskey, M. Oguss, G. Wills, Guan-Hua Chen, D. Dligach, M. Churpek, A. Mayampurath, Frank J. Liao, Cherodeep Goswami, Karen K. Wong, Brian W. Patterson, M. Afshar, 2025, medRxiv)
- Evaluating Small Open LLMs for Medical Question Answering: A Practical Framework(A. Buskila, 2026, arXiv.org)
- Attitudes toward large language model-based Artificial Intelligence systems as an information source for shared decision-making in radiation oncology(R. Moser, L. M. Buchecker, J. Nano, N. Mayr, Sophie T. Behzadi, Sophia Kiesl, S. Maier, L. Allwohn, J. Lammert, L. Adams, M. Tschochohei, S. Combs, K. Borm, 2025, The Oncologist)
- Catch the Patient, Not the AI: Collective Sensemaking in an Online Health Community(Feng He, 2026, arXiv.org)
AI辅助文献综述、证据整合与自动化事实核查的方法学
共同点:把AI当作“研究工具”而非信息源本身,包括用LLM聊天机器人加速文献综述并配套人工事实核查协议、以及基于强化学习与PubMed检索构建结构化事实核查报告生成系统。研究重点是为本主题提供可复用的工作流模板:如何让AI生成脑震荡RTL信息的评价结论并逐条回溯至原始证据,从而保证报告的可核验性。
- Using an Artificial intelligence chatbot to critically review the scientific literature on the use of Artificial intelligence in Environmental Impact Assessment(A. Bond, D. Cilliers, F. Retief, R. Alberts, C. Roos, J. Moolman, 2024, Impact Assessment and Project Appraisal)
- Generating Biomedical Fact-Checking Reports with RL-Enhanced Agentic Search(Jiong-Xiao Wang, Di Ma, Chao-Qun Ni, 2026, arXiv.org)
AI在运动性脑震荡管理与教育学习场景中的应用范式、可解释性与公平性框架
共同点:把视野从“评价既有模型输出”拓展到“AI在特定领域的应用范式与评价框架”——运动性脑震荡全路径AI应用(检测诊断、监测、预后、风险建模)的范围综述、AIED中基于优势(asset-based)而非缺陷取向的学习者支持与评估范式、低资源语言与多元文化群体的健康虚假信息检测治理框架,以及孤独症谱系等特殊人群AI干预中的可解释性与伦理考量。研究重点是为面向返校学生群体设计AI信息评价提供场景化、可解释与公平性视角,并覆盖被前组忽略的特殊人群与低资源语种议题。
- Artificial intelligence applications in sport-related concussion: an updated scoping review.(M. Vecchiato, Filippo Dal Ponte, E. Zanardo, Stefano Palermi, Marco Bazo, D. Neunhaeuserer, A. Ermolao, S. Bressan, V. Baioccato, 2026, Journal of Science and Medicine in Sport)
- Toward Asset-based Instruction and Assessment in Artificial Intelligence in Education(Jaclyn L. Ocumpaugh, Rod D. Roscoe, Ryan S. Baker, Stephen J. Hutt, S. Aguilar, 2024, International Journal of Artificial Intelligence in Education)
- Evaluating Health Misinformation in Low-Resource Languages: Integrating Small Language Models with a Culturally-Sensitive Responsible NLP Framework (Bangla as a Case Study)(Farnaz Farid, Raihan Alam, Ala Al-Areqi, Farhad Ahamed, Muhammad Hassan Khan, S. Hossain, Irena Veljanova, A. Hossain, 2026, arXiv.org)
- Use Artificial Intelligence in Autism Spectrum Disorder: a comprehensive review in assessment, intervention, and data collection(Muna Abualhoul, 2025, Journal of Information Technology, Cybersecurity, and Artificial Intelligence)
合并后共15个平行分组、覆盖82篇文献,形成三条相互支撑的证据链。第一条是“脑震荡与返校学习的医学—结局链”(第1—6组):从RTL的定义口径、时长与预测因素,到RTL与RTS并行进程及急性症状群,再到症状报告、临时住宿措施与学业结局,并以临床管理指南、症状恢复轨迹、神经认知与眼动评估技术,以及风险识别/预防、治疗与共病心理干预作为专业知识底座。第二条是“校园落地与制度链”(第7—9组):州级与校级立法及指南文本分析、健康公平与社会决定因素缺位,实施科学评估与跨专业协作机制,教师、管理者与家长的认知、角色体验与培训转化,共同界定AI信息的使用者、执行情境与实际障碍。第三条是“AI信息评价方法链”(第10—14组):面向患者的LLM输出安全性/准确性/共情横断面比较,以DISCERN/EQIP/JAMA/GQS与可读性指数为核心的信息质量评价及专科化、非英语模型证据,个性化生成与多语言语域敏感适配,LLM-as-a-Judge自动评价、可复现性与用户信任,以及AI辅助文献综述与自动化事实核查的方法学模板。第15组则把视野拓展到AI在运动性脑震荡管理与教育学习场景中的应用范式、可解释性与公平性框架。相较三份初始化分组,本次合并补充/显化了此前分散的论点:RTL与RTS耦合进程、神经心理测评向课堂住宿的转化、AI评价的读者适读性与语域敏感、自动评价的可复现性缺口、用户与患者社群对AI确定性的消解,以及特殊人群与低资源语种的公平性议题。三条证据链的交汇处即是研究空白:目前尚无直接针对“脑震荡后返校学习”这一特定场景的AI信息质量评价研究,需以前两条链确立专业正确性与可执行性基准,以第三条链确立评价维度、测量工具与治理流程。
共同点:均以RTL这一核心结局变量为因变量,采用系统综述、回顾性队列与大样本登记数据,界定RTL的定义差异(任何形式返校 vs 无特殊照顾的完全返校)、量化返校天数,并检验性别、初始症状严重度、精神心理病史、恐惧回避信念、学校层级与运动级别等预测因素。研究重点是为“AI给出的返校时间与返校准入门槛是否与真实恢复进程一致”提供循证对照基准。相关文献:A. Bishay et. al, 2025 等 8 篇文献
共同点:不再把返校视为单一时间点,而是把RTL与RTS作为相互牵制、存在先后依赖的两条进程加以刻画;同时用急性期认知症状群、初始症状负荷预测恢复分支,并在中学现场实施系统的症状追踪与阶段性返校评估。研究重点是为AI信息中“分阶段、可监测、需动态调整”的返校建议提供依据,并揭示单纯症状消退不足以代表学业准备度。相关文献:R. Moran et. al, 2025 等 5 篇文献
共同点:聚焦“学生实际报告了什么症状、获得了什么支持、产生什么学业后果”的闭环,涵盖学生自评RTL护理方案的使用过程、个别化临时住宿措施的类型与落实、脑震荡对成绩与出勤的可量化影响、住宿措施是否真能加速恢复,以及中学后阶段可用支持与资源的系统梳理。研究重点是把学业端结局作为评价AI返校建议“可操作性”和“落地收益”的判据。相关文献:Erik B Philipson et. al, 2024 等 8 篇文献
共同点:提供返校决策的医学与神经科学底座——权威临床实践指南与分级推荐、基于组轨迹建模揭示的症状恢复异质性、认知控制相关脑功能改变,以及传统神经心理学测验的局限与眼动追踪、认知—运动整合指标、可解释机器学习等新型评估手段;并延伸到神经心理测评如何转化为具体课堂住宿建议、门诊多学科主动康复范式。研究重点是为判断AI信息的专业准确性、时效性与“能否落到具体住宿措施”提供知识标尺。相关文献:B. Eapen et. al, 2022 等 8 篇文献
共同点:从风险识别与健康教育侧切入目标人群——运动性脑震荡风险检查表/评估表编制与三维预防策略、获得性脑损伤儿童的教育安置与学习支持、小儿脑震荡与球类运动脑震荡的中文临床教育类文献。研究重点是为AI信息评价提供“目标读者画像+专业术语底座+本土语境”,并揭示中文科普语料中常见的表述模糊与建议粗放问题。相关文献:Hao Yang et. al, 2026 等 5 篇文献
共同点:属于临床治疗与护理干预分支,包括病例系列总结的临床特征、以EEG脑生物标志物支持的高压氧治疗疗效评估,以及针对脑震荡后焦虑伴抑郁患者的心理护理对情绪与生活质量的作用。研究重点在于刻画“生理—心理”双重前置条件:返校学习能力不仅取决于症状消退,还受情绪共病与治疗进程制约,AI信息若忽略心理共病维度即构成内容缺陷。相关文献:Gui-Ling Wang et. al, 2013 等 3 篇文献
共同点:从政策与制度层面进行文本与内容分析——各州青少年脑震荡立法是否包含面向全体学生(而非仅运动员)的RTL条款、州卫生部门RTL指南的传播与可获得性、校园RTL政策分析、高校RTL团队建设倡议、性侵与运动脑震荡政策的交叉比较,以及社会决定因素(SDoH)在RTL研究中的系统性缺位。研究重点是为AI信息评价补充“制度合规性与公平性”维度:同一份AI建议对不同州、不同学段、不同资源条件学生的可执行性差异显著。相关文献:J. Howland et. al, 2021 等 7 篇文献
共同点:以实施科学与协作网络视角,考察RTL协议在中学与高校落地时的可行性、可接受性、适宜性与组织变革准备度,识别沟通链条断裂、教职员工培训不足、角色边界模糊等障碍,并呈现学业支持人员、教师运动代表、运动防护师之间的协作诉求与社区参与的转化模型。研究重点是为AI信息工具界定“谁在用、在什么流程节点用、需要与谁对接”,避免输出与实际工作流脱节的建议。相关文献:Joseph M. Fetta et. al, 2021 等 4 篇文献
共同点:以问卷调查与质性访谈/主题分析为主要方法,从一线执行者与信息使用者视角揭示RTL落地的真实障碍——对脑震荡与住宿措施的知晓度与误解、角色分工不清、政策知晓度低、沟通与资源不足带来的信心不足与挫败感,以及教师培训项目对知识、态度与实际支持行为的改善效果。研究重点是为“AI信息面向学校人员时应采用的表述方式、颗粒度与可信度提示”提供需求侧证据。相关文献:H. A. Shepherd et. al, 2023 等 8 篇文献
共同点:采用标准化患者提问—多模型输出—医师/专家盲评的范式,以Likert量表与安全性、准确性、共情沟通维度,系统比较ChatGPT、Gemini、Copilot、DeepSeek、Claude等模型与常规搜索引擎的应答差异,并检验不同诊断/健康场景下的表现波动与潜在风险。研究重点是可迁移到脑震荡返校信息评价的核心指标集与“流畅但可能不安全、共情不足”的共性缺陷判定。相关文献:Da-Lian Liang et. al, 2026 等 6 篇文献
共同点:以标准化信息质量工具与可读性指数为核心测量框架,横向比较不同大模型在信息完整性、来源透明度、可读性、共情与信任度上的表现,并扩展到中文语料、月经健康等专科教育场景以及微调/基座模型差异、患者与公众信任感知,检验微调数据质量对表现的放大或削弱作用。研究重点是为脑震荡返校信息建立“工具化、可复算”的评分规程与阈值经验。相关文献:Da-Lian Liang et. al, 2026 等 8 篇文献
共同点:关注“同一信息对不同读者是否讲得对”,通过读者画像驱动的个性化简明语言摘要实验,以及英语/法语/阿拉伯语多语域(专业、通俗、含糊、情绪化、信息不足)对照评价,揭示理解力提升与偏见强化、幻觉及安全失败之间的权衡。研究重点直接对应返校场景的核心难点:学生、家长、教师与校医的信息需求与语言水平差异极大,评价指标必须包含“适读性而非仅正确性”。相关文献:Joey Chan et. al, 2026 等 2 篇文献
共同点:从评价工具与可信度治理角度切入——以LLM-as-a-Judge对齐专家评分实现可扩展自动评价、以重复推断刻画小参数开源模型输出的可复现性缺口,并考察患者与医务人员对LLM参与信息获取与共同决策的态度与信任差异,以及患者社群如何接受、转译与消解AI输出的确定性。研究重点是为构建可持续、可审计、被使用者接受的AI信息评价机制提供方法学与治理路径。相关文献:E. Croxford et. al, 2025 等 4 篇文献
共同点:把AI当作“研究工具”而非信息源本身,包括用LLM聊天机器人加速文献综述并配套人工事实核查协议、以及基于强化学习与PubMed检索构建结构化事实核查报告生成系统。研究重点是为本主题提供可复用的工作流模板:如何让AI生成脑震荡RTL信息的评价结论并逐条回溯至原始证据,从而保证报告的可核验性。相关文献:A. Bond et. al, 2024 等 2 篇文献
共同点:把视野从“评价既有模型输出”拓展到“AI在特定领域的应用范式与评价框架”——运动性脑震荡全路径AI应用(检测诊断、监测、预后、风险建模)的范围综述、AIED中基于优势(asset-based)而非缺陷取向的学习者支持与评估范式、低资源语言与多元文化群体的健康虚假信息检测治理框架,以及孤独症谱系等特殊人群AI干预中的可解释性与伦理考量。研究重点是为面向返校学生群体设计AI信息评价提供场景化、可解释与公平性视角,并覆盖被前组忽略的特殊人群与低资源语种议题。相关文献:M. Vecchiato et. al, 2026 等 4 篇文献
总计 83 篇相关文献
目的:分析脑震荡后焦虑症伴抑郁症患者的临床护理干预,探讨应用心理护理的效果。方法:从本 院近三年收治的脑震荡患者中,选取2021.05-2024.05期间,确诊为焦虑症伴抑郁症患者50例,以随机数字表法进行 患者分组护理,对照组25例实施常规护理,观察组25例提供心理护理,对比两组心理状况、生活质量及护理满意 度。结果:在护理前,对两组心理、生活质量评估显示,焦虑评分、抑郁评分及WHOQOL评分差异均无统计意义, (p > 0.05);护理后两组不良心理情绪改善,生活质量提高,各评分均要比护理前更优,两组相比下观察组显著优于 对照组,(p < 0.05);调查两组患者对临床护理的满意度得知,观察组更满意,总满意率远远比对照组要高,(p < 0.05);观察组护理质量评分明显比对照组各评分更高,(p < 0.05)。结论:在脑震荡后焦虑症伴抑郁症患者中应用 心理护理,可有效缓解患者的焦虑及抑郁情绪,进一步改善其预后质量,值得推广。
本研究基于风险管理理论,以提升体育类专业大学生风险识别能力与预防意识为目标,编制运动性脑震荡风险检查表与评估表,采用文献资料法、德尔菲法等多种方法识别风险因子、确定核心风险并提出预防策略。结果显示,研究识别出4个一级指标、9个二级指标及36个三级风险因子,核心风险源为学生心理行为偏差、教师教学防护缺失等;构建了学生、教师、环境及其他三个维度的预防策略体系。结论表明,本研究明确了体育类专业大学生运动性脑震荡的风险构成与核心风险源,提出的三维预防策略可为高校相关防控教育提供理论与实证参考,也为运动性脑震荡公共卫生防控提供新思路。
Background The Translational Science Benefit Model (TSBM) was developed to broadly capture systematic measures of health and societal benefits from scientific research, beyond traditional outcome measures. We aimed to develop a systematic process for the application of the TSBM and to then provide an example of a novel application of the TSBM to an ongoing Return-to-Learn (RTL) after youth concussion project involving partnerships with community stakeholders. Methods We invited investigators, project advisory board, and participants of the RTL project to participate in a modified Delphi process. We first generated a list of potential translational benefits using the indicators of the TSBM as guideposts. We then prioritized the benefits on an adapted Eisenhower matrix. Results We invited 35 concussion care or research experts to participate, yielding 20 ranked translational benefits. Six of these recommendations were ranked high priority, six were regarded as investments, and eight were ranked as either low yield or low priority. Discussion This study found that activities such as education and training of stakeholders, development of policy and consensus statements, and innovation in dissemination, were perceived as higher priority than other activities. Our approach using a modified Delphi process and incorporating the TSBM can be replicated to generate and prioritize potential benefits to society from research studies.
Outpatient sports-related concussion (SRC) management continues to evolve as evidence emerges supporting a multidisciplinary approach to the clinical assessment of SRC. Early active rehabilitation has replaced strict cognitive and physical rest. With this paradigm shift in management, pragmatic approaches are highly sought by busy clinicians that provide direction to individualized treatment, which can potentially expedite symptom resolution. Treatment strategies that address domain-based symptom constellations continue to be developed by clinician researchers. Although the optimal timing and dose of these domain-specific therapies has yet to be determined, future directions of SRC treatment will answer these and other questions regarding SRC management.
OBJECTIVE Return-to-learn (RTL) recommendations and management protocols for athletes after sport-related concussion (SRC) remain heterogeneous. With respect to academics after SRC, the current systematic review evaluated 3 outcome measures in each included study: 1) RTL definition, 2) RTL duration, and 3) factors associated with longer RTL. METHODS Studies published before December 2022 that addressed RTL after an SRC were identified via a compressive literature search of the Cochrane Central Register of Controlled Trials, CINAHL, SPORTDiscus, Educational Resources Information Center, Ovid MEDLINE, and PubMed databases. Studies solely examining nonsport-related concussion/traumatic brain injury studies were excluded. Study quality was assessed using the Oxford Centre for Evidence-Based Medicine Criteria and the Modified Newcastle-Ottawa Scale for risk of bias. For each study included, the design, participants, interventions, outcomes, results, and conclusions were extracted. RESULTS Of 1838 studies identified, 21 met the inclusion criteria. These studies represented 8475 athletes (aged 5-27 years) who sustained an SRC. A total of 17/21 (81%) provided an RTL definition. Of the studies that defined RTL as any RTL, the median (range) was 5 (2.5-35.3) days, while the median (range) of those that defined RTL as full RTL without accommodations was 17.5 (3-49) days. Female gender (7/9 studies) and initial symptom severity (3/3 studies) were consistently associated with delayed RTL. Single studies identified anxiety, attention-deficit/hyperactivity disorder (ADHD), and integrated performance management as potential factors for delayed RTL. Age (0/5 studies), concussion history (2/6), depression history (0/1), migraine history (0/2), learning disability (0/1), school type (public vs private) (0/2), and early air travel after SRC (0/1) did not appear to have consistent effects on RTL. CONCLUSIONS Across 21 primary studies, RTL definitions varied greatly, and similarly, a wide range of days to RTL was seen at 3-35 days. Female gender and initial symptom severity were most consistently associated with prolonged RTL. Certain factors such as a history of ADHD/anxiety showed potential effects but require further investigation. Management guidelines should consider these findings while recognizing the need for more comprehensive research to confirm and quantify the impact of various factors on RTL duration.
No abstract available
No abstract available
While evidence supports the gradual return to participation in collegiate athletics following a concussion, limited evidence exists of the return to learn (RTL) procedures for a successful return to the higher education classroom. The objective of this study was to determine what academic supports and hindrances are present for NCAA student-athletes during the RTL process following a concussion from academic professionals’ perspectives. In this qualitative study with purposive recruitment strategies, academic support professionals (5=female, 4=male, age=44.2±13.9 years, experience=11.6±10.4 years) and faculty athletic representatives (5=female, 5=male, age=48.6±7.8 years, experience=3.9±5.6 years) volunteered to complete virtual interviews. We conducted semistructured one-on-one interviews with academic support professionals and faculty athletic representatives to understand better their lived experiences supporting student-athletes during the RTL process. Communication, multidisciplinary collaboration, and student-athlete support were described as essential components of streamlined RTL. Participants also highlighted the importance of stakeholder education on concussions and the RTL process to facilitate collaborative care. Our work highlights the need for strong communication and collaboration that results in student support and informed stakeholders, ultimately facilitating successful RTL practices for student-athletes in higher education. Clearly defined institutional policies and procedures are needed regarding RTL.
No abstract available
OBJECTIVE Prior studies have investigated associations between gender, symptom resolution, and time to return to play following sport-related concussion (SRC). However, there is a notable gap in research regarding the association between gender and return to learn (RTL) in adolescents. Therefore, this study 1) compared the patterns of RTL between boys and girls who are high school student athletes, and 2) evaluated the possible association between gender and time to RTL after adjusting for covariates. METHODS A retrospective cohort study of a prospective surveillance program that monitored concussion recovery of athletes in high schools throughout the state of Maine between February 2015 and January 2023 was performed. The primary independent variable was gender, dichotomized as boys and girls. The primary outcome was time to RTL, defined by the number of days for an athlete to return to school without accommodations. Mann-Whitney U-tests were used to compare RTL between the boys and girls. Each athlete's RTL status was dichotomized (i.e., returned vs had not returned) at several time points following injury (i.e., 1, 2, 3, and 4 weeks), and chi-square tests were performed to compare the proportions who achieved RTL between groups. Multivariable linear regression analyses were performed to evaluate the predictive value of gender on RTL. Covariates included age, number of previous concussions, history of learning disability or attention-deficit disorder or attention-deficit/hyperactivity disorder, history of a psychological condition, history of headaches or migraines, initial Sport Concussion Assessment Tool (SCAT3/SCAT5) score, and days to evaluation. RESULTS Of 895 high school athletes, 488 (54.5%) were boys and 407 (45.5%) were girls. There was no statistically significant difference in median [IQR] days to RTL between genders (6.0 [3.0-11.0] vs 6.0 [3.0-12.0] days; U = 84,365.00, p < 0.375). A greater proportion of boys successfully returned to learn without accommodations by 3 weeks following concussion (93.5% vs 89.4%; χ2 = 4.68, p = 0.030), but no differences were found at 1, 2, or 4 weeks. A multivariable model predicting days to RTL showed that gender was not a significant predictor of RTL (p > 0.05). Longer days to evaluation (β = 0.10, p = 0.021) and higher initial SCAT3/SCAT5 scores (β = 0.15, p < 0.001) predicted longer RTL. CONCLUSIONS In a cohort of high school athletes, RTL did not differ between boys and girls following SRC. Gender was not a significant predictor of RTL. Longer days to evaluation and higher initial symptom scores were associated with longer RTL.
We conducted a narrative review and content analysis to (i) determine the extent to which research on return to learn (RTL) following sport-related concussion (SRC) has considered social determinants of health (SDoH) or health equity and (ii) provide recommendations to enhance the incorporation of SDoH and health equity into considerations of how best to support concussed students as they RTL following injury. We coded 32 studies from a systematic review focused on clinical recovery following SRC, including RTL (Putukian et al., 2023). The review informed the latest Concussion in Sport Group consensus statement (Patricios et al., 2023). We examined studies to determine whether the SDoH domains and subcategories, derived from the US Department of Health and Human Services Healthy People 2020 and 2030 websites, were addressed. Pairs of raters coded articles independently and resolved disagreements through discussion. SDoH remain largely unrepresented in studies on return to school following concussion. Future research is needed to better characterize how SDoH and health equity factors influence return to school following concussion to ensure that concussed and recovering students are supported.
OBJECTIVE Patterns of clinical recovery and factors that may influence return to activity after concussion have been well studied; however, data regarding return-to-learn (RTL) duration after concussion and factors that influence this period have not been adequately explored. HYPOTHESIS We have hypothesized that sport level, sex, concussion history, and initial symptom burden will be associated with RTL duration after concussion. SETTING The study was conducted at US military service academies. STUDY DESIGN This was a prospective cohort study. PARTICIPANTS A total of 1509 service academy cadets (41% female, 20.2 ± 1.48 y, 174.9 ± 10.3 cm, 73.8 ± 14.0 kg) who sustained a concussion during the study period from December 2014 to March 2020 were included. MAIN MEASURES The primary outcome of interest was time elapsed between an incident injury and self-reported return to normal academic performance. All participants provided baseline demographics, including sex (male, female), sport level (varsity, non-varsity), and concussion history (yes, no). Participants completed standardized concussion assessments at regular time points. Initial symptom burden was collected within 48 hours after injury, and self-reported RTL duration was collected during the unrestricted return-to-activity (URTA) evaluation. RESULTS On average, RTL duration was faster in men (8.1 ± 11.2 days) than in women (13.6 ± 20.3 days). Univariate models demonstrated that participants with an initial symptom burden ≤6 returned sooner than those endorsing 7-15 symptoms (HR = 0.73, 95% CI = 0.64-0.83, P ≤ .001) and ≥16 symptoms (HR = 0.56, 95% CI = 0.48-0.65, P ≤ .001). Sex (HR = 0.70, 95% CI = 0.62-0.77, P ≤ .001), and sport level (HR = 0.66, 95% CI = 0.59-0.73, P ≤ .001) also contributed to RTL duration in the univariate model. In the multivariable model, participants with an initial symptom burden ≤6 returned 25%-41% sooner than those endorsing 7-15 symptoms (HR = 0.75, 95% CI = 0.65-0.86, P ≤ .001) and ≥16 symptoms (HR = 0.59, 95% CI = 0.51-0.69, P ≤ .001). Sex (HR = 0.71, 95% CI = 0.64-0.80, P ≤ .001) and sport level (HR = 0.75, 95% CI = 0.67-0.85, P < .001) also contributed to RTL duration in the multivariable model. CONCLUSIONS RTL duration for men, varsity athletes, and individuals with fewer initial symptoms after injury was significantly faster than for women, nonvarsity athletes, and individuals with a greater initial symptom burden.
OBJECTIVE Patterns of clinical recovery and factors that may influence return to activity after concussion have been well studied; however, data regarding return-to-learn (RTL) duration after concussion and factors that influence this period have not been adequately explored. HYPOTHESIS We have hypothesized that sport level, sex, concussion history, and initial symptom burden will be associated with RTL duration after concussion. SETTING The study was conducted at US military service academies. STUDY DESIGN This was a prospective cohort study. PARTICIPANTS A total of 1509 service academy cadets (41% female, 20.2 ± 1.48 y, 174.9 ± 10.3 cm, 73.8 ± 14.0 kg) who sustained a concussion during the study period from December 2014 to March 2020 were included. MAIN MEASURES The primary outcome of interest was time elapsed between an incident injury and self-reported return to normal academic performance. All participants provided baseline demographics, including sex (male, female), sport level (varsity, non-varsity), and concussion history (yes, no). Participants completed standardized concussion assessments at regular time points. Initial symptom burden was collected within 48 hours after injury, and self-reported RTL duration was collected during the unrestricted return-to-activity (URTA) evaluation. RESULTS On average, RTL duration was faster in men (8.1 ± 11.2 days) than in women (13.6 ± 20.3 days). Univariate models demonstrated that participants with an initial symptom burden ≤6 returned sooner than those endorsing 7-15 symptoms (HR = 0.73, 95% CI = 0.64-0.83, P ≤ .001) and ≥16 symptoms (HR = 0.56, 95% CI = 0.48-0.65, P ≤ .001). Sex (HR = 0.70, 95% CI = 0.62-0.77, P ≤ .001), and sport level (HR = 0.66, 95% CI = 0.59-0.73, P ≤ .001) also contributed to RTL duration in the univariate model. In the multivariable model, participants with an initial symptom burden ≤6 returned 25%-41% sooner than those endorsing 7-15 symptoms (HR = 0.75, 95% CI = 0.65-0.86, P ≤ .001) and ≥16 symptoms (HR = 0.59, 95% CI = 0.51-0.69, P ≤ .001). Sex (HR = 0.71, 95% CI = 0.64-0.80, P ≤ .001) and sport level (HR = 0.75, 95% CI = 0.67-0.85, P < .001) also contributed to RTL duration in the multivariable model. CONCLUSIONS RTL duration for men, varsity athletes, and individuals with fewer initial symptoms after injury was significantly faster than for women, nonvarsity athletes, and individuals with a greater initial symptom burden.
OBJECTIVE To evaluate the effectiveness of a newly implemented return to learn (RTL) concussion protocol in an adolescent athletic population.
Aim To understand academic support structures for Washington state public high school students with concussion during the COVID-19 pandemic. Materials & methods Prospective, repeated cross-sectional study of 21 schools in 2020 and 2021. Results About 28% of schools reported not providing any return-to-learn (RTL) accommodations for students with concussion throughout the COVID-19 pandemic. RTL accommodation provision was associated with larger student body size (β = 0.002) and higher graduation rate (β = 0.261) but was not associated with presence of RTL school policy. About 38.1% of schools received no guidance on how to provide RTL accommodations during the COVID-19 pandemic, and many reported that students with concussion struggled more. Conclusion Schools struggled to provide RTL accommodations for students with concussion during the COVID-19 pandemic, highlighting the need for evidence-based guidance and resource allocation to vulnerable schools.
OBJECTIVE Return-to-learn (RTL) after sport-related concussion (SRC) is an important yet understudied topic. The authors sought to do the following: 1) describe patterns of RTL among athletes by school level (i.e., middle school, high school, college); and 2) evaluate the predictive value of school level on RTL duration. METHODS A retrospective, single-institution cohort study of adolescent and young adult athletes 12-23 years old who sustained an SRC between November 2017 and April 2022 and who presented to a multidisciplinary specialty concussion clinic was conducted. The independent variable was school level, trichotomized into middle school, high school, and college. Time to RTL was the primary outcome and was defined as days from SRC to return to any academic activities. ANOVA was used to compare RTL duration across school levels. A multivariable linear regression was performed to evaluate for predictive value of school level on RTL duration. Covariates included the following: sex, race/ethnicity, learning disorder, psychiatric conditions, migraines, family history of psychiatric conditions/migraines, initial Post-Concussion Symptom Scale score, and number of prior concussions. RESULTS Of 1007 total athletes, 116 (11.5%) were in middle school, 835 (83.0%) were in high school, and 56 (5.6%) were in college. The mean RTL times (in days) were as follows: 8.0 ± 13.1 (middle school), 8.5 ± 13.7 (high school), and 15.6 ± 22.3 (college). One-way ANOVA showed a statistically significant difference between groups (F[2, 1007] = 6.93, p = 0.001). A Tukey post hoc test revealed a longer RTL duration in collegiate athletes when compared to middle school (p = 0.003) and high school (p < 0.001) athletes. Collegiate athletes had longer RTL duration compared to other school levels (β = 0.14, p < 0.001). There was no difference between middle school and high school athletes (p = 0.935). The subanalysis revealed a longer RTL duration in high school freshmen/sophomores (9.5 ± 14.9 days) when compared to juniors/seniors (7.6 ± 12.6 days; t = 2.05, p = 0.041), and being an older (junior/senior) high school athlete was predictive of shorter RTL duration (β = -0.11, p = 0.011). CONCLUSIONS When examining patients who presented to a multidisciplinary sport concussion center, RTL duration was longer in collegiate athletes when compared to middle and high school athletes. Younger high school athletes had longer time to RTL compared to their older counterparts. This study provides insight into how varying scholastic environments may contribute to RTL.
Objective: Although concussions affect millions of young adults annually, researchers have yet to assess factors that may affect future implementation of post-concussion academic supports within higher education. Therefore, we sought to evaluate preimplementation outcomes of the acceptability, feasibility, appropriateness, and readiness for change of the Post-Concussion Collegiate Return-to-Learn (RTL) Protocol among university stakeholders. Setting: An online survey. Participants: A convenience sample (N = 49; 63.3% female) of athletic trainers (ATs; n = 25, age = 30.1 ± 7.6 years) and university faculty/staff (n = 24, age = 38.3 ± 9.9 years) across the National Collegiate Athletic Association (NCAA) Power 5 Conferences from January to February 2022. Design: A cross-sectional study. Main Measures: To compare preimplementation outcome measures using the Acceptability of Intervention Measure (AIM), Feasibility of Intervention Measure (FIM), Intervention Appropriateness Measure (IAM), and Organizational Readiness for Implementing Change (ORIC) regarding the RTL protocol between ATs and university faculty/staff. Additional outcomes included internal and external barriers to implementing at their respective institutions. Statistical analyses were conducted using Mann-Whitney U tests, with effect sizes estimated using eta-squared coefficient (η2). Results: Quantitative analyses yielded no statistically significant group differences (Ps > .05) across the AIM, FIM, and IAM outcomes, indicating both groups perceived the protocol to be acceptable, feasible, and appropriate. Moreover, ATs reported higher agreement regarding motivation, desire, willingness to do “whatever it takes,” commitment, and determination to implement the novel protocol than faculty/staff. Further, ATs reported higher agreement regarding their institution's confidence to keep track of its progress, support adjustment, maintain momentum, manage institutional politics, coordinate tasks, encourage investment, and handle the challenges of future implementation of the RTL protocol. Conclusions: Preliminary findings suggest ATs and university faculty/staff across the NCAA Power 5 Conferences may perceive the RTL protocol to be acceptable, feasible, and appropriate for future use; however, noteworthy internal and external barriers may influence its uptake. Future research should utilize implementation frameworks to support the protocol's adoption and reach.
COPYRIGHT © 2023 Bevilacqua and McPherson. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Commentary: Establishing the college Return to Learn team for concussion: a practical approach
Youth concussion is common but there is a paucity of information on symptoms students report to school personnel and a gap in understanding what accommodations schools can provide.
Objective To define the time frames, measures used and modifying factors influencing recovery, return to school/learn (RTL) and return to sport (RTS) after sport-related concussion (SRC). Design Systematic review and meta-analysis. Data sources 8 databases searched through 22 March 2022. Eligibility criteria Studies with diagnosed/suspected SRC and interventions facilitating RTL/RTS or investigating the time and modifying factors for clinical recovery. Outcomes included days until symptom free, days until RTL and days until RTS. We documented study design, population, methodology and results. Risk of bias was evaluated using a modified Scottish Intercollegiate Guidelines Network tool. Results 278 studies were included (80.6% cohort studies and 92.8% from North America). 7.9% were considered high-quality studies, while 23.0% were considered high risk of bias and inadmissible. The mean days until symptom free was 14.0 days (95% CI: 12.7, 15.4; I2=98.0%). The mean days until RTL was 8.3 (95% CI: 5.6, 11.1; I2=99.3%), with 93% of athletes having a full RTL by 10 days without new academic support. The mean days until RTS was 19.8 days (95% CI: 18.8, 20.7; I2=99.3%), with high heterogeneity between studies. Several measures define and track recovery, with initial symptom burden remaining the strongest predictor of longer days until RTS. Continuing to play and delayed access to healthcare providers were associated with longer recovery. Premorbid and postmorbid factors (eg, depression/anxiety, migraine history) may modify recovery time frames. Though point estimates suggest that female sex or younger age cohorts take longer to recover, the heterogeneity of study designs, outcomes and overlap in CIs with male sex or older age cohorts suggests that all have similar recovery patterns. Conclusion Most athletes have full RTL by 10 days but take twice as long for an RTS. PROSPERO registration number CRD42020159928.
Return to learn (RTL) is the individualized process of coordinating cognitive care and reintegration for students into the academic setting after any sport and recreational-related concussion (SRRC). The guidelines for RTL are based on empirical evidence, however, implementation differs by institution. The purpose of the policy analysis is to evaluate RTL guidelines after SRRC of student-athletes in New England secondary school public school systems. A review of the six New England states’ policies surrounding RTL was conducted. The Comprehensive Analysis of Physical Activity Framework was referenced to identify the analytic components of existing legislation and because of the relatively new implementation of RTL-specific policy, a novel policy analysis tool was utilized. States with RTL-specific language scored on average 7.9 to 11.1 points higher when compared to states without RTL-specific language. This difference was associated with disparities in access to RTL resources for residents according to their geographic location. Lobbying efforts should be targeted toward states without RTL-specific language to provide equal care and opportunities for student-athletes to receive RTL services. RTL policy provides a responsibility to assist students who have suffered from an SRRC and can serve to improve health outcomes and academic achievement.
Introduction Current recommendations for concussed college students promote rapid implementation of accommodations (<3 days), however, it is unknown whether college accessibility staff find this recommendation feasible. Thus, this research aimed to gather the perceptions of accessibility staff regarding implementation of accommodations for students who have sustained concussions. Methods A cross-sectional design was used to distribute a nationwide survey to accessibility staff working on college campuses (n = 87). Distribution methods included a nationwide listserv along with hand searching institutional websites. Survey recruitment took place electronically across the United States between 1/2025–7/2025. Results Total sample mean and median feasibility scores were 7.88 and 8, respectively (0–10 scale). Seventy percent of participants reported the ability to accommodate a concussion in <3 days, with 89% accomplishing this task when provided 4 days. Limited student reporting (28%) and streamlined communication with medical providers (15%) were most frequently noted as variables impeding the feasibility of this task. No significant associations were found between feasibility scores and location (p = 0.840) or institution size (p = 0.094). No significant associations were found between perceived barriers and location (p = 0.091) or size of institution (p = 0.635). Discussion Most sampled accessibility staff across the Unted States find rapid (<3d) accommodations to be a feasible task, supporting our understanding and confidence in recent RTLU guidance. These data reveal an existing capacity to perform RTLU recommendations, encouraging feasibility.
No abstract available
No abstract available
No abstract available
Concussion or mild traumatic brain injury (mTBI) is a common phenomenon in the United States, with up to 3.6 million sport-related mTBIs diagnosed annually. Return to learn protocols have been developed to facilitate the reintegration of students into school after mTBI, however, the implementation of return to learn protocols varies significantly across geographic regions and school districts. An integrative review of the literature was performed using Whittemore and Knalf's methodology. A search of published literature was conducted using the PRISMA checklist. Database searches were conducted from March 2,019 to October 2,021 using the terms “mild traumatic brain injury” and “return to learn.” Twenty-eight publications were included. Three themes were derived from this review: lack of policy, poor staff education on concussion symptoms and stakeholder communication breakdown. The development of communication patterns and use of a return to learn protocol could facilitate a gradual return to full academic workload after concussion.
ABSTRACT Primary objective Part IV concludes this four-part review of ‘What Traditional Neuropsychological Assessment Got Wrong About Mild Traumatic Brain Injury,’ with a focus on clinical applications and future directions. Methods and procedures These reviews have highlighted the limitations of traditional neuropsychological assessment methods, particularly in the evaluation of the patient with mild traumatic brain injury (mTBI), and especially within the context of all of the 21st Century advances in neuroimaging, quantification and network neuroscience. Main outcome and results How advanced neuroimaging technology and contemporary network neuroscience can be applied to assessing the mTBI patient at this time along with neuroimaging of the future are reviewed. The current status of computerized neuropsychological test (CNT) development is reviewed as it applies to mTBI assessment. Likewise, how the future of various types of virtual reality (VR), artificial intelligence (AI), wearable sensors, and markerless gaming technology could enhance the mTBI CNT assessment tool box of the future is reviewed. Conclusions The review concludes with some aspirational statements about how improvements along with novel CNT methods could be developed and integrated with advanced neuroimaging technologies in the future to be tailored to meet the needs of the mTBI patient.
ABSTRACT Background Individuals recovering from mild traumatic brain injury (TBI) represent a heterogenous population that requires distinct treatment approaches. Identification of recovery trajectories improves our ability to understand the natural history of mild TBI recovery and develop targeted interventions. Objective To utilize group-based trajectory modeling (GBTM) to identify distinct patterns of symptom recovery following mild TBI in the first 6 months after mild TBI. Methods This study is comprised of 253 adults who presented to the emergency department with mild TBI and completed assessments for six-months post-injury. Patients were recruited for the prospective observational cohort study, HeadSMART. The primary outcome measure was the Rivermead Postconcussion Symptom Questionnaire. GBTM was used to identify longitudinal trajectories of recovery following mild TBI using Rivermead scores at baseline, one, three, and six months following diagnosis. Results Findings identified four distinct trajectories of symptom recovery follwing mild TBI including 9% of participants who were categorized with minimal acute symptoms that decreased over time, 45% with mild acute symptoms that decreased over time, 33% with relatively higher acute symptoms that decreased over time, and 13% with relatively higher acute symptoms that increased over time. Conclusions GBTM identified four distinct trajectories of recovery following mild TBI and GBTM may be useful for research interventions that can alter recovery trajectories.
ABSTRACT Objectives Fear avoidance is associated with symptom persistence after mild traumatic brain injury (mTBI). In this study, we investigated whether fear avoidance was associated with other outcomes such as return to work-related activity (RTW). Materials and Methods We analyzed associations between fear avoidance and RTW 6–9 months after mTBI, in two merged prospective mTBI cohorts. Adult participants aged 16 or over (n=175), presenting to outpatient services in New Zealand within 3 months of their injury, who were engaged in work-related activity at the time of injury, were included. Participants completed the Fear Avoidance Behavior after Traumatic Brain Injury (FAB-TBI) questionnaire at enrollment and 6 months later. Associations between FAB-TBI scores and RTW outcome were analyzed using multivariate approaches. Results Overall, 53% of participants had RTW by 6–9 months after mTBI. While early fear avoidance was weakly associated with RTW, persistent high fear avoidance between study assessments or increasing avoidance with time were associated with greater odds of still being off work 6–9 months after injury. Conclusions Pervasive and increasing avoidance of symptom triggers after mTBI were associated with lower rates of RTW 6–9 months after mTBI. Further research is needed to better understand transition points along the recovery trajectory where fear avoidance behaviors fade or increase after mTBI.
ABSTRACT Description In June 2021, the U.S. Department of Veterans Affairs (VA) and U.S. Department of Defense (DoD) approved a joint clinical practice guideline for the management and rehabilitation care for those who have symptoms in the post-acute period following mild traumatic brain injury (mTBI). This synopsis describes some of the clinically important recommendations. Methods In January 2020, VA and DoD leaders assembled a joint VA/DoD guideline development team of multidisciplinary clinical stakeholders that developed key questions, systematically searched and evaluated the literature, created two 1-page algorithms, and refined 19 recommendations using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system. The process closely conformed to the National Academy of Medicine’s tenets for trustworthy clinical practice guidelines. Recommendations This synopsis describes clinically important recommendations for the management and rehabilitation of mTBI. Outpatient primary care providers are the target audience for this synopsis and guideline. The current recommendations are an update from the 2016 VA/DoD Clinical Practice Guidelines for the Management of Concussion-Mild Traumatic Brain Injury.
There is a growing body of research showing that cerebral pathophysiological processes triggered by pediatric mild traumatic brain injury (pmTBI) may extend beyond the usual clinical recovery timeline. It is paramount to further unravel these processes, because the possible long‐term cognitive effects resulting from ongoing secondary injury in the developing brain are not known. In the current fMRI study, neural processes related to cognitive control were studied in 181 patients with pmTBI at sub‐acute (SA; ~1 week) and early chronic (EC; ~4 months) stages post‐injury. Additionally, a group of 162 age‐ and sex‐matched healthy controls (HC) were recruited at equivalent time points. Proactive (post‐cue) and reactive (post‐probe) cognitive control were examined using a multimodal attention fMRI paradigm for either congruent or incongruent stimuli. To study brain network function, the triple‐network model was used, consisting of the executive and salience networks (collectively known as the cognitive control network), and the default mode network. Additionally, whole‐brain voxel‐wise analyses were performed. Decreased deactivation was found within the default mode network at the EC stage following pmTBI during both proactive and reactive control. Voxel‐wise analyses revealed sub‐acute hypoactivation of a frontal area of the cognitive control network (left pre‐supplementary motor area) during proactive control, with a reversed effect at the EC stage after pmTBI. Similar effects were observed in areas outside of the triple‐network during reactive control. Group differences in activation during proactive control were limited to the visual domain, whereas for reactive control findings were more pronounced during the attendance of auditory stimuli. No significant correlations were present between task‐related activations and (persistent) post‐concussive symptoms. In aggregate, current results show alterations in neural functioning during cognitive control in pmTBI up to 4 months post‐injury, regardless of clinical recovery. We propose that subacute decreases in activity reflect a general state of hypo‐excitability due to the injury, while early chronic hyperactivation represents a compensatory mechanism to prevent default mode interference and to retain cognitive control.
Context: Return to Learn is a popular policy that institutions have adopted to help students re-integrate back into the classroom after suffering a concussion. Although a popular policy, and required by some universities, students do not always obtain accommodations for their concussions. While several factors could prevent students from receiving accommodation, one specific cited barrier has been the collegiate professor. Objective: To investigate the experiences of collegiate professors in providing academic accommodations for students who suffer concussions. Design: Qualitative studyPatients or Other Participants: Eleven collegiate professors, from two universities, with similar population demographics.Data Collection and Analysis: Participants choose to participate in one semi-structured interview, focus group, or both, to be recorded and transcribed for inductive coding and thematic analysis. Results: Findings promoted three major themes: Unaware of Policies, Lack of Confidence, and Frustration. Secondary themes identified deficiencies in communication, unprovided resources, and a lack of knowledge of concussion policy. All themes frame the various struggles and frustration professors have experienced when providing accommodations to students suffering a concussion. Conclusions: College professors are depended on to support their students and provide them with the accommodations they need; however, professors have not had positive experiences when called upon to assist. College professors have not been presented with or made aware of policies and procedures that would assist in providing accommodations to their students. This has led to frustration with the lack of support and lack of communication that would aid their confidence in providing the accommodations necessary for student success.
BACKGROUND An estimated 1 in 5 adolescents have sustained a concussion in North America. Teachers and school administrators are responsible for implementing academic accommodations and other supports for optimal return to learn following a concussion. The primary objective of this study was to describe the prevalence and feasibility of providing academic accommodations to students following concussion from the perspectives of middle and high school teachers and school administrators. METHODS A cross-sectional survey was administered to teachers and school administrators (grades 7-12) across Canada online via REDCap. Participants were recruited via word-of-mouth and social media sampling. Survey responses were analyzed descriptively using proportions. RESULTS The survey was completed by 180 educators (138 teachers and 41 school administrators), of whom 86% had previously provided academic accommodations to students following concussion, and 96% agreed that students should have access to accommodations following concussion. Some accommodations (eg, breaks, extra time) were provided more often and were more feasible to provide than others (eg, no new learning, reduced bright light). Educators reported limited preparation time and limited school personnel support to assist students following concussion. IMPLICATIONS FOR SCHOOL HEALTH POLICY, PRACTICE, AND EQUITY The most feasible accommodations should be prioritized, ensuring students are supported within the school environment. CONCLUSIONS Teachers and school administrators confirmed the importance of providing accommodations to students following concussion.
OBJECTIVE Many schools utilize academic accommodations to help athletes return-to-learn after sport-related concussion, yet little is known about the impact of accommodations on recovery. In a cohort of adolescent athletes with sport-related concussion, the authors sought to 1) describe academic accommodations, 2) determine predictors of receiving accommodations, and 3) determine how accommodations influenced recovery, as defined by days to return-to-learn, symptom resolution, and return-to-play. METHODS A retrospective survey study was undertaken that included all athletes between the ages of 12 and 24 years who were seen at a regional sport-related concussion center from April 1, 2020, to April 1, 2022. Demographic characteristics, past medical history, injury characteristics, school-related factors, and recovery were collected via a telephone-based survey and from medical charts. The independent variable was the use of academic accommodations by students. The outcome variables included days to return-to-learn, symptom resolution, and return-to-play. Univariable and multivariable logistic regression analyses were performed. RESULTS Of 300 athletes contacted, 96 consented to participate in this study (mean age ± SD 16.6 ± 2.6 years, 60.4% male). The mean return-to-learn was 9.0 ± 13.2 days. In total, 63.5% of athletes received some kind of school accommodation. The most common accommodations included extra time on tests (46/96 [47.9%] athletes) and extra time for assignments (43/96 [44.8%]), whereas preferential seating in class (0/96 [0.0%]) and reduced workloads (4/96 [4.2%]) were the least commonly used accommodations. Multivariable logistic regression revealed that White race (OR 0.17, 95% CI 0.02-0.71, p = 0.03) and higher initial Post-Concussion Symptom Scale (PCSS) score (OR 1.04, 95% CI 1.01-1.07, p = 0.02) were predictive of receiving accommodations. Univariable models showed that receiving accommodations was not predictive of return-to-learn (β = 4.67, 95% CI -0.92 to 10.25, p = 0.10), symptom resolution (β = 24.71, 95% CI -17.41 to 66.83, p = 0.26), or return-to-play (β = 13.35, 95% CI -14.56 to 41.27, p = 0.35). Regarding other predictors of each outcome, several factors emerged irrespective of accommodations. Multivariable analysis revealed that longer time to return-to-learn was associated with a history of psychiatric illness (β = 8.00, 95% CI 1.71-14.29, p = 0.02) and initial PCSS score (β = 0.14, 95% CI 0.06-0.23, p < 0.01). Finally, predictors of days to return-to-play included school personnel who were knowledgeable of concussion (β = -5.07, 95% CI -9.93 to -0.21, p = 0.04) and higher initial PCSS score (β = 0.67, 95% CI 0.04-1.29, p = 0.03). CONCLUSIONS The most common accommodation was extra time on tests/assignments, whereas reduced workload was the least common. White race and greater initial PCSS score were associated with receiving accommodations. Receiving accommodations did not significantly impact days to return-to-learn, symptom resolution, or return-to-play.
Aim: The purpose of this study was to examine academic achievement, absenteeism and school accommodations following adolescent sport-related concussion (SRC). Methods: A case-series was conducted among grade 8–12 students who suffered an SRC. The primary outcomes were change in pre- and post-concussion overall, core report card grade point average (GPA) and absenteeism due to concussion. The most helpful school accommodations were tabulated. Results: Pre- and post-concussion GPA was obtained from 33 students – 16 (48%) developed persistent post-concussion symptoms (symptoms lasting >4 weeks). There was no significant difference in pre- and post-concussion grades among students with a SRC for overall (p = 0.75) or core (p = 0.56) GPA. The median number of missed school days was 4 (interquartile range [IQR]: 2–8). Allowing for physical and cognitive rest was identified as the most helpful accommodation (30%). Conclusion: Larger studies should investigate the role that school accommodations and development of persistent post-concussion symptoms have on academic outcomes.
Purpose: Concussion experts now recommend some degree of both physical and cognitive rest to facilitate recovery from concussions. Illinois implemented the Youth Sports Concussion Safety Act for the 2016-17 school year, increasing requirements for schools to provide appropriate concussion management, including return-to-learn (RTL) protocols. It also extends coverage to elementary school children, private schools, and charter schools. Schools offering interscholastic athletics are required to form Concussion Oversight Teams (COT), which will establish return-to-play (RTP) and RTL protocols. This study sought to determine the prevalence of appropriate concussion policies in Illinois K-12 schools and whether certain school types required greater support in developing these policies. Methods: A link to a 26-question online …
Abstract An estimated 20 percent of adolescents in North America have sustained a concussion. Teachers and school administrators are responsible for supporting following concussion to return to school. The purpose of this study was to describe the perspectives of Canadian teachers and school administrators with supporting returning to school post-concussion. This qualitative study was guided by an interpretivist philosophy. Semi-structured interviews with grades 7–12 teachers (n = 13) and school administrators (n = 5) were coded inductively and analysed using reflexive thematic analysis. We organized the data into five themes: 1) Educator roles differ for administrators and teachers; 2) Students’ symptoms affect their learning; 3) Students should have access to academic accommodations; 4) Students benefit from social support, compassion, and empathy; and 5) Concussion education and management processes are lagging. In Canada, teachers and administrators have different roles when supporting students returning to school after a concussion and those roles influence their engagement with the students and their awareness of students’ needs. Lack of concussion education and concussion management processes at schools may affect how students are supported following a concussion. Our findings can inform the development and implementation of supports to facilitate return to school for students following a concussion.
BACKGROUND: Many students return to school after concussion with symptoms but without formal support. OBJECTIVE: To examine concussion symptoms and temporary academic accommodations during school use of a four-week student-centered return to learn (RTL) care plan. METHODS: Five public high schools used the RTL care plan and contributed student-level data after student report of concussion. Data on concussion symptoms, temporary academic accommodations corresponding to reported symptoms, and accommodations provided during RTL care plan use were examined. RESULTS: Of 115 students, 55%used the RTL care plan for three (34%) or four (21%) weeks. Compared to students whose symptoms resolve within the first two weeks, students who used the RTL care plan for three or four weeks reported more unique symptoms (P = 0.038), higher total severity score (P = 0.005), and higher average severity per symptom (P = 0.007) at week one. Overall, 1,127 weekly accommodations were provided. While least reported, emotional symptoms received corresponding accommodations most often (127/155 reports: 82%of occurrences). CONCLUSIONS: Use of an RTL care plan can facilitate the RTL of students with a concussion and may aid in the identification of students who are in need of longer-term support.
Objectives The objectives were 1) to compare the effects of adolescent sport-related concussion (SRC) and sport-related extremity fracture (SRF) on academic outcomes including change in school grades and school attendance; and 2) to determine which specific academic accommodations were most helpful during recovery from these injuries. Methods A prospective cohort study was conducted to compare changes in school grades, school attendance and academic accommodations among students (grades 8–12) with an SRC or SRF. School grades were extracted from student immediate pre- and post-injury report cards. Students completed attendance log books and an exit interview to determine which accommodations were helpful and how accommodating they perceived their school to be during their recovery. Results Overall, 124 students (92 with SRC and 32 with SRF) submitted both pre- and post-injury report cards. Students who sustained an SRC or SRF experienced similar decreases in grades post-injury (SRC: -1.0%; 95% CI: -2.1, 0.1 and SRF: -0.9%: 95% CI: -2.1, 0.3). Students with an SRC missed significantly more days of school compared to those with an SRF (median of 4 days [IQR: 1, 7] versus 1 day [IQR: 0, 4], p<0.0001). In total, 60/113 (53.1%) SRC students reported their school to be very accommodating while only 31/77 (40.3%) SRF students reported their school to be very accommodating (p = 0.082). Conclusions Students who sustain an SRC miss significantly more days of school but demonstrate similar changes in school grades post-injury compared to those with an SRF. Future studies are needed to identify the pre- and post-injury factors associated with poor academic functioning following concussion and identify measures that can be taken to help optimize academic outcomes in these patients.
No abstract available
Objective: Care after concussion is important for all patients, although especially critical in children and adolescents because of continued rapid brain growth and maturation. Postconcussion symptoms often lead to impaired school performance. Conflicting data regarding best return-to-learn practices make it difficult for school administrators to develop policies to best accommodate concussed students. We sought to assess high school principals' knowledge of concussion, the personnel responsible for implementing accommodations, and the overall willingness to enact recommended academic accommodations. Methods: In our cross-sectional study, we surveyed 410 Indiana high school principals (157 responded). Assistant principals were excluded. Results: One-third of the respondents received academic accommodations training for concussed students and more than 80% were somewhat or very comfortable with academic management. Greater than 90% were willing to provide accommodations as long as necessary. However, nearly 40% of responding principals were unlikely or unwilling to implement accommodations for standardized testing. National and state data suggest the median Indiana high school should expect 30 and more concussions per year; more than 90% of principals estimated that fewer than 30 concussions occurred each year at their school. Conclusions: The underestimation of concussion frequency highlights an opportunity for further education of high school principals to ensure all concussed students receive appropriate return-to-learn accommodations.
Abstract Objective The purpose of this pilot study was to determine the feasibility of implementing a concussion educational program to college faculty and to explore the effect of the program on faculty knowledge regarding concussion and the effect on learning and return to school. Methods: The study utilized a mixed method design, with a pretest and post-test to measure concussion knowledge before and after a didactic presentation on concussion and academic performance. Semi-structured interviews gathered information on faculty’s perception of concussion and academic performance. Results: Faculty demonstrated improved scores on post-test knowledge assessment of concussion and academic performance. Faculty reported that they would be more likely to observe for signs of concussion in students, and would be more willing to adhere to recommendations for academic accommodations. Conclusions: The results of the study indicate that a didactic presentation on concussion and academics may be a feasible method for training college faculty.
Abstract Mild traumatic brain injury or concussion is one of the most common injuries sustained by children and adolescents in the United States. A survey of 623 principals across 25 states assessed their perceptions on concussion prevalence and school responses. Principals generally underestimated concussion rates and lacked formal concussion management teams, though most emphasized the importance of staff awareness and believed they offered effective accommodations. Findings revealed significant gaps in concussion management, suggesting many students may not receive adequate support, potentially leading to negative outcomes in recovery and academic progress.
PURPOSE Examine concussion effects on academic outcomes, including student perspectives. METHODS This study included a systematic review and meta-analysis examining post-concussion school attendance, academic performance, perceptions of academic difficulty, and accommodations for students in elementary through college settings. The analysis considered pre- and post-injury factors, along with injury factors that contribute to post-concussion academic outcomes. RESULTS The systematic review showed that students with concussion miss more school days and perceive higher levels of academic difficulty, but results about academic performance varied. Meta-analysis yielded small concussion effects on school absence and academic performance and moderate effects on perceptions of academic difficulty. Female sex, older age, history of migraine, prior concussions, severe or persistent symptoms, vestibular-ocular motor, and cognitive disruptions are risk factors, but these moderators were not identified in the meta-analysis due to lack of effect sizes. IMPLICATIONS FOR SCHOOL HEALTH POLICY, PRACTICE, AND EQUITY This study confirmed negative concussion effects on academic absences, performance, and perceptions of academic difficulty. Identified contributing factors will guide future practices to support students returning to learn after concussion. CONCLUSIONS Negative impacts to academics from concussion may be amplified by complicating factors. Future investigations are needed to confirm risk factors and mitigating effects of early identification and post-injury supports.
Between 2009 and 2014, all 50 states and the District of Columbia passed legislation to improve the recognition and management of youth concussed in sports. These laws can include requirements for concussion training for school athletic personnel, concussion education for children and their parents, return-to-play (RTP) procedures, and medical clearance to for RTP. Concussion can impact academic learning and performance in children and adolescents. Postconcussion academic accommodations during recovery can be an important component of secondary prevention for mitigating the sequalae of head injury. Few state youth concussion laws, however, include provision of postconcussion return-to-learn (RTL) accommodations and most of those that do address RTL apply to student athletes only. Concussions may occur in youth who are not participating in organised sports (eg, falls, traffic crashes) and thus may not be subjected to RTL accommodations, even if the state mandates such procedures for athletes. Low income and students of colour may be more likely to have non-sports concussions than their more affluent and white peers, thus potentially creating demographic disparities in the benefits of RTL procedures. State youth sports concussion laws should be revised so that they include RTL provisions that apply to all students, athletes and non-athletes alike.
Abstract ObjectiveThe current paper presents a comparison of university policies and health and academic accommodations offered to undergraduate students following sexual assault (SA) and sports-related concussions (SRC). Procedures and protocols for universities to consider adapting from their SRC policies to their SA policies are detailed.Participants: The SRC and SA policies at the 50 United States public flagship universities were analyzed.Methods: The research team coded for a number of policy details including health referrals, academic and financial accommodations, and requirements for follow-up with university personnel.Results: Compared to SA polices, SRC policies at the public flagship universities offer more comprehensive academic accommodations and physical and mental health resources. Conclusions: Comprehensive policies for student SA survivors, like those available for students who suffer SRCs, would improve student health and academic outcomes and increase SA survivors’ likelihood of graduating college, thereby reducing individual and societal human capital loses.
Although all states have legislation pertaining to youth sports concussion, most of these laws focus on return-to-play procedures; only a few address return-to-learn (RTL) accommodations for students who have experienced a concussion. To address this gap in the legislation, some states and nongovernmental organizations have developed RTL guidelines to advise school personnel, parents, and health care providers on best practices for accommodating students' postconcussion reintegration into academic activity. In 2018, the Massachusetts Department of Public Health (MDPH) developed RTL guidelines which were disseminated to school nurses (SNs) at all public and nonpublic middle and high schools in the state. In 2020, the MDPH engaged the Injury Prevention Center at Boston Medical Center to survey Massachusetts SNs to assess the usefulness of the guidelines. The response rate was 63%; 92% found the booklet extremely useful or moderately useful; and 70% endorsed that the booklet fostered collaboration among stakeholders.
Neuropsychologists play an important role in the clinical evaluation of many children and adolescents who have sustained concussions. This paper discusses clinical approaches and concerns in conducting neuropsychological evaluations of middle and high school aged students during the acute, subacute, and more chronic phases of concussion recovery. Issues of baseline testing and validity assessment are reviewed. Typical clinical recommendations arising from the neuropsychological evaluation are outlined, with an emphasis on the relationship between test findings and common academic accommodations.
BACKGROUND: Postsecondary students with traumatic brain injuries (TBI) are a rapidly growing population, encompassing those who sustained injuries prior to attending postsecondary education and those who endure injuries during their postsecondary studies. Not only do these individuals face a broad range of symptoms, all of which can affect academic achievement, but they also do not achieve comparable academic outcomes to their peers without disabilities. OBJECTIVE: There is a need to develop and examine the effectiveness of available supports and resources to meet the needs of these students. METHODS: Twenty-three articles were systematically reviewed to illustrate what supports are currently described in the literature for postsecondary students with TBI and what research methods were used to assess the effectiveness of these supports. RESULTS: Three categories of supports emerged, including concussion management protocols, typical study/learning strategies and accommodations, and interventions developed specifically for this population. Findings also indicated a lack of rigorous research methods used to assess these interventions’ effects. CONCLUSIONS: Implications for future research include a need for additional study of all supports and resources described in this review, and assessment of whether or not education professionals, including postsecondary disability services professionals, are aware of and using the tools and strategies addressed in this review.
No abstract available
OBJECTIVES Sport-related concussion is a complex mild traumatic brain injury for which diagnosis, monitoring, and prognosis remain largely dependent on subjective clinical assessment. Artificial intelligence has emerged as a potential tool to enhance objectivity by integrating large, multimodal datasets across the concussion care pathway. DESIGN Scoping review. METHODS A systematic literature search was conducted across six databases (MEDLINE, EMBASE, SPORTDiscus, Scopus, Web of Science, and Cochrane Central) from inception to December 2025. Eligible studies were classified into four domains: Detection & Diagnosis, Monitoring & Surveillance, Prognosis & Recovery, and Prevention & Risk Modeling. RESULTS Fifty-five studies met the inclusion criteria. Detection & Diagnosis was the most represented domain, primarily leveraging electroencephalography, speech, motor, and multimodal clinical data. Monitoring & Surveillance studies focused on wearable sensors, mouthguards, and video-based impact detection to quantify exposure and reduce false-positive events. Prognosis & Recovery models examined recovery trajectories, persistent symptoms, and reinjury risk, while Prevention & Risk Modeling studies predominantly relied on biomechanical and finite element-derived data to estimate injury risk. Despite promising performance, studies were highly heterogeneous and frequently limited by small or imbalanced samples, inconsistent outcome definitions, limited external validation, and poor model interpretability. CONCLUSIONS Artificial intelligence shows growing potential to support sport-related concussion management across multiple clinical domains. However, current evidence supports its use primarily as a decision-support tool. Future research should prioritize large, multicenter studies, transparent labeling strategies, explainable artificial intelligence frameworks, and rigorous external validation to enable safe implementation.
Predicting NHL Performance from Cognitive-Motor Metrics After Concussion: An Explainable AI Approach
In this work, we introduce a novel approach using cognitive-motor integration (CMI) metrics and concussion history to predict NHL participation. Specifically, we leverage data from the BrDi Test—a validated neuroscience-based assessment involving visuomotor tracing tasks under both standard and non-standard conditions. Using data from the BrDi Test—a neuroscience-based cognitive-motor integration task this research explores the link between mild traumatic brain injury (mTBI), or concussion, and NHL participation. We emphasize both model accuracy and interpretability to support human scouting decisions by comparing explainable machine learning (XAI) models with traditional black-box approaches. We frame the problem as a binary classification task and evaluate a range of machine learning models. Despite the limitations of a small, class-imbalanced dataset, a class-weighted Decision Tree achieved the best average generalization performance, offering strong interpretability alongside competitive performance. Our contributions include the first application of BrDi data to NHL prediction, a comparative analysis of XAI and black-box models for talent evaluation, and a modular ML pipeline adaptable to small-sample, high-dimensional datasets. The results support the potential value of cognitive metrics in augmenting traditional scouting decisions and provide a foundation for future research as more data becomes available.
No abstract available
ABSTRACT There is considerable uncertainty about the role that Artificial Intelligence (AI) might play in Environmental Impact Assessment (EIA), including into research. AI large language model (LLM) chatbots have the potential to increase the efficiency of EIA research, but their outputs can create concerns. This paper investigates the potential time savings achievable using LLM chatbots to undertake a critical review of literature focussing on the use of AI in EIA. Using a combination of ChatGPT and Elicit, literature was reviewed to identify 12 key issues associated with the use of AI in EIA and this paper was prepared in three and a half days from initial conception. A protocol is developed to assist researchers in fact checking evidence delivered through Elicit (or other machine learning tools) which serves as a novel outcome of this research. Using comments from three peer reviewers allowed some more objective reflection on the credibility of the LLM chatbot-derived output, on the appropriateness of the time savings, and on the future research needed on the application of LLM chatbots in this context.
The artificial intelligence in education (AIED) community has produced technologies that are widely used to support learning, teaching, assessment, and administration. This work has successfully enhanced test scores, course grades, skill acquisition, comprehension, engagement, and related outcomes. However, the prevailing approach to adaptive and personalized learning has two main steps. First, the process involves detecting the areas of knowledge and competencies where students are deficient. This process also identifies when or how a student is considered “at risk” or in some way “lacking.” Second, the approach involves providing timely, individualized assistance to address these deficiencies. However, a considerable body of research outside our field has established that such deficit framing, by itself, leads to reactive and less productive strategies. In deficit-based frameworks, powerful student strengths, skills, and schemas—their assets—are not explicitly leveraged. In this paper, we outline an asset-based paradigm for AIED research and development, proposing principles for our community to build upon learners’ rich funds of knowledge. We propose that embracing asset-based approaches will empower the AIED community (e.g., educators, developers, and researchers) to reach broader populations of learners. We discuss the potentially transformative role this approach could play in supporting learning and personal development for all learners, particularly for students who are historically underserved, marginalized, and “deficit-ized.”
Recent studies emphasize online health information quality. However, little research focuses on Arabic content for emergency conditions like heart disease, hypertension, and stroke despite the high demand for this information. This study aims to address this gap by using artificial intelligence to create a benchmark dataset for Arabic health information and evaluate its quality. We assessed the quality of health information across three criteria: source quality, treatment quality, and content trustworthiness. The Kruskal-Wallis test was used to analyze quality differences across content types (General Health Information, Medical Advice, Treatment Description) and website categories (Government, Journalistic, Portal, Professional). Data augmentation techniques such as paraphrasing, back translation, and RandAugment were also employed to enhance quality assessment using the Arabic BERT model. The study also proposes a novel architecture termed the Mixture of Classification. In this approach, each health document is processed in parallel by three instances of an Arabic BERT model: the first identifies the type of health information, the second determines the category of the provider, and the third estimates a continuous quality score via regression. Significant quality differences were observed among website categories and content types. Portal sites achieved the highest mean score (11.64), while Journalistic sites scored the lowest (3.33). Treatment Descriptions had the highest score (18.74), while Medical Advice scored the lowest (4.73). These differences were statistically significant with large effect sizes (Cohen’s \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$f = 0.409$$\end{document} for categories; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$f = 0.288$$\end{document} for content types), indicating substantial practical impact. Paraphrasing showed the best performance, with 89% accuracy, 85% F1 score, 76% precision, and 82% recall with high confidence and minimal class imbalance. A notable 36.3% accuracy gap was identified between low-quality (95.01%) and high-quality (58.71%) content using RandAugment data. These findings highlight the importance of content type, provider category, and quality score for improving health information search rankings. Content type, provider category, and quality score are key factors in enhancing the ranking of Arabic health information. Paraphrased data augmentation contributes to improved model reliability in distinguishing between quality classes. Future research should extend this approach to other languages and health-related topics. However, a major challenge remains: achieving a balanced dataset, particularly for binary classification between high- and low-quality content, as well as for the new classification based on provider category, content type, and quality score. The goal is to ensure an equal distribution across all these categories.
Autism Spectrum Disorder (ASD) presents varied challenges across assessment, intervention, and behavioural monitoring. With recent advancements in Artificial Intelligence (AI), introduced tools have emerged that support clinicians, educators, and families in understanding and managing ASD more effectively. Significant innovations include explainable AI, mobile applications, computer vision, and wearable devices, which establish promising potential to increase accuracy, accessibility, and consistency of care. This review produces findings from recent literature to examine how AI contributes to the ASD care process. It highlights the benefits of AI while noting ethical and practical considerations. Also addresses challenges in data security, privacy, and ethical AI deployment within ASD care settings.
Background: Multimodal large language model (MLLM)-based systems capable of generating health-related information and diagnostic suggestions are increasingly used for health information retrieval; however, their accuracy, readability, and quality in oral healthcare remain unclear. Oral mucosal diseases comprise a heterogeneous group of conditions affecting the oral lining, ranging from benign and reactive lesions to potentially malignant and malignant disorders. Objective: This study evaluated and compared the diagnostic performance, readability, and information quality of MLLMs with traditional search engines included as comparator platforms, in diagnosing oral mucosal diseases. Methods: A cross-sectional observational study was conducted using 100 validated oral mucosal case scenarios representing benign, malignant, potentially malignant, infectious, and reactive oral lesions. Each scenario was entered into ChatGPT 3.5, ChatGPT 4.5 (Plus), Microsoft Copilot (smart), Grok (xAI), Claude (Sonnet 4.5), DeepSeek v3.1, and search engines Google, Bing, and Yahoo. Diagnostic accuracy, Positive Predictive Value (PPV), and Negative Predictive Value (NPV) were compared against reference diagnoses. Information quality was assessed using the DISCERN tool, and readability was evaluated using Flesch–Kincaid Reading Ease (FRES) and Grade Level (FKGL) scores. Statistical analyses included Cochran’s Q and McNemar tests (p < 0.05). Results: ChatGPT 4.5 demonstrated the highest overall diagnostic accuracy (88.5%), PPV (92%), and NPV (88%), followed by DeepSeek v3.1 and Claude (Sonnet 4.5). Traditional search engines performed poorly (accuracy 18–55%). MLLMs achieved higher DISCERN scores (2.84–3.20) but lower readability (FKGL = 11–14) than search engines (FKGL = 6–7). No platform met the recommended sixth-grade reading level for consumer health information. Conclusions: MLLMs, particularly ChatGPT Plus (GPT-4.5), outperformed conventional search engines in diagnostic accuracy and content quality but produced complex, less-readable text. Future AI development should prioritise improving clinical accuracy alongside readability and transparency to ensure equitable access to reliable oral health information.
When large language models (LLMs) are implemented in the field of health information quality evaluation, ensuring that their outputs align with human judgment, logic, and strategic preferences has become a focal point of current research on AI alignment. This study proposes the Health Information Quality Evaluation Framework based on AI Post-Alignment (HIQE-PA) to examine whether large language models remain consistent with expert judgment standards in health information quality evaluation tasks. Using expert evaluation results as the benchmark, the framework assesses the consistency, stability, and cross-source adaptability of model outputs through structured task implementation and multidimensional statistical indicators. In the experimental study, ERNIE-3.5 and ChatGLM2-6B-32K were selected as general-purpose models, and the corresponding ERNIE-3.5 + HIQE-PA model and ChatGLM2-6B-32K + HIQE-PA model were constructed to compare post-alignment performance under different model conditions. Under three post-alignment performance evaluation indicators, including deviation degree, predictability, and fitness, the LLMs’ post-alignment performance was examined through their outputs in the health information quality evaluation task. The results show that the general LLMs achieved a high level of post-alignment performance only in the dimensions of readability and completeness, while the LLMs + HIQE-PA improved post-alignment performance across all dimensions. In particular, the ERNIE-3.5 + HIQE-PA model performed prominently, producing evaluation outputs that were closer to expert consensus and maintaining consistency with the expert benchmark across different text sources. This study demonstrates that post-alignment examinations can provide empirical support for model selection and governance in the health information domain.
Electronic Health Records (EHRs) store vast amounts of clinical information that are difficult for healthcare providers to summarize and synthesize relevant details to their practice. To reduce cognitive load on providers, generative AI with Large Language Models have emerged to automatically summarize patient records into clear, actionable insights and offload the cognitive burden for providers. However, LLM summaries need to be precise and free from errors, making evaluations on the quality of the summaries necessary. While human experts are the gold standard for evaluations, their involvement is time-consuming and costly. Therefore, we introduce and validate an automated method for evaluating real-world EHR multi-document summaries using an LLM as the evaluator, referred to as LLM-as-a-Judge. Benchmarking against the validated Provider Documentation Summarization Quality Instrument (PDSQI)-9 for human evaluation, our LLM-as-a-Judge framework uses the PDSQI-9 rubric and demonstrated strong inter-rater reliability with human evaluators. GPT-o3-mini achieved the highest intraclass correlation coefficient of 0.818 (95% CI 0.772, 0.854), with a median score difference of 0 from human evaluators, and completes evaluations in just 22 seconds. Overall, the reasoning models excelled in inter-rater reliability, particularly in evaluations that require advanced reasoning and domain expertise, outperforming non-reasoning models, those trained on the task, and multi-agent workflows. Cross-task validation on the Problem Summarization task similarly confirmed high reliability. By automating high-quality evaluations, medical LLM-as-a-Judge offers a scalable, efficient solution to rapidly identify accurate and safe AI-generated summaries in healthcare settings.
Background/objectives Patients increasingly use large language model (LLM) interfaces for health information, but their safety and quality for public questions about interstitial cystitis/bladder pain syndrome (IC/BPS) remain uncertain. This study evaluated the safety, accuracy, empathic communication, information quality, reliability, and readability of five publicly accessible LLM interfaces. Methods In this CHART-guided cross-sectional comparative study, 58 public-facing IC/BPS questions were submitted once, in English, to ChatGPT, Gemini, Microsoft Copilot, DeepSeek, and Doubao using a standardized single-turn, zero-shot protocol. Three blinded senior urologists independently assessed safety, accuracy, empathic communication, DISCERN, EQIP, JAMA benchmark criteria, and Global Quality Score. Six readability indices were calculated. Results Inter-rater agreement was significant for all manually assessed metrics. Fleiss’ kappa for safety was 0.822, and ICC(2,1) values for other rater-assessed metrics ranged from 0.761 to 0.848. Unsafe responses occurred in all interfaces, ranging from 5.2% for ChatGPT to 8.6% for DeepSeek and Doubao, without a significant between-interface difference (Cochran Q = 1.000, p = 0.910). Accuracy and empathic communication differed significantly across interfaces (both p < 0.001). ChatGPT had the highest median accuracy score, whereas DeepSeek had the highest empathic communication score. Information-quality and reliability scores also differed significantly (all p < 0.001); ChatGPT achieved higher DISCERN, EQIP, and GQS scores, while Gemini achieved higher JAMA scores. Readability differed significantly across interfaces, but none met predefined patient-facing readability benchmarks. Conclusion Publicly accessible LLM interfaces showed domain-specific differences when answering IC/BPS-related public questions. Unsafe responses were uncommon but present in all interfaces, and no interface consistently outperformed the others. LLM interfaces may support general IC/BPS education and question preparation but should not replace clinician-led evaluation or individualized medical advice.
BACKGROUND Generative artificial intelligence chatbots are increasingly used for health information, including sensitive conditions, such as erectile dysfunction (ED), but their safety and quality remain uncertain. To compare six large language model chatbot interfaces for ED information across safety, accuracy, empathy, information quality, source-related transparency, readability and response length. METHODS Thirty-eight standardized ED questions were submitted once to six chatbot interfaces, yielding 228 responses. Five independent raters assessed Safety, Accuracy, Empathy, DISCERN, Ensuring Quality Information for Patients, the Journal of the American Medical Association benchmark and Global Quality Score. Readability and response length were also evaluated. Between-interface differences were tested using Cochran's Q or Friedman tests. RESULTS Twenty-six responses (11.4%) were safety-flagged. Safety-flag frequency differed across interfaces (Q = 11.765, P = 0.038), but the effect size was small (0.062; 95% CI, 0.026-0.148), and no pairwise difference remained significant after multiplicity correction. Accuracy was generally high (median 4.00/5 for all interfaces), but differed slightly overall (P = 0.010, Kendall's W = 0.080). Empathy (P < 0.001, W = 0.160), DISCERN (P = 0.003, W = 0.508), Ensuring Quality Information for Patients (P < 0.001, W = 0.576), and Journal of the American Medical Association benchmark scores (P < 0.001, W = 0.888) differed significantly, whereas Global Quality Score did not (P = 0.146). Readability and response length also differed (all P < 0.001). CONCLUSION Large language model chatbots generally provided accurate ED-related information, but performance varied across clinically relevant dimensions. They may support information seeking, but should not replace individualized clinical assessment.
Background Large language model (LLM) chatbots are increasingly used to obtain health information. However, fluent and clinically plausible responses may still contain safety-relevant omissions, inadequate source attribution and disclosure, or difficult-to-read text. Objective To evaluate the safety, accuracy, empathy, information quality, response-level source attribution and disclosure, overall quality, and readability of five LLM chatbot interfaces answering lay-oriented questions about herpes zoster. Methods This cross-sectional evaluation used 46 standardised English-language questions across seven herpes zoster domains. Each question was submitted once to ChatGPT-5.5 Instant, DeepSeek-V4-Pro, Doubao-Seed-2.0-Pro, Gemini 3.5 Flash, and Qwen3.7-Plus under consumer-access conditions on June 2–3, 2026. Five dermatologists independently assessed safety, accuracy, empathy, DISCERN, Ensuring Quality Information for Patients (EQIP), Journal of the American Medical Association (JAMA) benchmark criteria, and the Global Quality Score (GQS). Six readability indices were calculated. Paired comparisons and inter-rater agreement were evaluated using prespecified statistical methods. Results The five interfaces generated 230 complete responses. Under the study conditions, 66 responses (28.7%) were classified as potentially unsafe using the prespecified ≥3/5-rater majority threshold, with no significant difference between interfaces (Cochran’s Q = 1.661, df = 4, p = 0.798). Sensitivity analyses using alternative thresholds showed consistent results. Accuracy and empathy differed across interfaces (both p < 0.001; Kendall’s W = 0.127 and 0.211, respectively), although effect sizes were small. Information-quality and overall-quality measures also showed outcome-specific differences. All six readability indices differed across interfaces (all p < 0.001); DeepSeek-V4-Pro generally produced text estimated to be easier to read, whereas Gemini 3.5 Flash produced text estimated to be more difficult to read. Safety agreement was substantial (Fleiss’ κ = 0.682), and ICC (2,1) values for other manually rated outcomes ranged from 0.832 to 0.889. Conclusion In this standardised benchmark, LLM chatbot interfaces provided generally favourable accuracy and information-quality scores but showed clinically relevant safety limitations, sparse source attribution and disclosure, and readability challenges. No interface consistently performed best across all outcomes. These findings represent single first responses obtained under the tested conditions and do not establish response stability across repeated queries. LLM-generated herpes zoster information should therefore be interpreted cautiously and should not replace individualised professional assessment or professionally reviewed patient information.
Background Mental disorders are a growing global health burden, yet healthcare resources remain scarce. Large language models (LLMs) may support public mental health information seeking, but their information quality, readability, and empathy in psychiatric contexts require validation. Method We developed a test bank of 48 public mental health questions from literature, Google Trends (2004–2025), and clinical consultations. Eight LLM chatbots were compared for information quality, source transparency, readability, and empathy using established rating instruments, readability indices, and psychiatrist-rated and user-perspective empathy assessments. Statistical analysis used Kruskal–Wallis and Dunn's post-hoc tests, with Spearman correlations. Results Gemini 3.0 Pro and GPT-5.2 Think showed relatively higher information quality and source transparency scores, whereas spontaneous source transparency was poor across models, with median JAMA scores of 0. No model met the sixth-grade readability standard; Claude Sonnet 4.5 generated relatively more readable responses, whereas Claude Sonnet 4.5 Think produced responses with the highest reading difficulty. In the psychiatrist-rated empathy assessment, Gemini 3.0 Pro and DeepSeek-V3 showed the highest high-empathy response rates, at 54.2% and 43.8%, respectively. User-perspective empathy ratings were generally lower, with DeepSeek-V3 and Gemini 3.0 Pro showing the highest user-perspective high-empathy response rates, at 31.2% and 27.1%, respectively. Information quality and source transparency metrics showed only weak correlations with readability metrics. Conclusion LLMs face important challenges in psychiatric information delivery. Gemini 3.0 Pro and GPT-5.2 Think showed higher information quality and source transparency scores and better information structuring, whereas Claude Sonnet 4.5 generated relatively more readable responses. However, source opacity and high reading difficulty limit direct patient-facing use. Empathy varied across models and differed between psychiatrist-rated and user-perspective assessments, suggesting that empathic communication requires separate optimization and validation with intended users. Future work should balance information quality, source transparency, readability, traceability, safety, and emotionally appropriate responses.
Background Patients increasingly use large language models (LLMs) to seek medical information, but their performance in answering diabetic retinopathy (DR) questions remains insufficiently characterized. Objective To evaluate the safety, accuracy, patient-centered communication, information quality, transparency, and readability of diabetic retinopathy information generated by widely accessible large language models. Methods Fifty standardized DR patient questions were submitted to ChatGPT, Gemini, DeepSeek, Doubao, Copilot, and Perplexity. Responses were independently evaluated by five raters using guideline-based reference answer elements. Rater-based outcomes included Safety, Accuracy, Empathy, DISCERN, EQIP, JAMA benchmark criteria, and Global Quality Score (GQS). Readability was assessed using the Automated Readability Index, Coleman–Liau Index, Flesch–Kincaid Grade Level, Gunning Fog Index, SMOG, and Flesch Reading Ease Score. Inter-rater agreement and between-model differences were analyzed using appropriate agreement statistics and paired nonparametric tests. Results Inter-rater agreement was moderate for Safety (Fleiss’ κ = 0.581) and good for non-binary outcomes, with ICC (2,1) values ranging from 0.806 to 0.870. Most responses were classified as safe, with safe response rates ranging from 86.0 to 94.0%, and no significant between-model difference was observed. Accuracy scores were generally high across models, although between-model differences were statistically significant with a small effect size. Empathy varied more noticeably, with Copilot scoring highest and Doubao and Perplexity scoring lower. Perplexity achieved the highest DISCERN and JAMA scores, while Copilot performed best on EQIP. GQS differences were small. Readability differed significantly across models, with Gemini generally producing the most readable responses; however, all models exceeded the recommended sixth-grade reading level, and all Flesch Reading Ease Scores were below the recommended threshold. Conclusion No single LLM performed best across all domains. Although the evaluated models generally provided safe and accurate responses to DR patient questions, limitations remained in safety escalation, empathy, transparency, and readability. LLM-generated DR information may support general patient education but should not replace professional ophthalmic consultation.
Abstract Background The quality and accessibility of menstrual health education (MHE) in low- and middle-income countries, including India, remain inadequate due to persistent challenges (eg, poverty, social stigma, and gender inequality). While community-driven initiatives have sought to raise awareness, artificial intelligence offers a scalable and efficient solution for disseminating accurate information. However, existing general-purpose large language models (LLMs) are often ill-suited for this task, tending to exhibit low accuracy, cultural insensitivity, and overly complex responses. To address these limitations, we developed MenstLLaMA—a specialized LLM tailored to the Indian context and designed to deliver MHE empathetically, supportively, and accessibly. Objective We aimed to develop and evaluate MenstLLaMA—a specialized LLM tailored to deliver accurate, culturally sensitive MHE—and assess its effectiveness in comparison to existing general-purpose models. Methods We curated MENST—a novel, domain-specific dataset comprising 23,820 question-answer pairs aggregated from medical websites, government portals, and health education resources. This dataset was systematically annotated with metadata capturing age groups, regions, topics, and sociocultural contexts. MenstLLaMA was developed by fine-tuning Meta-LLaMA-3-8B-Instruct, using parameter-efficient fine-tuning with low-rank adaptation to achieve domain alignment while minimizing computational overhead. We benchmarked MenstLLaMA against 9 state-of-the-art general-purpose LLMs, including GPT-4o, Claude-3, Gemini 1.5 Pro, and Mistral. The evaluation followed a multilayered framework: (1) automatic evaluation using standard natural language processing metrics (BLEU [Bilingual Evaluation Understudy], METEOR [Metric for Evaluation of Translation with Explicit Ordering], ROUGE-L [Recall-Oriented Understudy for Gisting Evaluation-Longest Common Subsequence], and BERTScore [Bidirectional Encoder Representations from Transformers Score]); (2) evaluation by clinical experts (N=18), who rated 200 expert-curated queries for accuracy and appropriateness; (3) medical practitioner interaction through the ISHA (Intelligent System for Menstrual Health Assistance) interactive chatbot, assessing qualitative dimensions (eg, relevance, understandability, preciseness, correctness, and context sensitivity); and (4) a user study with volunteer participants (N=200), who evaluated MenstLLaMA in 15- to 20-minute randomized sessions, rating the system across 7 qualitative user satisfaction metrics. Results MenstLLaMA achieved the highest scores in BLEU (0.059) and BERTScore (0.911), outperforming GPT-4o (BLEU: 0.052, BERTScore: 0.896) and Claude-3 (BERTScore: 0.888). Clinical experts preferred MenstLLaMA’s responses over gold-standard answers in several culturally sensitive cases. In medical practitioners’ evaluations using the ISHA—the chat interface powered by MenstLLaMA—the model scored 3.5 in relevance, 3.6 in understandability, 3.1/5 in preciseness, 3.5/5 in correctness, and 4.0/5 in context sensitivity. User evaluations indicated even stronger results, with ratings of 4.7/5 for understandability, 4.3/5 for relevance, 4.28/5 for preciseness, 4.1/5 for correctness, 4.6/5 for tone, 4.2/5 for flow, and 3.9/5 for context sensitivity. Conclusions MenstLLaMA demonstrates exceptional accuracy, empathy, and user satisfaction within the domain of MHE, bridging critical gaps left by general-purpose LLMs. Its potential for integration into broader health education platforms positions it as a transformative tool for menstrual well-being. Future research could explore its long-term impact on public perception and menstrual hygiene practices, while expanding demographic representation, enhancing context sensitivity, and integrating multimodal and voice-based interactions to improve accessibility across diverse user groups.
Background Chinese large language models (LLMs) are increasingly used for HIV/AIDS health information, yet their multidimensional performance remains unevaluated. This study systematically assessed five mainstream Chinese LLMs across accuracy, caring, completeness, understandability, and actionability. Methods A cross-sectional evaluation was conducted using 50 standardized HIV/AIDS questions. Each model's responses were rated by three independent experts on five dimensions, including adapted PEMAT-based items. Latent profile analysis (LPA) and partial correlation network analysis were employed to identify quality typologies and interdimensional dependencies. Chi-square tests examined associations between profile membership and model identity or question domain. Results The four general-purpose LLMs showed high accuracy (mean > 4.5/5) and completeness (86%−98% full marks), with no significant differences among them (all P-FDR > 0.05). HuatuoGPT-II, the only medical-domain-specific model, scored significantly lower on both dimensions (all P < 0.001). Its responses were less consistent with current guidelines and included one instance of implicit stigmatizing language. All models showed systemic deficits in caring (median = 3.0/5) and actionability (max = 1.70/3). Understandability showed a ceiling effect and no significant profile-level differences, although small model-level differences were observed. LPA identified four profiles: High Accuracy and Completeness (50.0%), Moderate-Balanced (3.6%), Low-Quality (18.8%), and High Completeness and Actionability-Oriented (27.6%). Network analysis identified completeness as the most central node (strength = 0.813) and actionability as a key bridge (betweenness = 0.167). A weak negative partial correlation between accuracy and actionability (r = −0.083) suggested a potential trade-off between technical precision and practical guidance at the statistical level. Profile distributions differed by model (χ2 = 124.99, P < 0.001) but not domain (P = 0.625), indicating heterogeneity stems from model-specific characteristics. Yuanbao-Hy3 showed the most balanced humanistic-technical profile. DeepSeek-V4 excelled technically but lacked caring. Conclusions In this evaluation, the four general-purpose models outperformed the single medical-domain-specific model examined in factual accuracy and completeness. However, systemic deficits in empathetic communication and actionable guidance persist. Model selection should prioritize holistic profile distributions over single-dimension scores. Developers must integrate de-stigmatizing corpora and action-oriented alignment; clinicians should position AI as supplementary to professional care. Future evaluations should adopt person-centred, multi-turn, and real-user validated frameworks.
Introduction Since its introduction in November 2022, the artificial intelligence large language model ChatGPT has taken the world by storm. Among other applications it can be used by patients as a source of information on diseases and their treatments. However, little is known about the quality of the sarcoma-related information ChatGPT provides. We therefore aimed at analyzing how sarcoma experts evaluate the quality of ChatGPT’s responses on sarcoma-related inquiries and assess the bot’s answers in specific evaluation metrics. Methods The ChatGPT responses to a sample of 25 sarcoma-related questions (5 definitions, 9 general questions, and 11 treatment-related inquiries) were evaluated by 3 independent sarcoma experts. Each response was compared with authoritative resources and international guidelines and graded on 5 different metrics using a 5-point Likert scale: completeness, misleadingness, accuracy, being up-to-date, and appropriateness. This resulted in maximum 25 and minimum 5 points per answer, with higher scores indicating a higher response quality. Scores ≥21 points were rated as very good, between 16 and 20 as good, while scores ≤15 points were classified as poor (11–15) and very poor (≤10). Results The median score that ChatGPT’s answers achieved was 18.3 points (IQR, i.e., Inter-Quartile Range, 12.3–20.3 points). Six answers were classified as very good, 9 as good, while 5 answers each were rated as poor and very poor. The best scores were documented in the evaluation of how appropriate the response was for patients (median, 3.7 points; IQR, 2.5–4.2 points), which were significantly higher compared to the accuracy scores (median, 3.3 points; IQR, 2.0–4.2 points; p = 0.035). ChatGPT fared considerably worse with treatment-related questions, with only 45% of its responses classified as good or very good, compared to general questions (78% of responses good/very good) and definitions (60% of responses good/very good). Discussion The answers ChatGPT provided on a rare disease, such as sarcoma, were found to be of very inconsistent quality, with some answers being classified as very good and others as very poor. Sarcoma physicians should be aware of the risks of misinformation that ChatGPT poses and advise their patients accordingly.
Abstract Background Implementing structured shared decision-making (SDM) requires high-quality, reliable patient information. In radiation oncology, patients often have limited knowledge and misconceptions about therapy and side effects, affecting their decision-making. Large language model-based AI systems (LLMs) may help by providing evidence-based information in accessible language, but successful implementation depends on the willingness of patients and health care professionals (HCPs) to adopt these technologies. Methods A survey was conducted among patients undergoing radiation therapy and HCPs between 03/2024 and 02/2025. Data was collected using structured electronic questionnaires (32 items for patients, 35 for HCPs). The survey assessed sociodemographic characteristics, the status of SDM in oncology, sources of information relevant to SDM, and current and anticipated LLM applications. Data were analyzed using descriptive statistics and logistic regression analysis. Results The internet was the prime information source for patients (n = 400). Regarding current use of LLMs, a large discrepancy between patients and HCPs (n = 200) was observed (18.2% vs 69.5%). Although 77% of HCPs believed that patients will rely on LLMs in the future, only 29.1% of patients agreed. Most patients (65.8%) stated that even as LLMs improve, they will continue to trust physicians more; 46% of HCPs shared this view. Only 16.5% of patients were convinced that LLMs provide all relevant data for SDM in cancer care. Familiarity with technology was the strongest predictor of LLM use among patients. Conclusion Only a minority of radiation oncology patients currently use LLMs, and many remain skeptical about their future role—contrasting with the more optimistic expectations of HCPs.
Large language models are increasingly used in healthcare communication, yet most evaluations emphasize response quality while assuming that the user's concern has been interpreted correctly. We introduce HerHealthEval, a controlled evaluation framework for multilingual understanding of women's-health communication. For each clinical case, HerHealthEval provides matched versions in English, French, and Modern Standard Arabic using six communicative forms: canonical, clinical, layperson, indirect or hedged, emotionally concerned, and deliberately under-specified. The first five express the same underlying concern and retain the same clinical information, whereas the under-specified form intentionally omits relevant details to test whether the model recognizes that clarification is needed. We evaluate a multilingual instruction model and QLoRA-adapted variants on concern classification, risk calibration, clarification behavior, parse compliance, and cross-form consistency. Results reveal that aggregate accuracy and consistency can conceal safety-relevant failures. A multilingual adaptation model reaches 0.994 under-triage in French and Arabic under language-asymmetric risk supervision. A controlled re-adaptation using source-derived, language-invariant risk labels reduces under-triage to 0.572 and 0.558, respectively. These findings show that robust multilingual healthcare evaluation requires explicit testing of register variation, uncertainty handling, and the provenance and invariance of adaptation labels.
Automated fact-checking is essential for ensuring the reliability of public health information, yet the biomedical domain poses unique challenges. Validating biomedical claims requires rigorous interpretation of scientific literature, assessment of retrieved evidence, and comprehensive justification toward the conclusion. Although Large Language Models (LLMs) enhanced by Retrieval-Augmented Generation (RAG) and agentic search perform automated fact-checking in a retrieve-then-verify paradigm, current methods still output isolated prediction labels, lacking explanatory depth and offers limited utility for human understanding. To bridge this gap, we introduce an LLM-based agent named BioCheck Agent that generates structured biomedical fact-checking reports with agentic search. Rather than merely outputting supported or refuted labels, our agent synthesizes final conclusions with retrieved evidence and rigorous analysis. To ensure domain-specific accuracy, BioCheck Agent exclusively searches high-quality scientific literature in PubMed, utilizing advanced Boolean search operators. Recognizing that direct prompting often results in hallucinations and low-quality reports, especially for lightweight open-source models, we further propose the Evidence-Grounded Group Relative Policy Optimization (EG-GRPO) to perform reinforcement learning on BioCheck Agent with a task-specific reward that incentivizes advanced search behavior and high-quality evidence retrieval while penalizing hallucinations. Our experimental results show that compared to the base model Qwen3.5-4B, BioCheck Agent with EG-GRPO improves label prediction accuracy on SciFact by 9.95%. Furthermore, it achieves a 3.7% higher evidence quality score and a 19.63% lower evidence hallucination rate, demonstrating its ability to generate biomedical fact-checking reports with improved accuracy and quality.
Background: Large language models (LLMs) can improve healthcare information delivery in low-resource settings but may produce inaccurate or culturally inappropriate advice. This study evaluated domain-specific fine-tuning for maternal health and vaccination in Nigeria. Objective: To compare HelpMum's MamaBot-Llama and Vax-Llama with Meta's Llama-3.1-8B-Instruct for accuracy, safety, clarity, contextual appropriateness, and trustworthiness. Methods: We evaluated 200 healthcare questions, 100 each for maternal health and vaccination, across five subdomains per domain. MamaBot-Llama and Vax-Llama were fine-tuned using Low-Rank Adaptation on over 36,000 maternal health and 9,000 vaccination question-answer pairs, respectively. Two Nigerian licensed physicians independently rated responses using a 5-point Likert scale. Paired comparisons used Wilcoxon signed-rank tests. Results: Performance varied by domain. MamaBot-Llama significantly outperformed the base model across all criteria, with a 4.9% overall improvement (p<.001), including gains in clinical trustworthiness (+7%) and medical accuracy (+5%). Critical issues decreased by 50%, and clinicians preferred it in 78% of cases. In contrast, Vax-Llama showed a 5.2% overall decline (p<.001), with critical issues increasing by 192% and safety concerns by 400%. Conclusions: Domain-specific fine-tuning can improve healthcare LLM performance when based on high-quality, clinician-curated data, but may also degrade performance when dataset quality is inadequate. Rigorous domain-specific validation is essential before clinical deployment. Physician evaluators provided informed consent, and chatbot logs were anonymized. Keywords: Large language models; Fine-tuning; Maternal health; Vaccination; Healthcare AI; Low-resource settings; Nigeria; Model evaluation; LoRA; Medical accuracy
Artificial Intelligence (AI) technologies, while serving as a foundational enabler for modern social media and digital health services, exert a bivalent effect by simultaneously acting as a combatant against and a spread vector for misinformation. A prevalent challenge in mitigating this issue arises in non-English contexts and low socioeconomic classes, where limited data hinders the training of AI models for effective detection. Consequently, culturally and linguistically diverse (CALD) communities struggle to access trustworthy health information through AI-driven tools. Current AI tools underperform due to a lack of training data and are largely unable to consider language nuances and traditions in non-English contexts. This research addresses these gaps by proposing a CALD-friendly AI-based health misinformation detector and providing a dashboard for medical professionals to analyse this misinformation, a critical step toward mitigating a growing concern among CALD populations. To this end, we conduct a series of experiments using a Bangla-translated health misinformation dataset to evaluate the performance of various Small Language Models (SLMs). SLMs are particularly relevant in this context given the frequent underperformance of Large Language Models (LLMs), which often stems from insufficient domain-specific knowledge and the prohibitive costs of resource-intensive fine-tuning. The results demonstrate that Phi-4 is the superior model, achieving an ideal balance between precision and recall in claim extraction. Then, to mitigate the limitations of SLMs, we design and test a novel health misinformation detection framework grounded in Responsible Natural Language Processing (NLP), which incorporates cultural sensitivity, potential for harm, and communication quality, thereby providing a holistic lens for evaluating misinformation in low-resource languages.
Patients and caregivers increasingly use artificial intelligence (AI) tools to interpret medical reports, weigh care decisions, and seek emotional support. Yet most research treats patient-facing AI as a private exchange between a user and a system. This study examines how AI-related content is taken up once users carry it back into the peer communities, using data from House086, China's largest online community for lymphoma patients and caregivers. We identified roughly 400 publicly accessible threads (2014-2026) through keyword searches and manual screening, extracted them into structured case profiles using a schema-prompted large language model, and conducted mixed-method analysis. After quality control, the verified analytic sample comprised 337 post-ChatGPT records. Members most often reported using AI for informational support, followed by second opinions and psychosocial support. Although members often introduced AI favorably, roughly one in six described feeling overwhelmed by AI output. When other members responded, they frequently engaged the poster's underlying medical or emotional intent while leaving the AI dimension unaddressed. This tendency persisted even in threads seeking triangulation between AI output and other information sources. When members did discuss the AI, they were more often cautious than endorsing. Rather than systematically auditing AI output, the community more often worked to deflate the false certainty it produced, placing a single AI answer back among multiple sources of judgment. This study argues that AI does not replace the interpretive work of online health communities, nor is it systematically audited by them. Instead, it shifts the locus of sensemaking downstream, so that the community continues to catch the person even when it does not catch the AI.
Plain Language Summaries (PLS) aim to make research accessible to lay readers, but they are typically written in a one-size-fits-all style that ignores differences in readers'information needs and comprehension. In health contexts, this limitation is particularly important because misunderstanding scientific information can affect real-world decisions. Large language models (LLMs) offer new opportunities for personalizing PLS, but it remains unclear whether personalization helps, which strategies are most effective, and how to balance personalization with safety. We introduce ReLay, a dataset of 300 participant--PLS pairs from 50 lay participants in both static (expert-written) and interactive (LLM-personalized) settings. ReLay includes user characteristics, health information needs, information-seeking behavior, comprehension outcomes, interaction logs, and quality ratings. We use ReLay to evaluate five LLMs across two personalization methods. Personalization improves comprehension and perceived quality, but it also raises the risk of reinforcing user biases and introducing hallucinations, revealing a trade-off between personalization and safety. These findings highlight the need for personalization methods that are both effective and trustworthy for diverse lay audiences.
Incorporating large language models (LLMs) in medical question answering demands more than high average accuracy: a model that returns substantively different answers each time it is queried is not a reliable medical tool. Online health communities such as Reddit have become a primary source of medical information for millions of users, yet they remain highly susceptible to misinformation; deploying LLMs as assistants in these settings amplifies the need for output consistency alongside correctness. We present a practical, open-source evaluation framework for assessing small, locally-deployable open-weight LLMs on medical question answering, treating reproducibility as a first-class metric alongside lexical and semantic accuracy. Our pipeline computes eight quality metrics, including BERTScore, ROUGE-L, and an LLM-as-judge rubric, together with two within-model reproducibility metrics derived from repeated inference (N=10 runs per question). Evaluating three models (Llama 3.1 8B, Gemma 3 12B, MedGemma 1.5 4B) on 50 MedQuAD questions (N=1,500 total responses) reveals that despite low-temperature generation (T=0.2), self-agreement across runs reaches at most 0.20, while 87-97% of all outputs per model are unique -- a safety gap that single-pass benchmarks entirely miss. The clinically fine-tuned MedGemma 1.5 4B underperforms the larger general-purpose models on both quality and reproducibility; however, because MedGemma is also the smallest model, this comparison confounds domain fine-tuning with model scale. We describe the methodology in sufficient detail for practitioners to replicate or extend the evaluation for their own model-selection workflows. All code and data pipelines are available at https://github.com/aviad-buskila/llm_medical_reproducibility.
The escalating global mental health crisis, marked by persistent treatment gaps, availability, and a shortage of qualified therapists, positions Large Language Models (LLMs) as a promising avenue for scalable support. While LLMs offer potential for accessible emotional assistance, their reliability, therapeutic relevance, and alignment with human standards remain challenging to address. This paper introduces a human-grounded evaluation methodology designed to assess LLM generated responses in therapeutic dialogue. Our approach involved curating a dataset of 500 mental health conversations from datasets with real-world scenario questions and evaluating the responses generated by nine diverse LLMs, including closed source and open source models. More specifically, these responses were evaluated by two psychiatric trained experts, who independently rated each on a 5 point Likert scale across a comprehensive 6 attribute rubric. This rubric captures Cognitive Support and Affective Resonance, providing a multidimensional perspective on therapeutic quality. Our analysis reveals that LLMs provide strong cognitive reliability by producing safe, coherent, and clinically appropriate information, but they demonstrate unstable affective alignment. Although closed source models (e.g., GPT-4o) offer balanced therapeutic responses, open source models show greater variability and emotional flatness. We reveal a persistent cognitive-affective gap and highlight the need for failure aware, clinically grounded evaluation frameworks that prioritize relational sensitivity alongside informational accuracy in mental health oriented LLMs. We advocate for balanced evaluation protocols with human in the loop that center on therapeutic sensitivity and provide a framework to guide the responsible design and clinical oversight of mental health oriented conversational AI.
Access to reliable maternal healthcare information is a major challenge in rural India due to limited medical resources and infrastructure. With over 830 million internet users and nearly half of rural women online, digital tools offer new opportunities for health education. This study evaluates large language models (LLMs) like ChatGPT-4o, Perplexity AI, and GeminiAI to provide reliable and understandable pregnancy-related information. Seventeen pregnancy-focused questions were posed to each model and compared with responses from maternal health professionals. Evaluations used semantic similarity, noun overlap, and readability metrics to measure content quality. Results show Perplexity closely matched expert semantics, while ChatGPT-4o produced clearer, more understandable text with better medical terminology. As internet access grows in rural areas, LLMs could serve as scalable aids for maternal health education. The study highlights the need for AI tools that balance accuracy and clarity to improve healthcare communication in underserved regions.
No abstract available
No abstract available
No abstract available