For the CardioCare program, phase 1 focused on the development of the system, which was conducted at the Health Management Center of the First Affiliated Hospital of Sun Yat-sen University in Guangzhou, China. This center conducts routine health examinations for residents. Phase 1 activities included system design and software development, followed by a pilot implementation with real patients and providers to assess usability and integration into the clinical workflow. The primary goals were to ensure the system’s technical functionality in a real-world setting and to gather preliminary data on user engagement prior to a future randomized trial. This implementation study is reported in accordance with the iCHECK-DH (Guidelines and Checklist for the Reporting on Digital Health Implementations) statement to ensure transparency, completeness, and reproducibility of digital health reporting [6].

The CardioCare system is a custom digital health software that integrates cardiovascular risk screening with a personalized cardiovascular risk management clinic. Key components of the system include a risk prediction module and personalized health management tools. The CardioCare system incorporates a 10-year CVD risk prediction model based on the Guangzhou Biobank Cohort Study (GBCS) [7]. This validated model uses 6 patient variables (age, sex, systolic blood pressure, use of antihypertensive medication, smoking status, and diabetes status) to estimate the 10-year risk of major cardiovascular events [8]. The model was chosen for its demonstrated predictive performance in a contemporary Chinese population and its practicality (not requiring laboratory cholesterol measurements). During checkups, relevant data are obtained and input into the CardioCare system, either by importing from the electronic health record or by manual entry, after which the system automatically calculates the 10-year CVD risk score (Figure 1). To enhance transparency and replicability, Table 1 summarizes the operational data flow of the CardioCare system, including data inputs (automated vs manual), processing steps, and output delivery mechanisms. Illustrative examples of key system outputs are provided in Multimedia Appendix 1.

The CardioCare system is fully integrated into the hospital’s internal health information system and operates within its centralized big data platform. All checkup data, including laboratory test results, clinical history, and medication use, are automatically available to the system without requiring external interface standards. Physical measurements such as blood pressure and anthropometrics may be manually confirmed or entered by clinic staff as needed. The system comprises a web-based front-end accessible to clinicians and a back-end server hosted within the hospital’s secure internal network. Patient data are encrypted during storage, and access is strictly controlled using a role-based permission model. All user activity is logged and routinely audited to ensure data integrity. During the pilot, the system maintained an uptime of more than 99.5%. Minor issues such as brief delays in data loading during peak hours were promptly resolved through server resource optimization. Within 5 working days after the checkup, based on the calculated risk and the individual’s health profile, the system generates tailored recommendations. For example, a patient with hypertension and moderate risk might receive weekly reminders to check blood pressure and daily prompts to take antihypertensive medication. The content and schedule of these messages are customized to each patient’s risk factors and readiness to change.

The system was designed for seamless integration into the existing clinical workflow at the checkup center. Role-based access control is used to ensure data security and patient privacy, meaning that only authorized clinicians (eg, checkup physicians, nurses, and cardiologists involved in follow-up) can access or input data. The user interface was co-designed with input from frontline health care providers to be intuitive and to fit within the time constraints of a busy checkup clinic.

After internal testing with historical patient data to validate system outputs, the CardioCare system was developed in routine clinical encounters over a 6-month pilot period. Eligible participants were adults aged ≥40 years attending routine health checkup at the study site, with an estimated 10-year CVD risk ≥10% and no prior diagnosis of CVD. A 10-year predicted CVD risk of ≥10% was used to define high risk, based on prior validation of the GBCS model [8] and in alignment with risk thresholds used in international guidelines (eg, the World Health Organization and the European Society of Cardiology) for identifying individuals requiring follow-up intervention [9]. These individuals were selected as they represent individuals at elevated risk who are not typically managed in specialist cardiac care settings.

After the checkup visit, the examining physician used the CardioCare system to calculate the patient’s risk and reviewed the result with the patient. Patients above the 10% risk threshold received a personalized CVD risk report and counseling. At the end of the visit, they were formally invited to enroll in a “cardiovascular risk management clinic”—a follow-up and ongoing care program facilitated by the system. Invitations were delivered via system-generated messages outlining the rationale for follow-up and instructions for enrollment. The implementation was governed by a multidisciplinary project team that included epidemiologists, cardiologists, public health experts, clinical managers, and software engineers. Oversight was coordinated by a core leadership group led by the principal investigators, who held monthly project meetings to review progress, resolve operational challenges, and approve any updates to the system.

Health care providers (physicians and nurses) received training on the CardioCare system before the pilot launch. A total of 9 staff members (5 physicians, 2 nurses, and 2 administrative assistants) participated in structured training sessions led by the research team and software developers. Training included hands-on demonstrations, role-based walk-throughs, and case-based troubleshooting scenarios. Throughout the pilot, research staff monitored system performance, including any software stability and data integration, and provided on-site technical support. Key implementation metrics included the number of high-risk individuals identified and their uptake of the cardiovascular risk management clinic invitation. Usability was assessed via a focus group involving 4 physicians, 2 assistants, and 11 participants. Participants were recruited through poster announcements in the clinical workspace and direct invitation from participants at the checkup center. The session was guided by a semistructured discussion outline developed by the research team, which covered topics such as ease of use, perceived usefulness, communication with patients, and barriers to implementation. The discussion was recorded with participants’ consent, and detailed notes were taken by 2 researchers. Audio recordings were transcribed verbatim and analyzed using an inductive thematic approach. Two members of the research team independently reviewed and coded the transcripts to identify recurrent concepts and patterns. Initial codes were generated directly from the data without a predefined framework. Codes were then grouped into broader themes through iterative discussion among the research team. Discrepancies in coding were resolved through consensus. Given the pragmatic implementation focus of the pilot, analysis was conducted using a rapid thematic approach without the use of qualitative analysis software.

Several adaptations were made during deployment based on feedback and real-time performance monitoring. These included optimizing server performance during peak hours to reduce loading delays, simplifying certain data-entry fields to improve workflow efficiency, and updating patient-facing risk reports to enhance clarity and comprehension.

The primary implementation outcomes included feasibility, adoption, acceptability, and engagement. Feasibility referred to the ability of the system to operate reliably in the clinical environment, integrate into routine workflow, and function without major technical issues. Adoption was defined as clinicians’ use of the system for eligible checkup participants following training and deployment. Acceptability encompassed clinicians’ and patients’ perceptions of usability, clarity, and fit within routine care, as explored through provider focus groups and patient feedback. Engagement referred to whether high-risk individuals acted on the follow-up invitation by attending the cardiovascular risk management clinic, with quantitative engagement outcomes reported in the Results section.

All data were stored on secure servers within the hospital’s internal network. Data transmission occurred over encrypted hospital intranet channels, and role-based access control restricted access to authorized personnel only. All access was logged and audited regularly to ensure data integrity and privacy.

This project was reviewed and approved by the institutional review board of the School of Public Health of Sun Yat-sen University and classified as a quality improvement initiative (institutional review board-2024‐110). All participants provided written informed consent. The study complied with institutional data protection protocols and relevant guidelines for digital health implementation.

A multidisciplinary team of epidemiologists, cardiologists, software engineers, and public health experts successfully developed the CardioCare system over a 6-month period. The final system was deployed in the Health Management Center without major technical difficulties. In April 2023, the risk assessment module in the CardioCare system (including the 10-year cardiovascular risk questionnaire and risk scoring system) was officially launched for routine use in the health checkup center. Subsequently, on August 29, 2023, the hospital opened a dedicated postcheckup cardiovascular risk management clinic to offer follow-up services for individuals identified as high risk by the system. During internal testing on retrospective patient records, the automated risk calculation yielded results identical to manual calculations using the published GBCS risk equation, confirming the accuracy of the implementation. No data security or privacy issues arose during this phase; role-based access controls functioned as intended, and all access was logged without breaches.

The CardioCare system operated smoothly in the clinical setting. Clinicians were able to complete the risk assessment and review the results within the usual duration of the checkup consultation (no significant delays were reported). The auto-generated patient risk reports and recommendation printouts were successfully produced for all high-risk patients. All clinicians adhered to using the system for every eligible patient, indicating good acceptance. There were a few minor technical issues noted, such as occasional slow loading of patient data during peak clinic hours, but no crashes or loss of data occurred. These issues were resolved by adjusting the server configuration.

Feedback from health care providers was largely positive. Clinicians who used the system reported that having an objective risk score improved how they communicated risk to patients, for example, showing a patient that they had “a 20% chance of heart attack or stroke in 10 years” helped prompt discussions about lifestyle changes. However, clinicians noted that the manual process for delivering personalized messages (phone calls and mailing printouts) was labor intensive. Providers suggested automating the patient communication (via mobile app or SMS) as a priority for future development—a feature that was already in our road map (specifically, incorporating an automated chatbot for messaging).

Patient feedback was available only from a small sample (181/2069, 8.7%). Most of these patients reported that the initial risk explanation and the written report were clear and informative. They appreciated the tailored advice. Nevertheless, common themes for why many peers might not follow through included the following: lack of perceived urgency (“If I feel okay, I didn’t think I needed the special heart program”), family or time constraints, and fear or denial about illness—consistent with known barriers to care seeking in this population. A few also mentioned that a phone-based or online format for follow-up might be preferable to in-person clinic visits, especially for those living farther from the hospital.

Between January 2024 and April 2025, the CardioCare system was used to assess cardiovascular risk among 9148 individuals who submitted data for evaluation during routine health checkups. Of these, 300 were excluded due to a prior diagnosis of CVD, and 8848 individuals underwent CVD risk assessment using the system. Risk stratification results showed that 5179 (58.5%) individuals were classified as low risk, 1600 (18.1%) as moderate risk, and 2069 (23.4%) as high risk (defined as ≥10% 10-year CVD risk). All individuals classified as high risk were invited to attend the cardiovascular risk management clinic for follow-up care. Across the reporting period, a total of 181 high-risk individuals attended the clinic at least once, yielding an overall follow-up rate of 8.7%. Most of these patients (n=152, 83.9%) had only 1 follow-up visit, while a smaller proportion returned for multiple visits. Table 2 summarizes the main implementation outcomes and lessons learned from the pilot deployment of the CardioCare system.