A guideline-based artificial intelligence system for hypertension in primary care: a single-arm pre–post study
Hypertension control remains suboptimal in primary care, particularly in resource-constrained settings lacking clinical decision support. This retrospective cohort study evaluated the real-world changes associated with a Guideline-based Digital System (GDS) implemented in a primary care network in Chengdu, China. Data from 515 hypertensive patients with complete baseline and 12-month follow-up data were analyzed. The primary outcome was office blood pressure control rate (< 140/90 mmHg, or < 130/80 mmHg for patients with diabetes); secondary outcomes included changes in systolic/diastolic blood pressure, fasting blood glucose, lipid profiles, and medication adherence (MMAS-8). After 12 months, systolic blood pressure decreased by 8.86 mmHg (95% CI: -10.67~-7.04; P < 0.001) and diastolic by 3.42 mmHg (95% CI: -4.60~-2.26; P < 0.001). The control rate increased from 60.39% to 76.70% (+ 16.31% points; P < 0.001). MMAS-8 score changed from 4.98 to 6.18 ( P < 0.001). Significant changes were also observed in fasting blood glucose and lipid profiles (all P < 0.05). Implementation of the GDS system was associated with pre-post changes in blood pressure control, cardiovascular risk factors, and medication adherence, supporting the potential scalability of AI-based guideline-digitalized systems in resource-constrained primary care settings. However, due to the lack of a control group, high exclusion rate, and potential selection bias, these findings cannot establish causality and should be presented as preliminary associations.
Authors
- Zhilin Gou
- Ying Xiao
- Bo Zhou
- Xinyu Zhang
- Yan Yang (ORCID: https://orcid.org/0009-0007-6042-0452)
- Yubin Liu
Publication Details
- Journal
- Scientific Reports
- Published
- 2026-11-19
- DOI
- https://doi.org/10.1038/s41598-026-81038-3
- Primary Topic
- Blood Pressure and Hypertension Studies
- Type
- article
- Field-Weighted Citation Impact
- 0.00