Multidisciplinary team for antithrombotics management and clinical outcomes in older adults with atrial fibrillation: a target trial emulation

Abstract

Background

Recent international guidelines emphasize a multidisciplinary, patient-centered approach to managing atrial fibrillation (AF), particularly regarding antithrombotic (antiplatelet and anticoagulant) management. These guidelines advocate establishing multidisciplinary AF teams, but the clinical benefits of this approach for high-risk, clinically complex subgroups—particularly among very old and frail patients—remain uncertain. Our objective was to evaluate the impact of a hospital multidisciplinary team meeting dedicated to antithrombotics management on a composite measure of all-cause death, major thromboembolic events, or major or clinically relevant bleeding within 6 months in older adults with AF.

Methods

A prospective multicenter cohort study was conducted in five acute geriatric departments in the Paris area between May 2021 and January 2024. Using a target trial emulation approach (cloning, censoring, weighting strategies), outcomes were analyzed in patients aged ≥ 75 years with AF or atrial flutter, followed in the geriatric departments (via outpatient consultation or hospitalization; > 98% were hospitalized). Participants were followed for 6 months or until death. The primary exposure was a hospital multidisciplinary team meeting within 45 days of inclusion, involving geriatricians, cardiologists, neurovascular specialists, and hemostasis experts. Cumulative incidences were estimated using the reverse Kaplan–Meier method. 

Results

The study included 818 patients, 138 (16.9%) in the hospital multidisciplinary-team meeting arm (median age 89 (Q1Q3 84–93), 57% female). The 6-month cumulative incidence of the primary composite outcome was 35.3% (95% CI, 29.6 to 41.8) in the multidisciplinary-team meeting arm and 36.2% (95% CI, 32.2 to 40.1) in the control arm (risk difference − 0.9 (95% CI, − 7.5 to 6.0); p = .79). The 2 arms did not differ in individual events within the composite measure.

Conclusions

A hospital multidisciplinary team meeting dedicated to antithrombotics management in older adults with AF was not associated with a reduction in all-cause death, major thromboembolic events, or major or clinically relevant bleeding within 6 months. These findings should be interpreted with caution due to the observational design and potential for residual confounding.

Trial registration

ClinicalTrials.gov registration: NCT04932603. 

Supplementary Information

The online version contains supplementary material available at 10.1186/s12916-025-04291-9.

Background

Atrial fibrillation (AF) is the most common sustained arrhythmia worldwide, with a rising prevalence that affects approximately 10% of individuals over the age of 80 [1, 2]. Compared to two decades ago, contemporary patients with AF present with increasingly complex clinical profiles, marked by a higher burden of comorbidities at diagnosis and rising overall incidence and prevalence of the condition [1, 2]. 

To address the multifaceted healthcare needs of patients with AF, international guidelines have endorsed a holistic, integrated care approach. Initially conceptualized through the Atrial Fibrillation Better Care (ABC) pathway, this strategy has evolved into the AF-CARE model, which is now recommended in the most recent guidelines as the updated framework for comprehensive AF management [3–5]. These frameworks emphasize three core pillars of management: (A) avoid stroke and thromboembolism (B), better symptom control through rate or rhythm management, and (C) cardiovascular and comorbidity risk factor management. Adherence to the ABC pathway has been associated with improved clinical outcomes in observational real-world studies [2–4], findings that have been corroborated by the prospective, randomized mAFA-II trial [6], a second cluster randomized trial (MIRACLE-AF) [7], and a recent systematic review and meta-analysis [8].

Despite these encouraging results, there remains a critical evidence gap regarding the effectiveness of the “ABC pathway” in high-risk, clinically complex subgroups—particularly among very old and frail patients. Patients described as “complex” in clinical trials or meta-analyses often differ substantially from those encountered in geriatric practice. For instance, in the meta-analysis by Romiti et al., which included over 285,000 patients, the median age ranged from 56.7 to 75.2 years, suggesting limited representation of the oldest and most vulnerable populations.

While managing comorbidities is a core competency in geriatric care, decision-making around anticoagulation in older adults is often substantially more nuanced. In frail older patients, overlapping risk factors for thromboembolic and bleeding events—such as advanced age, hypertension, chronic kidney disease, polypharmacy, and poor treatment adherence—make therapeutic decisions particularly challenging and highlight the need for truly individualized prescribing strategies [9, 10].

To address these complexities, five geriatric departments in the Paris area, in collaboration with cardiology and neurovascular teams, established a hospital-based multidisciplinary team meeting dedicated to antithrombotics management for older adults with AF in 2016. This meeting was specifically designed to guide the management of antithrombotic therapy in older, clinically complex adults with AF. This study aimed to evaluate the impact of this hospital multidisciplinary team meeting on a composite outcome including all-cause mortality, major thromboembolic events, or major or clinically relevant bleeding in patients aged 75 years and older with AF.

Methods

General considerations and ethics

This study used the target trial emulation framework, a method to estimate associations as close as possible to causal effects of interventions by using observational data [11]. Following a 2-step strategy, we designed a target trial (i.e., a randomized experiment that could theoretically answer the causal question), then the trial was explicitly emulated with observational data (see Additional file 1: Table 1 [12]). The study was registered at ClinicalTrials.gov, and a statistical analysis plan was written before database locking and registered on the OSF website [13]. The reporting of the study complied with the STROBE statement[14] (see Additional file 2: Table 2).

Table 1.

Baseline characteristics of the study population with and without antithrombotics multidisciplinary team (AMT) discussion

Characteristics	Total cohort N = 818	AMT within 45 days from admission N = 138	No AMT within 45 days from admission N = 680	Absolute standardized mean difference
Age, years, n(%)
75–84	225 (27)	37 (27)	188 (28)	.06
85–89	295 (36)	53 (38)	242 (36)
≥ 90	298 (36)	48 (35)	250 (37)
Age, years
Median (Q1Q3)	89.0 (84.0–92.8)	88.0 (84.0–92.0)	89.0 (84.0–93.0)	.001
Female, n (%)	463 (57)	78 (56)	385 (57)	.002
Hypertension, n(%)	583 (71)	104 (75)	479 (70)	.11
Diabetes, n(%)	199 (24)	34 (25)	165 (24)	.009
BMI, kg/m2
Median (Q1Q3)	24.0 (21.2–27.3)	23.7 (20.6–26.7)	24.1 (21.3–27.5)	.21
NA	210 (26)	20 (14)	190 (28)
Coronary artery disease, n (%)	233 (28)	34 (25)	199 (29)	.10
AF already known, n(%)b	708 (87)	120 (87)	588 (86)	.01
Valvular AF, n(%)c	15 (2)	2 (1)	13 (2)	.04
Heart failure, n (%)	163 (19)	36 (26)	127 (19)	.18
Peripheral arterial pathology, n (%)	294 (36)	59 (43)	235 (35)	.17
Thromboembolic venous disease, n (%)	138 (17)	24 (17)	114 (17)	.02
Stroke or TIA, n (%)	333 (41)	77 (56)	256 (38)	.37
Renal insufficiency, n (%)d
No	585 (71)	100 (72)	485 (71)	.11
Yes	192 (23)	35 (25)	157 (23)
Dialysis	9 (1)	3 (2)	6 (1)
NA	32 (4)	0 (0)	32 (5)
Dementia, n (%)e	412 (50)	66 (48)	346 (51)	.06
Solid cancer, n (%)	188 (23)	37 (27)	151 (22)	.11
CCI
Median (Q1Q3)	7 (6–8)	7 (6–8)	7 (6–7)	.04
HASBLED score
Median (Q1Q3)	2 (2–3)	3 (2–4)	2 (2–3)	.52
CHA2DS2-VASC score
Median (Q1Q3)	5 (4–6)	5 (4–6)	5 (4–6)	.35
Previous bleeding, n (%)f	328 (40)	104 (75)	224 (33)	.94
Anticoagulant accidentg	280 (34)	86 (62)	194 (28)	.72
Anticoagulant use, n (%)	573 (70)	81 (59)	573 (70)	.29
DOAC	478 (58)	63 (47)	415 (61)	.31
VKA	76 (9)	11 (8)	65 (10)	.06
LMWH	19 (2)	7 (5)	12 (2)	.18
Antiplatelet use, n (%)	137 (17)	30 (22)	107 (16)	.15
Combination of antithrombotics, n (%)h	44 (5)	7 (5)	37 (5)	.02
Total number of long-term medications taken per day
Median (Q1Q3)	9 (6–11)	9 (6–11)	9 (6–11)	.06
ADL score
Median (Q1Q3)	5.0 (3.5–6.0)	5.0 (3.5–6.0)	5.0 (3.5–5.5)	.08
NA, n (%)	8 (1)	2 (1)	6 (1)
Lifestyle, n (%)
Lives alone	393 (48)	59 (43)	334 (49)	.15
Married/cohabitating	328 (40)	64 (46)	264 (39)
Institution	97 (12)	15 (11)	82 (12)
Walking, n (%)
Autonomous	307 (37)	64 (46)	243 (36)	.27
Cane	225 (27)	29 (21)	196 (29)
Walker	177 (22)	32 (23)	145 (21)
Does not walk	109 (13)	13 (9)	96 (14)
Repeated falls, n (%)i	247 (30)	43 (31)	204 (30)	.02

In the absence of the “NA” (not available) field, the variable contains no missing data

Abbreviations: Q1Q3 first and third quartile, BMI body mass index, AF atrial fibrillation, PE pulmonary embolism, DVT deep vein thrombosis, TIA transient ischemic attack, CCI Charlson Comorbidity Index, HASBLED Hypertension, Abnormal renal and liver function, Stroke, Bleeding, Labile international normalized ratio, Elderly, Drugs or alcohol, CHA2DS2-VASc, Congestive heart failure, Hypertension, Age ≥ 75 years, Diabetes mellitus, prior stroke or thromboembolism, Vascular disease, Age 65–74 years, Sex, DOACdirect oral anticoagulant; VKA, vitamin K antagonist, LMWH low molecular weight heparin, ADL activities of daily living, TIA transient ischemic attack

^aPercentages may not total 100% because of rounding

^bBefore the start of the hospital stay

^cOnly if severe or “very tight” or described as such or with specific follow-up, or presence of a prosthesis

^dWith a Cockcroft glomerular filtration rate (GFR) < 30 ml/min. Creatinine is the reference creatinine, noted in the history, or the discharge creatinine after resolution of acute problems

^eMini-Mental State Examination score ≤ 24 performed outside an acute context or notion of “cognitive disorders” or “memory disorders” noted in previous medical reports

^fSee Supplemental eTable 3

^gMajor bleeding occurring while the patient is on anticoagulant therapy

^hTwo antiplatelet drugs with or without an anticoagulant, or the combination of an antiplatelet and anticoagulant drug

ⁱ ≥ 2 per year

For more details, see Supplemental eTable 3

This study was approved by the Institutional Ethics Committee on May 25, 2021 (no. CER-2021–026). All enrolled patients or their legal representative gave their non-opposition for the use of their medical data.

Setting and participants

The study was conducted in 5 acute geriatric wards in the Paris area (see Additional file 3: Table 3). From May 2021 to January 2024, consecutive patients ≥ 75 years old with confirmed AF or atrial flutter [5, 15], whether pre-existing or newly diagnosed, were screened and included prospectively after providing informed consent. Patients who did not meet eligibility criteria or declined the use of their data were not included. All patients were followed from the date of consent and eligibility (at admission or during hospitalization for later AF diagnosis) for up to 6 months.

The hospital multidisciplinary team meeting dedicated to antithrombotics management

The hospital multidisciplinary team meeting dedicated to antithrombotics management was designed to address complex therapeutic decisions in older patients with AF, either in-person or remotely. It was implemented as part of routine clinical practice before the study was conceived. These meetings involved geriatricians, cardiologists, neurovascular specialists, and hemostasis experts, with at least one member affiliated with a professional society. Attendance was voluntary and unpaid. The decision to present a patient’s antithrombotics treatment at these meetings was at the discretion of the referring geriatrician.

Before each meeting, the geriatrician completed a standardized form that was developed in agreement with all experts and highlighted key factors influencing antithrombotics treatment decisions. From a literature review (primarily European and American guidelines) [5, 15, 16], the form included details on age, sex, comorbidities, medications, cognitive and functional status (Mini-Mental State Examination score [17], activities of daily living score [18], mobility level and fall frequency), nutritional status (weight, albumin level), renal function (Cockcroft-Gault clearance), test results (e.g., cardiac echocardiography, brain magnetic resonance imaging), thromboembolic and hemorrhagic risk scores (e.g., CHA2DS2-VASc and HAS-BLED, recommended at that time) [16], a brief clinical history (including prior bleeding or thromboembolic events), and the main clinical issue for discussion.

At the hospital multidisciplinary team meeting, joint decisions were made about the antithrombotics regimen based on current guidelines and expert opinion, specifying drug type(s), dosage(s), duration(s), monitoring, and follow-up plans. The referring geriatrician was responsible for implementing the team’s decision and informing the general practitioner via a discharge letter. All recommendations were documented in the patient’s electronic medical record.

In our analysis, patients that were discussed during the meeting were classified in the intervention group (see below for the statistical analysis). No patients in the control group received any component of the antithrombotics multidisciplinary team intervention.

Objectives and outcomes

The objective was to evaluate whether a hospital multidisciplinary team meeting dedicated to antithrombotics management in older adults with AF was associated with the occurrence of all-cause death, major thromboembolic events, or major or clinically relevant bleeding (whichever occurred first). Major thromboembolic events included acute coronary syndromes, ischemic stroke, transient ischemic attacks, arterial embolism, and venous thromboembolic disease such as deep vein thrombosis and pulmonary embolism. Major or clinically relevant bleeding events were defined according to the International Society on Thrombosis and Haemostasis criteria [19] (details in Additional file 4: Table 4).

The primary outcome was the composite major thromboembolic event, major or clinically relevant bleeding events, or all-cause death measured at 6 months. Secondary outcomes included the same composite outcome measured at 3 months and each individual event within the composite at 3 and 6 months. Subgroup analyses for the primary outcome were conducted according to age at hospitalization and history of AF. Adherence to the hospital multidisciplinary team’s recommendations (e.g., treatments, dosage) at 3 and 6 months was also assessed in the intervention arm.

Data collection and management

All data were collected prospectively by using an electronic case report form that was tested and validated by all investigators (see Additional file 4: Table 4). Follow-up was conducted by medical consultation, with a detailed, dated report available in the patient’s chart, or by a phone call to the patient’s general practitioner or referring physician at 3 and 6 months. Particular attention was paid to ongoing treatments and adherence to the hospital multidisciplinary team’s recommendations, if applicable.

Participants were followed from inclusion until death, lost to follow-up, or 6 months (administrative censoring), whichever occurred first. For patients lost to follow-up, vital status was obtained from the French National Institute of Statistics and Economic Studies registry.

Statistical analysis

Cloning, censoring, and weighting were used to emulate the target trial, with a grace period of 45 days to receive the hospital multidisciplinary-team meeting recommendations after hospital admission, as proposed elsewhere [20, 21]. Patients who died during the 45-day grace period without initiating the intervention were not excluded; instead, their deaths were counted as events in both treatment arms, since their data remained compatible with both strategies and were never censored post-cloning. This approach is illustrated in Additional file 5: Fig. 1 (patient #4). This allowed for simultaneously accounting for baseline confounding and immortal time bias at the same time. First, we duplicated each patient’s observation in the cohort (cloning): one clone was assigned to the hospital multidisciplinary-team meeting arm (intervention arm) and the other clone to the control arm. Therefore, the clones’ baseline characteristics were identical, with balanced arm sizes, similar to a 1:1 randomized experiment. Indeed, in the cloning, censoring, and weighting framework, each patient is assigned to both strategies at baseline through cloning, ensuring identical groups at baseline for measured but also unmeasured confounders. Second, we censored the clones if they deviated from their assigned strategy. If a patient had not received recommendations from the hospital multidisciplinary team meeting, its clone assigned to the intervention arm was censored at the end of the grace period. Similarly, if a patient received recommendations from the hospital multidisciplinary team, its clone assigned to the control arm was censored on the day of the meeting. Events after the artificial censoring date were not considered. To account for informative censoring introduced by design, we used a strategy of inverse probability of censoring weighting [22] for each clone in the final analysis. No patients in the control group received any component of the antithrombotic multidisciplinary team intervention. The cloning, censoring, and weighting approach we used is specifically tailored to estimate this per-protocol estimand, as clones are censored when they deviate from their assigned strategies, and it is not compatible with an intention-to-treat framework (see Additional file 6: Methods–Additional information). 

Fig. 1.

Detailed study diagram. Abbreviations: AT, antithrombotics; AF, atrial fibrillation

Finally, cumulative incidences for each arm were estimated by using the inverse-weighted Kaplan–Meier estimator, based on the previously computed weights. Absolute risk differences between the arms at 3 and 6 months were estimated from these curves. To account for correlation between patient observations, we computed robust variances with 1000 bootstrap replicates.

The same analytical strategy was used with the time-to-event secondary outcomes. Compliance with hospital multidisciplinary team recommendations was defined as the proportion of participants who adhered to the recommended treatment regimen after the hospital multidisciplinary team meeting in the intervention arm.

Subgroup analyses of the primary outcome were performed according to age (75–84, 85–90, ≥ 90 years) and history of AF (recent-onset vs pre-existing before hospitalization).

Missing data

We used multiple imputations with chained Eqs. (30 imputed datasets) in all analyses to impute missing covariables (confounders and prognostic factors). Regarding time-to-event outcomes, the full procedure of cloning, censoring, and weighting was applied to each imputed dataset, and the estimators of intervention effects were averaged across all datasets following Rubin’s rule [23]. Regarding the secondary outcome adherence rate, participants who died on the recommended treatment were considered adherent. All results are presented after imputation of missing variables.

Sensitivity analyses

Sensitivity analyses included a modification of the grace period to 30 and 60 days, a complete-cases analysis, and use of a propensity score to account for confounding at baseline with a landmark time of 45 days [24] (see Additional file 6: Methods–Additional information).

All tests were two-tailed with a 0.05 type 1 error rate. We did not correct for multiple comparisons regarding secondary outcomes, which remain exploratory. Analyses involved using R v4.3.3 with, in particular, the mice (3.16.0), survival (3.7–0, Hmisc (5.1–1), tidyverse (2.0.0) and dplyr (1.1.4) packages. The other packages used were readr, gtsummary, survival, labelled, rlang, lubridate, simstudy, Table 1, smd, flextable, officer, ggmice, finalfit, cobalt, MatchThem, ggplot2, survminer, boot, doFuture, survey, and GenBinomApps.

Results

Characteristics of the study population

We included 818 patients (Fig. 1); the median age was 89 years (Q1Q3 84–93), and 463 (57%) were women (Table 1, end of the document). Patients had multiple comorbidities (median Charlson Comorbidity Index: 7 (6–7)) and were taking multiple medications at baseline (median number of medications: 9 (6–11)). The median activities of daily life score was 5.0 (3.5–6.0), and 11.9% of patients were in an institution. We included 138 patients (16.9%) in the intervention arm and 680 (83.1%) in the control arm. The median follow-up was 182 days.

Patient characteristics before cloning are described in Table 1. As compared with the control arm, the intervention arm had a higher median HAS-BLED score (3 (2–4) vs 2 (2–3), standardized mean difference (SMD) = 0.519), a more frequent bleeding history (75.4% vs 22.9%, SMD = 0.941), and more anticoagulant-related complications (62.3% vs 28.5%, SMD = 0.721). In addition, they had a higher prevalence of prior ischemic stroke or transient ischemic attack (55.8% vs. 37.6%, SMD = 0.370).

Primary outcome

A total of 285 primary outcome events occurred (see Additional file 7: Table 5 and Additional file 8: Table 6). The 30-day mortality rate was 10.5%, and the 6-month rate was 30.3% in the overall population. In the same population at 6 months, the incidence of thromboembolic events was 2.8%, ischemic stroke or transient ischemic attack 1.5%, and major or clinically relevant bleeding 8.4%.

The 2 arms did not significantly differ in the primary outcome: the cumulative incidence of death, major thromboembolic, major or clinically relevant bleeding events within 6 months was 35.3% (95% CI, 29.6 to 41.8) in the intervention arm versus 36.2% (95% CI, 32.3 to 40.1) in the control arm (risk difference − 0.9 (95% CI, − 7.5 to 6.0; p = 0.79) (Table 2 and Fig. 2).

Table 2.

Cumulative incidences and risk differences for primary and secondary outcomes

	3-month cumulative incidence, % (95% CI)	p-value	6-month cumulative incidence, % (95% CI)	p-value
Major or clinically relevant bleeding events, major thromboembolic event or death
AMT arm	24.6 (20.3–29.5)	.73	35.3 (29.6–41.8)	.79
Control arm	25.4 (22.1–28.9)		36.2 (32.3–40.1)
Difference	− 0.8 (− 5.1–4.4)		− 0.9 (− 7.5–6.0)
Death
AMT arm	22.2 (17.7–27.1)	.62	30.2 (24.2–36.9)	.66
Control arm	23.4 (20.1–26.7)		31.7 (28.2–35.4)
Difference	− 1.2 (− 5.6–4.0)		− 1.5 (− 7.5–5.8)
Major or clinically relevant bleeding events
AMT arm	5.0 (2.6–8.0)	.24	12.7 (7.5–18.4)	.06
Control arm	3.4 (2.2–7.4)		7.3 (5.3–9.6)
Difference	1.6 (− 1.0–4.4)		5.4 (0.4–11.6)
Major thromboembolic event
AMT arm	2.2 (0.9–4.0)		3.2(1.2–5.6)	.47
Control arm	1.3 (0.6–2.4)		2.4 (1.3–3.7)
Difference	0.9 (− 0.2–2.3)	.16	0.8 (− 1.3–3.4)

This table shows cumulative incidences and risk differences for primary and secondary outcomes according to intervention arms and risk differences between arms at 3 and 6 months

All results are presented after multiple imputation of missing variables

Abbreviations: CI confidence interval, AMT antithrombotics multidisciplinary team

Fig. 2.

Weighted cumulative incidence curves for the primary outcome by study arm. The light color area represents the 95% confidence intervals of the curves. All results are presented after multiple imputation of missing variables

Secondary outcomes

At 3 months, the 2 arms did not significantly differ in the composite outcome: 24.6% (95% CI, 20.3 to 29.5) in the intervention arm versus 25.4% (95% CI, 22.1 to 28.9) in the control arm (risk difference − 0.8 (95% CI, − 5.1 to 4.4); p = 0.73). The 2 arms did not differ at 3 and 6 months in separate analyses of death, major thromboembolic events, or major or clinically relevant bleeding events (Table 2; Additional file 9: Fig. 2, Additional file 10: Fig. 3, Additional file 11: Fig. 4). Although not statistically significant, the bleeding outcomes, particularly at 6 months, show a notable absolute risk difference (5.4%, 95% CI 0.4–11.6; p = 0.06, Table 2).

Fig. 3.

Weighted cumulative incidences and absolute risk differences of the primary outcome across various subgroups. In the “history of AF” subgroup, “recent” means that AF was discovered during the hospital stay, and “known” means that it was already known before hospitalization, based on objective evidence. All results are presented after multiple imputation of missing variables. Abbreviations: AMT, antithrombotics multidisciplinary team; Cum. Incidence, cumulative incidence; CI, confidence interval

Adherence to the hospital multidisciplinary team’s recommendations regarding anticoagulant type and dose was 83.1% at 3 months and 87.3% at 6 months (Additional file 12: Table 7).

Subgroup analyses did not reveal any significant differences by age group or history of AF diagnosis (Fig. 3; Additional file 13: Fig. 5, Additional file 14: Fig. 6) and all sensitivity analyses yielded similar results (see Additional file 15: Table 8, Additional file 16: Table 9, Additional file 17: Table 10; Additional file 18: Fig. 7, Additional file 19: Fig. 8, Additional file 20: Fig. 9, Additional file 21: Fig. 10).

Discussion

This study, which aimed to assess the impact of a hospital multidisciplinary team meeting dedicated to antithrombotics management on clinical outcomes in older adults with AF compared to usual care, found no significant differences between the arms in the occurrence of all-cause death, major thromboembolic events, or major or clinically relevant bleeding at 6 months, whether assessed as a composite outcome or as separate events. These findings were observed despite high adherence to multidisciplinary team recommendations.

To our knowledge, no prior study has evaluated multidisciplinary assessments of complex cases to determine the optimal therapeutic strategy regarding antithrombotics for frail, older adults with AF. Romiti et al. recently investigated the impact of adherence to the ABC pathway in a contemporary cohort of clinically complex AF patients [25]. Their analysis drew upon data from the ESC-EHRA EURObservational Research Programme (EORP) Atrial Fibrillation General Long-Term Registry—a large, prospective, multicenter observational registry conducted across 250 cardiology centers in 27 countries. Among the 9966 AF patients enrolled, 8289 (83.1%) were classified as clinically complex. In this subgroup, adherence to the ABC pathway was associated with significant reductions in the risks of all-cause mortality (adjusted hazard ratio (aHR): 0.72, 95% CI: 0.58–0.91) and major adverse cardiovascular events (MACE; aHR: 0.68, 95% CI: 0.52–0.87). However, in that study, the median age of clinically complex patients was 72 years (IQR: 64–78), compared to 89.0 years (84.0–92.8) in our population. Only 109 patients (1.3%) had major neurocognitive disorders, compared to 412 patients (50%) in our cohort. Even among those classified as having “high clinical complexity,” the median age was 74 years (67–79), with a median daily medication count of 6 (5–7), notably lower than the 9 (6–11) observed in our population. These findings highlight the substantial gap between the “clinically complex” patients included in large cardiology-led cohorts and the highly vulnerable patients typically encountered in geriatric settings.

Our cohort had a particularly high mortality rate. The 1- and 6-month mortality rates were 10.5% and 30.3%, respectively, which is consistent with previous studies. For example, in a cohort of 1808 patients with AF who were ≥ 75 years old and hospitalized in the acute geriatric unit of the University Hospital of Pisa, the 10-month mortality rate was 45.2% [26]. The risks of ischemic stroke and bleeding were elevated in our population, with 6-month thromboembolic and major or clinically relevant bleeding rates of 2.8% and 8.4%, respectively, including a 1.5% risk of ischemic stroke or transient ischemic attack, higher than previously reported in the literature. In a cohort study of more than 30,000 older Japanese patients with AF (mean age 81.5 ± 4.8 years) from the All Nippon AF in the Elderly (ANAFIE) registry, the 2-year incidence rates were 3.01% for stroke/systemic embolic events and 2.00% for major bleeding [27]. The PREFER in AF trial (461 hospitals from 7 European countries) reported a 3.4% annual stroke/systemic embolic event rate in 6412 patients > 75 years old [28]. This discrepancy may be explained in part by the higher risk profile of our patients than in the previous studies.

In such patients, outcomes may be predominantly driven by underlying irreversible decline, leaving limited room for the measurable impact of any single intervention, including structured multidisciplinary consultations. Recent observational studies provide valuable perspectives on this dilemma. For instance, in a multicenter cohort of hospitalized geriatric patients with AF [29], Brunetti et al. found no significant difference in clinical outcomes between patients who continued or discontinued oral anticoagulation, despite a 1-year mortality exceeding 50% in the deprescribed group. In contrast, Bucci et al. [30] observed that anticoagulation discontinuation was associated with increased all-cause mortality (HR 1.22, 95% CI 1.11–1.35) and cardiovascular events (HR 1.38, 95% CI 1.15–1.53), with an excess bleeding risk concentrated in the early post-discontinuation period (HR 1.51, 95% CI 1.24–1.83). Furthermore, an ancillary analysis of the mAFA-II trial [31] showed that the benefits of integrated care based on the ABC pathway were markedly attenuated in patients aged ≥ 80 years and those with greater clinical complexity. Similarly, a meta-analysis by Romiti et al. reported a diminishing benefit of ABC pathway adherence on stroke prevention in patients with higher CHA₂DS₂-VASc scores. Together, these findings suggest that beyond a certain threshold of frailty and comorbidity, the effectiveness of guideline-recommended interventions may be limited. This underscores the need for highly individualized management strategies in this population—taking into account not only thromboembolic and bleeding risks, but also life expectancy, functional and cognitive status, and patient-centered goals of care.

Our study has several strengths. Despite describing a very old, comorbid population that is particularly vulnerable to iatrogenic issues and typically excluded from RCTs, we emulated a target trial to conduct the observational analysis. This approach helps identify and mitigate biases. Our use of cloning, censoring, and weighting strategies avoided introducing immortal time bias into the analysis by aligning the time of eligibility assessment, intervention assignment during the grace period, and the start of follow-up, as explained elsewhere [24]. This design also helped to account for indication bias by cloning observations at baseline, thus ensuring that the groups being compared were identical in terms of measured and unmeasured confounders. The internal validity of our study was also strengthened by the completeness of our data collection, with minimal missing data.

We must acknowledge several limitations. Firstly, our intervention was deliberately focused on the decision-making process regarding anticoagulant therapy and does not constitute a comprehensive integrated care model as proposed by the ABC or AF-CARE approaches [3–5]. While elements such as comorbidity management are indeed part of routine geriatric practice [32, 33], we acknowledge that this does not ensure systematic, protocol-driven optimization of all relevant domains, including cardiovascular risk factors. Similarly, although rate and rhythm control often follow standardized algorithms [5], these strategies are frequently underutilized in real-world settings due to concerns about side effects, polypharmacy, and limited access to interventional options such as ablation. As such, our study should be interpreted as evaluating a targeted intervention within a broader continuum of AF care, and its results cannot be generalized to more comprehensive care models. Secondly, although the follow-up period was limited to 6 months, this duration remains clinically meaningful in such a vulnerable population. Indeed, more than one-third of patients died during this short timeframe, underscoring the high mortality risk and the urgency of optimizing care strategies in this group. Additionally, adverse events such as thromboembolism and major bleeding frequently occur soon after anticoagulant initiation or discontinuation, a pattern supported by current guidelines and large-scale observational data [5, 15, 30]. The numerically higher bleeding risk observed in the intervention group at 6 months may reflect more frequent treatment modifications in this actively managed population. Thirdly, while inclusion of all-cause mortality may attenuate the specific effects of the intervention on thromboembolic or bleeding events, this choice reflects both the high early mortality in this frail population and the difficulty of reliably attributing cause of death in routine clinical settings. Consequently, all-cause mortality is commonly used, alone or in combination, in geriatric and AF studies [2, 30, 31]. Fourthly, geriatricians may have gained experience over time from multidisciplinary meeting outputs, which may have influenced care. This learning curve and potential contamination bias could explain the lack of observed differences between the groups, thus limiting the external validity of our findings. Indeed, despite the publication and widespread dissemination of evidence-based practice guidelines for managing all aspects of AF care, underdiagnosis and inadequate treatment, especially among non-cardiovascular clinicians, remain prevalent worldwide [10]. Failure to adhere to evidence-based practice guidelines leaves patients vulnerable to avoidable morbidity and mortality. Fifthly, the time required to consult the multidisciplinary team varied among patients. We chose an empirical 45-day grace period, covering more than 90% of patients discussed by the multidisciplinary team. Although this threshold was not based on formal clinical guidelines [20], it was deemed clinically relevant: the meetings were scheduled monthly, and delays could occur when a high number of cases were pending review. Consequently, patients who received the intervention after day 45 were considered as following the control strategy (see Additional file 5: Fig. 1 (patient #1)). Importantly, sensitivity analyses using 30- and 60-day cut-offs produced consistent results, which reinforces confidence in our findings. Finally, at this stage, patients, families, and other healthcare professionals (e.g., nurses, pharmacists, general practitioners) were not involved in the multidisciplinary discussions.

Outcomes could be further improved by building on this initial experience, expanding these meetings to include community-based physicians, rethinking the format (e.g., videoconferencing), and incorporating structured patient and caregiver education to enhance treatment adherence through collaborations with advanced practice nurses. 

Conclusions

In this multicenter target emulation trial of 818 patients with confirmed AF who were ≥ 75 years old from May 2021 to January 2024, we found no significant differences in the occurrence of major thromboembolic events, major or clinically relevant bleeding, or all-cause death between patients whose cases were reviewed and not reviewed by a hospital multidisciplinary team. However, as this study remains observational, residual biases may explain the observed results, and these results need to be confirmed in a randomized trial.

Abbreviations

ABC

Anticoagulation, better control, comorbidity

AF

Atrial fibrillation

AF CARE

Comorbidity, avoid stroke and thromboembolism, reduce symptoms by rate and rhythm control, evaluation and dynamic reassessment

MACE

Major adverse cardiovascular events

RCT

Randomized controlled trials

Authors’ contributions

LZ and NST contributed equally as co-last authors. They were involved in the conception and design of the study, the development of the research protocol, and the manuscript writing. NST also played a significant role in the statistical analysis. o Concept and design: MP, NST, LZ o Acquisition, analysis, or interpretation of data: MP, NST, LZ, AC, LA o Drafting of the manuscript: MP, NST, LZ o Critical review of the manuscript for important intellectual content: MP, EB, EP, NL, PA, CF, MV, NST, LZ o Statistical analysis: MP, NST, AC o Obtained funding: MP o Supervision: NST, LZ o All authors read and approved the final manuscript

Funding

This work was supported by a Master grant (MP was awarded a research grant “Année Recherche” by the Regional Health Agency (“Agence Régionale de Santé, Ile-de-France”), November 2023–October 2024. The funders did not play a role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

Data availability

The datasets used during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Ethics Committee on May 25, 2021 (no. CER-2021–026). All enrolled patients or their legal representative gave their non-opposition for the use of their medical data.

Consent for publication

Not applicable.

Competing interests

o EP received honoraria for participation in expert meetings on low-molecular-weight Heparin (Léo Pharma), vitamin K antagonist (Merck) and direct oral anticoagulant (Boehringer Ingelheim, Bayer Healthcare AG, and Bristol Myers Squibb-Pfizer) o MV received honoraria for a lecture at a scientific meeting in 2021 (Bayer) o NST received support for attending meeting in 2025 (Sanofi) o LZ received honoraria for a lecture at a scientific meeting in 2023 and for participation in 2 expert meetings in 2024 (Bristol Myers Squibb) o The remaining authors declare no competing interests.