Opportunistic screening for atrial fibrillation among frail older patients, little effort for a high diagnostic yield. Outcomes of the Dutch-GERAF study

Abstract

Background and Aims

Frailty increases the risk of atrial fibrillation (AF) and its complications. This study investigated the feasibility and diagnostic yield of an eHealth screening for the detection of new AF, in frail older patients.

Methods

Patients referred to the Geriatric Medicine outpatient clinics were eligible. A Frailty Index (FI) was calculated. Patients were screened for AF with electrocardiograms (ECGs) at baseline and a smartphone photoplethysmography (PPG) application, during 6 months.

Results

Nine hundred fifty-two patients (median age 79 years) were included, mean FI of 0.16, 311 were frail (33%) and 751 had sinus rhythm (79%) at baseline. Six hundred forty-one patients (85%) performed PPG recordings (median 2), 295 (39%) at least 3 recordings. Twenty (2.7%) new cases of AF were found, 10 at baseline and 10 during follow-up. Among 16 (2%) patients, additional irregular PPG recordings were acquired, but no confirmatory ECG took place.

Conclusion

The screening strategy proved feasible in very old and frail patients. A diagnostic yield of 2.7% was found by ECG, and an additional 0.9% of new AF cases were suspected on PPG recordings. The non-binding approach of the strategy might be disadvantageous for the patient category. Future PPG AF screening programmes for very old and frail patients should strictly organise their means of AF confirmation.

Key Points

• An at-home eHealth screening strategy for atrial fibrillation is feasible among frail older people.

• Adding an electrocardiogram (ECG) to the work-up for frail older people on the emergency department will identify new cases of atrial fibrillation.

• Opportunistic screening for atrial fibrillation at geriatric medicine outpatient clinics reaches a very high diagnostic yield.

Introduction

The prevalence and incidence of atrial fibrillation (AF) is expected to rise steeply in the coming decades [1, 2]. A relevant proportion of stroke is thought to be due to undetected, and often asymptomatic AF [3]. To prevent strokes as a consequence of AF, timely detection and initiation of anticoagulation is paramount [2]. Furthermore, AF is associated with the development of heart failure, depression, hospitalisations and cognitive decline, which are especially relevant to the older population [1, 2]. The current guideline on the diagnosis and management of AF recommends opportunistic screening for all people 65 years and older, and considers systematic screening for patients 75 years and older at a high risk for stroke [1]. Frail older people are both at high risk for the development of AF, and stroke, but are underrepresented in large randomised controlled trials [1, 4, 5]. Frailty can be defined as a biological syndrome of diminished reserve and resistance to stressors, resulting from cumulative declines across multiple physiologic systems, increasing the risk for adverse outcomes [6]. Rockwood and colleagues propose the accumulation of deficits model, based on the Comprehensive Geriatric Assessment (CGA), which, besides categorising patients as frail, also quantifies the extent of frailty [7–9].

With the availability of handheld electrocardiogram (ECG) devices and wearables, screening for AF has become easily accessible for both patients and professionals. These strategies also detect AF that has not been uncovered or validated clinically, also referred to device detected, or subclinical AF (SCAF) [10, 11]. The recent LOOP study, ARTESIA trial and NOAH AFNET 6 trial have shown that initiation of oral anticoagulation upon the detection of SCAF does not lead to fewer strokes in all patient populations, and might even be harmful in some, especially in patients with a very low burden of AF [10, 12–15].

Opportunistic screening for clinical AF among geriatric outpatients can result in a very high diagnostic yield of 5.5% and appears to be very cost-effective [5, 16]. In the Dutch-GERAF study (GERAF study) we investigate the effectivity, and feasibility of screening by ECG at the outpatient visit complemented by an eHealth opportunistic interval screening approach, aimed at early detection of clinical AF among frail older patients [17].

Methods

This study was an investigator-initiated, prospective multicentre cohort study, conducted at 6 sites in the Netherlands, of which the design has been published recently [17]. In the GERAF study the recommendation to opportunistically screen patients of 65 years and older is applied [1]. With a pragmatic design, we intended to allow as much consecutive patients referred to the Geriatric Medicine outpatient clinics as possible to participate in the screening programme, as this has shown to be very cost-effective [16]. The primary outcome was the rate of newly detected clinical AF during 6 months. The secondary objectives were to investigate the feasibility of the eHealth screening strategy in frail older patients. Future studies will describe the 1 and 3 year outcome data concerning unplanned hospital admissions, stroke, major bleeding, progression of cognitive disorders and mortality [17].

Inclusion criteria were an age of 65 years or older, and to be able to perform at least 3 PPG recordings within 6 months. Patients were excluded if they had an implanted cardiac device, severe tremor (in which case reliable PPG recordings were expected to be too challenging) or a severe dementia, defined as a Montreal Cognitive Assessment (MoCA) score, or Mini-Mental State Examination (MMSE) score, of 15 points or less [18, 19].

Included patients without known AF at baseline participated in an opportunistic screening strategy for AF. The strategy consisted of a baseline ECG followed by photoplethysmography (PPG) smartphone application recordings, performed with the Heart Rhythm Software Development Kit, developed by Happitech (Rotterdam, The Netherlands) [20–22]. The sensitivity and specificity of this PPG application on a recording of sufficient quality have been reported to be 98.1% [21].

All patients underwent a CGA assessing frailty, functional status, cognitive function, mobility and an evaluation of polypharmacy (the use of five or more prescription drugs), within the context of all available medical information on co-morbidities and previous investigations of the patient [23]. If necessary, the CGA includes a full neuropsychological assessment. A CGA-based Frailty Index (FI) was calculated, following the Accumulation of Deficits model [5, 9, 24]. This FI consists of 46 factors, if absent factors are scored 0, if present factors are scored either 1 or 2, and add up to a maximum score of 51 points, and has been applied in our previous studies [5, 25, 26]. The FI is calculated as the sum of factors present divided by 51, and patients were classified as robust if the FI is below 0.18, moderately frail with a FI from 0.18 to 0.25, and severely frail with a FI above 0.25.

The opportunistic screening strategy was pragmatically designed to be easily integrated into usual care. At study entry, all patients underwent a 12 lead ECG as part of the CGA. Then a PPG recording was performed, preferably using patients’ own smartphone to optimise familiarity with the application [21]. During the 6-month follow-up, patients repeated PPG recording at outpatient visits and were encouraged to perform additional recordings at home, in order to reach at least 3 recordings within 6 months. In case the PPG application detected an irregular rhythm, patients were automatically prompted to run a second confirmatory PPG recording, and if both showed an irregular rhythm, the application alarmed the patient to contact their physician to perform a confirmatory ECG. Patients were instructed to call the outpatient clinic after the confirmatory ECG would be organised, and evaluated either by the treating Geriatrician, Geriatrician on call or Cardiologist/Resident Cardiology on call.

To avoid classifying short-lasting episodes of SCAF as clinical AF, only ECG-confirmed AF was considered newly detected AF in this study. The study conforms to the Declaration of Helsinki, and the protocol was approved by the Medical Ethics Committee Oost Nederland, with reference number 2019–5889. All participating patients gave written informed consent.

Statistical analysis

Statistical analyses were performed with SPSS for Windows, version 20. P values ≤.05 is considered statistically significant. Newly detected cases of AF are reported as the rate of new cases among all patients with sinus rhythm (SR) at study entry. Feasibility of the screening strategy was evaluated by assessing the rate of participation, the proportion of patients willing to run a second or third attempt to acquire a PPG recording of sufficient quality, the proportion of patients that performed 1–2 independent PPG recordings, and proportion of patients that performed 3 or more PPG recordings. Adherence to the study protocol was classified as, no PPG recordings, 1–2 PPG recordings, or 3 or more PPG recordings. For descriptive purposes, differences between the baseline characteristics of patients known with AF and those with SR were analysed with an independent t-test for continuous variables, and proportions were compared using Fisher’s exact test. Factors associated with adherence to the study protocol were analysed by logistic regression with adjustment for age, sex and frailty, and reported as Hazard Ratios (HR) with 95% Confidence Intervals (95% CI). Whether adherence to protocol was associated with the detection of new AF was analysed by multinomial logistic regression with adjustment for age, sex and frailty.

Results

Patients were included between the first of January 2021 and 30 October 2022, at Geriatric outpatient clinics of six hospitals in the Netherlands. There were 1258 patients eligible for participation, 1001 met the inclusion criteria and gave informed consent. Thirty-one patients were excluded, and 18 withdrew their consent. In total, 952 patients participated in the GERAF study.

The baseline demographic characteristics are shown in Table 1. The cohort consisted of very old patients, with a median age of 79 ± 6.2 years at study entry, a rate of polypharmacy of 52%, a high prevalence of cardiovascular diseases (77% of patients had at least 1 cardiovascular disease, and 53% at least 2), high prevalence of cognitive disorders of 43% and a large proportion of patients either had a gait disturbance or reported falls (42% and 41% respectively). Frailty was normally distributed in the cohort, with a mean FI of 0.16 ± 0.08. Overall, 175 patients (18%) were moderately frail, and 136 (14%) severely frail.

The baseline prevalence of AF was 21% (201 patients). There were 751 (79%) patients with SR at baseline who participated in the screening strategy.

Feasibility of opportunistic PPG screening for AF

The 751 patients with SR at baseline performed a total of 3474 PPG recordings, with a median of 2 recordings per patient, of which 638 (18%) were made at baseline. Table 2 gives an overview of the utilisation of the PPG application. Of these 3474 recordings, 596 (17%) could not be interpreted due to various reasons, such as movement artefacts, insufficient signal, interruptions or incorrect placement of the fingertip. There were 295 patients (39%) that performed at least 3 recordings, 346 (46%) performed 1 or 2 recordings and 110 (15%) made no PPG recordings at all. As a first attempt, 2519 (73%) recordings were of sufficient quality and 374 (58%) patients made an immediate second attempt 726 times, either after the first attempt was insufficient (631 recordings) or after the application requested a confirmatory recording in case an irregular rhythm was detected (95 recordings). After a second failed attempt, 181 (28%) patients performed a third attempt 422 times. Table 3 shows the association between geriatric features and adherence to the study protocol. Characteristics associated with a lower likelihood of performing at least 3 PPG recordings were a higher age (HR 0.96, 95% CI 0.92–0.99), male sex (HR 0.59, 95% CI 0.37–0.93) and frailty (HR 0.89, 95% CI 0.84–0.94). Living with a partner (HR 1.59, 95% CI 1.02–2.47), and being able to execute tasks of daily living independently increased the likelihood of performing at least 3 PPG recordings (HR 4.94, 95% CI 1.16–21.10). The presence of cognitive disorders, either Mild Cognitive Impairment (HR 0.72, 95% CI 0.43–1.18) or dementia (HR 0.79, 95% CI 0.43–1.46), was not significantly associated with the execution of PPG recordings. After adjustment for age, sex and frailty, a lower likelihood of executing 1 or 2 PPG recordings remained in men (HR 0.52, 95% CI 0.33–0.82), and the likelihood of performing at least 3 PPG recordings decreased with increasing frailty (HR 0.90, 95% CI 0.85–0.96). Of note, the definition of frailty incorporates all relevant morbidities and disabilities (including cognitive function and dependence on others for tasks of daily living).

AF case finding

During the 6-month screening, AF was detected in 20 (2.7%) patients and a direct oral anticoagulant was initiated in 18 (90%). The means of identification of new cases is shown in Fig. 1. There were 44 (6%) patients who produced 56 PPG recordings indicative of an irregular heart rhythm. Of these, 27 occurred at study entry, followed by a 12-lead ECG and 4 (0.5%) cases of new AF were subsequently found. Another 6 (0.8%) patients had a regular PPG recording at baseline but showed AF on the study entry ECG. During the 6-month follow-up, another 29 irregular PPG recordings were collected by 24 patients, of which 8 patients performed a 12 lead ECG within 48 hours. All these eight ECGs showed SR, with or without premature atrial or ventricular complexes. In 16 (2%) patients with 19 irregular PPG recordings, no confirmatory ECG was performed, and AF could not be definitely confirmed. The validation study of the applied algorithm behind the PPG application had a sensitivity of 98.1%, and specificity of 98.1%, when the PPG recording was run simultaneously with a 12 lead ECG [21]. In the screening strategy applied in this study, the PPG recording and ECG were not performed simultaneously but could have up to 48 hours in between them. Taking into account the paroxysmal nature of AF, a positive PPG recording today, but SR on an ECG tomorrow still does not rule out the presence of short-lasting paroxysmal AF. An additional visual assessment of the Poincaré plots and tachogram was performed by two electrophysiologists (M.H. and R.R.) and showed that in 7 (0.9%) patients AF was likely, and in 9 the irregular PPG recording was the consequence of artefacts. Three (0.4%) patients had irregular PPG recordings, but AF was confirmed by other means than a consecutive confirmatory 12 lead ECG, 1 with a Holter registration ordered because of multiple irregular PPG recordings, and 2 with an ECG performed during routine care. Finally, 7 cases of AF (0.9% of the screened population and 35% of new cases) had only regular PPG recordings, but new AF was detected during an emergency department visit. There was no association between a better adherence to protocol and a higher detection rate of clinical AF, adjusted for age, gender and frailty, HR 0.39, 95% CI 0.11–1.42, P = .16 (see Table 4).

Discussion

This study investigated the diagnostic yield for case finding of new clinical AF of a repeated eHealth PPG screening strategy, and its feasibility among frail patients.

With the combination of a baseline 12-lead ECG, incidental 12-lead ECGs at emergency department visits and PPG recordings, a diagnostic yield for new AF of 2.7% was found. Unexpectedly, however, most cases were identified through opportunistically performing a 12-lead ECG as part of usual care. Ten cases were found on the study entry ECG (50% of new cases), 7 (35% of new cases) at incidental emergency department visits and 2 (10%) during routine care. A 6-month incidence of 2.7% is relatively high in comparison to other studies applying opportunistic screening [22, 27–29]. In the GERAF study, an even higher incidence of AF of an additional 0.9% could possibly have been found, as 7 out of the 16 patients who did not perform a confirmatory ECG within 48 hours of an irregular PPG recording were very likely to be AF. The high baseline prevalence of 21% and high incidence of AF of 2.7%–3.6% reflects the high-risk profile for cardiovascular outcomes of frail older patients and supports the recommendation for opportunistic screening for AF [1, 2]. Although the Dutch Guideline on the CGA recommends performing a 12-lead ECG for every newly referred patient, in daily clinical practice an ECG is not performed in all patients [23]. The results of this study strongly enforce that recommendation, and additionally shows that all clinical encounters such as emergency department visits offer an opportunity to identify new cases of clinical AF, and initiate treatment.

A very important finding of the GERAF study is that it is possible for frail older patients to perform a remote smartphone PPG screening strategy for case finding of AF. Despite the high prevalence of frailty, dependency on others and cognitive disorders, the majority of patients (85%) performed PPG recordings, 83% of the PPG recordings were of sufficient quality for analysis and 73% were successful at the first attempt. Noteworthy was the persistence of patients, 58% of patients performed an immediate second attempt if prompted by the application and 28% also a third if necessary. In the unadjusted regression analysis, the presence of cognitive disorders was not associated with a worse adherence to the screening strategy. And in the adjusted regression analysis, correcting for all geriatric features including cognitive disorders, only a higher FI lowered the likelihood of good adherence to the study protocol. These results connect very well with a recently published study by Mant et al. that convincingly showed that old age per se does not form an impairment for participation in a screening programme [30]. Those patients were selected from primary care practises solely based on age, and were likely to be less frail than the GERAF cohort. Overall, the results of the GERAF study show that a remote eHealth screening strategy utilising smartphone-based PPG heart rate recordings is a feasible approach even in frail older people.

The pragmatic design of the study offered patients the opportunity to screen at home and periodically during outpatient visits to their physician, but did not include any reminders or pressure to screen. Consequently, the number of PPG recordings performed at home was low, with a median of 2 measurements in a 6-month period. A higher age, higher frailty, or male sex was associated with a lower likelihood of performing at least 3 PPG recordings. The low absolute number of new AF cases restricts the possibility of a detailed multivariate analysis, into factors predictive for AF detection. In the STROKE-STOP study an increasing number of single-lead ECG recordings was associated with a higher likelihood of AF case detection [31], we cannot rule out that the same holds for frail older patients. A non-binding approach might have been disadvantageous for our patient category, and the number of home recordings may have been higher when a more stringent protocol was followed. Future screening programmes directed at very old and frail patients should offer more guidance and reminders for both performing the repeated recordings and 12-lead ECG confirmation if necessary. At study entry, the treating physician or outpatient nurse would help with the installation and first use of the application. In our view, this would also be a feasible approach on the long term. A bigger challenge however would be the oversight of at-home PPG recordings and timely invitation of patients after an irregular recording. Preferably this would be integrated into other telemonitoring systems that are already in use.

Strengths and limitations

One of the strengths of this study is its pragmatic design, which contributed to the high rate of participation. It resulted in a cohort of very old patients, with relevant proportions of patients with frailty, cognitive disorders and dependence on others for their daily life, and in this aspect forms a relevant addition to randomised clinical trials. All newly referred patients were invited to participate in the study, reducing the likelihood of having selected only people with interest and access to health screening programmes. The results of the GERAF study are limited to PPG heart rate detection in frail older people, and not to older people in general, or to PPG heart rhythm assessment as a technique. Improvement of the diagnostic yield might be possible with an increasing number of recordings per patient, and extra emphasis on performing the confirmatory ECG or applying a (single lead) ECG technique that allows expert confirmation of the heart rhythm directly. Currently, an irregular PPG recording alone is not sufficient for a diagnosis of AF which requires a confirmatory ECG, and the results of the GERAF study do not suggest otherwise [1].

An important limitation of the GERAF study is that a non-binding approach proves to hold a relevant risk of new cases remaining unconfirmed in day-to-day clinical practice. Future screening programmes directed at frail older patients should take the issue of confirming clinical AF into account and should consider reminders and guidance to screen. Another limitation is that, despite the considerable size of the cohort, the number of patients with new AF was limited. Hence, it is not possible from this data set to infer any effects of the pragmatic screening strategy on cardiovascular outcomes, such as stroke, major bleeding and death. However, the cost-effectiveness analysis of van Hulst and colleagues showed that opportunistic screening at a Geriatric Medicine outpatient clinic was very cost-effective, and even was with certainty cost-saving [16]. The 1 and 3 year outcome data of the GERAF study will provide more insight into the incidence of stroke, major bleeding and mortality of this geriatric cohort.

Conclusion

Home-based screening for AF with single-time PPG recordings is feasible within a very old and frail cohort with a large proportion of patients suffering from cognitive disorders. The screening strategy of the GERAF study showed a diagnostic yield of 2.7% and potentially identified an additional 0.9% of unconfirmed cases. Confirmation with ECG after possible detection of AF by PPG was often not followed up, possibly underestimating the diagnostic yield of PPG screening in this cohort. Future PPG AF screening programmes for very old and frail patients should strictly organise their means of AF confirmation.

Acknowledgements:

We would like to sincerely thank all geriatricians, nurses and residents for their contribution to the study, and inclusion of patients. We would like to especially thank L. Voorhout and S. Peerlings for their help with the data collection.

Declaration of Conflicts of Interest:

L.A.R. Zwart, J.S. Spruit, R.W.M.M. Jansen, R. Pisters, L. Louter, K. de Vries, D.G. Taekema and J.F.H. Wold declare to have no competing interests.

R.K. Riezebos is a founding board member and minority shareholder of Happitech.

J.R. de Groot is Supported by the Dutch Heart Foundation and Dutch CardioVascular Alliance, 01-002-2022-0118 EmbRACE; ZonMW/ERA4Health. Received research grants through institutions from Atricure, Bayer, Boston Scientific, Daiichi Sankyo, Johnson&Johnson, Medtronic and Philips. Received speaker fees from Atricure, Bayer, Berlin Chemie, Daiichi Sankyo, Johnson&Johnson, Menarini, Medtronic, Novartis and Servier.

M.E.W. Hemels received speakers fees from Roche Diagnostics, Bayer, BMS/Pfizer, Boehringer Ingelheim and Daiichi Sankyo.

Declaration of Sources of Funding:

The study was supported by unrestricted grants from Bayer Nederland, Boehringer Ingelheim, Daiichi Sankyo and Roche Diagnostics.

Data Availability Statement:

The data within the GERAF study is not publicly available, anonymised data can be shared upon reasonable request, and after unanimous approval of the GERAF investigators.

References