Global implementation and evaluation of atrial fibrillation screening in the past two decades – a narrative review

Atrial fibrillation (AF) is increasing due to population aging, increases in AF risk factors such as obesity, and partly due to increased AF detection through screening. The 2019 global burden of AF was estimated at 59.7 million people (95% confidence interval (CI) 45.7–75.3 million) and is double the estimates made in 1990^1,2,3,4. AF is associated with a five-fold higher risk of stroke⁵, and AF-related ischemic stroke is more likely to recur and almost twice as likely to be fatal^6,7. The total annual cost of AF treatment (including hospitalizations, inpatient cost, outpatient treatment, and prescription medications) was estimated to be US$6.65 billion in the United States of America (USA) in 2005⁸. Patients with AF utilize healthcare services more than people without AF, and in the USA, the annual average total healthcare costs for patients with AF was US$25,006 (95% CI $24,357–$26,912) more than that for patients without AF (p < 0.001)⁹. In Australia, the estimated cost of AF treatment was AUD881 million in 2015–2016, including hospital services, outpatient treatment, emergency departments, prescription medications and general practitioner (GP) services¹⁰.

The increasing burden of AF on healthcare systems highlights the need for effective AF identification, prevention, and management strategies. The current recommendations for AF screening vary among Guidelines (Fig. 1). The European Society of Cardiology (ESC)/European Association for Cardio-Thoracic Surgery (EACTS)¹¹, Canadian Cardiovascular Society (CCS)/Canadian Heart Rhythm Society (CHRS)¹², National Heart Foundation of Australia (NHFA)/Cardiac Society of Australia and New Zealand (CSANZ)¹³ and Asia Pacific Heart Rhythm Society (APHRS)¹⁴ recommend opportunistic screening among people aged ≥65 years by pulse palpation and using electrocardiographic (ECG) devices, including the use of single-lead rhythm traces. In addition, the ESC guidelines and APHRS recommend systematic screening of people aged ≥75 years^11,14. However, in 2019, the United Kingdom (UK) National Screening Committee issued a statement of not recommending a national population screening programme for AF because it is not clear if different types of AF have the same risk for strokes, the effectiveness of treatment for screening-detected AF is unknown, and it is not known if screening is more beneficial than the current approach to detection and management¹⁵. Recently, the American College of Cardiology (ACC)/American Heart Association (AHA) Joint Committee stated that it is not yet established that patients at high risk of developing AF would benefit from screening and interventions to improve rates of ischemic stroke, systemic embolism, and survival¹⁶. Similarly, the United States Preventive Services Task Force (USPSTF) stated that there is not enough evidence to recommend screening for adults aged 50 years or older given the potential harms of therapy – primarily bleeding from anticoagulants¹⁷, but the USPSTF notes that AF patients likely would benefit from contextualized management of modifiable risk factors^17,18,19,20.

ESC/ EACTS The European Society of Cardiology/European Association for Cardio-Thoracic Surgery Guidelines, CCS/ CHRS Canadian Cardiovascular Society/Canadian Heart Rhythm Society Guidelines, NHFA/CSANZ National Heart Foundation of Australia/Cardiac Society of Australia and New Zealand (CSANZ) Guidelines, APHRS Asia Pacific Heart Rhythm Society Guidelines; UK United Kingdom, USPSTF United States Preventive Services Task Force, ACC/AHA/ACCP/HRS American College of Cardiology/American Heart Association/American College of Chest Physicians/ Heart Rhythm Society Guidelines.

Full size image

Screening for AF enables early diagnosis and allows patients to gain the potential benefits of prevention management. In many countries, cardiovascular and other chronic health condition screenings have been the domain of primary care providers (i.e., general practitioners), as has the ongoing management of high-risk AF patients. This review aims to examine and synthesize findings from AF screening studies across various settings to gain insight into screening approaches, technology use, AF detection rates, implementation, and evaluation of the screening process.

The PubMed database was searched from 1st January 2000 to 18th January 2024. The search terms included “atrial fibrillation”, “auricular fibrillation”, “atrial flutter”, “screen”, “screening”, “search”, “searching”, “detect”, “detecting”, “detection”, “diagnose”, “diagnosing”, “diagnosis”, “diagnoses”, “identify”, “identifying”, and “identification”. All searches were restricted to human subjects and articles published in English (see the literature search strategies in Supplementary Table 1). Articles that described the implementation of screening programs, including the target population, screening methods (e.g. observational studies, interventional studies with or without a comparator/control) and primary outcomes, were included. Editorial, commentary, engineering articles (e.g. analysis of ECG signal processing, device accuracy studies), basic science articles (e.g. AF pathophysiology, biomarkers), and studies on post-stroke rhythm monitoring/treatments/complications were excluded (Fig. 2). The following information was extracted from the included articles and documented in Tables 1 and 2: country where the study was conducted, first author’s name, year of publication, year commenced AF screening, sample size, age eligibility (target population), risk factors included as enrollment criteria, screening methods and devices, primary outcomes, and the screening initiatives (e.g. primary care/pharmacy/remote screening or hospital and other community initiatives). The countries or regions where the studies were conducted were graphically identified on a world map²¹.

ECG electrocardiogram, n number of articles.

Full size image

Opportunistic screening was defined as individuals visiting a setting or healthcare facility for other purposes but were opportunistically invited to participate in AF screening, while systematic screening was defined as systematically identifying individuals from a source (e.g. registry, clinical database, population database) and inviting them to screening.

Studies that reported any of the following activities were considered having evaluated the implementation: surveys, interviews or some forms of evaluation and reported acceptability, usability, use of health services triggered by notifications of potential screening abnormalities, feasibility (e.g. barriers and enablers), or process evaluation framework (e.g. the Medical Council Guidance on Process Evaluation²², Realist approach²³, Critical Realism approach²⁴, or the “Reach, Effectiveness, Adoption, Implementation, and Maintenance of interventions, RE-AIM” approach²⁵). The results of this review were narratively summarized.

A total of 1767 abstracts were screened, 138 full-texts were reviewed, and 87 studies were identified as meeting the inclusion criteria from the last 24 years, with the majority of these studies conducted after 2010 (Fig. 3). Based on World Bank classifications²⁶, 78 (90%) studies were conducted in high-income economies, 7 (8%) in upper-middle-income economies, and 2 (2%) in lower-middle-income economies (Fig. 4).

The circles’ sizes denote the study’s sample size. RCT randomised controlled trial, denoted by *. Study without a comparator is denoted by ~. Studies that applied handheld devices are denoted by ^, wearable or patch devices are denoted by #, and implantable devices are denoted by @.

Full size image

High-income countries/regions are indicated in red font, upper-middle-income countries/regions are in blue, and lower-middle-income countries/regions are in purple.

Full size image

Study design and screening initiatives

Fifteen of the studies (17%) were either randomized control trials (RCT) or had a comparator group (of these, 8 of 9 studies in primary care were RCTs, and all 6 remote monitoring studies were RCTs) (Table 1), and 72 (83%) were without a comparator group (Table 2).

Most studies examined community screening initiatives conducted in primary care/ general practice (n = 31)^{27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61}, pharmacy (n = 11)^{62,63,64,65,66,67,68,69,70,71,72,73}, and community centers and villages (n = 10)^{74,75,76,77,78,79,80,81,82,83,84}. Several community initiatives studies were classified as ‘remote self-screening’ (n = 30)^{85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117}, indicating that they were not connected to a community health infrastructure and relied on participants self-screening at home or elsewhere in the community and sending their ECG data to a monitoring center. The first population-based remote self-screening study (n = 7173) was conducted in 2015 (STROKESTOP)¹¹². Some initiatives were very large, including nearly half a million individuals^102,105 and 2.8 million individuals⁹⁵, but only 6 of the 30 remote self-screening studies had comparator groups. A very small number of studies described screening initiatives in institutions of hospitals (n = 4)^{118,119,120,121} and nursing home (n = 1) (Fig. 3)¹²².

Study population and sampling approach

The majority (n = 72/87, 83%) of studies identified their target population by a specific age group, 3 studies targeted people from more than one age group based on the presence of risk factors^70,123 and geographical location⁷⁶, and 12 studies did not specify age criteria. Among the 72 studies, 45 (63%) targeted older people with minimal eligible age from 65 to 80 years, while 27 (37%) targeted people with minimal eligible age from 18 to 60 years (see Supplementary Fig. 1).

Twenty-one studies used one or more risk factors, in addition to age criteria, to further target their AF screening program, with all of these studies using hypertension (21/21) as one of the risk factors, and many required patients to have one or more risk factors from a list including heart failure (18/21), diabetes (18/21), stroke or transient ischemic attack (13/21), peripheral vascular diseases (8/21), ischemic heart disease/ coronary artery disease (7/21), sleep apnea (4/21), valvular heart disease (3/21), and chronic obstructive pulmonary disease (3/21). The sampling method fitted an opportunistic approach in 56 studies (64%) and a systematic approach in 31 (36%) studies (Tables 1 and 2).

Screening device

The screening devices used in the studies were handheld devices (n = 72, 83%), wearable devices (n = 13, 15%) and implantable devices (n = 2, 2%). Most primary care studies used handheld devices. Wearable and implantable ECG devices were mainly used in remote self-screening studies and mostly after the year 2015 (Fig. 3). Handheld devices included: single-lead rhythm trace devices (n = 43), 12-lead ECG (n = 13), sphygmomanometer embedded with AF-detecting algorithm (n = 10), 3-lead ECG (n = 1), photoplethysmography (PPG) app interface with smartphones (n = 3), and Holter (n = 2). Wearable devices included: ECG patch (n = 6); watches/ wristbands using PPG sensors (n = 4); ECG vest/band (n = 3); and implantable device/ loop recorder (n = 2) (Tables 1 and 2). The use of handheld and wearable screening devices by screening initiatives is shown in Table 3.

AF detection rates

Of the 15 studies with intervention and comparator groups targeting participants ≥50 years, 9 were conducted in general practice (GP) or primary care settings, and 6 were remote self-screening studies. Six of the GP setting studies reported that the intervention screening had higher AF detection rates compared with usual care^{29,31,32,39,46,124}, and three studies reported an insignificant difference in AF detection rates between intervention and control (usual care) groups^42,44,60. Systematic screening approaches that targeted people ≥65 years through general practice yielded higher AF detection rates compared with usual care^29,31,41, while there were insignificant differences in AF detection rates in opportunistic screening at general practice compared with usual care^42,44,60. All 6 RCTs that were remote self-screening studies reported superior AF detection rates compared to control groups (i.e., usual care^{92,99,107,111,112} or delayed screening group¹¹⁰). The duration of remote self-screening ranged from using handheld ECG devices intermittently over 2 weeks¹¹² to 12 months⁹⁹, or wearing an ECG patch for continuous monitoring over 2 weeks⁹² to 4 weeks¹¹⁰, or having an implantable cardiac device continuous monitoring for a median follow-up of 64.5 months¹¹¹ (Table 1).

Among the 72 studies without a comparator group, a wide range of new AF detection rates were reported: 18 studies reported new AF detection rates below 1%, 48 studies reported 1–10%, 5 studies reported above 10%, and one study reported AF incidence rate of 2.25% patient-years (95% CI 2.03–2.48) (Table 2).

As illustrated in Fig. 5, higher AF detection rates were generally more common among older screening groups and those that targeted their screening with additional risk factors. Very high AF detection rates were reported by a few studies: Svendsen JH and colleagues reported the LOOP study with an AF detection rate of 31.8% enrolled hospital patients aged 70–90 years with at least one risk factor of stroke (n = 6004) and monitored them with continuous implanted loop recorders over a median duration of 64.5 months compared with usual care¹¹¹. Somerville S and colleagues conducted a small study (n = 86) at a general practice with a detection rate of 30%, recruiting and screening an equal number of participants with and without a history of AF selected from the medical records; the nurses performed pulse palpation followed by 12-lead ECG if an irregular rhythm was detected⁵⁷. One other small general practice study conducted by Orchard J and colleagues reported that receptionists and nurses screened 88 patients using handheld single-lead ECGs and found that 19% of the participants had AF, but all of them had been previously diagnosed with AF⁴⁷.

n = 86 (65 studies reported mean age, 11 minimal eligible age, and 10 median age). Red crosses indicate studies that included risk factors (in addition to age criteria) in the enrollment criteria, and blue crosses indicate studies with age criteria as the only enrollment criteria. SJ: Svendsen et al.¹¹¹ screened hospital patients with at least one risk factor of stroke using continuous implanted loop recorders compared with usual care; SS: Somerville et al.⁵⁷ included participants who had been previously diagnosed with atrial fibrillation selected from medical records at a general practice; OJ: Orchard et al.⁴⁷ screened patients presented at a general practice and all detected atrial fibrillation patients had been previously diagnosed.

Full size image

Among 70 studies that reported the date when AF screening began, 64 (91%) studies began from 2010 onwards, and the AF detection rates were mostly randomly distributed between 1% and 10% without a trend of temporal variation (Fig. 6).

n = 70. SJ: Svendsen et al.¹¹¹ screened hospital patients with at least one risk factor of stroke using continuous implanted loop recorders compared with usual care; OJ: Orchard et al.⁴⁷ screened patients presented at a general practice and all detected atrial fibrillation patients had been previously diagnosed.

Full size image

Study implementation and evaluation

Practice staff (e.g., practice nurses, GPs, and receptionists) facilitated general practice (GP)/primary care screening initiatives, with cardiologists’ support in ECG interpretation^30,39,40,41. Participants’ rhythm traces were transmitted into the patient’s medical records for GPs to discuss with the patients^37,48. There was a lack of description of the clinician’s experience reviewing rhythm traces and the process and utilization of cardiologist services in confirming and managing clinically significant rhythm abnormalities, including AF.

In pharmacy screening initiatives, participants’ rhythm traces were recorded and then remotely interpreted by cardiologists. The study team contacted participants with possible AF to see GPs for further evaluation^67,70,72. A study involved pharmacists performing opportunistic AF screening in UK general practice during influenza vaccination⁶⁹. A few studies involved pharmacy students performing AF screening in health promotion fairs in the community; in these studies, students handed participants a card with the results of the device’s automatic interpretation to take to their usual healthcare providers^62,64.

Atrial fibrillation screening initiatives in the community involved setting up screening facilities in community centers^74,75,79,81. The research team recorded participants’ rhythm traces, which were stored in a virtual portal and reviewed by the researchers 2–4 weeks later. Then, the researchers notified the participants of their AF diagnoses^74,75. A few community screening initiatives involved sending health workers (or trained personnel) to remote rural villages to screen people using mobile technologies^82,84. The participants’ rhythm traces were interpreted by cardiologists, and AF findings were transmitted to the participants’ GPs to consider initiation of therapy⁸¹.

In hospital screening initiatives, patients admitted to the hospitals were opportunistically recruited to participate in AF screening^118,120,121. In addition, screening stations were set up in hospitals to screen eligible individuals from the community recruited through an AF awareness campaign advertised on social media. Participants with screen-diagnosed AF were referred to consult their general practitioner or cardiologist. However, the process and mode of communication were not detailed¹¹⁹. In hospital hypertension clinics and cardiologist offices, clinicians opportunistically performed triaging screening on patients, providing the screening devices to patients to perform remote screening and then re-assessing patients upon returning the screening devices. These clinician-initiated patient self-screening studies were categorized as remote self-screening because the patients performed self-screening remotely¹⁰⁹. Generally, remote self-screening involved the use of various screening devices where rhythm traces were remotely reviewed by the research team^{85,87,89,103,112,116}, or the research team subcontracted the ECG interpretation to paid services by another organization⁹⁹.

Nineteen (19/87, 22%) of the studies reported an evaluation of the implementation of the screening programs, mainly (12 studies) using survey questionnaires^{30,38,41,44,69,94,97,99,105,108,109,117,118} and 7 studies conducted interviews^{45,47,48,49,64,66,67,68}. Surveys were mainly applied in remote self-screening studies evaluating participant acceptability, usability and satisfaction with the screening, while interviews were mainly conducted in general practice and pharmacy screening initiatives (Table 3). Surveys in Apple Heart Study¹⁰⁵ and Fitbit Heart Study^102,125 found that 57% of participants in Apple Heart Study¹⁰⁵ and 24% of respondents in Fitbit Heart Study¹⁰² visited their usual healthcare providers outside the research study when they received irregular rhythm notifications. Generally, surveys of remote self-screening reported good acceptability and usability of the screening^94,99,108. For example, researchers surveyed participants and obtained a System Usability Score¹²⁶ of 85 (a score above 68 is generally considered acceptable), indicating high usability of the ECG patch worn by participants aged 65 years and older¹⁰⁸.

In the included studies, researchers applied thematic analysis to semi-structured interviews and identified barriers, e.g. perception that pharmacist’s roles do not include screening, lack of time to engage with customers and difficulty in integrating screening in the workflow, and enablers, e.g. incorporating AF screening in pharmacy health promotion events and training pharmacists to use the screening devices⁶⁷ and appointing a local “champion”, i.e. someone to facilitate the implementation of AF screening in pharmacy⁶⁶. In addition, in assessing how to make AF screening at pharmacies sustainable, researchers identified that pharmacy students could provide sustainability of AF screening in pharmacies⁶⁴. In general practice/primary care studies, researchers applied the Realist evaluation approach²³ in analyzing qualitative semi-structured interviews¹²⁷. The Realist approach has provided a more robust framework for assessing participants’ perceptions of and experience with the screening processes and generating themes from the qualitative interviews²³. The main themes in the GP setting included a lack of resources and referral pathways in busy clinical settings as screening barriers, while enablers included the appointment of a designated person to lead and facilitate AF implementation at the point-of-care setting^{45,47,48,49,68}.

This review found substantial literature on AF screening, mainly from higher-income countries/ regions, that mostly described opportunistic approaches to screening (64% of studies) targeting older patients aged 65 years and older in line with many current Guidelines^11,12,13,14. However, over this 24-year period, less than 20% of studies had a comparator group reporting a measure of the effectiveness of the screening approach, only 14 studies were RCTs (Table 1) and only about 20% reported on some implementation measures (Table 3) with very few providing data to inform how to scale up AF screening. The studies described initiatives across various settings (primary care, pharmacy, community centers, hospital and remote screening). Handheld ECG devices were commonly used, and the screening initiative was most commonly in a location associated with existing healthcare with access to direct health professional supervision, with some describing remote self-screening (Table 3).

Screening approaches, enablers, and barriers in primary care

Many studies (n = 31) were general practice/primary care screening initiatives. Primary care services are common in many countries. Among 38 countries in the Organization for Economic Co-operation and Development (OECD), on average, approximately 80% of individuals aged 15 or older reported visiting a doctor (using primary care service) at least once a year, ranging from around 65% in Sweden and the United States to 89% in France¹²⁸. In Australia, a recent national health survey found that 90% of Australians visited their general practitioners (GPs) at least once a year¹²⁹. This healthcare utilization was even higher among older people. In 2019–2020, 95% of people aged ≥65 years saw a GP at least once yearly¹³⁰. GP visits provide occasions for opportunistic AF screening, such as palpating the patient’s pulse and acquiring an ECG if clinically indicated^41,124. However, the frequency of opportunistic AF screening occurring in general practice is uncertain. A survey of GPs in Australia found that raising GPs’ awareness of AF screening and improving their confidence in ECG interpretation may increase AF screening¹³¹. In the SAFE Study conducted in the UK⁴¹, GPs and practice nurses in the intervention group received education about the importance of AF screening and ECG interpretation, and they palpated patients’ pulses, followed by acquiring 12-lead ECGs if irregular pulses were detected. However, the SAFE Study reported only a modest increase in AF detection rate (difference of 0.59%, 95% CI 0.20–0.98%) between the intervention and control (usual standard care) groups.

Most studies in this review utilized mobile devices. Some research indicates that using single-lead handheld ECG devices to detect AF is more accurate than pulse palpation. A recent study involving 6159 participants reported that the sensitivity of pulse palpation was 78.6% and positive predictive value (PPV) was only 4.8%¹³², while handheld single-lead ECG devices generally have high accuracy (sensitivity of 90% and specificity of 99%) in detecting AF¹³³. However, existing data also indicate that the uptake of AF screening among GPs is low, partly due to time pressure and lack of resources^47,127. Some studies have described potential enablers such as identifying practice staff resources and including AF screening in their roles^43,47,49, and designating “AF-screening champions”⁴⁹ in practice. When practice administrative staff were encouraged to be involved, some studies suggested their reluctance because tasks were beyond their traditional scope of duties⁴⁷. For practice nurses, AF screening was more easily integrated into their workflow⁴⁹. While several studies have focused on GPs (primary care providers) implementing AF screening, this may not be practicable in some countries. Primary care is increasingly under strain. Nonetheless, it may be feasible for primary care providers to focus screening of AF in the highest risk groups, such as older patients, as part of an overall health check. Notably, financial viability was reported as important to implementation, with some indicating the need for financial incentives to make screening feasible^42,48,134.

This review provided additional insights into other novel approaches to implementing AF screening within health practice. For example, optimizing the use of patients’ waiting time was examined in one study³⁷ involving the setting up self-screening stations in the waiting room. Patients placed their fingers on the designated contact points of single-lead Kardia ECG devices; the automated ECG results were transmitted to the patient’s medical records, and the GP could discuss the results with them. This study reported the general acceptability of this screening approach. However, some older patients reported needing assistance to use the devices and touching shared screening devices could be a barrier (particularly in the context of the COVID-19 pandemic)³⁷. This approach might be feasible but requires a sustainable process for GPs to interpret unclassified ECGs (i.e., ECGs that could not be interpreted by the device’s automated algorithms), suspected AF and other abnormalities, and to have timely access to referral pathways when an abnormality was detected. A qualitative evaluation with GPs conducted in the UK recently highlighted that the lack of an established protocol or a referral pathway to manage ECG abnormalities was a barrier to implementing AF screening¹³⁵. A survey of 588 healthcare providers (77% from Europe, 14% from Asia/ Oceania and 8% from North or South America) conducted by the AF-SCREEN International collaboration reported that there is a need to better define an appropriate mechanism for managing screen-detected AF¹³⁶.

Pharmacies are commonly visited venues by the public. Providing handheld single-lead ECG devices to pharmacists, training them to perform AF screening and tasking them to refer suspected abnormalities to GPs, were reported to be feasible in a study involving 1000 pharmacy consumers⁶⁸. This approach optimises the interaction between pharmacies and GPs, which is essential in many aspects, such as medication reviews and detecting high blood pressure at pharmacies. Nonetheless, for large-scale implementation, the issues of resources (training large numbers of pharmacists and providing them with screening devices) and providing financial incentives for pharmacies to sustain AF screening would all need to be addressed. However, it was encouraging to read that AF screening was included in the pharmacy curriculum of a university, and their pharmacy students were trained to perform AF screening using handheld single-lead ECG devices as a pharmacy community service^62,64.

Remote self-screening

A significant number of studies (30/87) focused on self-screening and remote monitoring approaches to AF screening; most of these were in high-income countries, with roughly half using a mobile handheld device and the other half using a wearable device (Table 3) and 6 studies compared the self-screening approach to usual care, and all showed superior AF diagnosis rates compared to usual care (Table 1). For example, in Sweden, the STROKESTOP study involved 7173 people aged 75–76 years in self-recording ECGs using handheld ECG devices remotely twice daily over 2 weeks, and the study reported a net benefit of screening and treatment compared to usual care¹¹². This population-based screening systematically identified eligible participants and invited them to participate, which was different from opportunistic point-of-care screening. This type of systematic screening offers equitable access to eligible populations, while opportunistic point-of-care screening is subjective to the context of each health visit and the interaction between the clinician and patient.

Remote monitoring typically involves participant-led self-screening, and a central team (including trained personnel/ technicians/ research nurses and cardiologists) that remotely monitors, reviews and interprets the participants’ ECGs^{87,106,111,112,137}. One study subcontracted ECG interpretation to an external service provider⁹⁹, and other studies incorporated a two-stage diagnostic approach. When an irregular pulse was detected by the smart devices’ photoplethysmography sensors, participants were sent an ECG patch to be worn for a week, which was analyzed upon return^102,105. This type of market-driven wearables and screening of broader, often low-risk populations, potentially stresses the already busy general practitioners who must address potential abnormalities, which are often false positives. Additionally, it poses challenges to providing integrated continual care, including how to fund remote ECG interpretation services, ensure patients with abnormal findings access the necessary care, and integrate the screening results with patient medical records in their usual medical practice.

Remote self-screening as a triage to further investigations was an approach applied in the “mobile Atrial Fibrillation App” (mAFA) I and mAFA II trials”^{95,96,97,98,138}, Apple Heart Study¹⁰⁵, and Fitbit Heart Study^102,125. In the mAFA I & II trials, participants with “possible AF” detected by the Huawei watches or wristbands were further investigated using ECG or Holter monitors through the designated mAFA telecare team. The integrated management included anticoagulants for stroke prevention, better symptom management, cardiovascular and comorbidity risk management, and educational programs. However, modest rates of effective follow-up were reported (Table 1)^{94,98,138,139}. In Apple Heart Study¹⁰⁵ and Fitbit Heart Study^102,125, when the wearable devices detected irregular heart rhythms, the participants were provided with a telemedicine visit and mailed an ECG patch to be worn for a week. The researchers reported many participants visited their usual healthcare providers outside the research study when they received irregular rhythm notifications, implying that this could be patients’ usual health-seeking behavior when a potential abnormality with their health was detected. Therefore, the remote monitoring systems should be integrated into the patient’s usual primary care providers to ensure continuity of care. Despite the surveys showing the general acceptability and usability of this screening approach^94,99,108, there was a lack of in-depth qualitative process evaluation using a robust framework^22,23,24,25 to evaluate this model of screening.

Reaching out to the communities

In 10 (12%) studies, community screening occurred in various settings and involved a dedicated in-person screening team (e.g., screening booths in communities or screening campaigns in villages). In Hong Kong, screening stations in community centers were set up to screen eligible people in the community and the rhythm traces were interpreted by the researchers remotely^74,75. This screening model could be feasible in some settings for people who can access screening venues, but it may miss out on people in more remote areas. Some studies did describe approaches to integrating AF screening in rural health facilities to make screening more accessible to rural residents, including Indigenous communities^58,77. In a study in India, health workers visited and screened villagers using handheld single-lead ECG devices⁸⁴, and in Denmark, caregivers visited older people at homes and screened them using mobile devices⁸².

Evaluation of screening implementation

Only a few studies included some components of a structured implementation evaluation. This was most commonly in the form of implementation evaluation surveys among intervention screening program participants, addressing the acceptability and usability of the screening programs^94,99,108. However, most survey evaluations did not report the contextual enablers and barriers to implementation because they did not have an evaluation framework^22,25,140. Only a small percentage (7/87, 8%) of studies included in-depth interviews of participants and fewer in-depth interviews of personnel involved in the screening program implementation. Such qualitative evaluations could have elicited further information on enablers and barriers to implementing AF screening programs^{45,47,48,49,68}. The Realist²³ or Critical Realism²⁴ evaluation approach, where the implementation process, context, mechanism of impact and outcomes were examined, could be adopted as in some selected studies^{48,127,141,142}. Generally, there is a lack of application of evaluation frameworks, such as the Medical Research Council Guidelines on Process Evaluation²² and the RE-AIM²⁵, and a lack of transparency of refinements to interview processes¹⁴¹.

This review did not aim to examine screening costs, which entails a different literature search strategy. Cost-effectiveness is important in considering the future implementation of AF screening programs. This review did identify a spectrum of screening devices ranging from handheld devices to wearable patches being used, and the unit costs of the various devices do vary¹⁴³, and the cost implications are contextual to the country and application settings¹⁴⁴. In a study using nurses to perform pulse palpation followed by a 12-lead ECG when an irregular rhythm was detected, Morgan S et al. estimated £6 per consultation with a practice nurse, resulting in a cost per AF case detected of £186 (95% CI = £138–£300) for a yield of 4.5% AF detection rate using systematic screening compared with 1.3% employing opportunistic screening⁴⁶. Using a similar pulse palpation approach, Hobbs et al. estimated £337 for each additional AF case detected by opportunistic screening compared to usual care, for a yield of 1.63% (opportunistic screening) versus 1.04% (usual care) AF detection rates⁴¹. Using handheld single-lead ECG devices to screen participants ≥65 years old twice weekly over a year yielded an almost fourfold increase in AF detection compared with usual care, and the estimated cost was £8255 per additional AF diagnosis⁹⁹. Studies that applied devices that could continuously screen for months, such as the study using implantable loop recorders over a median follow-up of 64.5 months, would entail providing devices and interrogating large amounts of data and, as such, would be far more costly¹¹¹. Continuous remote monitoring could yield a higher AF detection rate at a higher cost but may identify more low-risk AF patients. Moreover, there is a lack of reports on the cost-effectiveness of such studies.

Strengths and limitations

A structured literature search strategy (Supplementary Table 1) was used, yielding a spectrum of AF screening across various settings (general practice, pharmacy, community centers, villages, hospitals, and remote screening) over a 24-year period. This provided insights into the utilities of various technologies and screening approaches. However, the search was limited to one database (PubMed). Systematic review approaches of replication of search and data extraction using multiple databases were not applied.

The current review has identified and synthesized the findings of 87 studies on AF screening approaches, technology, and program implementation. The findings demonstrate a large breadth of approaches to implementing AF screening but, somewhat surprisingly, lack strong data on the effectiveness of approaches with few RCTs or studies with comparator groups and very limited information to inform implementation or scale-up beyond the research setting. Despite the recommendations in many guidelines, the relative dearth of information on how to implement AF screening may be a reason for the lack of widespread implementation of AF screening. This review suggests that primary care is a feasible location for implementing AF screening, at least in some existing healthcare systems, with additional resources including relatively low-cost mobile devices and designated personnel to lead AF screening that targets individuals 65 years and older. Yet further research to examine other community approaches, especially for settings in which primary care services are not integrated or accessible, is needed. More recent studies show that a potential alternative to primary care screening is the implementation of remote screening programs, but remote self-screening programs need careful evaluation amongst older populations. However, there is limited remote screening research with comparator groups that target older people. Similarly, there is limited evaluation of the potential adverse impact of unguided population-wide screening on an already over-burdened healthcare system (and especially the primary care sector). The lack of integration of these remote screening initiatives into existing healthcare systems could hinder their implementation. Finally, this review provides limited data on the financial models and cost-effectiveness of AF screening initiatives, and further research is needed to establish cost-effective approaches to scale up AF screening. AF screening is recommended by many guidelines, and this review synthesises what we know from the last 24 years but also highlights the practical information gaps that will help implement AF screening and integration into existing health systems in the future.