Ischemic Stroke
Conditions
Keywords
Artificial Intelligence, Stroke Prevention, Primary Prevention
Brief summary
This study will evaluate whether an artificial intelligence (AI)-driven dynamic health management strategy can help prevent ischemic stroke in adults at high risk of stroke. Participants will be identified through community-based screening in Beijing using the AI-ExpoStroke model together with established stroke risk factors. Communities will be randomly assigned to either an AI-driven health management group or a usual community-based health management group. Participants in the AI-driven group will receive continuous health management supported by a digital platform, mobile applications or WeChat-based tools, wearable-device data when available, personalized health guidance, and remote support from community health care providers. Participants in the usual-care group will receive routine community health services, including health examinations, health education, chronic disease follow-up, and medication guidance. Participants will be followed for 36 months. The main goal is to determine whether AI-driven health management reduces the occurrence of first-ever ischemic stroke. The study will also evaluate transient ischemic attacks, stroke-related disability, mortality, control of major vascular risk factors, adherence to health management, and health economic outcomes.
Detailed description
This is an investigator-initiated, multicenter, open-label, stratified cluster-randomized, parallel-group clinical study conducted in community settings in Beijing, China. The study is designed to evaluate the effectiveness, safety, and health economic value of an AI-driven dynamic health management strategy for the primary prevention of ischemic stroke in adults identified as being at high risk of stroke. Potential participants will be identified from prospective community-based screening programs. Eligibility will be determined using the AI-ExpoStroke risk assessment model together with established stroke "8+2" high-risk factors. The AI-ExpoStroke model was developed and externally validated as part of preceding observational research and is used in the present interventional study primarily for identification and enrollment of individuals at high risk of stroke. Randomization will be performed at the community level rather than at the individual participant level. Communities will be stratified according to area type, baseline risk-factor profile, and community health service resources, and will then be randomly assigned in a 1:1 ratio to the intervention group or control group. The random allocation sequence will be generated by an independent statistician using SAS or R. Because of the nature of the intervention, the study is open label. Participants in the intervention group will receive AI-driven remote follow-up and continuous dynamic health management through a stroke prevention and management cloud platform, mobile applications or WeChat-based tools, wearable-device interfaces when applicable, and coordinated support from community health care providers. Participants will generally be encouraged to report health information such as blood pressure, body weight, medication use, and lifestyle-related information at least monthly. The platform will provide individualized risk-management targets, health reminders, lifestyle recommendations, and remote guidance from community health care providers. AI-generated recommendations will be used as supportive management tools and will not replace routine clinical decision-making by physicians. Participants in the control group will receive usual community-based health management for individuals at high risk of stroke. This includes routine health examinations, basic health education, standard chronic disease follow-up, and medication guidance. Participants in the control group will not receive the dynamic AI-based management service. All participants will be followed for 36 months, with formal follow-up assessments at 12, 24, and 36 months. Suspected stroke, transient ischemic attack, hospitalization, and death will be evaluated when they occur. Endpoint events will be verified using relevant clinical information and, when applicable, neuroimaging findings and medical records according to a predefined endpoint adjudication process. The primary objective is to determine whether AI-driven dynamic health management reduces the 3-year cumulative incidence of first-ever ischemic stroke compared with usual community-based health management. Secondary evaluations include transient ischemic attack, stroke-related disability, mortality, vascular risk-factor control, adherence to intervention and follow-up, improvement in stroke-prevention knowledge, and health economic outcomes. The study does not involve an investigational drug, an invasive investigational device, or collection of additional research-specific biological samples. Laboratory and clinical information used in the study will primarily be obtained from examinations performed as part of routine clinical care, routine physical examinations, or standard community health management.
Interventions
OTHERUsual Community-Based Health Management
Usual community-based health management for adults at high risk of stroke, including routine health examinations, basic health education, standard chronic disease follow-up, and routine medication guidance according to community health care practice. Participants do not receive the AI-driven dynamic health management service provided through the AI-ExpoStroke platform.
OTHERAI-Driven Dynamic Health Management
A comprehensive AI-supported health management strategy for adults at high risk of stroke. The intervention integrates the AI-ExpoStroke risk assessment system with a digital stroke prevention and management platform, mobile applications or WeChat-based tools, wearable-device interfaces when applicable, and coordinated community health care. Participants generally report blood pressure, body weight, medication use, and lifestyle-related information at least monthly. The platform supports dynamic risk assessment, individualized risk-factor management targets, health reminders, lifestyle recommendations, health education, and remote guidance from community health care providers. AI-generated recommendations are intended to support health management and do not replace clinical decision-making by physicians.
Sponsors
Xuanwu Hospital, Beijing
Study design
Allocation
RANDOMIZED
Intervention model
PARALLEL
Primary purpose
PREVENTION
Masking
NONE
Masking description
blank
Intervention model description
This is a stratified cluster-randomized, parallel-group study. The unit of randomization is the community rather than the individual participant. Communities are stratified according to area type, baseline risk-factor level, and community health service resources, and are then randomly assigned in a 1:1 ratio to the AI-driven dynamic health management group or the usual community-based health management group. The random allocation sequence is generated by an independent statistician using SAS or R.
Eligibility
Sex/Gender
ALL
Age
30 Years to 80 Years
Healthy volunteers
No
Inclusion criteria
* Age 30 years or older, with no restriction on sex. * Permanent resident of Beijing. * No previously diagnosed stroke based on prospective community screening. * Identified as being at high risk of stroke by the AI-ExpoStroke model in combination with the established stroke "8+2" high-risk factors. * Able to comply with study follow-up and willing to provide informed consent.
Exclusion criteria
* Previous diagnosis of ischemic stroke or hemorrhagic stroke. * Severe cognitive impairment. * Severe organic disease, including malignant tumors, New York Heart Association (NYHA) class III or higher heart failure, renal failure, or other severe conditions. * Psychiatric or language impairment that prevents completion of study questionnaires or follow-up. * Inability to obtain complete follow-up data or unwillingness to permit access to relevant study data.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| 3-Year Cumulative Incidence of First-Ever Ischemic Stroke | From randomization through 36 months | The proportion of participants who experience a first-ever ischemic stroke during the 36-month follow-up period. Ischemic stroke will be confirmed based on clinical symptoms and signs together with neuroimaging evidence, including computed tomography or magnetic resonance imaging, according to the predefined endpoint adjudication process. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Cumulative Incidence of First-Ever Transient Ischemic Attack | From randomization through 36 months | The proportion of participants who experience a first-ever transient ischemic attack (TIA) during the follow-up period, based on clinical evaluation and relevant medical records according to the predefined endpoint adjudication process. |
| Proportion of Participants With Disability Following Incident Stroke | From randomization through 36 months | The proportion of participants with an incident stroke who are assessed as having stroke-related disability during follow-up. Disability status will be evaluated at endpoint assessment, and the modified Rankin Scale (mRS) will be recorded for participants with stroke-related disability. |
| All-Cause Mortality | From randomization through 36 months | The proportion of participants who die from any cause during the 36-month follow-up period. Deaths will be verified using available medical records, discharge summaries, death certificates, or other relevant records. |
| Cardiovascular and Cerebrovascular Mortality | From randomization through 36 months | The proportion of participants who die from cardiovascular or cerebrovascular causes during the 36-month follow-up period. Cause of death will be determined using available clinical records and the predefined endpoint adjudication process. |
| Proportion of Participants Achieving Blood Pressure Control | At 12, 24, and 36 months | The proportion of participants who meet the predefined blood pressure control target at follow-up assessments. |
Outcome results
None listed