Smartphone-based mobile health impact on oral anticoagulation adherence, quality of life, and healthcare utilization in adult atrial fibrillation

Abstract

BACKGROUND

Atrial fibrillation (AF) is a prevalent arrhythmia associated with increased stroke risk, morbidity, and healthcare burden. Oral anticoagulation (OAC) is essential for stroke prevention, but adherence is often suboptimal. Smartphone-based mobile health (mHealth) applications offer potential to enhance adherence, quality of life (QoL), and reduce healthcare utilization.

AIM

To systematically review randomized controlled trials (RCTs) evaluating the impact of smartphone-based mHealth interventions on OAC adherence, health-related QoL, and healthcare utilization in adults with AF.

METHODS

Following PRISMA guidelines and a PROSPERO-registered protocol (CRD420251169904), databases (PubMed, Scopus, Web of Science, EMBASE, Cochrane CENTRAL, CINAHL, ClinicalTrials.gov) were searched from January 1, 2000, to July 30, 2025. Eligible studies were RCTs of mHealth apps for AF management in adults (≥ 18 years). Two reviewers screened records, extracted data, and assessed risk of bias using Cochrane risk of bias 2. Narrative synthesis was used due to heterogeneity.

RESULTS

Of 968 records, 16 RCTs (n = 209-3324 participants) were included, primarily from China, United States, Taiwan, and South Korea. Integrated interventions (e.g., mAFA-II aligning with AF better care pathway) improved OAC adherence (e.g., higher self-reported scores, P < 0.05; increased OAC persistence, P < 0.001), enhanced QoL (e.g., EQ-5D improvements, P < 0.001), and reduced healthcare utilization (e.g., rehospitalizations hazard ratio = 0.32, P < 0.001) without increasing bleeding risks. Simpler apps showed inconsistent benefits. Heterogeneity in designs, adherence measures, and follow-up (3 months to 2 years) limited generalizability. Risk of bias was high in 10 studies due to open-label designs and post-hoc analyses.

CONCLUSION

mHealth apps, especially comprehensive ones, can improve OAC adherence, QoL, and reduce utilization in AF patients. Future research should focus on long-term efficacy, cost-effectiveness, and equitable implementation for diverse populations.

Key Words: Atrial fibrillation; Mobile health; Oral anticoagulation; Adherence; Quality of life; Healthcare utilization; Systematic review

Core Tip: Atrial fibrillation (AF) is a prevalent arrhythmia associated with increased stroke risk, morbidity, and healthcare burden. Oral anticoagulation (OAC) is essential for stroke prevention, but adherence is often suboptimal. Smartphone-based mobile health (mHealth) applications offer potential to enhance adherence, quality of life (QoL), and reduce healthcare utilization. mHealth apps, especially comprehensive ones, can improve OAC adherence, QoL, and reduce utilization in AF patients. Future research should focus on long-term efficacy, cost-effectiveness, and equitable implementation for diverse populations.

INTRODUCTION

Atrial fibrillation (AF) is the most widespread sustained arrhythmia in the world with a population of more than 37 million cases and a rate that is predicted to increase with the ageing of the population and increase in cardiovascular disease burden[1,2]. AF is characterized by the chances of ischemic stroke that are almost five times more significant, with mortality, and healthcare expenses are higher[3,4]. Therefore, stroke prevention has taken the center stage in AF management, which is mostly done by the use of oral anticoagulation (OAC) therapy[5]. Both the non-vitamin K oral anticoagulants and vitamin K antagonists are very efficient in prevention of thromboembolic events and better outcomes[6,7].

However, in the actual world there exists obstacles that do not allow optimal utilization of OACs. There is often a low level of medication adherence and persistence, which is explained by the complication of the regimen, the lack of empowerment of the patient, fear of bleeding, and the absence of monitoring[8,9]. Poor compliance leads to significant increase in the risks of thromboembolism, hospital admission and fatality[10]. Formalized anticoagulation services have not succeeded in sealing gaps in therapy persistence; even[11]. In addition to these complications, AF causes a severe adverse effect on QoL due to physical symptomology, emotional distress, and treatment burden[12,13]. The use of healthcare in AF is also too high, which is mainly motivated by high hospitalization and outpatient visits[14,15]. To sum up, adherence, QoL, and avoidable resources reduction are essential general aims in AF care, which should be developed at the same time.

To respond to these challenges, digital health solutions, specifically mobile health (mHealth) applications, which can be implemented in smartphones, are new. Application The use of specific apps can provide patient-centered facilities-including medication prompts, symptom diaries, remote dynamic capability, and patient education resources-to facilitate better patient self-management and engagement in care[16,17]. The results of the chronic cardiovascular diseases management suggest that mHealth intervention could positively influence medication adherence, healthy behavior, and reduction in hospital rehospitalization[18,19]. In AF in particular, smartphone-based apps that are individually targeted at anticoagulation adherence have been the subject of RCTs, some of which have shown clinically meaningful and significant improvements in both adherence and patient-reported outcomes (PROs)[20]. Nevertheless, results are not the same in trials, especially in terms of the benefits of QoL and healthcare.

Since the aspects of digital health adoption are growing swiftly and some guideline requests to include the use of validated mHealth solutions in cardiovascular care are also under revision[21], it is reasonable to conduct a comprehensive overview of the currently existing RCT literature. This systematic review will aim at synthesizing the evidence regarding the impact of mHealth applications (integrated into a smartphone) that have been implemented through RCTs or other forms of clinical trials on OAC adherence, health-related quality of life (HRQoL), and health care utilization in adult AF individuals. This review will bring clarity to the effectiveness of digital interventions, areas of knowledge that need further investigation, and future research and clinical approaches to maximize AF management in a more digital healthcare environment by synthesizing the existing evidence.

MATERIALS AND METHODS

Protocol and reporting standards

This methodological review was systematic and based on an already specified methodology in terms of assessing the effectiveness and safety of smartphone application based interventions in the integrated management of patients with AF. Our approach is anchored on a prospectively registered protocol in the PROSPERO (CRD420251169904) international database that provides total transparency, specification of purpose, inclusion criteria, and the method of analysis. In the course of reviewing, we followed the PRISMA guideline[22]. The study selection methodology was organized in terms of the following Population, Intervention, Comparator, and Outcome (PICO) framework.

Participants (P): We considered randomized controlled trials (RCTs), such as individually randomized, cluster-randomized studies, that involved adults (> 18 years) with a known diagnosis of AF (paroxysmal, persistent, or permanent).

Interventions (I): Trials were included when they tested a smartphone-/tablet-based mHealth application as the major intervention, and the application was created to aid in managing AF. Functionalities that were eligible included: OAC alerts, medication nonadherence implementation, patient education module, risk assessment/decision support, symptom diary, or tele-consultation functionality.

Comparators (C): Comparators were usual care, defined as standard clinical follow-up, non-digital educational materials, or generalized health applications not tailored to AF.

Outcomes (O): The review focused on primary and secondary outcomes that reflected efficacy, patient experience, and safety.

Primary outcome

OAC Adherence professionals use to measure the level of adherence allows individuals to select an appropriate filler, an oral anticoagulant, and to stay sufficiently adherent to that filler to reduce the risk of warfarin and blood clots (Carlton, 2003). Adherence red operation; Takes place when an individual (or patient) is given instructions to take an OAC, typically daily, referred to as oral anticoagulation; Within the healthcare domain, OAC is widely used to prevent and treat thromboembolic disorders such as AF–related stroke and venous thromboembolism. Its safe use requires individualized risk–benefit assessment, regular monitoring or adherence support, and integration of patient education to minimize bleeding complications.

Secondary outcomes

HRQoL: Assessed by validated instruments, such as the AF effect on quality of Life (QoL), the three-level form of questionnaire (EQ-5D-3 L), a five-item self-report tool, and the short form health survey. The other related PROs were the brief coping orientation to problems experienced scale (approach and avoidance coping strategies), EuroQol visual analog scale, and treatment satisfaction questionnaire for medication.

Healthcare utilization: It will include all-cause or cardiovascular rehospitalization, emergency visits, unplanned outpatient visits, or length of stay. Readmission events will tend to have follow-up until a period of 2 years.

Safety endpoints: Major bleeding (defined by the criteria of the International Society of Thrombosis and Hemostasis) and thromboembolic events.

Information sources and search strategy

We systematically searched MEDLINE (Ovid), EMBASE, Scopus, Web of Science, Cochrane CENTRAL, CINAHL, and ClinicalTrials.gov from January 1, 2000, until July 30, 2025, without language restrictions. The lower limit reflected the era of smartphone-enabled interventions. Search terms combined AF (“atrial fibrillation”, “AFib”), mHealth (“mobile health”, smartphone, app, telemedicine), and RCT filters. Search terms combined AF (“atrial fibrillation”, “AFib”), mHealth (“mobile health”, smartphone, app, telemedicine), and RCT filters. We also manually searched the reference lists of relevant studies for any further available literature.

Study selection

All the records retrieved were imported in Rayyan. Titles and abstracts were screened independently by two reviewers against the acceptance criteria and possibly relevant records were screened against these criteria in full-text. The disputes were solved either via a disagreement or by arbitration with a third reviewer. Identification, screening, eligibility and inclusion procedure was all recorded in a PRISMA flow diagram, including clear justification for exclusion during the full-text phase.

Data extraction

Data were elicited based on a structured protocol to ensure that all the necessary data were captured to be used in the synthesis. The data were extracted as follows: Trial design and setting, Sample size and patient demographics (mean age, sex distribution, AF type, comorbidities), as Intervention characteristics (application name, specific functionalities, integration with wearables, frequency/intensity, training/onboarding) as Comparator description as adherence definitions and measurement methods as QoL tools and healthcare utilization metrics and as follow-up time.

Risk of bias assessment

The quality of the methods employed in the listed trials was evaluated with the Cochrane risk of bias 2 tool. The bias was measured in the following domains: Randomization process, absence of intended intervention, lack of outcome data, measure of the outcome, and the reported result. The evaluation has taken into account the constraints of digital health trials, including the nature of interventions as an open-label trial (D2/D4), methodological challenges associated with cluster randomization (D1).

Data synthesis and analysis

A strong narrative synthesis was selected as the main technique of investigation because of a high level of clinical and methodological heterogeneity in interventions (between simple reminders and complex care pathways) as well as high statistical dependency on numerous additional analyses derived from the one big mAFA II study. This method placed more emphasis on the distribution of results based on the outcome in comparison with cross-studies instead of quantitative pooling (meta-analysis).

RESULTS

Study selection and characteristics

We conducted a systematic search that identified 968 records first. Once it had been stripped of 84 duplications, 884 distinct records were put through a title and abstract screening. Out of those 128 full-text articles were evaluated during the eligibility process. Finally, our review included 16 RCT[23-38] out of the 1009 trials (Figure 1). These were cross-national studies, conducted in China, Taiwan, South Korea, and the United States, and with the number of participants differing between smaller studies, including 100 patients each, and large cluster-randomized trials involving 3324 participants each. Table 1 highlights the study features of 16 studies included in the study, focusing on their design, patient features, intervention elements, and measuring outcomes.

Figure 1 PRISMA-2020 flow diagram for systematic review. RCT: Randomized controlled trial.

Table 1 Characteristics of included studies (n = 16).

Ref.	Country/setting	Study design and focus	Trial/program name	Sample size (intervention/control)	Key participant characteristics	Intervention description	Follow-up duration	Primary outcomes assessed
mAFA-II program cluster (integrated care via ABC pathway)
Guo et al[23], 2020	China; 40 centers	Cluster RCT; efficacy of integrated care	mAFA-II	3324 (1646/1678)	Mean age 67-70 years; 38% female	mHealth-supported ABC pathway (risk scores, OAC guidance, education)	Mean 262-291 days	Composite (stroke/TE, death, rehospitalization)
Romiti et al[24], 2023	China; 40 centers	Post-hoc analysis; win ratio method	mAFA-II	3324	(Subset of core trial population)	ABC pathway (as in core trial)	Mean 291 days	Composite outcome analyzed via win ratio
Yao et al[25], 2021	China; 40 centers	Ancillary analysis; patients with multimorbidity	mAFA-II	1890 (subgroup with ≥ 2 comorbidities)	Mean age 72-73 years; 34%-42% Female	ABC pathway	Mean 419-457 days	Composite outcome, thromboembolism, bleeding
Guo et al[26], 2022	China; 40 centers	Subgroup analysis; patients ≥ 75 years	mAFA-II	1163 (subgroup ≥ 75 years)	Mean age 827 ± 5.2 years	ABC pathway	Median about 340-395 days	Composite outcome, rehospitalization
Guo et al[27], 2023	China; 40 centers	Post-hoc analysis; patients with heart failure	mAFA-II	714 (subgroup with HF)	Mean age 727 ± 13.1 years; 40% Female	ABC pathway	Median about 281-284 days	Composite outcome, bleeding, non-fatal CV events
Corica et al[28], 2025	China; 40 centers	Post-hoc analysis; latent class analysis (phenotypes)	mAFA-II	3324	Mean age 685 ± 13.9 years	ABC pathway	Median 295 days	Interaction between phenotypes and treatment effect
Guo et al[29], 2023	China; 40 centers	Post-hoc analysis; Sex-stratified outcomes	mAFA-II	3324 (2062 male/1262 fem ale)	Mean age male: 67.5, female: 70.2 years	ABC pathway	6 and 12 months	Composite outcome stratified by sex
Guo et al[30], 2020	China; 2 hospitals	Ancillary study; dynamic bleeding risk	mAFA (pilot/II)	1,793 (intervention arm only)	Mean age 640 ± 24.0 years	Dynamic HAS-BLED reassessment via app	Dynamic monitoring (up to 365 days)	Change in high bleeding risk, incident bleeding
Guo et al[31], 2017	China; 2 hospitals	Pilot cluster RCT; feasibility	mAFA pilot	209 (113/96)	Mean age 67-71 years	mAF app (management, self-care)	3 months	Adherence, quality of life, satisfaction
Independent trials
Yoon et al[32], 2024	South Korea; Multicenter	RCT; adherence via alerts	ADHERE-app	498	Mean age 657 years; 68% male (Edoxaban users)	App with daily push alerts + BP/HR monitoring	6 months	Adherence (pill count)
Turakhia et al[33], 2021	United States; 25 sites	RCT (terminated early); digital/human support	SmartADHERE	139 enrolled	Mean age 65 ± 9.6 years; 30% female (Rivaroxaban users)	App + text messaging + phone counseling	6 months (planned)	PDC
Caceres et al[34], 2020	United States; single-center	RCT; ECG monitoring post-procedure	iHEART	238 (post-ablation/cardioversion)	Mean age 613 years; 77% male	Smartphone ECG (Kardia) + motivational texts	6 months	Quality of life (AFEQT, etc.), symptom severity
Hsieh et al[35], 2021	Taiwan; single-center	Single-blind RCT; web-based management	Web-based program	231	Mean age 731 ± 11.7 years; 50% male	Web-based program (education, consultation)	6 months (outcomes) + 2 years (readmission)	Adherence, quality of life, readmission events
Xu et al[36], 2024	China; 5 hospitals	RCT; app-based education and support	Alfalfa app	113	Mean age 617 ± 11.0 years; 62% male	Alfalfa app (education, remote consultation)	12 months	Adherence (MMAS-8), knowledge, satisfaction
Magnani et al[37], 2025	United States; rural pennsylvania	RCT; intervention for rural patients	Rural AF intervention	270 (rural cohort)	Median age 731 years; 60% female	Relational agent app + KardiaMobile monitor	12 months	Adherence (PDC), quality of life, healthcare utilization
Tran et al[38], 2022	United States; single-center	RCT; Pill organizer + monitoring	BOAT-OAR	100 (Apixaban users)	Mean age 708 ± 8.0 years; 50% female	Smartphone ECG + pill organizer	Median 5.9 months	Medication compliance (pill count), clinical events

RCT: Randomized controlled trial: ABC: Atrial fibrillation better care; OAC: Oral anticoagulant; TE: Thromboembolism; AS-BLED: Reassessment score–hypertension, abnormal renal/Liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly, drugs/alcohol; mHealth: Mobile health; ADHERE-app trial: Application-driven heart rhythm education and empowerment trial; SmartADHERE trial: Smartphone-based adherence to anticoagulation and heart rhythm education trial; PDC: Proportion of days covered; ECG: Electrocardiogram; iHEART: Interactive heart health application for rhythm tracking; AFEQT: Atrial fibrillation effect on quality-of-life questionnaire; MMAS-8: 8-item morisky medication adherence scale: AF: Atrial fibrillation; BOAT-OAR: Bleeding outcomes and thromboembolism in oral anticoagulant-related trial.

Primary outcome: OAC adherence

Eight studies[30-33,35-38] assessed the effect of mHealth interventions on OAC participation. The outcomes were weak and highly reliant on the mode of measuring adherence. Investigations using PROs or assessing general OAC uptake all presented considerable improvements. Web-Based Integrated Management Program showed a much better increase in Medication Adherence Rating Scale (MARS) scores after 6 months than control (P = 0.001). Likewise, Alfalfa App[36] and the mAFA Pilot Trial[31] indicated much greater adherence scores after 1 and 3 months with MMAS-8 and 3-item Adherence Estimator, respectively (all P < 0.05). As did the Rural AF Intervention[37], which did not show any difference in objective proportion of days covered (PDC), but which demonstrated significantly better self-reported adherence at 4 and 8 months (odds ratio = 1.89). On the systems-wide level, a secondary study of the mAFA-II trial[30] revealed that the use of OAC significantly rose in the intervention group and declined by 25% in the usual care group after 12 months (P < 0.001), which signifies better treatment adherence. Conversely, studies with objective measures often had no statistically significant group difference. In both the intervention (0.86) and control (0.88) arm, the SmartADHERE Trial[33] reported high PDC scores (P = 0.62). Median compliance in terms of pill counts was found to be near-perfect in both groups (100 vs 99.7, P = 0.247) according to the BOAT-OAR Study[38]. No difference in 12-month PDC was also observed in Rural AF Intervention[37]. A subtle exception was given in the ADHERE-App Trial[32]. Although it did not find a difference in the ongoing median pill count (98% vs 97.5%, P = 0.15), it was also able to prove that a substantially larger percentage of intervention participants had reached a clinically significant level of adequate adherence (> 95%) at 6 months (73.9% vs 61.0%, P = 0.007). Table 2 shows the effect of mHealth Interventions on Oral Anticoagulant Adherence.

Table 2 Oral anticoagulant adherence.

Ref.	Intervention/trial name	Adherence metric	Follow-up	Key finding	Interpretation
Studies reporting a positive effect on adherence
Yoon et al[32], 2024	ADHERE-app (Alerts + monitoring)	Pill count (≥ 95% threshold)	6 months	Proportion with adequate adherence (≥ 95%) significantly higher in intervention (73.9% vs 61.0%; P = 0.007)	Intervention increased the likelihood of achieving high adherence
Hsieh et al[35], 2021	Web-based integrated program	MARS (self-report scale)	6 months	Significantly greater improvement in adherence vs control over time (GEE analysis, P = 0.001)	Sustained improvement in self-reported adherence
Xu et al[36], 2024	Alfalfa app (education + consultation)	MMAS-8 (self-report scale)	3 months	Medication compliance significantly higher at 1 and 3 months (all P < 0.05)	Short-term improvement in self-reported adherence
Guo et al[31], 2017	mAFA pilot (mAF app)	3-item adherence estimator	3 months	Drug adherence significantly better at 1 and 3 months (all P < 0.05)	Early improvement in adherence with integrated app
Magnani et al[37], 2025	Rural AF intervention (agent + monitor)	3-item self-report instrument	4 and 8 months	Self-reported adherence significantly higher (OR = 1.89)	Positive subjective effect, despite no objective difference
Guo et al[30], 2020	mAFA-II (dynamic management)	OAC usage/uptake	12 months	OAC use increased in intervention (63.4% to 70.2%) but decreased in control (P < 0.001)	Intervention improved OAC persistence at a population level
Studies reporting no significant difference in objective adherence
Yoon et al[32], 2024	ADHERE-app (alerts + monitoring)	Pill count (continuous median)	6 months	No difference in median adherence (98% vs 97.5%; P = 0.15).	High baseline adherence in both groups limited improvement
Turakhia et al[33], 2021	SmartADHERE (blended support)	PDC	6 months	No difference in mean PDC (0.86 vs 0.88; P = 0.62) or PDC ≥ 80%	Exceptionally high adherence in control group (“ceiling effect”)
Tran et al[38], 2022	BOAT-OAR (ECG + organizer)	Pill Count (median)	5.9 months	No difference in median compliance (100% vs 99.7%; P = 0.247)	Near-perfect adherence in both groups
Magnani et al[37], 2025	Rural AF intervention (agent + monitor)	PDC	12 months	No difference in 12-month PDC (median 0.97 both groups)	Objective measure did not confirm self-reported benefits

ADHERE: Smartphone-based adherence to anticoagulation and heart rhythm education; MMAS-8: 8-item morisky medication adherence scale; AF: Atrial fibrillation; OR: Odds ratio: ADHERE-app: Application-driven heart rhythm education and empowerment; BOAT-OAR: Bleeding outcomes and thromboembolism in oral anticoagulant-related; PDC: Proportion of days covered; ECG: Electrocardiogram; MARS: Medication adherence report scale; GEE: Generalized estimating equations.

PROs and HRQoL

The effectiveness of the digital interventions on PROs was directly linked to the thoroughness of the intervention. Integrated management programs in all cases presented beneficial outcomes, and interventions concentrating on monitoring alone did not present any significant impact compared to normal care. The Web-Based Integrated Management Program[35] enhanced both generic HRQoL (EQ-5D-3 L, P < 0.001) and adaptive coping strategies significantly (P < 0.001), as shown in Table 3. In a similar vein, the mAFA Pilot Trial[31] and the Alfalfa App Study[36] showed considerable QoL, anxiety, and treatment satisfaction improvements (all P < 0.05). However, other studies, such as the iHEART Study[34] and the Rural AF Intervention[37], focused on rhythm monitoring and reported no statistically significant differences in AF-specific or generic HRQoL between groups, but both groups in the iHEART study experienced a statistically significant difference between baseline and post.

Table 3 Impact of mobile health interventions on patient-reported outcomes.

Ref.	Intervention/trial name	PRO domain and measurement tool	Follow-up	Key finding
Studies reporting significant improvements
Hsieh et al[35], 2021	Web-based integrated program	HRQoL: EQ-5D-3 L and EQ-VAS	6 months	Significantly higher scores in intervention group at 3 and 6 months (e.g., EQ-5D, β = 0.19, P < 0.001 at 6 months)
Hsieh et al[35], 2021	Web-based integrated program	Coping: Brief COPE scale	6 months	Significantly improved approach coping and reduced avoidance coping (P < 0.001)
Guo et al[31], 2017	mAFA pilot trial	HRQoL: EQ-5D-Y; satisfaction: ACTS	3 months	Significantly increased QoL, reduced anxiety/depression, and higher satisfaction (all P < 0.05)
Xu et al[36], 2024	Alfalfa app study	Satisfaction: Treatment satisfaction questionnaire for medication	3 months	Scores for effectiveness, side effects, convenience, and global satisfaction significantly higher (P < 0.001)
Studies reporting no significant difference
Caceres et al[34], 2021	iHEART study (ECG + texts)	HRQoL: AFEQT, SF-36, EQ-5D	6 months	No significant differences between groups. Clinically meaningful improvement observed in both arms
Magnani et al[37], 2025	Rural AF intervention	HRQoL: AFEQT, PROMIS-29	12 months	No significant differences in disease-specific or general quality of life scores

PRO: Patient-reported outcomes; HRQoL: Health-related quality of life; iHEART: Interactive heart health application for rhythm tracking; AF: Atrial fibrillation; AFEQT: Atrial fibrillation effect on quality-of-life questionnaire; SF-36: Short form health survey; EQ-5D: EuroQol 5-dimension; EQ-VAS: EuroQol visual analogue scale; COPE scale: Coping orientation to problems experienced scale; ACTS: Anti-clot treatment scale; PROMIS-29: Patient-reported outcomes measurement information system–29.

Clinical efficacy, healthcare utilization, and safety

The mAFA-II integrated care model showed that there were major clinical events and hospitalization reductions of significant and consistent effect, and that specific subgroups associated with high risk have benefited more in safety outcomes. The mAFA-II intervention resulted in a significant primary composite outcome, stroke, thromboembolism, death, and rehospitalization [hazard ratio (HR) = 0.39, P < 0.001], mainly comprising of a significant reduction in rehospitalizations (HR = 0.32, P < 0.001). The effect was strong among the analytical techniques, comprising a win ratio analysis (win ratio = 2.78, P < 0.001) and high-risk populations, like patients with multimorbidity (HR = 0.37, P < 0.001). Readmissions were also reduced dramatically during the two years with the help of the Web-Based Integrated Program[35] (OR = 0.41, P = 0.03). In the area of safety, in the overall mAFA-II trial, no significant difference in the bleeding rates appeared (HR = 0.95, P = 0.85); however, in an ancillary analysis, dynamic bleeding risk monitoring in the app was observed to decrease the incidence of incident bleeding events over time significantly (P < 0.001). Besides, thromboembolic events were also significantly reduced by the intervention (HR = 0.17, P = 0.002) in the high-risk multimorbidity cohort. Other smaller trials recorded no considerable differences in clinical endpoints, probably because they were followed in a short period and the events were few (Table 4).

Table 4 Impact of mobile health interventions on clinical events and healthcare utilization.

Ref.	Intervention/trial name	Outcome metric	Result (intervention vs control)	Effect estimate (95%CI)	P value
Clinical efficacy and healthcare utilization
Guo et al[23], 2020	mAFA-II trial	Primary composite (stroke/TE, death, rehospitalization)	1.9% vs 6.0%	HR 0.39 (0.22-0.67)	< 0.001
Guo et al[23], 2020	mAFA-II trial	Rehospitalization (any cause)	1.2% vs 4.5%	HR 0.32 (0.17-0.60)	< 0.001
Romiti et al[24], 2023	mAFA-II (win ratio)	Composite (prioritizing death)		Win ratio 2.78 (1.85-4.17)	< 0.001
Yao et al[25], 2021	mAFA-II (multimorbidity)	Primary composite outcome		HR 0.37 (0.26-0.53)	< 0.001
Hsieh et al[35], 2021	Web-based program	Readmission (2-year follow-up)	11 vs 23 events	OR 0.41 (0.18-0.93)	0.03
Safety (bleeding and thromboembolism)
Yao et al[25], 2021	mAFA-II (multimorbidity)	Thromboembolism (stroke/TE)	0.5% vs 2.9%	HR 0.17 (0.05-0.51)	0.002
Guo et al[23], 2020	mAFA-II trial	Extracranial bleeding	1.5% vs 1.6%	HR 0.95 (0.54-1.66)	0.85
Guo et al[30], 2020	mAFA-II (dynamic)	Incident bleeding events (rate)	Rate decreased over time (1.2% to 0.2%)		< 0.001

TE: Thromboembolism; OR: Odds ratio; HR: Hazard ratio.

Intervention processes, knowledge, and feasibility

The results of process analyses of process outcomes revealed a high feasibility and identified critical playing mechanisms, such as improvement of patient knowledge and self-management capabilities, which were likely to mediate the clinical benefits gained. The interventions were characterized by good user engagement as the Rural AF Intervention showed high fidelity (median 84% daily use) and the mAFA-II program maintained active communication between patients and clinicians at 12 months (P < 0.001). More importantly, educational applications were effective in promoting proximal outcomes: The Alfalfa App was very effective at increasing the patient's knowledge of the concept of anticoagulation (Jessa Atrial fibrillation Knowledge Questionnaire score, P < 0.001), and the Web-Based Integrated Program demonstrated more adaptive coping mechanisms (P < 0.001). This patient activation has led to more optimal treatment in that the mAFA-II program has achieved an increase in OAC uptake and long-term retention, contrary to a considerable drop in the standard care group (P < 0.001) (Table 5).

Table 5 Mechanisms of action and feasibility of mobile health intervention.

Ref.	Intervention/trial name	Process outcome domain	Key finding
Feasibility and engagement
Magnani et al[37], 2025	Rural AF intervention	Intervention fidelity: Median use of relational agent over 120 days	Excellent fidelity; participants used the agent a median of 84% of days
Guo et al., 2020[23]	mAFA-II (dynamic risk)	Communication fidelity: Frequency of patient/doctor messages	Significant, sustained message frequency over 12 months (P < 0.001)
Hsieh et al[35], 2021	Web-based program	Intervention delivery	Implemented via face-to-face training to ensure user competency
Patient activation and knowledge
Xu et al[36], 2024	Alfalfa app study	Anticoagulation knowledge (JAKQ)	Knowledge significantly higher in app group at 1 and 3 months (e.g., 78.1% vs 25.0% at 3 months; P < 0.001)
Hsieh et al[35], 2021	Web-based program	Coping strategies (brief COPE scale)	Significantly improved approach coping and reduced avoidance coping (P < 0.001)
Treatment optimization
Guo et al[23], 2020	mAFA-II (dynamic risk)	OAC treatment uptake/persistence	OAC use increased in intervention arm (63.4% to 70.2%) but decreased by 25% in usual care (P < 0.001)

AF: Atrial fibrillation; COPE scale: Coping orientation to problems experienced scale; OAC: Oral anticoagulant; COPE scale: Coping orientation to problems experienced scale; JAKQ: Japanese version of the anticoagulation knowledge questionnaire.

Risk of bias within studies

The analysis of the methodological quality demonstrated that there were serious issues with the internal validity of the evidence. According to Table 6, with the general risk of bias high, 10 studies[23-29,31,33,34] out of 16 sources of use were identified as contributing to the high risk of bias, mainly due to the massive mAFA program cluster and 2 underpowered individual trials. The mAFA-II trial together with its ancillary reports, was always rated as high risk because of some inherent limitations of the cluster-randomized design that resulted in imbalances on baselines and the inability to blind the subjects and staff in a multifaceted, mHealth-based intervention. Moreover, most of the analyses of this cohort were post-hoc exploratory studies, hence a high likelihood of selective reporting bias. The risk of bias was more divergent among the independent trials. The SmartADHERE and iHEART trials were classified as high risk because of the premature termination/under-powering, and lack of missing data. The other five trials that were not randomized (e.g., ADHERE-app, Web-Based Program) were rated as some concerns, mainly due to the fact that they failed to provide blinding in order to determine PROs (e.g., adherence scales, QoL), thus, otherwise sound methodologies (Figure 2).

Figure 2 Risk of bias assessment.

Table 6 Risk of bias assessment.

Ref.	D1	D2	D3	D4	D5	Overall
Guo et al[23], 2020	Some concerns	High	Low	High	Low	High
Romiti et al[24], 2023	Some concerns	High	Low	High	Low	High
Yao et al[25], 2021	Some concerns	High	Low	High	Some concerns	High
Guo et al[26], 2022	Some concerns	High	Low	High	Some concerns	High
Guo et al[27], 2023	Some concerns	High	Low	High	Some concerns	High
Corica et al[28], 2025	Some concerns	High	Low	High	High	High
Guo et al[29], 2023	Some concerns	High	Low	High	High	High
Guo et al[30], 2020	No information	No information	Low	Low	Low	Some concerns
Guo et al[31], 2017	Some concerns	High	Low	High	Low	High
Yoon et al[32], 2024	Low	Some concerns	Some concerns	Low	Low	Some concerns
Turakhia et al[33], 2021	Low	Some concerns	High	Some concerns	Low	High
Caceres et al[34], 2020	Low	Some concerns	High	Some concerns	Some concerns	High
Hsieh et al[35], 2021	Low	Some concerns	Low	Some concerns	Low	Some concerns
Xu et al[36], 2024	Low	Some concerns	Low	Some concerns	Low	Some concerns
Magnani et al[37], 2025	Low	Some concerns	Low	Some concerns	Low	Some concerns
Tran et al[38], 2022	Low	Some concerns	Low	Some concerns	Low	Some concerns

DISCUSSION

This systematic review summarizes the accumulating evidence of the efficacy of mHealth applications as effective adjuncts in dealing with patients with AF, especially in the attributes of increasing OAC adherence, improving QoL and cutting down on healthcare utilization. The aggregate results show that multidimensional digital interventions, particularly the ones that are based on the AF better care (ABC) pathway, allow a greater positive effect on medication compliance and result in better clinical outcomes without presenting a higher risk of safety. The implications of these insights are significant in the modern AF management since the healthcare systems all over the world are trying to find an approach to a multifaceted burden of AF, which is scalable and patient-centered.

Contextual interpretation of reading existing literature

Our results support and derive the inferences drawn by the other systematic reviews on the use of digital interventions in the management of cardiovascular diseases. In line with meta-analyses carried out by other researchers[39-41], the present review confirms that integrated multifaceted mHealth tools (including education, reminders, and self-monitoring) will be of great use in increasing medication compliance in diverse chronic conditions (including AF). It can be seen in the significant decrease in the risk towards composite clinical outcomes reported in the mAFA-II programme[23] where the incorporation of structured care frameworks like the ABC pathway into these digital platforms serves as an indication of the value of comprehensive and guideline-driven interventions, as opposed to digital notifications.

On the contrary, the benefits of single-function mHealth solutions were less reliable. As an illustration, those applications that were based on the active monitoring of symptoms or the rhythm but did not include proactive patient education or clinical support did not show significant changes in the QoL or adherence indicators, which are reflected in the reviews by Kitsiou et al[42] and Desteghe et al[43]. This implies that passive forms of data collection can perhaps not be enough to make any kind of meaningful change to influence behavior without customized responses and clinician incorporation.

It may be noteworthy that some emerging evidence indicates that its efficacy varies between subgroups of patients, and smaller effect sizes have been reported to be observed in elderly, multimorbid, or heart failure patients[26,28]. These trends demonstrate that it is necessary to implement the adaptive designs of interventions that would take into account cognitive, physical, and social complexities, and possibly involve caregiver engagement and simplified interfaces as optimal accessibility strategies.

In terms of safety, our review report established that the incorporation of mHealth platforms does not make any adverse events like bleeding or thromboembolism more likely, and effective methods of providing supportive assistance will not disrupt patient safety, which is again consistent with previous literature[35,44].

In economic terms, even though the majority of the researches do not include the formal health economic analysis, an intuitive cost-savings as a result of hospitals, emergency visits, and readmissions reduction is large. This goes in line with observational modelling in cardiac telemonitoring, making it possible to believe in potential cost-effectiveness, once there is optimization of the system integration and adherence to the user[45]. Still more stringent prospective economic appraisals entrenched in practical tests are critical to support these advantages to policy makers and payers.

The review strengths and limitations

This is one of the main strengths of the review as it considers the recent high-quality RCT and large-scale cluster studies involving diverse populations and various healthcare settings. The utilization of stringent methodology facilitated the incorporation of studies that had strong design, like the mAFA-II trial that included more than three thousand patients, thereby making the results widely generalizable[23]. In addition, through the evaluation of a variety of clinically relevant outcomes involving adherence, QoL indicators, and healthcare use, this review offers a comprehensive view of the effectiveness of mHealth interventions.

Interpretation is, however, tempered by a number of limitations. Heterogeneity of the interventions is a challenge since the included studies differed in the smartphone features included, the degree of components in patient engagement, and the follow-up periods, which extended from a few months up to two years. This heterogeneity makes it difficult to compare things directly and to pool them in a centrifuge. Also, most of the studies comprise relatively short-term follow-up, which does not provide information on long-term adherence sustainability and clinical benefit sustainability. Open-label designs also raise methodological issues in digital health trials as they may facilitate bias through awareness in patients and doctors to change behaviour and reporting[33,36]. There were numerous studies based on self-reported measures of adherence, which, although validated, is prone to social desirability and recall bias. Moreover, cost-effectiveness and economic impact information are still limited, making it difficult to draw final conclusions about the value of mHealth integration to the health system in general. The literature included offered little discussion of patient subgroups, including multimorbidity, older age, cognitive impairment, and different socioeconomic settings, which needs to be bridged to understand the implications of equity and accessibility. It is not yet clear whether it can be generalized to low-resource and rural settings where the burden of AF is disproportionate. Further research is advised to focus on various inclusion and barriers to implementation in the real world.

Clinical and research implications

These results are clinically encouraging and will provide the necessary support to consider the use of a validated mHealth app in routine AF care pathways, especially those that provide multidomain interventions applied through the ABC-based approach. The platforms have the capability to support patient-centered education, promote prolonged adherence to anticoagulation, early triggers of symptoms, and integrate the management of comorbidities. To avoid leaving out or withholding digital tools, clinicians must focus on patient choice, technological literacy, and cognitive abilities and provide them with suitable training and support to optimize their use.

Also, the equity considerations should be taken into account because digital health innovation has a high risk of widening the disparity in health services when the underserved groups do not have access to or knowledge of using necessary devices and connectivity[46]. Creating culturally relevant material and addressing socioeconomic inequalities are critical issues that need to be handled to have inclusive benefits.

Research-wise, the future research must be on the long-term efficacy, practical viability, and comparative efficacy of various digital interventions with head-on trials investigating the most optimal elements. Population heterogeneity should be used in the design of pragmatic trials that will enhance generalizability. Moreover, digital biomarkers and personalized suggestions based on artificial intelligence represent promising opportunities in the improvement of intervention accuracy and effectiveness. There should also be focused attention to health economic appraisals that assess the cost-expenses and budget implications in order to inform the decision-making on the system-level adoption. Finally, effective data governance, privacy, and data interoperability terms are a precondition of safe, scalable implementation.

CONCLUSION

This systematic review suggests that mHealth applications on the smartphone could effectively improve OAC adherence and better clinical outcomes in patients diagnosed with AF who took anticoagulant. The stroke, death, and rehospitalization rates were reduced greatly in integrated interventions along the pathway of ABC. Although medication adherence did not improve similarly, apps that had extensive functionalities such as symptom monitoring and comorbidity management proved to be the most effective. The effects of QoL and the effects of healthcare use were mixed. Notably, no greater safety hazards were reported. There are still problems with outreach to the older generations and dissimilar populations. The long-term effects, cost-effectiveness, and fair implementation are areas that research needs to concentrate on in the future to reap the benefits of digital health solutions in the management of AF.