Integrated care management and comprehensive geriatric assessment using a mHealth-based approach in older multimorbid patients with atrial fibrillation: the AFFIRMO cluster-randomised trial

Summary

Background

The AFFIRMO trial aimed to evaluate the efficacy of a mobile-health (mHealth) structured intervention adapting the Atrial Fibrillation Better Care (ABC) pathway, combining the comprehensive geriatric assessment (CGA), to enhance guideline-directed, holistic AF management and reduce the risk of clinical outcomes in older patients with atrial fibrillation (AF) and multimorbidity.

Methods

In a multicentre, open-label, cluster-randomised study conducted in six European countries, clusters were allocated to usual care or intervention (iABC system) (ClinicalTrials.govNCT06775028, https://clinicaltrials.gov/study/NCT06775028). In usual care clusters, patients were managed according to clinical guidelines.

Findings

We enrolled 1260 patients (mean [SD] age 74 [6] years; 44.4% (560/1260) female; median CHA₂DS₂-VASc 4 [IQR 3–5]), with 634 (50.3%) assigned to intervention and 626 (49.7%) to usual care. Overall, 41.0% (517/1260) had >2 comorbidities, with a high prevalence of anticoagulation (1226, 97.3%) and guideline-recommended therapies. CGA in the intervention arm documented a highly preserved function. The primary endpoint of all-cause unplanned hospitalisation occurred in 108 (17.1%) and 114 (18.2%) patients in the intervention and usual care groups [adjusted OR 0.95, 95% CI 0.61–1.49; p = 0.84]. Use of the AFFIRMO Mobile App was suboptimal, with 68.5% (433/632) of patients using it for only one-third of the observation time.

Interpretation

In a setting of high baseline guideline adherence and low event rates, the iABC system did not reduce unplanned hospitalisations compared with usual care. These findings suggest no incremental benefit of mHealth integrated care interventions when applied to optimally managed populations, particularly in the context of suboptimal intervention uptake.

Funding

European Union's Horizon 2020 Research and Innovation Program (grant agreement 899871).

Research in context.

Evidence before this study

Epidemiological data show how contemporary patients with atrial fibrillation (AF) are increasingly older and more clinically complex, often with multimorbidity and frailty. This leads to a higher risk of non-cardiovascular events and increased hospitalisations.

The ‘Atrial fibrillation Better Care’ (ABC) pathway was therefore proposed as a framework to streamline implementation of an integrated care approach for optimising the guideline-adherent management of AF patients.

We searched the PubMed database for studies in English published from inception to 30th June 2026, using all fields: “Atrial fibrillation Better Care”[All Fields] AND (“randomized controlled trial ”[Publication Type] OR “randomized controlled trials as topic ”[MeSH Terms] OR “randomised clinical trial ”[All Fields] OR “randomized clinical trial ”[All Fields]) AND (“outcome ”[All Fields] OR “outcomes ”[All Fields]). Out of 14 results retrieved, we found two interventional clinical trials testing the ABC pathway in patients with AF.

Despite being limited to non-European patients, the 2 prior clinical trials demonstrated that such an intervention significantly improved guideline adherence and reduced the risk of adverse clinical outcomes, particularly hospitalisations, with consistent effects in older and multimorbid adults.

Added value of this study

The AFFIRMO trial is the first European cluster-randomised controlled trial to evaluate whether an integrated care approach based on the ABC pathway, also integrating the tools of comprehensive geriatric assessment (CGA), conveyed through a mHealth-based approach (AFFIRMO Mobile App and AFFIRMO Clinician Dashboard), would reduce the risk of unplanned all-cause hospitalisation and other major clinical events, in older adults (≥65 years) with AF and multimorbidity conducted in six European Countries.

Implications of all the available evidence

In a setting of high baseline guideline adherence and low event rates overall, the iABC system did not reduce unplanned hospitalisations compared with usual care. There was no incremental benefit of mHealth integrated care interventions when applied to already optimally managed population, with low use of the mHealth system and limited efficacy of CGA in the context of a highly preserved population.

Introduction

Atrial fibrillation (AF) is highly prevalent amongst older adults, and management is challenging and often suboptimal.¹ Consideration of various aspects associated with complexity in ageing, such as frailty, multimorbidity and polypharmacy, is needed when managing these patients, potentially impacting the clinical course.^2,3 Comprehensive geriatric assessment (CGA)⁴ tools have been advocated for the assessment and management of older adults with AF.³

Multimorbidity is associated with a higher risk of adverse clinical outcomes, more pronounced symptomatic status and is less likely to be managed with oral anticoagulant (OAC) drugs.³ Hence, more holistic or integrated approaches have been advocated by recent guidelines and patient care pathways.⁵

The Atrial Fibrillation Better Care (ABC) pathway has been proposed as a structured approach to streamline the implementation of integrated care in AF patients.⁶ Several retrospective and prospective cohort analyses have shown the positive impact of the ABC pathway adherence on reducing major clinical outcomes in the AF clinical course, including mortality, stroke and bleeding.⁷ Thus far, two independent randomised clinical trials (RCTs) conducted in Asia, which were conducted in different healthcare settings (one using an App in largely urban areas,⁸ and the second was in rural villages with village doctors and telehealth support⁹) have shown that the ABC pathway intervention was be effective in reducing the risk of clinical outcomes, compared to usual care. In both cluster randomised trials, guideline adherent holistic management was suboptimal, and the implementation of ABC pathway use improved guideline adherence and reduced adverse outcomes, including hospitalisations.^8,9 The latter has a major impact on healthcare costs, which can be mitigated by ABC pathway implementation.¹⁰

No RCT has specifically evaluated the impact of the ABC pathway in older, multimorbid patients with AF within European healthcare systems. Subgroup analyses from the two prior RCTs suggest that older AF patients and multimorbid subgroups derived benefit from optimisation of guideline therapy, leading to improved outcomes.^11,12

Therefore, the ‘Atrial Fibrillation integrated approach in FraIl, multimoRbid and polyMedicated Older people’ (AFFIRMO) trial was designed to evaluate the efficacy of a mobile-health structured clinical management intervention, adapting the ABC pathway by integrating the tools of the CGA compared to usual care, in older multimorbid European patients with AF to enhance uptake of guideline-directed holistic AF management and reduce the risk of clinical outcomes, focused on hospitalisations.⁴

Methods

Detailed descriptions of the study design and methods have been reported elsewhere.⁴ Briefly the AFFIRMO trial is a European multicentre open-label cluster-randomised study, conducted in six different European countries (Bulgaria, Denmark, Italy, Romania, Serbia, and Spain), designed to test the efficacy of a mobile-health supported intervention on optimising the clinical management of older adults with AF and multimorbidity against usual care, on the reduction of the occurrence of unplanned all-cause hospitalisation over a 12-month follow-up observation period.

Participating countries and sites selection

Participating centres, overseen by National Coordinating Centres and Coordinators, were chosen based on interest, adequacy of facilities, resources, and adequate patient volume for the study. Centres completed feasibility forms evaluated by the Central Coordinating Centre (CCC). Initially, each National Coordinator selected up to 14 sites, with final selections to ensure an even distribution of clusters across countries. In each country, after ethical approval, centres were randomised into intervention or control groups, stratified by the presence of an interventional electrophysiology facility.

Patients and inclusion/exclusion criteria

Study participants were outpatients aged 65 years or older with AF and multimorbidity, defined as having at least one other cardiovascular or non-cardiovascular chronic clinical condition.⁴ All patients were informed in detail of all study procedures and of the intervention's structure and all provided individual consent to participate in the study. Of note, the use of the AFFIRMO Mobile App could be assisted by a formal or informal caregiver.

The main exclusion criteria were the presence of a mechanical prosthetic heart valve or moderate-to-severe mitral stenosis; refusal to be enrolled or to sign the written informed consent; or inability to understand the study procedures or to attend the follow-up. Additionally, all patients enrolled in the intervention clusters were excluded if they did not have a suitable electronic device to download and run the AFFIRMO Mobile App. Inclusion and exclusion criteria are detailed in the study design paper.⁴

Study intervention

The mHealth study intervention involved the iABC system, comprising the AFFIRMO mobile app and the AFFIRMO clinician dashboard, which facilitated the implementation of the ABC pathway for each patient as per the standard definition.⁶ The app helped patients address all three ABC pillars by collecting data on vital signs, OAC adherence, AF symptoms, and other chronic conditions, while offering health tips to improve lifestyle habits and manage comorbidities based on baseline assessments. For each chronic condition at baseline, a specific questionnaire and health tips were activated in the app. During follow-up, using the AFFIRMO app, patients could also report new diagnoses, among those evaluated at baseline. If confirmed by the physician, the relevant disease-specific questionnaire was activated within the app.

Moreover, at the time of mobile app activation, each patient completed a questionnaire assessing overall patient empowerment and disease awareness, the Patient Health Engagement scale (PHE-s).¹³ Accordingly, each patient received specific, informative materials about AF and its clinical consequences, tailored to their level of disease awareness.

Furthermore, each patient enrolled by the iABC system clusters received a complete CGA at the baseline visit to identify the relevant functional dimensions that needed clinical management. Following this evaluation, specific interventions were implemented by the study physicians in order to address each functional domain found to be managed.⁴ Meanwhile, the AFFIRMO clinician dashboard gave study physicians an overview of AFFIRMO app usage and patient-entered data, organised by the ABC pathway pillars, to inform follow-up consultations and optimise management of chronic conditions. Patients followed in the centres randomised to the control arm were managed by their usual health-care providers, following international and local clinical guidelines. No restrictions are imposed on AF management.⁴

Clinical assessment and follow-up procedures

At the baseline visit, the following data were collected: demographic characteristics, vital signs, clinical history including definition of comorbidities and AF characteristics, neurosensory disabilities, ECG, relevant laboratory values, and prior and ongoing specific pharmacological and non-pharmacological treatments. Furthermore, the level of frailty was assessed at baseline using the FRAIL scale for all the patients.⁴ A complete CGA had to be performed in patients from clusters randomised to the intervention group. Also, quality of life was assessed according to the EuroQOL EQ-5D-5L questionnaire.⁴ Levels of patients’ empowerment measured with the Patient Health Engagement Scale (PHEs®)¹³ were also collected at enrolment.

Follow-up visits were conducted in both arms at 3, 6, and 12 months, during which the following data were collected: vital signs, ECG, incident clinical events or new chronic clinical conditions diagnosed since the last visit, safety profile, and prior and ongoing specific pharmacological and non-pharmacological treatments. Additionally, all patients assigned to intervention-randomised clusters received recommendations to reinforce app use at each follow-up visit. At 12-months, all baseline measures were repeated. All data were collected through a centralised electronic case report form (eCRF), which can be found published elsewhere.⁴

Study outcomes

The primary outcome of the study was the occurrence of all-cause unplanned hospitalisation over the 12-month follow-up period. As detailed in the study design paper, a list of secondary outcomes was considered.⁴ Safety assessments included monitoring and recording predefined safety aspects, including bleeding episodes, all serious adverse events (SAEs), and regular vital sign measurements.

All serious clinical events were recorded in the eCRF, and supporting documentation was required from study investigators for hospitalisations and deaths. An event validation committee (EVC), with one member in each country coordinated by the CCC, adjudicated all primary outcome events and the causes of death. To prevent assessment bias, the EVC was blinded to the participants' intervention.

Ethical and regulatory aspects

This clinical study was designed, implemented and reported following the International Conference on Harmonisation (ICH) Harmonised Tripartite Guidelines for Good Clinical Practice, with applicable local regulations (Directive 2001/20/EC—Clinical Trials Directive, Regulation EU No 536/2014- Clinical Trials Regulation) and with the ethical principles laid down in the Declaration of Helsinki. All patients’ information were managed according to the purposes of article 13 of the EU Regulation no. 679 of 2016, and the “Guidelines for Data Processing within the Framework of Clinical Drug Trials” issued by the Italian Data Protection Authority on the 24th of July 2008. The protocol, along with the study documents, was reviewed and approved in each country by the Competent Authority, if requested, and by the Central and/or Local Institutional Review Boards/Ethical Committees, in accordance with local national regulations. Hence, we do not have a unique single approval number for the study. All the information regarding country- or site-specific approval documents and numbers is retained by the CCC at Italian Heart Care Foundation (HFC) and available upon reasonable request. The study was also registered on ClinicalTrials.gov platform (NCT06775028, https://clinicaltrials.gov/study/NCT06775028).

An independent Data and Safety Monitoring Committee (DSMC) constantly oversaw and monitored the trial activities throughout the study period. No major deviations from the original protocol have been made throughout the study conduct. Furthermore, no protocol amendments have been made affecting the recruitment and conduct of the study. We slightly changed the analytical approach regarding the adjustments made to the regression analysis for the primary outcome, moving from an analysis based on variables significantly different at baseline between the study arms, to a pre-specified adjustment model, as detailed in the Statistical Methods below and available in the final Study Analysis Plan, following reviewers’ comments to the study protocol paper.⁴

Role of the funding source

The study was sponsored, promoted and managed by the HFC, and funded by the European Union's Horizon 2020 research and innovation programme under the grant agreement No. 899871 as part of the AFFIRMO project consortium activities.⁴ Horizon 2020 research and innovation programme was not involved, either directly or indirectly, in the management of the study, nor did it influence its conduct or the interpretation of the results.

Statistical methods

The sample size was determined assuming a primary endpoint rate of 30% in the usual care arm within an average follow-up duration of one year, and a 25% relative reduction with the iABC intervention (absolute decrease from 30% to 22.5%). To identify this difference with 80% power and a two-sided alpha error level of 0.05, assuming a low variation between clusters (coefficient of variance between clusters K ≤ 0.1) a total of 50 clusters with approximately 25 patients per cluster (i.e. a total of 1250 patients) were estimated as the necessary sample size.

All continuous data were reported as mean and standard deviation (SD) or median and interquartile range (IQR) and compared between groups using the t-test or Mann–Whitney U test, as appropriate. Ordinal and categorical variables were reported as counts and percentages and compared between groups using the chi-square test.

The primary analyses of efficacy and safety were conducted according to the intention-to-treat (ITT) principle. The ITT population comprised all enrolled patients within the randomised cluster, irrespective of whether they were treated according to the corresponding strategy. Two patients who withdrew consent (one in each treatment arm) were excluded from the primary endpoint and clinical secondary endpoints analyses.

A multi-level logistic regression model, incorporating random intercepts for centre and, where supported by the data, random slopes and adjusted for a set of clinically relevant baseline covariates. Covariates to be included in the regression model were prespecified a priori based on their known or plausible association with the outcome, and the stratification covariates included in the randomisation procedure were used to assess the effect of the iABC system. A model including random-slopes specification was explored but showed convergence instability (non–positive-definite Hessian), likely reflecting the limited number of clusters relative to model complexity. Therefore, the prespecified random-intercept multilevel model was retained as the primary analytical approach because it provided stable and parsimonious estimates while appropriately accounting for clustering.

Comparison of the treatment groups for the primary efficacy endpoint was reported as odds ratios (OR) and 95% confidence intervals (CI). The following variables were a priori selected to adjust the logistic regression model: i) country of enrolment and presence of an interventional electrophysiology in the enrolling centre were considered as the stratification variables included in the randomisation procedure; ii) age, gender, number of comorbidities (considering heart failure, coronary heart disease, peripheral artery disease, hypertension treated with at least 2 drugs, prior stroke or transient ischaemic attack, moderate-severe chronic kidney disease or renal dialysis, chronic obstructive pulmonary disease, diabetes mellitus, prior major bleeding, malignancy and categorised in the model as 1 vs 2 vs >2), previous hospitalisation in the <6 months, education level (<8 years vs ≥8 years vs not specified), living status (alone vs not alone), frailty categories (according to the Frail scale and categorised in the model as robust vs pre-frail vs frail). In addition, sensitivity analyses for the primary endpoint included a minimally adjusted model (including only treatment group and random effects), a Cox proportional hazards model, a multi-level logistic model for the combined endpoint of all-cause unplanned hospitalisation or death, and a multivariable multi-level negative binomial regression model for the number of unplanned hospitalisations.

For the Cox proportional hazards sensitivity analysis, robust standard errors clustered by centre were used to account for within-centre correlation. The proportional hazards assumption was assessed using Schoenfeld residuals, and no evidence of violation was observed (global test p = 0.91). All models, except the minimally adjusted, included the same covariates of the primary analysis.

Subgroup analyses were also performed according to some pre-specified groups⁴: i) age categorised according to the median value; ii) sex; iii) number of comorbidities categorised as 1 vs 2 vs >2; iv) number of medications categorised as <5 vs 5–9 vs ≥10); v) time from first AF diagnosis categorised as <6 vs 6–12 vs >12 months; and vi) country of enrolment. For each subgroup, specific treatment effects were estimated using the same modelling approach adopted for the primary analysis, and an interaction term between treatment assignment and the subgroup variable was tested.

According to the study design as published elsewhere, a set of secondary outcomes was also tested⁴: i) all-cause death; ii) any hospitalisation; iii) the composite of all-cause death and any hospitalisation; iv) non-fatal ischaemic stroke or TIA or cardiovascular (CV) death; v) new diagnosis of heart failure or heart failure worsening requiring hospitalisation; vi) new occurrence of renal dysfunction or worsening renal function; vii) major bleeding (ISTH definition); and viii) intracranial bleeding.

Missing outcome data were minimal due to rigorous follow-up procedures. Information on the primary endpoint was available for 1255 of 1260 enrolled participants (99.6%), while mortality status was available for all participants. Missing data were limited to a small proportion of baseline covariates included in adjusted analyses (educational level, living status and frailty category) and multiple imputation was not performed. Participants with missing educational information were retained through a separate “not specified” category, whereas missing values for living status and frailty category were incorporated into existing categories to preserve model stability.

In all the analyses, a two-tailed p-value <0.05 was considered statistically significant. All analyses were performed with SAS version 9.4 (SAS Institute Inc.) and R (version 4.4.1; R Foundation for Statistical Computing, Vienna, Austria) software.

Results

Overall, the study enrolled 1260 patients from April 2024 to January 2025, with 634 (50.3%) assigned to the iABC system intervention arm and 626 (49.7%) to the usual care arm. Baseline characteristics of the overall population have been previously reported elsewhere.⁴

Table 1 shows baseline characteristics according to the assigned study intervention arm. Patients in the iABC mHealth system were younger than those in the usual care arm, with broad similarities in comorbidities and risk profile. Compared to usual care arm, there was a slightly higher prevalence of first diagnosed AF, coronary heart disease and hypertension treated with at least two drugs in the iABC arm, with a lower prevalence of peripheral artery disease and malignancy.

Table 1.

Baseline characteristics according to randomised strategy.

Variables	iABC	Usual care
	N = 634	N = 626
Age, years mean (SD)	73 (6)	74 (6)
Age, years median [IQR]	73 [69–77]	74 [70–78]
Age ≥80 years, n (%)	99 (15.6)	113 (18.1)
Female sex, n (%)	265 (41.8)	295 (47.1)
Education level, n (%)
Less than high school or equivalent (<8 years)	101 (15.9)	167 (26.7)
High school graduate or equivalent (8–12 years)	310 (48.9)	255 (40.7)
College/University (>12 years)	208 (32.8)	144 (23.0)
Unknown	15 (2.4)	60 (9.6)
Living status, n (%)
Alone	129 (20.4)	149 (23.8)
With spouse	371 (58.5)	353 (56.4)
With children	45 (7.1)	48 (7.7)
With spouse & children	77 (12.2)	38 (6.1)
Nursing home	0 (0.0)	3 (0.5)
Unknown	12 (1.9)	35 (5.6)
Time from first AF diagnosis, n (%)
<6 months	143 (22.5)	106 (16.9)
6–12 months	60 (9.5)	50 (8.0)
>12 months	421 (66.4)	449 (71.7)
Unknown	10 (1.6)	21 (3.4)
First diagnosed, n (%)	96 (15.1)	70 (11.2)
Type of first diagnosed AF, n (%)
Paroxysmal	63 (65.6)	40 (57.2)
Not paroxysmal	22 (22.9)	15 (21.4)
Still under observation	11 (11.5)	15 (21.4)
Not first diagnosed, n (%)	538 (84.9)	556 (88.8)
Type of not first diagnosed AF, n (%)	239 (44.4)	230 (41.4)
Paroxysmal	120 (22.3)	114 (20.5)
Persistent	20 (3.7)	14 (2.5)
Long-standing persistent	154 (28.6)	197 (35.4)
Permanent	5 (1.0)	1 (0.2)
Unknown
Symptomatic AF, n (%)	301 (47.5)	361 (57.7)
EHRA score, n (%)
1	322 (50.8)	257 (41.1)
2a	126 (19.8)	171 (27.3)
2b	103 (16.3)	117 (18.7)
3	64 (10.1)	68 (10.9)
4	8 (1.3)	5 (0.8)
Not evaluated	11 (1.7)	8 (1.3)
CHA2DS2-VASc score, median [IQR]	4 [3–5]	4 [3–5]
HAS-BLED score, median [IQR]	1 [1–2]	1 [1–2]
Number of baseline comorbiditiesa, median [IQR]	2 [1–3]	2 [1–3]
N. of baseline comorbiditiesa, n (%)	177 (27.9)	182 (29.1)
1	193 (30.4)	191 (30.5)
2	264 (41.6)	253 (40.4)
>2
Heart failure, n (%)	267 (42.1)	252 (40.3)
Coronary heart disease, n (%)	208 (32.8)	168 (26.8)
Prior STEMI, n (%)	57 (9.0)	45 (7.2)
Prior NSTEMI, n (%)	56 (8.8)	34 (5.4)
Prior Angina, n (%)	111 (17.5)	98 (15.7)
Prior PCI, n (%)	129 (20.4)	97 (15.5)
Prior CABG, n (%)	32 (5.1)	20 (3.2)
Peripheral artery disease, n (%)	32 (5.1)	52 (8.3)
Hypertension treated with at least two drugs, n (%)	516 (81.4)	476 (76.0)
Systolic BP, mmHg mean (SD)	127 (16)	132 (17)
Prior stroke or TIA, n (%)	80 (12.6)	83 (13.3)
Prior Stroke, n (%)	67 (10.6)	65 (10.4)
Prior TIA, n (%)	25 (3.9)	19 (3.0)
Moderate-severe CKD, n (%)	96 (15.1)	108 (17.3)
Renal dialysis, n (%)	2 (0.3)	2 (0.3)
COPD, n (%)	59 (9.3)	64 (10.2)
Diabetes mellitus, n (%)	187 (29.5)	182 (29.1)
Prior major bleeding, n (%)	29 (4.6)	22 (3.5)
Malignancy, n (%)	70 (11.0)	93 (14.9)
Any previous hospitalisation (<6 Months), n (%)	237 (37.4)	225 (35.9)
Frail scaleb, median [IQR]	1 [0–2]	1 [0–2]
Frailty categoriesb, n (%)
Robust (0)	307 (48.5)	266 (43.0)
Pre-frail (1–2)	255 (40.3)	252 (40.8)
Frail (≥3)	71 (11.2)	100 (16.2)
Oral anticoagulant therapy, n (%)	617 (97.3)	609 (97.3)
VKAs	39 (6.2)	38 (6.2)
DOACs	578 (91.2)	571 (91.2)
Antiplatelet therapy, n (%)	51 (8.0)	53 (8.5)
Beta-blockers, n (%)	490 (77.3)	493 (78.8)
Anti-arrhythmic drugs, n (%)	225 (35.5)	214 (34.2)
Rate control strategy, n (%)
No	298 (47.0)	294 (47.0)
Yes	316 (49.8)	317 (50.6)
Planned	2 (0.4)	1 (0.2)
Still Under Evaluation	18 (2.8)	14 (2.2)
Rhythm control strategy, n (%)
No	261 (41.2)	337 (53.8)
Yes	334 (52.7)	250 (39.9)
Planned	19 (3.0)	20 (3.2)
Still under evaluation	20 (3.1)	19 (3.1)
Weekly physical activity, n (%)
No	104 (16.4)	178 (28.4)
Only light	342 (53.9)	268 (42.8)
Vigorous at least 20 m. 1/2 a week	86 (13.6)	73 (11.7)
Vigorous at least 20 m ≥3 times/week	99 (15.6)	109 (16.3)
Don't know/unsure	3 (0.5)	5 (0.8)
Number of drugs prescribed, n (%)
<5 (No polypharmacy)	127 (20.0)	110 (17.6)
5–9 (Polypharmacy)	435 (68.6)	477 (76.2)
≥10 (Excessive polypharmacy)	72 (11.4)	39 (6.2)

AF = Atrial Fibrillation; BP = Blood Pressure; CABG = Coronary Artery By-pass Graft; CKD = Chronic Kidney Disease; COPD = Chronic Obstructive Pulmonary Disease; DOACs = Direct Oral Anticoagulants; EHRA = European Heart Rhythm Association; IQR = Interquartile Range; NSTEMI = Non-ST Elevation Myocardial Infarction; PCI = Percutaneous Coronary Intervention; SD = Standard Deviation; STEMI = ST Elevation Myocardial Infarction; TIA = Transient Ischaemic Attack; VKA = Vitamin K Antagonists.

^aConsidering the following comorbidities: Heart Failure, Coronary Heart Disease, Peripheral Artery Disease, Hypertension treated with at least 2 drugs, Prior Stroke or TIA, Moderate-Severe CKD or Renal dialysis, COPD, Diabetes Mellitus, Prior Major Bleeding, Malignancy.

^bAvailable for 1251 patients.

Symptomatic AF and no reported weekly physical activity were more common in the usual care arm. Overall, 28.5% had one other chronic long-term condition other than AF, while 41.0% had >2 comorbidities, and a high stroke risk profile (median [IQR] CHA₂DS₂-VASc 4 [3–5]), with no difference between the two groups. Pre-frail and frail status were evident in 40.5% and 13.7%, respectively, with the proportion of ‘frail’ higher in the usual care arm at baseline (Table 1).

Baseline pharmacological treatments showed a high prevalence of anticoagulation therapy use (97.3%) overall, with similar proportions of VKAs and DOACs in both groups. Other drug therapies were broadly similar between the two arms (Table S1). Overall, polypharmacy and excessive polypharmacy were evident in 72.4% and 8.8%, respectively, with a higher proportion of excessive polypharmacy in the iABC arm. At the end of the enrolment period, there were no differences in rate control strategy between iABC and usual care arms, while a rhythm control strategy was more common in the iABC patients (52.7 vs 39.9%).

Comprehensive geriatric assessment

Overall, patients enrolled in the intervention arm exhibited a high level of autonomy in basic and instrumental activities of daily living, with a high prevalence of normal cognitive function (95.7%) (Table 2). Regarding motor function, 15.3% of patients had mild mobility impairment, while 6.5% had severe mobility impairment. A large proportion of patients had normal nutritional status (83.9%) and mood (83.6%).

Table 2.

Comprehensive geriatric assessment results in patients assigned to intervention arm.

	iABC
	N = 634
ADL, median [IQR]	6 [6–6]
IADL, median [IQR]	8 [7–8]
SPMSQ, median [IQR]	0 [0–0]
SPMSQ categories, n (%)
Normal cognition (0–2 errors)	607 (95.7)
Mild cognitive impairment (3–4 errors)	26 (4.1)
Moderate cognitive impairment (5–7 errors)	1 (0.2)
Severe cognitive impairment (8–10 errors)	0 (0)
SPPB componentsa
Balance, median [IQR]	4 [2–4]
Walking Test, median [IQR]	3 [2–4]
Sit to Stand, median [IQR]	3 [2–4]
SPPB total scorea, median [IQR]	9 [7–12]
SPPB categoriesa, n (%)
Severe mobility impairment (0–3)	39 (6.5)
Moderate mobility impairment (4–6)	92 (15.3)
Mild mobility impairment (7–9)	195 (32.4)
No mobility impairment (10–12)	275 (45.8)
MNA® total scoreb, median [IQR]	14 [12–14]
MNA® categories, n (%)
Normal nutrition (12–14)	529 (83.9)
At risk of malnutrition (8–11)	90 (14.3)
Malnourished (0–7)	11 (1.8)
GDS total score, median [IQR]	0 [0–1]
GDS categories, n (%)
Not depressed (0–1)	530 (83.6)
Depressed (2–4)	104 (16.4)

ADL = Activities of Daily Living; IADL = Instrumental Activities of Daily Living; GDS = Geriatric Depression Scale; MNA® = Mini Nutritional Assessment®; SPMSQ = Short Portable Mental Status Questionnaire; SPPB = Short Physical Performance Battery.

^aPerformed by 601 patients.

^bPerformed by 630 patients.

Engagement levels and quality of life

At baseline, patients assigned to the usual care arm had lower PHE-s scores and a higher prevalence of low psychological empowerment compared to the iABC arm (19.2% vs 14.6%). In the iABC arm, the prevalence of the eudaimonic project was higher than in the usual care arm (39.2% vs 32.2%) (Table 3).

Table 3.

Baseline patient engagement levels and quality of life by treatment arm.

PHE-s®	iABC	Usual care
	N = 615	N = 605
PHE-s® scores, mean (SD)	5.7 (1.2)	5.4 (1.3)
Psychological empowerment level, n (%)
Low	90 (14.6)	116 (19.2)
High	525 (85.4)	489 (80.8)
PHE-s® engagement positions, n (%)
Blackout	6 (1.0)	12 (2.0)
Arousal	84 (13.7)	104 (17.2)
Adhesion	284 (46.2)	294 (48.6)
Eudaimonic project	241 (39.2)	195 (32.2)

EQ-5D-5L	iABC	Usual Care
	N = 624	N = 614
Mobility, n (%)
No problems	379 (60.7)	290 (47.2)
Slight problems	132 (21.2)	145 (23.6)
Moderate problems	82 (13.1)	123 (20.0)
Severe problems	28 (4.5)	48 (7.8)
Unable to walk	3 (0.5)	8 (1.3)
Self-care, n (%)
No problems	532 (85.3)	480 (78.2)
Slight problems	65 (10.4)	82 (13.4)
Moderate problems	19 (3.0)	39 (6.4)
Severe problems	5 (0.8)	6 (1.0)
Unable to wash or dress	3 (0.5)	7 (1.1)
Usual activities, n (%)
No problems	429 (68.8)	373 (60.7)
Slight problems	132 (21.2)	112 (18.2)
Moderate problems	46 (7.4)	90 (14.7)
Severe problems	10 (1.6)	26 (4.2)
Unable to perform usual activities	6 (1.0)	13 (2.1)
Pain/discomfort, n (%)
No pain/discomfort	354 (56.7)	286 (46.6)
Slight pain/discomfort	178 (28.5)	163 (26.5)
Moderate pain/discomfort	74 (11.9)	133 (21.7)
Severe pain/discomfort	15 (2.4)	27 (4.4)
Extreme pain/discomfort	3 (0.5)	5 (0.8)
Anxiety/depression, n (%)
No anxiety/depression	380 (60.9)	330 (53.7)
Slight anxiety/depression	157 (25.2)	170 (27.7)
Moderate anxiety/depression	68 (10.9)	85 (13.8)
Severe anxiety/depression	15 (2.4)	25 (4.1)
Extreme anxiety/depression	4 (0.6)	4 (0.7)
Overall self-rated health (VAS), mean (SD)	72.3 (16.0)	69.4 (19.0)

PHE-s® = Patient Health Engagement Scale; SD = Standard Deviation; VAS = Visual Analogue Scale.

For quality of life, patients in the iABC arm demonstrated a lower prevalence of impairment in all five dimensions explored, compared to the usual care arm (Table 3). The overall self-rated health according to VAS was higher in the iABC arm than in the usual care arm (mean 72.3 ± 16.0 vs 69.4 ± 19.1, respectively; p = 0.002).

Follow-up and iABC mobile app use

After a mean (SD) follow-up of 367 days (39) [median 366, IQR 362–376], there were two withdrawals of consent, one in each arm, and no patients were lost to final visit [Fig. 1]. In the iABC arm, the median [IQR] number of days in which the patient accessed the App at least once was 56 [15–163], with the median [IQR] percentage of days over the study period in which the patient accessed the App at least once being 15.1% [4.1–44.9%]. Of note, 68.5% of patients used the AFFIRMO mobile app for only one-third of the total days they were included in the study.

Fig. 1.

Trial profile and patient disposition.

Outcomes

The primary endpoint of all-cause unplanned hospitalisation occurred in 108 (17.1%) and 114 (18.2%) patients in the iABC and usual care arms, respectively, during the 12-month follow-up, [adjusted OR 0.95, 95% CI 0.61–1.49; p = 0.84] [Table 4 and Fig. 2], with consistency on sensitivity analyses (Table S2), which showed in all the various models tested neutral effect of the study intervention. There was no significant pattern in the number of unplanned hospitalisations (Table S3), nor any statistical interactions between the prespecified subgroups [Figure S1]. Given the limited statistical power for interaction testing, these subgroup analyses should be interpreted as exploratory. There was no significant difference in the primary outcome according to the distribution of days of AFFIRMO App use (Table S4), nor were there significant differences in any of the secondary outcomes (Table 3).

Table 4.

Study Outcomes over a follow-up period of 12 months.

N (%)	iABC	Usual care	OR [95% CI]a	p
	N = 633	N = 625
Primary endpoint	108 (17.1)	114 (18.2)	0.95 [0.61–1.49]	0.84
Patients with all-cause unplanned hospitalisation
Clinical secondary endpoints
All-Cause Deathb	15 (2.4)	11 (1.8)	2.22 [0.93–5.29]	0.07
Patients with all-cause hospitalisation	163 (25.8)	180 (28.8)	0.89 [0.62–1.26]	0.51
Patients with all-cause death or hospitalisation	169 (26.7)	185 (29.6)	0.91 [0.67–1.25]	0.57
Patients with non-Fatal Ischaemic Stroke/TIA/CV Death	10 (1.6)	11 (1.8)	1.24 [0.48–3.21]	0.66
Patients with new diagnosis/worsening HF requiring hosp.	48 (7.6)	43 (6.9)	1.11 [0.65–1.91]	0.70
Patients with a new CKD/worsening CKD	68 (10.7)	52 (8.3)	1.32 [0.76–2.29]	0.33
Patients with a major bleeding (ISTH Definition)c	6 (0.9)	3 (0.5)	1.94 [0.42–9.06]	0.40
Patients with an intracranial bleeding	0	0	–	–

CI = Confidence Interval; CKD = Chronic Kidney Disease; CV = Cardiovascular; HF = Heart Failure; ISTH = International Society of Thrombosis and Haemostasis; OR = Odds Ratio; TIA = Transient Ischemic Attack.

^aAdjusted for country, presence of an interventional electrophysiology facility in the hospital of the enrolling centre, age, gender, number of comorbidities, previous hospitalisation in the <6 months, education level, living status, frailty categories.

^bDerived from a model without adjustment for country due to non-convergence of the full model.

^cDerived from the minimally adjusted model due to non-convergence of the full model.

Fig. 2.

Cumulative incidence of unplanned hospitalisations by treatment arm.

In Table S5, we reported the rate of medical treatments at the end of follow-up. We found a similar rate of OAC treatment in both the overall cohort and the two study arms. Furthermore, while generally the uptake of all the medical treatments remained high, we recorded a numerically higher rate of mineralocorticoid receptor antagonists (27.8 vs 24.5%, respectively at final visit and at baseline), lipid-lowering drugs (70.2 vs 67.7%) and SGLT2is (38.0 vs 34.5%) at the final visit, which were all more pronounced in the intervention arm. Moreover, the rate of anti-arrhythmic drugs remained higher in the iABC arm than in the control arm (34.1 vs 27.2%) (Table S5). Additionally, the rate of PPIs was higher in the intervention arm, as well as the proportion of patients with ten or more drugs prescribed (Table S5).

Regarding safety outcomes, the overall rate of major bleeding events was very low (9 events, 0.7%), with no difference between the study arms (Table 3). Regarding non-fatal SAEs which did not require hospitalisation, a total of 31 (2.5%) patients reported an event, with 39 events in total (Table S6). Among events reported more than once, the most frequent were malignancy (8 events) and minor bleeding (5 events).

As additional safety parameters, we investigated the rate of uncontrolled blood pressure (>160 mmHg) and the evidence of increased heart rate at 12-leads ECG (>100 bpm) at the end of follow-up. While no difference was found regarding the increased heart rate, we noted that patients randomised to iABC System reported a significantly lower rate of uncontrolled blood pressure (1.5 vs 4.0%, p = 0.01).

Discussion

In this prospective cluster RCT, the use of the iABC System to improve adherence to holistic or integrated care management of AF did not reduce the primary endpoint of all-cause unplanned hospitalisations or result in significant differences in any of the secondary endpoints. The results were consistent across sensitivity analyses and predefined subgroups.

The ABC pathway is a framework for optimising guideline-directed holistic or integrated care management of patients with AF. The apparent lack of benefit of the iABC System in the present study reflects that the patients in the usual care arm of AFFIRMO were very well treated with evidence-based treatments, for example 97% overall were anticoagulated at baseline. This contrasts with the mAFA trial⁸ where the baseline proportions of OAC use in the usual care and mAFA App (implementing ABC pathway) arms were 48.4% and 66.1%, respectively. At follow-up, the proportions of high-risk patients (CHA₂DS₂-VASc ≥2 in males, ≥3 in females) receiving OAC were 90.5% in the mAFA App arm, compared to a decline in the usual care arm (45.3%). In the MIRACLE-AF trial,⁹ the baseline proportions of OAC use in the intervention and usual care arms were 10.9% and 10.7%, respectively; this increased to 85.2% and 20.8%, respectively, at 12 months.

Furthermore, AFFIRMO explicitly aimed to improve overall management of all relevant comorbidities, as well as the management of the main geriatric conditions. In this context, the iABC System was designed to guide patients and physicians to obtain the best guideline-recommended medical and clinical treatment(s), in the notable absence of specific guidance for the management of older multimorbid AF patients and in the need to take account of overall levels of frailty.³

In AFFIRMO, use of the Mobile App was suboptimal, with 68.5% of patients in the lowest tertile of use. This likely reflects implementation challenges at multiple levels, including system integration, workflow compatibility, and usability (in terms of digital literacy) in an older multimorbid population. These findings align with prior studies demonstrating limited engagement with digital health tools and suboptimal long-term adherence to therapies, highlighting the importance of co-designed and context-adapted digital interventions.¹⁴ A qualitative study also highlighted the difficulties of integration into complex health systems, with many patients not using App-based rehabilitation.¹⁵ Indeed, the effectiveness of clinical decision support (CDSS) interventions depends heavily on actual use. For example, the O'CAFÉ trial found that while CDSS availability yielded only modest non-significant improvements overall, when clinicians actively engaged with the tool (26.6%), anticoagulation rates increased to 75.6% vs 65.6% without use (p < 0.05).¹⁶ In this context, at the time when the AFFIRMO trial was designed, data from the mAFA II trial clearly showed a significant reduction of the risk of hospitalisation using the mobile App implementing the ABC pathway.⁸ Another aspect to raise is related to the fact that the Mobile App is only half of the system, which is completed by the Physician Dashboard. The suboptimal use of the Mobile App, and therefore of the overall system, is a clear determinant of the study intervention effectiveness. Future studies aiming to provide easier-to-use apps will likely improve the apps' performance.

Previous studies and expert opinion have already highlighted that several actions, at different system levels (authorities, developers, researchers, healthcare professionals, patients), are still needed to remove current barriers and promote the significant adoption of m-health solutions among chronic patients.¹⁷ These include more adequate regulation, better app design and usability, more adequate studies to test the interventions, better training in use for both professionals and patients, and easier access to the various available solutions.¹⁷ How tools should be designed and targeted for AF populations with already high guideline adherence represents a ‘systems’ challenge, especially if the main intention of the tools is to improve guideline-recommended treatment adherence.

The primary endpoint event rate was substantially lower than assumed in the AFFIRMO trial protocol, which assumed a primary endpoint rate of 30% with usual care arm, and a power calculation assuming a 25% reduction in the primary endpoint with iABC System intervention to 22.5%.⁴ Instead, the ‘usual care’ arm had a primary endpoint rate of 18.2%, compared to 17.1% in the iABC arm (6.0% relative reduction of the primary endpoint). Such evidence appears unexpected, looking at the epidemiological evidence that while the progressive uptake of OAC in the management of AF progressively reduces the risk of thromboembolic events,¹⁸ the risk of unplanned hospitalisations, particularly those non-cardiovascular-related ones,¹⁹ which are not affected by the use of OAC, is progressively increasing, with even some data showing an increased risk related to the use of OAC.²⁰

Such rate of unplanned hospitalisations is so low, due to very good adherence to guidelines, that several thousand patients would be required to demonstrate a significant reduction in the endpoint with such a strategy. This illustrates how difficult it would be to achieve a significant clinical reduction in the current clinical context, given the high level of guideline-adherent management and lower-than-expected event rates in the usual care arm. Also, it is important to consider that if AFFIRMO estimates the intervention treatment effect in a context of high guideline adherence and low residual risk, under these conditions, even interventions with already proven efficacy in lower-quality-care settings are unlikely to demonstrate clinically meaningful incremental benefit.

The ABC pathway was recommended in the 2020 ESC guidelines,²¹ together with the 4S-AF framework, to evaluate, confirm and characterise AF based on stroke risk, symptoms, severity of AF burden, and substrate severity.²¹ Given the extensive wide publicity and implementation efforts directed at European clinicians, this may reflect the high uptake of guideline-adherent management of AF in this trial. Indeed, the low primary endpoint rate is reflective of the usual care arm being very well managed and adherent to contemporary guidelines.

Implementation efforts to improve the quality of AF detection and management are evident from other healthcare settings.²² For example, the implementation of an outpatient clinical decision support system increased OAC prescription rates from 47.3% to 67.5%, with DOAC use rising from 34.6% to 64.4%, which was associated with significantly reduced major bleeding events.²³ In the IMPACT-AF trial, education and review of anticoagulation appropriateness resulted in a significant increase in the proportion of patients treated with OAC and a reduction in stroke rates in the intervention vs control group.²⁴ All these implementation efforts may have led to the overall decline in adverse event rates associated with this common arrhythmia over the last decade²⁵ in the six European countries participating in AFFIRMO, despite the diverse healthcare settings, while also the national AF guidelines in each country were based on the 2020 ESC guidelines, which were advocated when the AFFIRMO trial was designed. The 2024 ESC guidelines¹ introduced the AF-CARE framework, which is largely consistent with the ABC pathway components. Our results also align with the results of the recent STEEER-AF trial, in which a strategy of implementing ESC guidelines through an electronic-based education programme yielded only marginal improvements in the primary and secondary outcomes, given the overall high uptake of guideline recommendations at baseline in both the intervention and control groups.²⁶ Nevertheless, contamination between clusters and temporal Hawthorne effects cannot be excluded. Awareness of participation in an AF-focused trial and widespread dissemination of ESC guidelines may have improved care in usual care clusters, further diminishing detectable between-group differences. Additionally, high implementation of guideline-directed treatments can also be identified for other comorbidities evaluated at baseline. Indeed, approximately 40% of patients in each group were diagnosed with heart failure and looking at SGLT2is drugs, we found that approximately 32% of patients in each group were treated with these drugs. Similarly, if approximately 29% of patients were diagnosed with diabetes mellitus, approximately 25% of patients were treated with antidiabetic therapy.

Moreover, our descriptive data regarding the prescription of medical treatments at the end of follow-up observation suggests a numerically higher uptake of guideline-directed treatments in the iABC group, which, together with the evidence of a significantly lower prevalence of uncontrolled systolic blood pressure, may indicate some secondary effectiveness of the iABC System in improving the overall quality of medical management.

Another aspect that could have limited the efficacy of the AFFIRMO intervention on the primary endpoint is related to the potential limited impact of the CGA. As highlighted elsewhere,⁴ one of the aspects of the AFFIRMO trial is the implementation of CGA, as a multidimensional management strategy of main geriatric conditions, alongside the ABC pathway. Our results showed that despite more than half of patients exhibiting some degree of frailty, patients enrolled in the iABC arm reported low functional impairment, with just 20% reporting a significant reduction in mobility function. This distribution constrained the potential impact of CGA-driven interventions, as most domains exhibited ceiling effects at baseline. In such functionally preserved individuals, CGA may serve primarily a descriptive rather than an interventional role, which likely contributed to the absence of measurable clinical benefit. Indeed, the impact of CGA in robust patients shows little to no benefit.²⁷ These data also align with a clear ceiling effect of some CGA tools in high-functioning individuals.²⁸ Considering the results of the study and the unexpectedly low levels of functional impairment, such a ceiling effect represents a design limitation of the study.

Clearly, the successful implementation of holistic management will very much depend on the actual utilisation of tools (including Apps), the organisational context, and the addressing of multilevel barriers at the physician and patient levels, as well as on shared decision-making.²⁹ Notwithstanding these aspects, the higher uptake of some interventions, such as the implementation of rhythm control strategies, which was higher in the intervention group, both at baseline and at the end of follow-up observation, reflects the effectiveness of the iABC System to determine a proactive approach towards more complex AF patients. On the contrary, the trial had high completeness of follow-up data and event validation procedures, which allows for providing solid evidence to contribute to the current discussion about the implementation of integrated care in the management of AF patients. Also, compared to other studies using mHealth solutions in AF patients, the AFFIRMO trial provides solid evidence coming from a well-implemented RCT which differs from previous studies, which were often of low quality, with low numbers of patients and using simple interventions at a low level(s) of complexity.³⁰ The results of the AFFIRMO trial still do not disqualify the possibility that further optimisation of medical care is possible with the final aim of further improving clinical outcomes. The lack of efficacy we found in this trial is attributable to the various factors we have identified to explain the results, which—when combined—are likely to multiply in effect rather than simply add. A future study, maybe focused on a more selected population with a higher baseline risk, could better elucidate the potential impact of the iABC System.

Conclusion

In older multimorbid patients with AF, use of the mHealth iABC System did not reduce unplanned hospitalisations or secondary outcomes compared with usual care. These findings likely reflect high baseline guideline adherence, lower-than-expected event rates, suboptimal App utilisation, as well as limited scope for CGA-driven improvement in a largely functionally preserved population. AFFIRMO highlights the importance of targeting integrated care interventions to settings with greater unmet clinical need, with population-centric needs including use of mHealth or digital solutions or additional assessments such as CGA.

Contributors

Conceptualisation: All authors; Data Curation: DL, GF; Formal Analysis: GYHL, MP, DL, GF, SPJ, and APM; Funding Acquisition: GYHL, JA, GAD, GG, DAL, DL, GF, FM, MoF, MPc, TSP, RP, MT, DLV, SPJ, and APM; Investigation: MP, GAD, LF, GG, DAL, DL, GF, FM, MoF, MPc, TSP, RP, MT, and DLV; Methodology: GYHL, MP, JA, GG, DAL, DL, GF, MoF, MPc, DLV, SPJ, and APM; Project Administration: SPJ, APM; Software: JA, AS; Supervision: GYHL, MP, GG, DAL, DL, GF, MoF, MPc, DLV, SPJ, and APM; Writing–Original Draft: GYHL, MP, SPJ, and APM; Writing–Review & Editing: JA, GAD, LF, GG, DAL, DL, GF, FM, MoF, MPc, TSP, RP, AS, MT, and DLV. All the coauthors read and approved the final version of the manuscript. GYHL, MP, DL, GF, SPJ, and APM had full access to the data and verified all the underlying data. All the members of the AFFIRMO Consortium significantly contributed to the construction of the scientific background and conceptualisation of the clinical trial.

Data sharing statement

The data that support the findings of this study are available from the corresponding authors upon reasonable request and with permission from the Consortium/Steering Committee.

Declaration of interests

GYHL: Institutional consultancy and/or speaker roles for BMS/Pfizer, Boehringer Ingelheim, Anthos, Huawei (No fees are received personally). He is a National Institute for Health and Care Research (NIHR) Senior Investigator Emeritus and co-PI of the TARGET project on digital twins for personalised management of atrial fibrillation and stroke (grant agreement No 101136244) and ARISTOTELES project on artificial intelligence for management of chronic long term conditions (grant agreement No 101080189), which are all funded by the EU's Horizon Europe Research & Innovation programme; MP: Speaker activity for Pfizer and BMS/JJ Alliance and consultant activity for Regeneron Pharmaceuticals; GG: small speaker fees from Abbvie, Sanofi, Pfizer; research grants from Chiesi, Sanofi; consulting fees from Takeda; DAL: has received investigator-initiated educational grants from Bristol-Myers Squibb and Pfizer, outside the submitted work. She is a co-applicant on the ARISTOTELES project on artificial intelligence for management of chronic long-term conditions (Grant Agreement No. 101080189), and the TARGET project on digital twins for personalised management of atrial fibrillation and stroke (Grant Agreement No. 101136244), all of which are funded by the European Union Horizon Europe Research & Innovation programme; FM: Consultant and speaker for Boehringer Ingelheim, Bayer, Novartis, Daiichi Sankyo, and BMS/Pfizer; APM: personal fees for the participation in committees of clinical trial sponsored by Bayer, Astra Zeneca, Sanofi, Novo Nordisk, outside the present work. All the other authors have nothing to declare.