Development and application of artificial intelligence follow-up system in patients with coronary heart disease

Abstract

Background

Coronary heart disease (CHD) is a leading cause of mortality globally and poses significant public health challenges, particularly in China. Despite advancements in medical treatments, postdischarge management of patients with CHD remains inadequate, often resulting in poor medication adherence, low self-management, and increased readmission risks. Traditional manual follow-up methods are time-intensive, inefficient, and lack personalization, limiting their ability to address patients’ complex healthcare needs.

Objective

This study aimed to develop and evaluate an AI-assisted follow-up platform incorporating patient profiling and interactive voice response (IVR) technologies. The platform seeks to improve follow-up efficiency, enhance patient satisfaction, reduce the workload of medical staff, and provide a scalable model for intelligent follow up in clinical settings.

Methods

This study utilized a prospective cohort design. A total of 4000 patients with CHD discharged from Tongji Hospital were enrolled, with 2000 patients assigned to a manual follow-up group from September 2022 to August 2023 and 2000 from September 2023 to August 2024 to an AI-assisted follow-up group. Key performance indicators, including follow-up rates, feedback collection rates, patient satisfaction, and medical staff workload, were compared between the two groups.

Results

The follow-up rates were similar between the manual group (82.0%) and the AI-assisted group (81.3%) (P = 0.567).The feedback collection rates did not significantly differ between the two groups (7.2% vs. 6.1%, P = 0.163). However, the AI-assisted follow up significantly reduced time expenditure, saving an average of 13.2 hours per 100 patients compared to manual follow up. Patient satisfaction was significantly higher in the AI-assisted group than in the manual group (94.5% vs. 74.0%, P < 0.001). Binary logistic regression analysis identified follow-up method as the significant predictor of patient satisfaction (odds ratio (OR) = 5.993, 95% confidence interval (CI): 4.829–7.438, P < 0.001). In addition, the AI-assisted platform significantly reduced the workload of medical staff (P < 0.01).

Conclusions

The AI-assisted follow-up platform demonstrated comparable effectiveness to manual methods while significantly improving efficiency and satisfaction for both patients and healthcare providers. By integrating patient profiling and IVR technologies, the platform offers personalized, dynamic, and cost-effective follow-up services. These findings highlight the potential of AI-enabled systems in optimizing chronic disease management and resource utilization.

Keywords

Coronary heart disease AI-assisted follow up cost-effectiveness patient profiling interactive voice response

Introduction

Coronary heart disease (CHD) remains a leading cause of mortality worldwide and poses significant public health challenges, particularly in China.¹ With the aging population and lifestyle changes, the prevalence and burden of CHD continue to rise, placing immense strain on healthcare systems.^2,3 Challenges such as poor medication adherence, limited self-management capacity, and psychological distress are common among patients with CHD, often resulting in increased readmission rates and reduced quality of life.^4,5 The 2023 guidelines for the chronic management of coronary artery disease emphasize the importance of ongoing follow-up and periodic reevaluation to enhance therapeutic adherence and reduce the risk of adverse cardiovascular events.⁶ Traditional follow-up methods, primarily relying on manual phone calls and text messages, are labor-intensive, lack efficiency, and fail to deliver personalized care tailored to the diverse needs of patients.^7,8 These limitations underscore the urgent need for innovative, technology-driven follow-up solutions.^9,10

The rapid advancements in artificial intelligence (AI) and digital health technologies have paved the way for intelligent follow-up platforms that leverage tools like patient profiling and interactive voice response (IVR).^11,12 These technologies enable personalized, precise, and dynamic follow-up care, optimizing patient outcomes while reducing healthcare provider workload.¹³ By integrating AI with clinical practices, intelligent follow-up systems have the potential to address the inefficiencies of traditional models and transform the landscape of chronic disease management. Patient profiling technology is a crucial component, capable of integrating multidimensional information such as patients’ health status, medical records, lifestyle habits, and psychological conditions to construct personalized, tagged user models, enabling precise intervention.^14,15 Domestically, there have been research efforts to utilize patient profiling technology. For instance, a comprehensive intelligent service platform developed by Su Mingliang¹⁶ and colleagues generates an “individual health profile” through multilayered health monitoring devices. It has facilitated the early detection of diseases, real-time monitoring, and precise management. However, current research and application of patient profiling have primarily focused on relatively healthy populations within communities. There is a scarcity of practical applications and validations that address the complex needs of patients within hospitals. This gap limits the ability to provide truly personalized management support for patients with CHD.

Concurrently, the application of IVR technology in follow-up management has demonstrated significant potential. By integrating Natural Language Processing (NLP) and Speech Recognition technologies, IVR enables the automatic identification of patient needs and collection of follow-up feedback. This provides healthcare teams with precise data support, substantially enhancing the efficiency of the follow-up process.^17,18 In Europe, this technology has been widely implemented in cardiac rehabilitation follow ups, significantly reducing the duration of follow-up visits and enhancing patient engagement and satisfaction.¹⁹ Domestically, several medical institutions have begun to explore the use of AI technology for follow-up purposes, such as employing intelligent voice platforms to intervene in lifestyle modifications and monitor for abnormal vital signs.^20,21 However, the current state of voice interaction technology in our country falls short in terms of precision when dealing with complex patient feedback. It struggles to promptly comprehend and address the issues reported by patients, which constrains the further enhancement of follow-up care quality.^22–24

This study aims to develop and implement an AI-assisted follow-up platform for patients with CHD, utilizing patient profiling and IVR technologies to enhance follow-up efficiency, improve patient satisfaction, and reduce the burden on medical staff. Through a comparative analysis of traditional manual follow-up methods and AI-assisted approaches, this research seeks to validate the feasibility and effectiveness of intelligent follow-up systems in optimizing healthcare delivery and resource utilization for CHD management.

Methods

Development of the coronary heart disease intelligent follow-up platform

The CHD intelligent follow-up platform was developed in collaboration with a third-party technology provider and implemented at Tongji Hospital. The platform integrates patient profiling and IVR technologies to provide comprehensive follow-up services. Core functionalities include patient management, intelligent monitoring of vital signs, automated follow ups, satisfaction surveys, and dynamic rehabilitation plan adjustments. The platform employs anthropomorphic synthesis to precisely replicate the professional linguistic style of nurse-led follow ups while integrating a specialized CHD knowledge base. It's core technical advantages include accurate identification of multi-intent expressions and core demands within patient dialogues, dynamic generation of personalized follow-up content. The platform interfaces with the hospital's health information system (HIS) to enable seamless data integration and personalized patient care.

The patient profiling

Structured patient profiles were automatically generated through integration with the HIS. The system extracted key data from admission records, clinical notes, laboratory and imaging results, medication orders, and discharge summaries. Profiling variables included demographic information (e.g. age and gender), clinical diagnoses (e.g. acute coronary syndrome (ACS), chronic coronary syndrome (CCS)), treatment history, laboratory parameters (e.g. low-density lipoprotein cholesterol (LDL-C) and troponin), comorbidities (e.g. hypertension and diabetes), and self-reported lifestyle factors (e.g. smoking status and physical activity).

The constructed profiles were used to inform personalized follow-up pathways. Standardized follow-up modules—such as medication guidance, dietary advice, and rehabilitation support—were conditionally triggered based on profile attributes using rule-based logic. This ensured alignment with clinical guidelines while allowing tailored content delivery according to patient-specific risks and needs. Patient profiles were dynamically updated through continuous integration of real-time data from IVR systems, patient-reported outcomes, and connected monitoring devices. When abnormal physiological signals or elevated risk indicators were detected, the platform automatically adjusted the follow-up frequency or issued alerts to healthcare providers and patients, supporting a closed-loop precision management system.

Functions of the CHD intelligent follow-up platform

The CHD intelligent follow-up platform included six modules: patient management, doctor management, intelligent sign monitoring, IVR, satisfaction survey and optimization, and expansion and normalization application. Since September 2023, the CHD intelligent follow-up platform has expanded the modules and functions related to the follow up of patients with CHD (Table 1).

Table 1.

Modules and functions of the coronary heart disease intelligent follow-up platform.

Functional module	Specific functions	Function description
Patient Management Module	Patient Information Collection and Profiling Generation	The platform utilizes artificial intelligence technology to extract patient medical records, testing reports, and health archives, automatically generating patient profiling to provide foundational data for subsequent management
	Special Disease Management Path Construction	Rapidly constructs disease-specific management pathways for various patients with chronic disease based on domestic and international clinical guidelines, expert consensus, and high-impact factor papers
	Individual Management Path Generation and Execution	Automatically generates personalized management plans based on patient profiling and disease-specific management pathways, and executes them through the platform, such as medication reminders, exercise guidance, and dietary advice
	Intelligent Follow up and Abnormality Alerts	Conducts intelligent follow ups with patients through methods like AI Outbound Robots and WeChat mini-programs, collecting follow-up information and forming to-do reminders for doctors to pay attention to abnormal situations
Doctor Management Module	Dynamic Adjustment of Rehabilitation Plans	Physicians can dynamically adjust rehabilitation plans according to changes in patient conditions, ensuring individualized, and precise management pathways
	Risk Management and Identification	The platform automatically reminds patients of follow-up visits and provides appointment services to improve patient follow-up rates and medical efficiency
	Follow-up Reminders and Appointment Scheduling	The platform automatically reminds patients of follow-up visits and provides appointment services to improve patient follow-up rates and medical efficiency
Intelligent Vital Signs Monitoring Module	Vital Signs Data Automatic Upload	Patients can measure vital signs data regularly through intelligent vital signs devices and automatically upload it to the platform
	Regular Push of Assessment Reports	The platform regularly pushes vital signs assessment reports, including risk situations and health advice, to help patients better understand their health conditions
	Family Collaborative Management	Supports the involvement of patients’ families in management, jointly focusing on patient health status and providing safety assurance
Interactive Voice Response (IVR) Module	Round-the-Clock Response	The platform provides 24/7 Round-the-Clock response services, answering patients’ home care questions and improving patient satisfaction
	Tiered Response System	The platform ensures timely and professional responses to patient inquiries through a structured system involving health managers, physician assistants, and hospital doctors, providing a multilevel approach to patient care
Satisfaction Survey and Optimization Module	Satisfaction Survey	The platform regularly conducts satisfaction surveys to collect feedback from patients and doctors for platform function optimization and service improvement
Satisfaction Survey and Optimization Module	Report Generation and Data Analysis	Generates satisfaction reports based on survey data and conducts data analysis to provide decision support for platform optimization
Expansion and Normalization Application Module	Expansion to Internal Medicine Departments	The platform is not only suitable for a specific department but can also be expanded to other internal medicine departments for comprehensive online management
	Functional Normalization Application	The platform‘s functions have been stably applied in patient management, improving management efficiency and patient compliance through channels such as the Tongji Hospital public account

Patient enrollment and data collection

This study utilized a prospective cohort design. Inclusion criteria were: (1) diagnosed with CHD²⁵; (2) age over 18 years; and (3) willing to participate and provide informed consent. Exclusion criteria included: (1) cognitive impairment; (2) inability to complete follow up (e.g. no telephone access and severe hearing impairment); and (3) participation in other interventional studies.

The sample size was calculated based on an expected difference in patient satisfaction rates between the two groups. With a power of 80% and α=0.05, and assuming a satisfaction rate of 70% in the manual group and 80% in the AI-assisted group, a minimum of 279 patients per group was required. However, to maximize representativeness and make full use of available data, we included all eligible inpatients admitted to the cardiology department during the study periods. As a result, a total of 4000 patients with CHD were enrolled. Patients discharged between September 2022 and August 2023 were assigned to the manual follow-up group (n = 2000), while those discharged between September 2023 and August 2024 formed the AI-assisted follow-up group (n = 2000). Patient demographics, clinical data, and follow-up records were extracted from the HIS.

Ethical approval

The study was approved by the Ethics Committee of Tongji Hospital (approval number: TJ-IRB202502002). All participants were informed of the purpose, procedures, and confidentiality of the study, and provided written informed consent prior to participation.

Establishment of a multidisciplinary treatment team

A multidisciplinary treatment (MDT) team with nine members was organized, including one chief nurse of Internal Medicine, two head nurses of the cardiology department, two cardiologists, three cardiovascular nurses, and one follow-up information technology specialist. The chief nurse of Internal Medicine is primarily responsible for the construction and coordination of the design for the intelligent follow-up platform for CHD. The head nurses are tasked with establishing the management system related to follow ups and implementing project training. Cardiologists are available online from 14:00 to 16:00 daily to address and resolve medium to high-risk or abnormal pending issues arising from the intelligent follow-up process. Cardiovascular nurses are in charge of developing educational materials for disease awareness, conducting baseline follow ups, and setting follow-up protocols. The follow-up information technology specialist provides technical support for the information needs of patients with CHD who have undergone treatment at our hospital, as well as for the medical staff involved in the follow-up process. At the same time, in order to improve the scientificalness of the platform in clinical practice, we have also built a dedicated AI-assisted follow-up platform team based on internal factors and individual domain implementation strategies in the implementation of science. We have established a clear team member responsibility management system, clarified the main tasks and responsibilities of senior, middle, implementation promoters and supervisors, and implementation team members, and established corresponding incentive systems for innovative implementers, and conducted regular training for team members.

AI follow up

The interactive voice response telephone calls

Utilizing IVR, our platform enabled automated home-based follow ups. It leveraged speech recognition and natural language processing to understand and respond to patient needs, facilitating meaningful dialogues. This technology streamlined feedback collection, offering healthcare staff timely data to enhance care quality and treatment plans. Examples of IVR feedback messages were provided in Table 2.

Table 2.

Examples of automated interactive voice response feedback messages.

Artificial intelligence (AI) Outbound Robot: Hello, may I speak with #Patient's Name#? This is your coronary heart disease follow-up assistant. To better understand your health status, I will conduct a few inquiries.

Patient: Yes, go ahead.

AI Outbound Robot: You were previously hospitalized in our cardiology department. We would like to know how you've been feeling since your discharge. Have you experienced any discomfort such as palpitations, chest pain, or tightness in the chest?

Patient: I feel fine when I'm sitting and resting, but I do get a bit chesty and out of breath if I walk or exercise for too long.

AI Outbound Robot: Thank you for your response. If your symptoms become severe, it's best to revisit the cardiology clinic. Have you been taking your medication as prescribed since your discharge?

Patient: Yes, I take my medication every day as instructed by the doctor.

AI Outbound Robot: That's good to hear. It's important to continue as prescribed; stopping or adjusting your medication on your own can lead to disease recurrence. Have you experienced any side effects such as bruising, bleeding gums, bloody stools, or bleeding from other areas?

Patient: No, I haven't had any bleeding issues.

AI Outbound Robot: Excellent. Regarding your sleep, have you been getting enough rest? Have you had any trouble sleeping or poor sleep quality?

Patient: My sleep hasn't been great.

AI Outbound Robot: Sleep is crucial for recovery. Have you tried any methods to improve your sleep, such as relaxation before bed or avoiding caffeine?

Patient: I've tried relaxing before bed, but it hasn't helped much.

AI Outbound Robot: I suggest you try meditation before sleep or listening to soft music. Now, let's discuss your lifestyle habits. Do you smoke or drink alcohol?

Patient: I don't smoke, but I do drink occasionally.

AI Outbound Robot: It's noted that smoking and drinking can increase the risk of coronary heart disease. It's advisable to avoid alcohol whenever possible. In terms of diet, have you been mindful of low-salt and low-fat, light meals?

Patient: I do go out for hot pot with family and friends every few days.

AI Outbound Robot: It's recommended that you maintain a light diet, control your intake of cholesterol, fat, and sugar, and limit your daily salt intake to less than 6 g, about the amount that fits on a beer cap. Increase your protein intake with lean meats, eggs, and dairy, and consume plenty of fresh fruits and vegetables. Now, about your exercise routine, do you engage in any physical activity?

Patient: I go for a walk a few times a week.

AI Outbound Robot: Walking is an excellent form of exercise. Please keep it up. You can engage in moderate exercise 3 months postsurgery, but it's important to progress gradually and not overdo it. If you experience chest pain or shortness of breath, stop immediately. Should your chest tightness persist or other symptoms arise, seek medical attention promptly. Maintaining good sleep and dietary habits is also very important for your health.

Patient: Thank you, I will pay attention.

AI Outbound Robot: Lastly, do you have any comments or suggestions regarding our hospital's services?

Patient: No, I'm satisfied, thank you.

AI Outbound Robot: Thank you for your feedback. I've noted it down. We also recommend that you keep an electronic upper-arm blood pressure monitor at home to regularly check your blood pressure and heart rate, ensuring they stay within the target range. Please continue with regular check-ups. Thank you for taking the call, and we wish you good health. Goodbye.

(After concluding the conversation, the AI outbound robot will automatically present relevant “Coronary Heart Disease Dietary Recommendations” and “Quit Smoking and Drinking” public WeChat articles based on the content of this dialogue.)

Manual follow up

Follow up in the manual group involved direct phone calls. Information on patient conditions, medication adherence, and lifestyle was collected and manually recorded in the HIS. Feedback from patients were recorded by the operator and reported to doctors and nurses.

Assessment indicators

The assessment indicators were follow-up rate, Feedback collection rate, duration of calls and human resource input, Feedback composition, patient satisfaction, Medical Staff workload Key metrics were calculated as follows:

Follow-up rate = number of effective follow ups/(number of effective follow ups + number of invalid follow ups) × 100%; feedback collection rate = number of patients with effective feedback/number of effective follow ups ×100%;

read rate = number of read times/number of send times × 100%; patient satisfaction rate = number of patient satisfied with the platform/total number of patient surveyed ×100%.

Patients satisfaction assessment

Patient satisfaction was assessed using a structured questionnaire developed specifically for follow-up interaction assessment. The questionnaire design was informed by elements from validated satisfaction tools, and adapted to evaluate five dimensions: usefulness of the follow-up content, clarity of information, convenience of the follow-up method, timeliness of platform response, and overall satisfaction. Each item was rated on a five-point Likert scale (1 = very dissatisfied to 5 = very satisfied), yielding a total score ranging from 5 to 25. The instrument was reviewed by a panel of clinical and informatics experts and demonstrated good internal consistency (Cronbach's α = 0.88). According to clinical practice standards, a total score ≥20 was categorized as “satisfied,” while a score <20 was classified as “not satisfied.”

The NASA-Task Load Index questionnaire

Evaluation of nursing workload to assess nurses’ perceived workload associated with the follow up, this study used the Chinese version of the NASA Task Load Index (NASA-TLX),²⁶ which has been culturally adapted and psychometrically validated in nursing settings in China.²⁷ As this instrument is not disease-specific but measures multidimensional workload (mental, physical, temporal demands, effort, performance, and frustration), its validation in healthcare contexts is well-established. The scale includes six dimensions: mental demand, physical demand, temporal demand, performance, effort, and frustration. Each item is scored on a 0 to 100 scale, with higher scores indicating greater task load. In line with validated recommendations, we adopted a simplified scoring approach by calculating the arithmetic mean of the six item scores to represent the overall workload score. The Chinese version demonstrated good reliability (Cronbach's α = 0.707) and test–retest reliability (r = 0.806), and the content validity index (S-CVI/Ave) reached 1.0. This strong psychometric profile ensures confidence in measuring subjective workload and supports comparison of workload before and after the implementation of the AI-assisted follow-up system.

Statistical methods

Data analysis was performed using SPSS statistical software (version 26.0, IBM Inc). Age, gender, disease category, and read rate of health education materials were analyzed descriptively. Patients’ characteristics, follow-up rate, feedback collection rate, healthcare staff satisfaction rate in both groups were measured using Pearson's chi-square test. Binary logistic regression was used to analyze factors influencing patient satisfaction, paired t-tests were used to compare medical staff workload scores.

Results

Patient characteristics

Patient characteristics for the two groups are shown in Table 3. Group differences in age, gender and disease category were not statistically significant.

Table 3.

Comparison of basic information in manual and artificial intelligence-assisted follow-up groups.

Characteristics	Manual follow-up group	AI-assisted^a follow-up group	Chi-square	P value
Number of patients	2000	2000
Age(years)			0.414	0.937
<30	45	48
30–50	285	279
51–65	845	861
≥65	825	812
Gender			0.017	0.896
Male	1252	1248
Female	748	752
Disease category			1.062	0.303
ACS	178	197
CCS	1822	1803

ACS: acute coronary syndrome; AI: artificial intelligence; CCS: chronic coronary syndrome.

Cost-effectiveness

Table 4 summarizes the major outcomes for all participants. As shown in Table 4 there was no significant difference in either follow-up rate (manual: 1640/2000, 82%; AI-assisted: 1626/2000, 81.3%; P = 0.567). The feedback collection rate in the AI-assisted follow-up group was not significantly different between the two groups that in the manual follow-up group (manual: 144/2000, 7.20%; AI-assisted: 122/2000, 6.10%; P = 0.163).

Table 4.

Comparison of AI-assisted and manual follow-up group indicators.

Indicators	Manual follow-up group, n (%)	AI-assisted follow-up group, n (%)	Chi-square	P value	RD (95% CI)
Follow up			0.327	0.567	−0.7% (−3.1%,1.7%)
Number of effective follow ups	1640	1626
Number of invalid follow ups	360	374
Follow-up rate, %	82	81.3
Feedback collection			1.949	0.163	−1.1% (−0.45%,2.65%)
Number of patients with effective feedback	144	122
Number of effective follow ups + invalid follow ups Rate, %	1856	1878
Number of effective follow ups + invalid follow ups Rate, %	7.20	6.10
Time spent, hours per 100 patients	14	0.8
Minutes, the durations of calls	9817	7713

Effective follow ups: Follow ups with complete data collection (excluding parameters related to health education materials); invalid follow ups: Follow ups with missing or incomplete data; call duration: The time medical staff spent communicating with patients via telephone calls. AI: artificial intelligence; CI: confidence interval.

The total time needed to complete follow up was 16 hours compared with 280 hours in the manual follow-up group. The results showed that AI-assisted follow up significantly reduced follow-up duration. The average difference in follow-up duration was decreased by 264 hours. Time spent in the manual follow-up group included the time spent by nurses over the telephone to inquire about the patient follow-up form information. The time spent on the AI-assisted follow-up group by medical staff mainly consisted of the time needed to give telephonic feedback on any abnormal form submitted by the patient. Compared with manual follow up, the AI-assisted follow up were of shorter duration.

Feedback composition

The composition of feedback content is shown in Figure 1. The read rate of feedback content is shown in Figure 2. It's content is automatically generated by AI. In the AI-assisted follow-up group, 8.58% of the feedback was medical reminders and follow-up module, including follow-up appointments reminder, follow up of medical test results (2.32% and 6.26%, respectively). Feedback related to home care and rehabilitation management module, including rehabilitation objectives, rehabilitation expectation, essential rehabilitation measures, and home care services (0.80%, 2.16%, 7.13%, and 3.26%, respectively). Feedback related to healthy lifestyle advice module, including medication management guidance, dietary advice and physical activity recommendations (2.38%, 13.59%, and 40.52%, respectively). Feedback related to patient education and self-management module is about collecting suggestions (21.59%).

Figure 1.

The composition of feedback content.

Figure 2.

The read rate of feedback content.

Patients satisfaction

According to Table 5, the patients satisfaction degree of AI-assisted follow up was significantly higher than that of manual follow-up group (AI-assisted: 1890/2000, 94.5%; manual: 1480/2000, 74.0%; P < 0.001).

Table 5.

Comparison of patient satisfaction between groups.

Indicators	Manual follow-up group, n（%）	AI-assisted^a follow-up group, n (%)	Chi-square	P value
Satisfaction			316.70	<0.001
Number of patients satisfied with the platform	1480	1890
Number of total actual patients surveyed	2000	2000
Satisfaction, %	74	94.5

AI: artificial intelligence.

A binary logistic regression model was used to analyze the factors influencing patient satisfaction, with satisfaction (satisfied = 1 and not satisfied = 0) as the binary dependent variable. Variables including age (original input), gender (male = 1 and female = 0), disease category (ACS = 1 and CCS = 0), and follow-up method (AI-assisted = 1 and manual =0) were entered into the model, as shown in Table 6. The results showed that the follow-up method was the significant predictor of patient satisfaction (odds ratio (OR) = 5.993, 95% confidence interval (CI): 4.829–7.438, P < 0.001).

Table 6.

Binary logistic regression analysis of influencing factors for patients satisfaction (n = 4000).

Variable	B	P	OR	95%CI
Age	0.007	0.991	1.007	0.991,1.023
Gender	0.183	0.212	1.2	0.901,1.60
Disease category	−0.209	0.365	0.811	0.515,1.276
Follow-up method	1.791	＜0.001	5.993	4.829,7.438

CI: confidence interval; OR: odds ratio.

NASA-Task Load Index

As demonstrated in Table 7, the workload of medical staff, as measured by the NASA-Task Load Index (NASA-TLX), was significantly reduced after implementation of the AI-assisted follow-up platform. Specifically, scores for all six workload dimensions—including mental demand, physical demand, temporal demand, performance, effort, and frustration—decreased substantially after adoption of the platform. All improvements were highly statistically significant (P < 0.01), with the paired t-test demonstrating that each workload domain benefitted from the intervention. These results indicate a broad reduction in subjective nursing workload during follow-up work facilitated by the AI-assisted system.

Table 7.

A comparison of NASA-TLX scores between manual and AI-assisted groups (n = 16).

Dimension	Manual (x̄ ± SD)	AI-assisted (x̄ ± SD)	t	P
Mental demand	80.94 ± 8.60	45.31 ± 14.66	9.83	<0.001
Physical demand	76.88 ± 15.60	38.56 ± 14.47	7.64	<0.001
Temporal demand	81.56 ± 12.345	40.25 ± 14.47	10.05	<0.001
Performance	45.69 ± 11.98	75.44 ± 7.59	−9.45	<0.001
Effort	84.69 ± 8.85	38.81 ± 11.65	14.10	<0.001
Frustration	78.44 ± 11.79	35.63 ± 10.84	11.83	<0.001
Total scores	74.67 ± 9.25	45.67 ± 10.48	9.84	<0.001

AI: artificial intelligence; NASA-TLX: NASA Task Load Index; SD: standard deviation.

Discussion

Research on healthcare professionals’ adoption of digital health technologies suggests that the development and implementation of intelligent follow-up platforms face several challenges. These include poor team collaboration, unclear role definitions, inadequate supervision, limited attention to patient-centered needs, varying levels of operational proficiency, and a lack of innovation.²⁸ To address these challenges, the Consolidated Framework for Implementation Research (CFIR) provides a structured approach to identifying key factors that influence the implementation process. By tailoring strategies to specific contextual factors, CFIR facilitates the translation of evidence-based innovations into clinical practice.^29–31 In this study, we adopted CFIR and integrated it with theories and guidelines related to organizational settings and individual characteristics. This approach enabled us to establish a dedicated team to optimize the usability and effectiveness of the intelligent follow-up platform for patients with CHD.

Our efforts focused on several key areas: clarifying the organizational structure, defining individual roles and responsibilities, improving access to knowledge and resources, and introducing an innovative incentive system. Senior leadership played a critical role in defining and communicating team member responsibilities, while middle management provided ongoing supervision and support. Implementation facilitators and oversight committees ensured that the platform aligned with patient needs and preferences, enhancing its functionality and user satisfaction. Concurrently, execution team members conducted regular training sessions to improve operational proficiency and familiarity with the platform, addressing variability in team competence. To foster a culture of innovation, we implemented incentive mechanisms to motivate team members and encourage proactive engagement. These measures created a sustainable momentum for continuous improvement. By optimizing the organizational structure and processes from top to bottom, the team enhanced the platform's implementation efficiency and overall effectiveness, addressing both technical and user-centered challenges.

In this study, no statistically significant differences were found between the AI-assisted and manual follow-up groups in terms of follow-up rate (81.3% vs. 82.0%, P = 0.567) or feedback collection rate (6.1% vs. 7.2%, P = 0.163). Although these differences were not statistically significant, the comparable performance of the AI-assisted system indicates that it can achieve follow-up and feedback outcomes similar to those of traditional manual methods. These findings support the potential of AI-based follow up as a feasible and efficient alternative in routine clinical practice, especially in contexts demanding scalability and workforce optimization. The implementation of patient profiling technology within the intelligent follow-up platform demonstrated significant advantages, particularly in the composition and quality of patient feedback. The profiling logic relied on predefined rules, it systematically incorporated critical clinical variables to tailor content. By integrating multidimensional health data, such as medical records, lifestyle habits, and psychological conditions, patient profiling enabled the generation of personalized health management pathways.^32–34 The studies indicate that patient profiling technology, by analyzing relevant medication details and adverse reaction data within patient medical records, can construct personalized characteristic models. This enhances the early detection and prevention of adverse drug reactions, reduces healthcare costs, and improves patient prognosis.^35,36 Our findings indicated that this technology played a pivotal role in tailoring feedback and recommendations to individual patient needs. Analysis of the feedback composition highlighted the efficacy of this approach. Exercise guidance accounted for the largest proportion of feedback (40.52%) and achieved the highest read rate (94.3%), suggesting that patients were highly receptive to actionable and personalized recommendations. This finding aligns with existing research emphasizing the importance of aligning follow-up content with patient priorities, as well-structured exercise guidance directly addresses the rehabilitation needs of patients with CHD.

Similarly, dietary advice and medication management guidance also featured prominently, underscoring the effectiveness of patient profiling in addressing critical areas for secondary prevention. However, feedback related to rehabilitation goals (0.8%) exhibited lower proportions and read rates compared to other categories. This discrepancy may indicate gaps in communication or insufficient emphasis on long-term health objectives within the follow-up content. Future improvements to the platform should prioritize enhancing the clarity and relevance of feedback in this domain to maximize patient engagement and outcomes.^37,38 Patient profiling technology also streamlined the feedback collection process by reducing redundancy and allowing healthcare providers to focus on clinically significant issues.^39,40 The platform's capability to dynamically adjust health recommendations based on patient responses further optimized the utility of the collected data, demonstrating the potential of intelligent systems to improve the quality and precision of follow-up care.

The use of IVR technology, which integrates speech recognition, natural language processing, and speech synthesis, has proven effective in automating follow ups by identifying and understanding patient needs. The platform employs anthropomorphic synthesis to precisely replicate the professional linguistic style of nurse-led follow ups while integrating a specialized CHD knowledge base. It's core technical advantages include accurate identification of multi-intent expressions and core demands within patient dialogues, dynamic generation of personalized follow-up content, and intelligent conversation logic with adaptive question routing capabilities to eliminate redundancies and effectively integrate follow-up workflows.^41–43 This technology not only streamlines the process of collecting and uploading feedback but also provides valuable data support for healthcare teams.^44,45 Our study demonstrated that the AI-assisted follow-up platform, which incorporates IVR technology, significantly reduced the duration of human resource investment compared to manual follow ups. Specifically, the platform saved an average of 13.2 hours per 100 patients, showcasing its efficiency in managing follow-up tasks on a larger scale. The observed reduction in human resource requirements is attributed to the platform's ability to simultaneously follow up with multiple patients and automatically generate and store follow-up data. In contrast, manual follow ups require nurses to contact each patient individually, complete follow-up forms, and manually record data. This difference highlights the capability of IVR technology to optimize resource allocation and reduce the administrative burden on healthcare providers. In a healthcare environment like China, where nursing shortages are a significant concern, such time savings are crucial for maintaining service quality while addressing resource constraints.⁴⁶

Furthermore, patient satisfaction serves as a key indicator of healthcare quality and management efficiency. Table 5 showed that patient satisfaction in the AI-assisted follow-up group was significantly higher than in the manual follow-up group (94.5% vs. 74.0%, P < 0.001). A binary logistic regression analysis further confirmed that the follow-up method was the only significant predictor of patient satisfaction (OR = 5.993, 95% CI: 4.829–7.438, P < 0.001), whereas patient characteristics such as age, gender, and disease category showed no significant association. This suggests that the follow-up modality, rather than demographic or clinical factors, plays a critical role in shaping patient experience. This aligns with IVR studies in chronic disease management where consistency and 24/7 accessibility drive engagement, particularly for medication adherence and lifestyle advice delivery. This aligns with previous research indicating that IVR systems enhance patient satisfaction by providing timely, consistent, and personalized interactions. For example, patients responded positively to tailored health education materials delivered through the platform, which addressed specific needs such as medication adherence and lifestyle modifications. Additionally, the interoperability of IVR with other platform modules, such as patient profiling and automated alerts, further amplified its effectiveness.^47,48

In this study, the use of an AI-assisted follow-up platform led to significant reductions in healthcare providers’ perceived workload, as reflected by the multidimensional NASA-TLX evaluation. All assessed domains—including mental, physical, and temporal demands, as well as effort, performance, and frustration—showed marked improvement following implementation of the digital system. As can be seen from Table 7, the overall NASA-TLX workload score decreased by approximately 29 scores. All paired comparisons reached statistical significance suggesting that the platform not only reduces objective labor intensity but also alleviates subjective psychological strain, potentially enhancing job satisfaction and reducing the risk of burnout. These findings are consistent with growing evidence that digital tools can streamline chronic disease management tasks, allowing healthcare professionals to allocate effort more efficiently while maintaining care quality.^49–51 Specifically, the system's capability to generate and process follow-up data in batches eliminated repetitive manual tasks, enabling nurses to focus on higher-priority clinical responsibilities. By reducing routine administrative labor, IVR technology not only improved work efficiency but also alleviated the workload of healthcare professionals involved in follow-up management.

Conclusion

This study demonstrates that the follow-up outcomes of the AI-assisted follow-up group are comparable to those of the manual follow-up group, with significant advantages in efficiency, patient satisfaction, and reduced workload for healthcare staff. By applying a MDT team and implementation science strategies, the optimized intelligent CHD follow-up platform achieved personalized, dynamic, and precise follow-up services, improving cost-effectiveness and patient adherence while alleviating staff workload. However, the platform still faces challenges in practice. For example, patients with lower health literacy or limited digital access showed lower engagement. Future follow-up protocols should be tailored to meet diverse patient needs, such as focusing on weight management, improving medication adherence, or providing an option for direct human support for patients with specific needs. Additionally, the proportion of feedback related to long-term rehabilitation goals remains low, indicating a need to enhance the clarity and relevance of such feedback. Future research should explore whether AI-assisted follow up contributes to improved clinical outcomes, such as earlier detection of patient abnormalities or reductions in the incidence of adverse cardiovascular events, to further validate its long-term effectiveness and safety. In summary, the intelligent follow-up platform has shown great potential in improving efficiency and optimizing chronic disease management. Addressing its current limitations is essential for broader adoption.

Supplemental Material

sj-docx-1-dhj-10.1177_20552076251360858 - Supplemental material for Development and application of artificial intelligence follow-up system in patients with coronary heart disease

Supplemental material, sj-docx-1-dhj-10.1177_20552076251360858 for Development and application of artificial intelligence follow-up system in patients with coronary heart disease by Liping Zhang, Yanan Chen, Qiao Yang, Lijuan Lu, yishuang Cui, Miao Zhou, Xiaolu Jin, Shiyi Zhang and Xifei He in DIGITAL HEALTH

Footnotes

Acknowledgements

We would like to thank all of the healthcare providers and individuals who participated in this study from Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology.

ORCID iD

Shiyi Zhang

Ethical approval

The study was approved by the Ethics Committee of Tongji Hospital (approval number: TJ-IRB202502002).

Contributorship

SZ and XH were involved in study conception and design; QY, YC, and LZ in data collection; SZ, LZ, and YC in analysis and interpretation of results; QY, LL, YC, MZ, and XJ in first draft manuscript preparation; and YC and LZ contributed to writing the original draft.. All authors reviewed the results and approved the final version of the manuscript.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by Tongji Hospital (grant number: 2023C03) and Wuhan Nursing Association (WHHL202405).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

Due to the nature of the questions asked, the data is not publicly available.

Gurantor

Xifei He.

Informed consent

All participants were informed of the purpose, procedures, and confidentiality of the study, and provided written informed consent prior to participation.

Peer review

Peer review statements are managed by the journal's standard process.

Supplemental material

Supplemental material for this article is available online.

References

Stone

Libby

Boden

. Fundamental pathobiology of coronary atherosclerosis and clinical implications for chronic ischemic heart disease management—the plaque hypothesis: a narrative review. JAMA Cardiol 2023; 8: 192–201.

Qiu

Huang

, et al. Construction and application of an out-of-hospital health management program for patients with chronic heart failure. Chin J Nurs 2022; 57: 401–407.

Merghani

Maestrini

Rosmini

, et al. Prevalence of subclinical coronary artery disease in masters endurance athletes with a low atherosclerotic risk profile. Circulation 2017; 136: 126–137.

Meissner

Akhtar

Lehmann

. Nonischemic sudden tachyarrhythmic death in atherosclerotic heart disease. Circulation 1991; 84: 905–912.

Smith

JR. Benjamin

Bonow

, et al. AHA/ACCF secondary prevention and risk reduction therapy for patients with coronary and other atherosclerotic vascular disease: 2011 update: a guideline from the American Heart Association and American College of Cardiology foundation. Circulation 2011; 124: 2458–2473.

Wang

Liang

. Guidelines for chronic disease management of coronary heart disease in the elderly. Integr Med Res 2023; 15: 30–42.

PéREZ SUST

Solans

Fajardo

, et al. Turning the crisis into an opportunity: digital health strategies deployed during the COVID-19 outbreak. JMIR Public Health Surveill 2020; 6: e19106.

Hopstaken

Verweij

Van Laarhoven

, et al. Effect of digital care platforms on quality of care for oncological patients and barriers and facilitators for their implementation: systematic review. J Med Internet Res 2021; 23: e28869.

Chow

Thiagalingam

Santo

, et al. TEXT messages to improve MEDication adherence and secondary prevention (TEXTMEDS) after acute coronary syndrome: a randomised clinical trial protocol. BMJ Open 2018; 8: e019463.

10.

Fraund-Cremer

Hoffmann

Arndt

, et al. Long-term follow-up of patients with complex coronary artery disease treated with minimally invasive direct coronary artery bypass. Cardiol J 2023; 30: 1003–1009.

11.

Ying

Dai

, et al. Construction and application of an intelligent health follow-up management system for patients with chronic kidney disease. J Nurs 2024; 39: 11–15. 30.

12.

Mao

Luo

Wang

, et al. Construction and application of an intelligent interactive discharge follow-up platform. Chin J Nurs 2024; 59: 1812–1817.

13.

Fang

Idnay

Sun

, et al. Combining human and machine intelligence for clinical trial eligibility querying. J Am Med Inform Assoc 2022; 29: 1161–1171.

14.

Wang

. Current status and prospects of big data-based “user profiling” in improving patient experience. Chin Nurs Manag 2020; 20: 776–779.

15.

Alrashdi

. Fog-based deep learning framework for real-time pandemic screening in smart cities from multi-site tomographies. BMC Med Imaging 2024; 24: 123.

16.

Wang

. Research on an intelligent integrated service management platform for healthcare integration based on active health access technology. Medical Healthc Equip 2019; 40: 31–35.

17.

Reid

Pipe

Quinlan

, et al. Interactive voice response telephony to promote smoking cessation in patients with heart disease: a pilot study. Patient Educ Couns 2007; 66: 319–326.

18.

Reidel

Tamblyn

Patel

, et al. Pilot study of an interactive voice response system to improve medication refill compliance. BMC Med Inform Decis Mak 2008; 8: 46.

19.

Piette

. Interactive voice response systems in the diagnosis and management of chronic disease. Am J Manag Care 2000; 6: 817–827.

20.

Chen

Zhao

, et al. Construction and application of a post-discharge follow-up model for stroke patients based on an intelligent voice platform. Military Nurs 2022; 39: 55–57.

21.

Cheng

Yang

Liu

, et al. Construction and application of a health information management platform. Chin J Nurs Educ 2017; 14: 786–789.

22.

Zhao

, et al. Construction and application of an artificial intelligence follow-up system for multiple myeloma patients. Chin J Nurs 2023; 58: 1826–1830.

23.

Ling

Sheng

, et al. Application of a cloud-based intelligent follow-up system in home rehabilitation for children with burns. Nurs Rehabil J 2024; 23: 46–48.

24.

Ling

Mei

Zhu

, et al. Development of an intelligent follow-up system and its application in children with short bowel syndrome. Chin J Acute Crit Care Nurs 2023; 4: 887–892.

25.

Liu

Zhu

Zhou

, et al. Triglyceride-glucose index as a marker of adverse cardiovascular prognosis in patients with coronary heart disease and hypertension. Cardiovasc Diabetol 2023; 22: 133.

26.

Hart

Staveland

. Development of NASA-TLX (Task Load Index): results of empirical and theoretical research [M]//HANCOCK P A, MESHKATI N. Advances in Psychology. North-Holland. 1988: 139–83.

27.

Liang

Zhao

Deng

, et al. Translation and validation of the NASA-TLX scale in Chinese. Nurs Res 2019; 33: 734–737.

28.

Borges DO Nascimento

Abdulazeem

Vasanthan

, et al. Barriers and facilitators to utilizing digital health technologies by healthcare professionals. NPJ Digit Med 2023; 6: 161.

29.

Birken

Powell

Presseau

, et al. Combined use of the consolidated framework for implementation research (CFIR) and the theoretical domains framework (TDF): a systematic review. Implement Sci 2017; 12: 2.

30.

Damschroder

Reardon

Opra Widerquist

, et al. Conceptualizing outcomes for use with the consolidated framework for implementation research (CFIR): the CFIR outcomes addendum. Implement Sci 2022; 17: 7.

31.

Damschroder

Aron

Keith

, et al. Fostering implementation of health services research findings into practice: a consolidated framework for advancing implementation science. Implement Sci 2009; 4: 50.

32.

Zhang

, et al. Design and application of a pictorial patient fall risk assessment form. J Nurs 2023; 38: 56–59.

33.

Lyu

Zhao

. Behavioral profiling of patients with mental disorders in online health communities. Chin Mental Health J 2022; 36: 385–390.

34.

Shi

Yan

Lin

, et al. Scoping review of user profiling applications in health management for elderly patients with chronic diseases. Chin J Nurs 2025; 60: 244–250.

35.

Dai

Shao

Yao

, et al. Construction and application of a nursing platform for spinal cord injury rehabilitation based on patient profiling. Chin J Nurs 2024; 59: 2693–2699.

36.

Salas

Petracek

Yalamanchili

, et al. The use of artificial intelligence in pharmacovigilance: a systematic review of the literature. Pharmaceut Med 2022; 36: 295–306.

37.

Zhang

Song

, et al. User profiling study on body mass management in breast cancer patients undergoing chemotherapy. J Nurs 2022; 37: 33–36.

38.

Chang

Zhang

Guan

, et al. Research progress on the application of health profiling in chronic disease management. J Nanjing Medical Univ (Social Sci) 2023; 23: 546–550.

39.

Rosano

GMC

Moura

Metra

, et al. Patient profiling in heart failure for tailoring medical therapy. A consensus document of the heart failure association of the European Society of Cardiology. Eur J Heart Fail 2021; 23: 872–881.

40.

Passias

Jalai

Line

, et al. Patient profiling can identify patients with adult spinal deformity (ASD) at risk for conversion from nonoperative to surgical treatment: initial steps to reduce ineffective ASD management. Spine J 2018; 18: 234–244.

41.

Heapy

Higgins

Goulet

, et al. Interactive voice response-based self-management for chronic back pain: the COPES noninferiority randomized trial. JAMA Intern Med 2017; 177: 765–773.

42.

Ashjian

Yoo

Piette

, et al. Implementation and barriers to uptake of interactive voice response technology aimed to improve blood pressure control at a large academic medical center. J Am Pharm Assoc (2003) 2019; 59: S104–S9.e1.

43.

Vogel

Block

Kefalianos

, et al. Feasibility of automated speech sample collection with stuttering children using interactive voice response (IVR) technology. Int J Speech Lang Pathol 2015; 17: 115–120.

44.

Dunn

Arakawa

Greist

, et al. Assessing the onset of antidepressant-induced sexual dysfunction using interactive voice response technology. J Clin Psychiatry 2007; 68: 525–532.

45.

Pichayapinyo

Saslow

Aikens

, et al. Feasibility study of automated interactive voice response telephone calls with community health nurse follow-up to improve glycaemic control in patients with type 2 diabetes. Int J Nurs Pract 2019; 25: e12781.

46.

Elmagboul

Coburn

Foster

, et al. Comparison of an interactive voice response system and smartphone application in the identification of gout flares. Arthritis Res Ther 2019; 21: 160.

47.

Liu

Yang

Zhang

, et al. Comparison of telephone follow-up and an intelligent SMBG management platform in the management of outpatients with type 2 diabetes. Modern Med Health 2020; 36: 122–124.

48.

Cohn

Elmasry

Ehlke

. Utilization, receptivity and reactivity to interactive voice response daily monitoring in risky drinking smokers who are motivated to quit. Tob Induc Dis 2018; 16: 25.

49.

Liu

Zhang

Shi

, et al. Implementation of smart nursing based on a smart ward system in respiratory and critical care medicine. J Nurs 2022; 37: 5–9.

50.

Zhang

Sheng

, et al. Relationship between nurses’ professional deviant behaviors and their perception of organizational fairness and work stress. J Nurs 2020; 35: 87–90.

51.

Shuai

Liu

, et al. Application of a home-based remote symptom monitoring and intelligent decision support system based on a cloud follow-up platform in patients undergoing chemotherapy for gynecological malignancies. Evid Based Nurs 2024; 10: 3135–3139.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.11 MB