Supplementary Files for "Large language models underrepresent quality-of-life burden across sociodemographic contexts in atopic dermatitis counseling"

Supplementary File 1: Standardized zero-shot "first-consultation" prompt template, the 18 verbatim scenario prompts (2 ages × 3 ethnic backgrounds × 3 socioeconomic settings), and all 90 verbatim large language model outputs. Supplementary File 2: Readability results (Flesch Reading Ease) and the complete statistical output (Friedman tests with Bonferroni adjustment) supporting the analyses. Supplementary File 3: Domain-level quality-of-life representation coding (✓/✗) for every model output, by scenario and model, across the pediatric (10-domain) and adult (12-domain) frameworks. Supplementary File 4: Independent scores from two blinded dermatologists for each model output across the five evaluation dimensions (clinical coverage/completeness, accuracy/safety, applicability/contextuality, respect/nonstigmatizing communication, empathy), reported as 5-point Likert scores and linear percentage conversions, by scenario and model.补充文件1：标准化零样本（Zero-shot）“首次问诊”提示词模板、18条逐字场景提示词（按2个年龄组×3个种族背景×3个社会经济场景划分），以及全部90条逐字大语言模型（Large Language Model，LLM）输出结果。 补充文件2：可读性评分结果（弗莱施阅读易度（Flesch Reading Ease）），以及支撑本研究分析的完整统计输出结果（采用邦费罗尼校正的弗里德曼检验（Friedman tests with Bonferroni adjustment））。 补充文件3：按场景与模型分组的、针对所有模型输出的领域级生活质量表征编码（✓/✗），涵盖儿科（10个领域）与成人（12个领域）两套评估框架。 补充文件4：由两名盲法皮肤科医师针对所有模型输出，在5项评估维度（临床覆盖度/完整性、准确性/安全性、适用性/语境契合性、尊重性/无污名化沟通、共情性）上给出的独立评分；评分以5级李克特（Likert）量表得分与线性百分比转换值的形式呈现，按场景与模型分组报告。