Profile-associated financial and access-related framing in LLM-generated pediatric asthma referral plans: a factorial audit of seven large language models
By
Zhendong Liu
Xiaoping Yang
Yu Zhang
Yujing Xu
Yue Xiang
Hongyan Wang
July 7, 2026
Objective: To evaluate whether name and bundled geographic-access signals are associated with differences in LLM-generated pediatric asthma referral plans.
Approach: Method: labelMethod: textKey Findings: The DeShawn name signal was associated with a higher financial-access term rate (IRR, 1.47; p < 0.001). The bundled geographic-access signal was also associated with a higher financial-access term rate (IRR, 2.40; p < 0.001). No positive multiplicative synergy was found between name and geographic signals (IRR, 0.79; p = 0.048). Variability was observed in SDOH Recognition Depth, Location-Friction Acknowledgment, Navigator Recommendation, and Empathy/Subjectivity by profile. Interpretation: LLM-generated pediatric asthma referral plans varied in financial-access, geographic-access, navigation, SDOH-recognition, and tone-related framing.
Limitations: The study does not establish downstream harm or positive synergistic interaction. Human validation showed moderate endpoint-specific reliability. Conclusion: The findings support the need for evaluating structural and access-related framing alongside biomedical content in clinical LLM audits.