Development of a Novel Algorithm to Detect Atrial Fibrillation Using an Automated Blood Pressure Monitor With an Irregular Heartbeat Detector

Abstract

Background: Atrial fibrillation (AF), which contributes to an increased risk of stroke, frequently remains undetected, suggesting an unmet need for easier and more reliable AF screening. The reports on screening AF using an Omron blood pressure (BP) monitor with an irregular heartbeat (IHB) detector show inconsistent results, so the aim of this study was to develop a novel algorithm to accurately diagnose AF with 3 BP measurements using an Omron automated BP monitor with IHB detector.

Methods and Results: In total, 303 general cardiac patients were included. Real-time single-lead ECG revealed AF in 44 patients. BP measurement was performed 3 times per patient using the Omron BP monitor HEM-907, and the number of IHBs detected was recorded. Based on these data, we developed the following algorithm: ≥1 IHB is detected during at least 2 of 3 BP measurements and the maximum number of IHBs detected is ≥2. Using this algorithm, we achieved a sensitivity of 95.5% and specificity of 96.5%, for diagnosing AF.

Conclusions: The novel algorithm with 3 BP measurements using the Omron automated BP monitor with IHB detector showed high sensitivity and specificity for diagnosing AF in general cardiac patients.

Atrial fibrillation (AF) is one of the most common cardiac arrhythmias.¹ It contributes to an increased risk of not only ischemic stroke but also heart failure, dementia, and death.^2–4 As society has aged, the prevalence of AF has increased. Moreover, AF is often underdiagnosed because many patients are asymptomatic, or paroxysmal AF does not always occur while the patient is visiting the clinic.⁵ Therefore, an accurate screening device for AF is an unmet medical need worldwide.

It has been proposed that automated blood pressure (BP) monitors be used to screen AF, because they are already widely used. Many elderly people have hypertension, and regularly perform automated BP measurements at home. If AF can be detected while BP measurements are being taken, it will increase the chance of opportunistic detection of asymptomatic AF patients, and they will be prompted to visit the hospital. Kabutoya et al confirmed a good association between the RR intervals on ECG and the pulse wave intervals detected with an automated BP monitor, and successfully demonstrated that an automated BP monitor could potentially be effective for screening for AF.⁶ Wiesel et al reported that an automated BP monitor made by Microlife, which contains an algorithm for diagnosing AF, could screen for AF with 3 BP measurements.^7–9 Based on those results, the UK primary care guidelines recommend using the Microlife BP monitor for home BP measurements by the elderly.¹⁰

Some models of automated BP monitors made by Omron have an additional function that detects irregular heartbeats (IHBs) during BP measurements. An IHB is defined as a rhythm that is ≤25% or ≥25% than the average rhythm of a pulse wave detected during BP measurement. There are several reports examining the utility of the Omron BP monitor with an IHB detector in screening for AF, but the results have not been consistent, possibly because all past studies used only 1 BP measurement.^11,12

In this study, we evaluated the ability to diagnose AF in general cardiac patients when 3 BP measurements were performed with an Omron automated BP monitor equipped with IHB detector. Furthermore, we developed a novel algorithm to accurately diagnose AF with an Omron BP monitor based on data from the number of IHBs detected during 3 BP measurements.

Methods

Subjects

Patients who were hospitalized or visited outpatient clinics at Kagawa University Hospital or who entered welfare-based nursing homes for the elderly in Kagawa were recruited. The exclusion criteria were age under 20 years, the presence of a pacemaker, and/or an implanted defibrillator and refusal to participate. Patients who could not remain at rest while seated, who could not have BP measured in the upper arm, such as dialysis patients, and any who were judged unsuitable for the study by their doctor were also excluded. If the Omron BP monitor stopped automatically and showed an error sign while measuring BP, we did not perform additional BP measurements in order to avoid unnecessary pain and thus also excluded those patients from the analysis.

In total, 345 consecutive patients were recruited; 9 patients were excluded because a medical history of pacemaker implantation was found after they provided informed consent; 31 patients (including 9 AF patients) showed an error sign during BP measurements; 2 patients whose data could not be collected because of technical or mechanical difficulties were excluded. Therefore, data from the remaining 303 patients were analyzed (Figure 1). Medical information of all patients was collected from their medical records.

Figure 1.

Flow chart of patient recruitment and participation. BP, blood pressure.

Procedure

The automated oscillometric BP monitor used for this study was an Omron HEM-907 (OMRON Healthcare, Kyoto, Japan), which is designed for clinical use in a physician’s office or in hospital. This device has an additional function that detects the number of IHBs (as defined earlier). When the BP monitor detects 1, 2 or ≥3 IHBs during BP measurement, the monitor shows 1, 2 or 3 dots, respectively. This device has passed the Association for the Advancement of Medical Instrumentation validation protocol and has been used in other clinical studies.^13,14

BP measurement was performed 3 times for each patient with the Omron HEM-907 in accordance with the Japanese Society of Hypertension Guidelines for the Management of Hypertension 2014.¹⁵ After at least 5 min of seated rest, the 1st BP measurement was performed, and subsequent measurements were taken at 1 min intervals. The systolic BP, diastolic BP, pulse rate, and the number of IHBs detected were recorded for each BP measurement.

Real-time single-lead ECG was recorded continuously during all BP measurements using the HealthPatch MD (VitalConnect Inc., CA, USA), ReadMyHeart (TRYTECH Co. Ltd., Tokyo, Japan), or Duranta (ZAIKEN Co. Ltd., Tokyo, Japan). The recorded ECGs were analyzed by 2 board-certified cardiologists who were blinded to the results of the BP measurements. The patients with AF were classified into the “AF” group and all patients without AF were classified into the “Non-AF” group. Patients in the “Non-AF” group were further classified into an “Other arrhythmias” group with arrhythmias except AF and the “Sinus” group with normal sinus rhythm. All patients diagnosed with sinus arrhythmia were classified as “Sinus” group because respiratory sinus arrhythmia was judged as being effectively harmless; 2 patients required additional reference to their medical records and a 12-lead ECG to diagnose their rhythm.

Taking the single-lead ECG as the reference method for definitive diagnosis of AF, the diagnostic values (sensitivity, specificity, and accuracy) for diagnosing AF were calculated in every case, with both the number of IHDs detected and the number of BP measurements in which IHBs were detected. Furthermore, we analyzed the data based on the number of IHDs detected during 3 BP measurements for the purpose of developing a novel algorithm that could diagnose AF more accurately.

The association between the RR intervals on the real-time single-lead ECG and the pulse wave intervals detected by the Omron BP monitor was checked with 50 intervals from the first 3 AF patients in the same way as reported by Kabutoya et al.⁶

The protocol was approved by the Kagawa University Ethics Committee, and signed informed consent was given by all patients.

Statistical Analysis

The baseline characteristics data are expressed as the mean±standard deviation (SD). The association between the RR intervals on ECG and the pulse wave intervals recorded by the Omron BP monitor was assessed using Pearson’s product-moment correlation coefficient. The sensitivity, specificity, and accuracy for the diagnosis of AF were calculated by generating 2 contingency tables with the single-lead ECG diagnosis as the reference. A significant association between 2 variables on a contingency table was determined using the Chi-square test, and P<0.05 was considered significant. The 95% confidence intervals for the sensitivity, specificity and accuracy were calculated using the Agresti-Coull method.¹⁶ All calculations were performed with SAS software (University Edition; SAS Institute Inc., Cary, NC, USA).

Results

Among the 303 patients, 252 (83%) visited the hospital’s outpatient clinics or were hospitalized in Kagawa University Hospital, and 51 patients (17%) were admitted to welfare-based nursing homes for the elderly. The mean age of the subjects was 72.2 years; 124 patients (40%) were over 75 years old, 23 patients (7%) were over 90 years old, and the oldest patients were 99 years old (Table 1).

Table 1. Baseline Characteristics of the Study Patients

	All patients (n=303)	AF (n=44)	Non-AF
			Sinus (n=211)	Other arrhythmias (n=48)
Age (years)	72.2±13.5	76.3±10.3	71.1±14.5	73.6±11.2
Male (%)	61.7	77.3	58.8	60.4
Body mass index (kg/m2)	23.5±4	22.7±3.4	23.7±4.2	23.5±3.3
Hypertension (%)	61.7	54.0	60.7	70.8
Diabetes mellitus (%)	29.0	15.9	31.3	29.2
Congestive heart failure (%)	21.5	45.5	15.2	25.0
TIA/cerebral infarction (%)	15.5	11.4	14.7	20.8
Vascular disease (%)	28.7	18.2	32.2	22.9
Systolic BP (mmHg)	131.0±18.8	124.8±14.3	132.3±19.0	129.6±20.0
Diastolic BP (mmHg)	71.9±13.1	70.4±14.3	72.6±12.6	69.9±13.9
Pulse pressure (mmHg)	59.0±14.6	54.4±12.4	59.6±14.9	59.6±13.8
Pulse rate (beats/min)	71.0±12.0	77.4±12.8	70.4±11.1	68.4±13.5

Data are expressed as mean±SD. AF, atrial fibrillation; BP, blood pressure; TIA, transient ischemic attack.

Based on the results of the ECG diagnosis, 44 patients (14%) had AF and were classified as the “AF” group. Among the 259 patients in the “Non-AF” group, 211 (70%) were classified as the “Sinus” group and 48 (16%) were classified as the “Other arrhythmias” group. The “Sinus” group included 15 patients with sinus arrhythmias. Among the 48 patients in the “Other arrhythmias” group, 29 had atrial premature contractions, 15 had ventricular premature contractions, 2 had both atrial and ventricular premature contractions, 1 had an atrial flutter and 1 had an atrial tachycardia.

More than 60% of the patients included in the analysis had hypertension. Compared with the patients in the “Non-AF” group, those in the “AF” group were older, had a higher prevalence of compensatory heart failure and a lower prevalence rate of diabetes and vascular disease. No differences in BP were observed between the “AF” and “Non-AF” groups, but the pulse rate was higher in the “AF” group than in the “Non-AF” group (Table 1).

A representative pulse wave detected by the Omron BP monitor and the real-time ECG recorded at the same time are shown in Figure 2, showing clearly the good association (r=0.991, P<0.01) between the RR interval on ECG and the pulse wave interval detected by the Omron BP monitor.

Figure 2.

Association between (A) the pulse wave interval recorded by the Omron BP monitor and (B) the RR interval on real-time single-lead ECG record corresponding to the pulse wave in an AF patient is shown in (C). Time is shown in milliseconds. AF, atrial fibrillation.

The number of IHBs detected during the 1st, 2nd, and 3rd BP measurement using the Omron BP monitor HEM-906 with an IHB detector was analyzed, and the diagnostic values (sensitivity, specificity, and accuracy) for diagnosing AF were calculated in each case (Supplementary Table). When at least 1 IHB was detected during a BP measurement (1st, 2nd or 3rd), the diagnosis of AF had a sensitivity within the range of 86.4–88.6% and a specificity within the range of 90.3–93.8% (Table 2, Supplementary Table). When at least 1 IHB was detected during 3 BP measurements, the sensitivity for diagnosing AF reached 100% but specificity was only 79.5% (Table 3). When at least 1 IHB was detected during at least 2 of 3 BP measurements, the diagnosis of AF had a sensitivity of 95.5% and a specificity of 93.8%. Furthermore, when at least 1 IHB was detected during at least 2 of 3 BP measurements and the maximum number of IHBs during 3 BP measurement was ≥2, the diagnosis of AF using the novel algorithm had sensitivity, specificity and accuracy of 95.5%, 96.5% and 96.4%, respectively. Importantly, all patients in the “Sinus” group were excluded by this novel algorithm. In the “Non-AF” group, 39 of 48 patients were also excluded by this algorithm, and the remaining 9 patients, who had frequent premature beats, were not distinguished from patients with AF.

Table 2. 2×2 Contingency Tables

	BP monitor reading	ECG reading
		AF	Non-AF
(A) At least 1 IHB detected during 3 BP measurements	Positive	44	53
	Negative	0	206
(B) At least 1 IHB detected during 2 of 3 BP measurements	Positive	42	16
	Negative	2	243
(C) Novel algorithm with 3 BP measurements	Positive	42	9
	Negative	2	250

(A) Positive reading was defined as when at least 1 IHB was detected during 3 BP measurements. (B) Positive reading was defined as when at least 1 IHB was detected during at least 2 of 3 BP measurements. (C) Positive reading was defined as when at least 1 IHB was detected during at least 2 of 3 BP measurements and the maximum number of IHBs detected during 3 BP measurements was ≥2. AF, atrial fibrillation; BP, blood pressure.

Table 3. Diagnostic Values for AF

	Sensitivity	Specificity	Accuracy
At least 1 IHB detected during 3 BP measurements	100% (95% CI 90.4–100)	79.5% (95% CI 74.2–84.0)	82.5% (95% CI 77.8–86.4)
At least 1 IHB detected during 2 of 3 BP measurements	95.5% (95% CI 84.0–99.6)	93.8% (95% CI 90.1–96.2)	94.1% (95% CI 90.8–96.3)
Novel algorithm with 3 BP measurements	95.5% (95% CI 84.0–99.6)	96.5% (95% CI 93.4–98.3)	96.4% (95% CI 93.5–98.0)

AF, atrial fibrillation; BP, blood pressure; CI, confidence interval; IHB, irregular heartbeat.

Discussion

The present study showed diagnostic values for AF when 3 sequential BP measurements using an Omron automated BP monitor with an IHB detector were performed. When these measurements using this monitor were performed and the novel algorithm was subsequently applied, the diagnosis of AF achieved a sensitivity of 95.5%, specificity of 96.5% and accuracy of 96.4% in general cardiac patients.

AF Detection Algorithm With Automated BP Monitor

Diagnostic performance for AF detection with an automated BP monitor has been demonstrated in recent years.¹⁷ Kabutoya et al reported that both the sensitivity and the specificity for diagnosing AF reached 100% when the definition of an IHB was changed from “25% variation” to “20%” or “15%” in 20 sinus and 16 AF patients using an automated BP monitor made by A & D.⁶ It was suggested that AF could be distinguished from sinus rhythm easily by tightening the criteria of an IHB, but their study subjects did not include patients with non-AF arrhythmias. There is concern that the more stringent the IHB criterion, the greater the number of false positives, especially in patients with non-AF arrhythmias.

Wiesel et al reported that the Microlife BP monitor could diagnose AF with 90–100% sensitivity and 89–99% specificity.^7–9 The Microlife BP monitor had the original algorithm, which has a coefficient of variation (CV=SD/M) obtained from the mean (M) and standard deviation (SD) of the pulse wave interval, and CV ≥0.06 was a positive result for detecting AF.⁷ Interestingly, that algorithm attempts to prevent premature contractions from becoming false positives by excluding pulse waves that vary ≥25% from the mean RR interval; however, some AF patients with large variations in their RR intervals may become false negatives.^18,19 This is the major difference with the Omron BP monitor, which only detects an IHB of ≤25% or ≥25% than the average rhythm during a BP measurement.

Improvement in Diagnostic Performance for AF With Multiple BP Measurements Using Automated BP Monitor

There are several studies that have evaluated the diagnostic performance for AF detection when BP measurement was performed 3 times with an automated BP monitor made by companies other than Omron.^7,9,11 Those reports defined AF when positive results were found during at least 2 of 3 BP measurements, and in this way the best diagnostic performance for AF could be obtained. Consistent with previous reports, our study with the Omron BP monitor also showed similar optimal AF detection when positive results were found during at least 2 of 3 BP measurements. When IHBs were detected during at least 1 of the 3 BP measurements, the diagnosis of AF had high sensitivity but low specificity. In that case, patients with sinus arrhythmia or extra systoles easily became false positives. In contrast, when IHBs were detected in all 3 BP measurements, the diagnosis of AF had high specificity but low sensitivity. In that case, false negatives may easily occur, failing to diagnose AF with subtle fluctuations. The condition of “positive results detected during at least 2 of 3 BP measurements” may be appropriate for diagnosing AF, which has characteristic pulse wave irregularities that randomly increase or decrease.

Novel Algorithm for Diagnosing AF Using Omron BP Monitor With IHB Detector

In this study, we developed a novel algorithm that was defined as the condition of IHBs being detected during at least 2 of 3 BP measurements and the maximum number of IHBs was ≥2 using an Omron automated BP monitor with an IHB detector. Using this novel algorithm, all 9 false-positive patients had non-AF arrhythmias, and none had respiratory sinus arrhythmia. In addition, even in patients with non-AF arrhythmias, only a minority showed a maximum value of ≥2 IHBs detected during at least 2 of 3 BP measurements, because the pulse waves in patients with sporadic premature atrial or ventricular contractions are not as consistent as in AF. Thus, a certain number of patients with other arrhythmias could be distinguished from those with AF. The novel algorithm was limited in its ability to distinguish AF from some non-AF arrhythmias such as persistent bigeminy or trigeminy; however, the novel algorithm could be applied in clinical practice, given the high level of diagnostic accuracy for AF demonstrated in this study.

Study Limitations

Firstly, a total of 29 patients, including 9 AF patients, were excluded because of an “error” sign on the Omron automated BP monitor. Almost all automated BP monitors using the oscillometric method have a function to stop BP measurements when IHBs appear at the same time as the values of systolic and diastolic BP are being determined and will display an “error” sign. This function is needed to measure the correct BP, but might decrease the sensitivity and specificity to detect AF. Secondly, the pulse rate of the AF group was slightly high, and 106 of 303 patients including 12 AF patients were taking β-blockers for the treatment of hypertension, heart failure, and arrhythmia. As this study was an observational study under daily care, we were unable to make detailed evaluations about the effects of pulse rate or medications on the diagnostic detection of AF. However, it is likely that detecting AF was not influenced by pulse rate unless it was severe tachycardia, because the good association between the RR interval on ECG and the pulse wave interval detected by the automated Omron BP monitor was observed independently of the RR interval (Figure 2). Lastly, the number of AF patients was nearly equal to that of other arrhythmia patients, whereas the prevalence of other arrhythmia patients was overwhelmingly high compared with that of AF patients in society. Further study examining another population would be necessary to prove the usefulness of this method in the real world.

Conclusions

This study showed that a novel algorithm using 3 BP measurements by an Omron automated BP monitor equipped with an IHB detector had high sensitivity and specificity for the diagnosis of AF in general cardiac patients.

Acknowledgments

This work was supported by a research grant from the OMRON HEALTHCARE company (Kyoto, Japan). We appreciate the assistance of Tanii Terue and Watanabe Satoko as clinical research coordinators. We also acknowledge Dr. Toshifumi Takabatake, the director of Takabatake Hospital, and Dr. Takayoshi Matsui, the director of Matsui Hospital, for their cooperation in recruiting patients and providing as place for performing the measurements.

Supplementary Files

Please find supplementary file(s);

http://dx.doi.org/10.1253/circj.CJ-19-0349

References

• 1.   Chugh SS, Havmoeller R, Narayanan K, Singh D, Rienstra M, Benjamin EJ, et al. Worldwide epidemiology of atrial fibrillation: A Global Burden of Disease 2010 Study. Circulation 2014; 129: 837–847.
• 2.   Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: The Framingham Study. Stroke 1991; 22: 983–988.
• 3.   Singh-Manoux A, Fayosse A, Sabia S, Canonico M, Bobak M, Elbaz A, et al. Atrial fibrillation as a risk factor for cognitive decline and dementia. Eur Heart J 2017; 38: 2612–2618.
• 4.   Wang TJ, Larson MG, Levy D, Vasan RS, Leip EP, Wolf PA, et al. Temporal relations of atrial fibrillation and congestive heart failure and their joint influence on mortality: The Framingham Heart Study. Circulation 2003; 107: 2920–2925.
• 5.   Senoo K, Suzuki S, Otsuka T, Sagara K, Matsuno S, Kano H, et al. Progression to the persistent form in asymptomatic paroxysmal atrial fibrillation. Circ J 2014; 78: 1121–1126.
• 6.   Kabutoya T, Imai Y, Hoshide S, Kario K. Diagnostic accuracy of a new algorithm to detect atrial fibrillation in a home blood pressure monitor. J Clin Hypertens 2017; 19: 1143–1147.
• 7.   Wiesel J, Fitzig L, Herschman Y, Messineo FC. Detection of atrial fibrillation using a modified microlife blood pressure monitor. Am J Hypertens 2009; 22: 848–852.
• 8.   Wiesel J, Abraham S, Messineo FC. Screening for asymptomatic atrial fibrillation while monitoring the blood pressure at home: Trial of Regular versus Irregular Pulse for Prevention of Stroke (TRIPPS 2.0). Am J Cardiol 2013; 111: 1598–1601.
• 9.   Wiesel J, Arbesfeld B, Schechter D. Comparison of the Microlife blood pressure monitor with the Omron blood pressure monitor for detecting atrial fibrillation. Am J Cardiol 2014; 114: 1046–1048.
• 10.   Willits I, Keltie K, Craig J, Sims A. WatchBP Home A for opportunistically detecting atrial fibrillation during diagnosis and monitoring of hypertension: A NICE Medical Technology Guidance. Appl Health Econ Health Policy 2014; 12: 255–265.
• 11.   Stergiou GS, Karpettas N, Protogerou A, Nasothimiou EG, Kyriakidis M. Diagnostic accuracy of a home blood pressure monitor to detect atrial fibrillation. J Hum Hypertens 2009; 23: 654–658.
• 12.   Marazzi G, Iellamo F, Volterrani M, Lombardo M, Pelliccia F, Righi D, et al. Comparison of Microlife BP A200 Plus and Omron M6 blood pressure monitors to detect atrial fibrillation in hypertensive patients. Adv Ther 2012; 29: 64–70.
• 13.   White WB, Anwar YA. Evaluation of the overall efficacy of the Omron office digital blood pressure HEM-907 monitor in adults. Blood Press Monit 2001; 6: 107–110.
• 14.   SPRINT Research Group, Wright JT Jr, Williamson JD, Whelton PK, Snyder JK, Sink KM, Rocco MV, et al. A randomized trial of intensive versus standard blood-pressure control. N Engl J Med 2015; 373: 2103–2116.
• 15.   Shimamoto K, Ando K, Fujita T, Hasebe N, Higaki J, Horiuchi M, et al. The Japanese Society of Hypertension guidelines for the management of hypertension (JSH 2014). Hypertens Res 2014; 37: 253–390.
• 16.   Agresti A, Coull BA. Approximate is better than ‘exact’ for interval estimation of binomial proportions. Am Stat 1998; 52: 119–126.
• 17.   Verberk WJ, de Leeuw PW. Accuracy of oscillometric blood pressure monitors for the detection of atrial fibrillation: A systematic review. Expert Rev Med Devices 2012; 9: 635–640.
• 18.   Chan PH, Wong CK, Pun L, Wong YF, Wong MM, Chu DW, et al. Diagnostic performance of an automatic blood pressure measurement device, Microlife WatchBP Home A, for atrial fibrillation screening in a real-world primary care setting. BMJ Open 2017; 7: e013685.
• 19.   Chan PH, Wong CK, Pun L, Wong YF, Wong MM, Chu DW, et al. Head-to-head comparison of the AliveCor Heart Monitor and Microlife WatchBP Office AFIB for atrial fibrillation screening in a primary care setting. Circulation 2017; 135: 110–112.