Quality in Primary Care Open Access

Keywords

atrial fibrillation, flu vaccination clinics, general practice, primary care, screening

Introduction

Atrial fibrillation (AF) confers a fivefold risk of stroke.[1] Anticoagulation reduces stroke risk by 68%,[2] making diagnosis worthwhile. However, estimates suggest half of those who could benefit from treatment are not receiving it[1] due to failure in diagnosis of asymp-tomatic patients and initiating treatment. Action is essential given the projected increase in strokes sec-ondary to AF with ageing[3] and increasingly obese populations.[4]

The Royal College of Physicians Edinburgh suggest opportunistic screening of those aged 65 years by pulse check, and 12-lead electrocardiography (ECG) for those with pulse irregularity,[5] whilst the British Heart Foundation[6] and NHS Improvement[7] advocate pulse screening at influenza vaccination (flu) clinics. The SAFE study (a large multicentre randomised con-trolled trial comparing targeted and total population screening versus routine practice for the detection of AF in patients 65 years old by pulse assessment at 50 primary care centres across the West Midlands) concluded that screening was more effective at detecting new cases of AF than routine care, opportunistic and systematic screening being equivalent, and systematic screening being more expensive.[8]

Moorlands Medical Centre (MMC) used electronic blood pressure (BP) devices that did not identify pulse irregularities,[9] and the practice computer template did not prompt pulse palpation. This raised concerns regarding the effectiveness of opportunistic screening.

Subsequent analysis of MMC’s practice records revealed an AF prevalence of 1.7% (Table 1), whilst the Quality Outcomes Framework (QOF) visit from 2009 reported an AF prevalence of 2.1% (A Foster, personal communication, September 2011). The reduced prevalence given an ageing population warranted investigation.

The SAFE study revealed similar AF prevalence in most age and gender categories (Table 1) compared with the MMC data. Noticeable differences existed amongst men and women aged 65–74 years, with prevalence of 6.9 and 4.1%,[8] as opposed to 4.77 and 2.24%, respectively. The SAFE study’s assertion sup-porting opportunistic screening depends on high rates of pulse assessment (70% of eligible patients were screened opportunistically). Was this level of oppor-tunistic pulse palpation missing at MMC?

Systematic screening increases AF detection by 60%.[6] More systematic ‘opportunistic’ screening of patients attending flu clinics has been conducted as pilots and local enhanced services (LES).[10] Reports concluded that this was an effcient method to screen opportunistically, with minimal disruption to surgery practice and effective use of resources and time.[11,12] The largest scheme allowed ECG evaluation of irregular pulses immediately or later at the practice’s discretion.[11]

This study assessed whether screening for atrial fibrillation without additional Funding amongst those aged 65 years attending flu clinics was effective, practical and feasible. Satisfied by discovering undiag-nosed cases, estimating the undiagnosed AF burden, assessing the accuracy of a second-year General Practice Specialty Trainee (GPST2) and interpretative software at diagnosing AF on ECG, completion without dis-rupting routine practice, estimating cost-effectiveness and guiding future practice screening activities for AF.

Methods

MMC has 8500 patients, and provide primary medical care in Leek, UK. Screening was approved by the partners and patients’ participation group. Ethical ap-proval was deemed unnecessary. Pulses were assessed by a GPST2 and a medical student. Administrative staff recalled patients for ECGs conducted by a nurse.

Radial pulse was assessed for 30 seconds, and regularity rather than rate was determined. Patients were categorised according to pulse as normal or abnormal. Simple irregular and regular categorisation offers better sensitivity, but correspondingly lower specificity than looking for irregularly irregular pulses (sensitivity 91% and specificity 74%, as opposed to 54 and 98%, respectively).[13]

Patient selection

Patients aged 65 years attending flu clinics during 2011 were offered pulse assessment post flu vacci-nation. Patients’ screening suitability (sole inclusion criterion age 65 years, exclusion criterion age 64 years) was assessed by the GPST2 and medical student before they left the surgery post vaccination. The purpose of screening was explained, verbal consent obtained and refusals noted. Past medical history was not known or sought. Patients with pulse abnormali-ties were informed that they required an ECG and would receive an appointment by post. An informa-tion leaflet was offered outlining pulse assessment, AF and the need for subsequent ECG. Patients failing to attend ECG appointments were phoned and/or sent another letter.

ECG

ECGs were conducted over four months by a practice nurse. A Biolog 3000 was used for all ECGs, except five when the machine broke; the second machine was a CardiofaxQ. The interpretative software used was CardioView[1] (software version 5.0.0.110, QRS Diag-nostics, 2009).

Interpretation of ECGs

ECGs were interpreted by the GPST2, who was blind to the interpretative software report, but aware that the indication was pulse irregularity. Subsequently, all ECGs were faxed to Broomwell HealthWatch TeleMedical Monitoring Services where they were interpreted by a cardiac physiologist or nurse specialist with Peer Review by a cardiologist (taken as reference standard). They were not blinded to the software report, but were aware of the indication and blinded to the GPST2’s interpretation.

Patients identified with new AF after ECG (primary end point) were reviewed by practice doctors.

Results

Practice records analysis

Analysis of MMC practice records revealed 1714 registered patients aged 65 years (Table 1). A total of 797 patients attended flu clinics during 2011. Of these, 573 were assessed as eligible for screening, being years). Subsequently, 573 patients had a pulse assess-ment, correspondingly screening 33% of the practice list aged 65 years.

Pulse assessment, yield and prevalence estimates

A total of 573 patients were assessed, of whom 95 had an irregular pulse, and from these 2 new cases of AF were diagnosed. Of the 95 patients with irregular pulses, 21 were excluded after medical records revealed prior diagnosis of AF, 5 were excluded as analysis of records revealed they were aged < 65 years (all had ECGs conducted, none revealed AF); another was excluded after admission to hospital with a myocar-dial infarction (MI; his ECG in hospital did not show AF). Subsequently, 68 ECG invitations were sent. Thirty-nine patients attended over four months after a second set of reminders with appointments were sent out. In total, 57% of the patients identified as having an irregular pulse attended for an ECG.

Two ECGs were rejected by the cardiologist due to their quality. Five were excluded as the ECGs were done on a different machine, with different interpretative software, and the cardiologists found the printouts illegible when faxed. Therefore 32 ECGs were inter-preted fully, revealing 2 new cases of AF (Figure 1).

Of the MMC screened population of patients with irregular pulses, 25.5% had AF (by subsequent records analysis or ECG). Both ‘new AF’ cases were aged 75 years. CHA₂DS₂-VASc[14] scores were 3 and 7, confer-ring an annual stroke risk of 3.2 and 9.6%, respectively. Both patients were counselled by practice clinicians regarding stroke risk and advising oral anticoagu-lation, but both chose aspirin.

The yield of patients with AF was 4%. The yield with ‘new AF’ was 0.35% (Box 1).

Age stratification of patients having ECGs

Time constraints prevented the collection of age data for all patients screened, and only those having ECGs were categorised (Table 2).

Cardiologist rhythm analysis

In total, 36% (14/39) of ECGs confirmed potentially irregularly irregular rhythms. Two showed AF; the others showed ectopic beats.

ECG interpretation

The GPST2 identified all cases of AF on ECG as did the interpretative software. No cases were mistakenly diagnosed as AF, correspondingly sensitivity, speci-ficity, PPV and NPV was 100% for both GPST2 and interpretative software for ECG interpretation regard-ing diagnosing of AF compared with cardiology opinion.

Discussion

Discovery of undiagnosed cases

Two ‘new AF’ cases were identified, a new case being discovered for every 286 patients screened. We as-sessed 573 patients, and 95 had abnormal pulses (16.6%). Thirty-nine of the 95 patients had an ECG (57%), 23 had AF (24.2%), and two were new cases (8.7%). In comparison, the SAFE study assessed 3278 patients in their opportunistic arm, 361 patients had abnormal pulses (11%), 238 had an ECG (65.9%), 84 had AF (35.3%), and 31 were ‘new’ AF diagnoses (36.9%). One new case was discovered for every 106 patients screened.[8] Clearly, opportunistic screening seems more effcient. Did this reflect the fact that MMC patients were already well screened opportunistically, that we conducted the screening badly, or both?

In the SAFE study, 61.8% of pulses were taken by GPs and 27.8% by practice nurses. In our screening, 58.8% of pulses were taken by the medical student and 41.2% by the GPST2. The MMC assessors’ inexperi-ence may account for more abnormal pulses being identified (16.6 vs. 11%) and less AF discovered (24.2 vs. 35.3%). The former is more likely; sensitivity is less likely to be lost than specificity (supported by the number of MMC ECGs showing ectopics).

The SAFE study reported a false-negative rate of 2% and a false-positive rate of 70% for pulse assessment.[8] The false-positive rate in MMC screening was 74% (Box 1), comparing favourably. Accepting satisfactory screening quality, this suggests that the MMC pro-vided a good opportunistic service in comparison with other practices, supported by the baseline prevalence of AF in those 65 years being 7.7% as opposed to 6.9% in the SAFE study.[8] Consequently, the majority presenting with AF on palpation of a radial pulse had been diagnosed previously (21 of 23). However, this reinforces the fear that patients who do not attend the practice are not being screened effectively for AF and we were re-sampling a previously screened population (only 2 of 23 pulses confirmed to have AF were new cases of AF).

Estimated practice prevalence

Assuming a representative sample and robust screening process, the MMC potentially has another four un-diagnosed cases of AF (Box 2). However, the sample appeared skewed, with the largest proportion coming from the 75–84 years group (3.3%), with those aged

65–74, 85–89 and 90 years poorly represented (Table 2). When considering the increasing prevalence of AF with age (Table 1), coupled with the previously noted lower recorded prevalence in the 65–74-year-old category at the MMC, and the missed diagnoses due to delay in conducting an ECG (for paroxysmal AF) or non-attendance for ECG, four additional patients seems a conservative estimate.

The SAFE study’s AF prevalence amongst those aged 65–75 years was 4.6%, and amongst those aged 75 years was 10.3%.[8]  Pre-screening data at the MMC revealed a prevalence of 3.5% in those aged 65–75 years, and 12.2% in those aged 75>years, rising to 12.4% after screening. Screening would be more fruitful amongst those 75 years old.  

Concerns regarding patient selection

Those patients aged 85 years may be poorly rep-resented because frailty or transport diffculties made attending surgery diffcult. A similar pattern was observed in the SAFE study, with 61% and 60% of men and women aged < 75 years attending screening; for those aged 75 years attendance fell to 50% and 39%, respectively.[8] MMC practice nurses visit older patients to give flu vaccines; subsequently, these patients do not attend flu clinics. The 65–74-year-old group’s poor attendance may represent poor uptake of the vaccine, or that flu vaccines are given opportunistically to patients attending routine ap-pointments, or at work, local chemists or supermar-kets. Given that the sample had only 39 patients, the reliability of these assertions is uncertain. Interest-ingly, concern regarding the applicability of screening to actual primary care populations due to poor par-ticipation rates among the elderly has been raised previously.[15] The MMC experience suggests that those most at risk of AF-related stroke (older patients who do not attend regularly, therefore missing opportun-istic pulse screening) are least likely to undergo screening at flu clinics.

Accuracy of ECG interpretation

Interpretation of ECGs at MMC was accurate com-pared with interpretation by cardiologists (taken as reference standard). However, with only two cases of AF and a small sample (32 fully interpreted ECGs) it is diffcult to be confident in the validity of the accuracy of the GPST2 and interpretative software. In addition, the cardiologist was not blinded to the interpretative software report, and this introduces possible bias.

Learning points for future screening

As noted by prior studies, the low specificity of pulse assessment made ECG confirmation of diagnosis an essential, but time- and resource-consuming, aspect of screening.[8,13] No patient refused pulse assessment, evidencing the acceptability and convenience of screening at flu clinics. However, uptake of the ECG appointment was poor (57%), even after additional appointments were offered. Because of non-attend-ance, the last ECG was not conducted until four months after the pulse check. The poor uptake of ECGs and delay in conducting ECGs are the main limitations of this study. This offers clear learning points. First, reinforcing to patients the value of interventions to prevent stroke more clearly might have increased ECG uptake. Second, the MMC com-pares poorly with previous trials which had higher ECG uptake rates of 73.3[13] and 74%.[16] However, in these, ECGs were done on the same day as the pulse assessment. This could not be provided without ad-ditional finances at the MMC, or without disrupting the flu clinic and routine clinical appointments. A similar exercise involving invitation for ECGs achieved a comparable response rate (56%).[15]

Fitzmaurice et al noted that a third of AF in populations aged > 65 years is either self-limiting or paroxysmal.[17] Clearly, ECGs must be conducted im-mediately after pulse assessment to increase ECG uptake and identify paroxysmal AF (which carries the same stroke risk as permanent AF).[18] Additional Funding is required to provide the capacity to conduct large numbers of ECGs in a short period, which is a necessity for screening at flu clinics.

Completing ECGs immediately after pulse palpation during screening at MMC could have identified four further new cases of AF, giving a number needed to screen to identify a new case of AF of 95 rather than 286, and providing a yield of new AF cases identified by screening of 1% rather than 0.35% (see Box 2).

Resource use and cost-effectiveness

The exercise required four afternoons assessing pulses (incurring no direct costs), 13 hours of nursing time conducting ECGs, costs incurred (and potential rev-enue lost from alternative remunerated work) due to wasted appointments subsequent to non-attendance, ECG interpretation and cardiology review (no add-itional cost, cardiology reporting was a pilot service), and disposables (letters, postage, ECG) provided by the practice and clinic costs (not charged).

Identifying a new case of AF cost approximately £234 (see Box 3); in comparison an LES reported a cost of £372.[11] However, as neither patient was anticoag-ulated, the investment was futile. The patients were counselled appropriately and made their decisions. The cost to prevent a stroke in year one assuming anticoagulation was approximately £9911 (see Box 4), again comparing favourably with published LES values of £11 594 to £17 534.[11]

The SAFE study estimated that the minimum worthwhile change in detection rate for screening versus routine practice was 1%.[8] Routine practice at MMC identified an AF prevalence of 3.7% in the screened sample 65 years. Post screening, the preva-lence rose to 4%. Clearly screening for AF at MMC flu clinics should not be repeated.

Should other practices screen at flu clinics?

Yields of new AF depend on the disparity between the true and known prevalence of AF in the population sampled. Practices with a low prevalence should review screening practices and improve opportunistic pro-vision (aiming for 70% annual screening) before screening at flu clinics. Screening populations at flu clinics that have already been well screened oppor-tunistically wastes resources and reduces the cost-effec-tiveness of screening.

Indiscriminately screening large populations (in-cluding various practice subpopulations with a variable undiagnosed burden of AF) may identify significant numbers of new patients with AF, making the exercise appear cost-effective. However, patients may be screened unnecessarily (if already screened opportunistically), and screening low-risk subpopulations (given lower prevalence) will identify fewer cases, with lower indi-vidual stroke risks, leading to poorer cost-effectiveness. Targeting screening to practices with lower diagnosed prevalence or lower rates of opportunistic screening, whilst investigating the value of targeted screening of patients with medical conditions commonly coexistent with AF, causing or perpetuating AF, or representing global cardiovascular risk factors, seems prudent. This may increase screening yield, and improve cost-effec-tiveness of screening and treatment by greater risk reduction with treatment for high-risk patients (Box 5).

The prevalence and incidence of AF increase dra-matically with age. Annual opportunistic assessment for AF as endorsed by the SAFE study[8] may be appropriate for younger patients, but more frequent assessment may be prudent as patients age and de-velop comorbidities increasing stroke risk.[3,14,19]

Conclusion

The MMC experience highlights the challenge of identifying people with asymptomatic AF and initi-ating treatment, thereby preventing strokes. Two new cases of AF eligible for anticoagulation were identified, but neither patient consented to anticoagulation. On-going education for professionals and patients is warranted, highlighting the clear benefits of identifying and treating AF with oral anticoagulation rather than aspirin.[18]

Patients aged 65–74 and 85 years were not effectively screened by routine practice or screening at flu clinics, posing ongoing challenges regarding screening these groups. Screening at flu clinics was acceptable to patients and feasible, but conducting ECGs immediately after pulse palpation is essential.

Screening targeted according to practice demo-graphics and lower than expected AF prevalence may offer better cost-effectiveness than indiscrimi-nately screening at all flu clinics. Practices should insert pulse checks into chronic disease templates and prompts for patients aged 65 years who attend the surgery.

Our experience and the literature suggest that clinical commissioning groups should encourage practices to reach 70% annual opportunistic screening rates amongst those aged 65 years,[8] before employing alternative, resource-demanding screening strategies such as flu clinics for practices with lower than expected diagnosed AF prevalence.

Novel fingertip devices[20] and newer blood pressure devices[21,22] offer encouraging alternative screening strategies that generate significantly fewer false posi-tives when identifying people with potential AF com-pared with palpation of the pulse (thereby requiring fewer confirmatory ECGs).