When trust does not translate into service preference: a cross-sectional study of patient responses to DeepSeek-enabled hospital intelligent services in a Chinese public hospital
4 affiliations
- 1School of Management, Xuzhou Medical University
- 2Jiangsu Provincial Key Research Center for Cross-Regional Health Governance
- 3Affiliated Hospital of Xuzhou Medical University
- 4School of Medical Artificial Intelligence, Xuzhou Medical University
Abstract
Objective: This study aimed to examine the current pattern of patients’ overall trust judgment regarding DeepSeek-enabled hospital intelligent services, analyze its association with service preference, and explore important considerations related to patients’ service mode choices.
Methods: A cross-sectional survey was conducted in a Chinese public hospital during May–June 2025. Convenience sampling was used to recruit 214 patients who had used the intelligent service. Data were collected via face-to-face questionnaires. Because the principal variables were ordinal and subgroup sizes were imbalanced, nonparametric tests were applied, including Spearman’s rank correlation, the Fisher–Freeman–Halton exact test, and the Kruskal-Wallis test.
Results: Patients showed a baseline level of trust in hospital intelligent services, with 46.7% reporting positive trust attitudes and 13.6% reporting distrust. Spearman rank correlation analysis demonstrated weak positive relationships between higher perceived accuracy of department recommendations ( ρ = −0.140, p < 0.05) and treatment suggestions ( ρ = −0.204, p < 0.01) and more favorable overall trust judgments (lower trust scores represented more favorable judgments), whereas no consistent associations were found for other indicators. The Fisher–Freeman–Halton exact test showed a significant association between overall trust judgment and service preference ( p < 0.001, Cramer’s V = 0.610), although this did not translate into a straightforward preference for more AI-led service modes: only 6.1% of patients preferred pure intelligent services, while 67.8% still preferred traditional human services. Data security protection received the highest importance rating ( M = 3.99, SD = 1.00) and was associated with lower preference for service modes involving greater AI participation. The Kruskal-Wallis test indicated a significant overall difference in data security protection ratings across service preference groups ( χ ² = 42.175, df = 2, p < 0.001). Further Spearman rank correlation analysis showed that AI-led service preference was significantly negatively correlated with the importance attached to data security protection ( ρ = −0.328, p < 0.001).
Conclusion: Patients hold basic trust in DeepSeek-based intelligent services, yet trust does not convert into preference for pure AI services. Data-security considerations were associated with lower preference for AI-involved service modes. Strengthening data-security governance, institutional transparency, and service reliability may support the responsible expansion of patient-facing intelligent services.
Related
- Comparison of ChatGPT, Gemini, and DeepSeek responses to questions on medication-related osteonecrosis of the jaw (MRONJ): a comparative analysis of expert-rated clinical quality, readability, response time, and clinical safety risk.
- Comparative Evaluation of ChatGPT, DeepSeek, and Physician-Generated Acupuncture Treatment Protocols Across Chinese and English Language Settings: Cross-Sectional Evaluation Study.
- Research protocol for the diagnostic accuracy of intelligent portable impulse oscillometry across different respiratory diseases: a multicentre, cross-sectional study
- Factors Associated with Anchor-Number Preference Among Orthopedic Surgeons Performing Bankart Repair in Saudi Arabia: A Cross-Sectional Survey.
- Adaptive Intelligent Microneedles for Sustained Wound Healing: Translating Microenvironment Signals into Stage-Specific Therapeutic Responses
BibTeX
@article{wen2026when,
title = {{When trust does not translate into service preference: a cross-sectional study of patient responses to DeepSeek-enabled hospital intelligent services in a Chinese public hospital}},
author = {Zongliang Wen and Beibei Fu and Yiheng Qin and Zhuying Gao and Yan Zhu},
journal = {Frontiers in Public Health},
year = {2026},
month = nov,
doi = {10.3389/fpubh.2026.1855398},
url = {https://www.frontiersin.org/articles/10.3389/fpubh.2026.1855398/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - When trust does not translate into service preference: a cross-sectional study of patient responses to DeepSeek-enabled hospital intelligent services in a Chinese public hospital
AU - Zongliang Wen
AU - Beibei Fu
AU - Yiheng Qin
AU - Zhuying Gao
AU - Yan Zhu
PY - 2026
DA - 2026/11/04
JO - Frontiers in Public Health
DO - 10.3389/fpubh.2026.1855398
UR - https://www.frontiersin.org/articles/10.3389/fpubh.2026.1855398/full
AB - Objective: This study aimed to examine the current pattern of patients’ overall trust judgment regarding DeepSeek-enabled hospital intelligent services, analyze its association with service preference, and explore important considerations related to patients’ service mode choices. Methods: A cross-sectional survey was conducted in a Chinese public hospital during May–June 2025. Convenience sampling was used to recruit 214 patients who had used the intelligent service. Data were collected via face-to-face questionnaires. Because the principal variables were ordinal and subgroup sizes were imbalanced, nonparametric tests were applied, including Spearman’s rank correlation, the Fisher–Freeman–Halton exact test, and the Kruskal-Wallis test. Results: Patients showed a baseline level of trust in hospital intelligent services, with 46.7% reporting positive trust attitudes and 13.6% reporting distrust. Spearman rank correlation analysis demonstrated weak positive relationships between higher perceived accuracy of department recommendations ( ρ = −0.140, p < 0.05) and treatment suggestions ( ρ = −0.204, p < 0.01) and more favorable overall trust judgments (lower trust scores represented more favorable judgments), whereas no consistent associations were found for other indicators. The Fisher–Freeman–Halton exact test showed a significant association between overall trust judgment and service preference ( p < 0.001, Cramer’s V = 0.610), although this did not translate into a straightforward preference for more AI-led service modes: only 6.1% of patients preferred pure intelligent services, while 67.8% still preferred traditional human services. Data security protection received the highest importance rating ( M = 3.99, SD = 1.00) and was associated with lower preference for service modes involving greater AI participation. The Kruskal-Wallis test indicated a significant overall difference in data security protection ratings across service preference groups ( χ ² = 42.175, df = 2, p < 0.001). Further Spearman rank correlation analysis showed that AI-led service preference was significantly negatively correlated with the importance attached to data security protection ( ρ = −0.328, p < 0.001). Conclusion: Patients hold basic trust in DeepSeek-based intelligent services, yet trust does not convert into preference for pure AI services. Data-security considerations were associated with lower preference for AI-involved service modes. Strengthening data-security governance, institutional transparency, and service reliability may support the responsible expansion of patient-facing intelligent services.
SN - 2296-2565
ER -