Appendix

Appendix

Contents

Table 1 Overview of Lau et al. (2013) study

Study component

Description

Objectives/hypotheses

To assess the accuracy of an iPhone ECG (using the AliveCor Heart Monitor), compared with a 12‑lead ECG interpreted by a cardiologist, as a diagnostic screening tool to detect AF.

Study design

Diagnostic accuracy evaluation. The AliveCor Heart Monitor reading was taken within 6 hours of the 12‑lead ECG. The AliveCor Heart Monitor reading for the first group of people (n=109) was interpreted by 2 cardiologists and a software programme (a developmental model of the AliveECG app). After analysis of these results, the software was updated and the new version was used to report readings for the second group of people in the study (n=204).

The updated software was used to build diagnostic functionality into the AliveECG app, but not before being recoded to work on compatible mobiles and tablets.

The 12‑lead ECG formed the reference test and was interpreted by a third cardiologist.

Setting

AliveCor Heart Monitor reading and 12‑lead ECG conducted within an Australian clinical setting.

Inclusion/exclusion criteria

Patients recruited from a single centre: 39 people in Group 1 and 48 in Group 2 had a diagnosis of AF at entry.

Primary outcomes

Sensitivity, specificity, Kappa statistic and overall accuracy of AliveCor Heart Monitor.

Statistical methods

Not disclosed.

Patients included

2 groups were recruited sequentially. Results from Group 1 (n=109) were used to assess and update the software. The results from Group 2 (n=204) were used only to assess the diagnostic accuracy of updated software.

Results

Outcome of 12‑lead ECG treated as definitive diagnosis, and used to define sensitivity and specificity of the AliveCor Heart Monitor.

Learning set (n=109)

Original software

  • Sensitivity: 87.2% (34/39; 95% CI 72.6%–95.7%)

  • Specificity: 97.1% (68/70; 95% CI 90.1%–99.7%)

  • Overall accuracy: 93.6% (102/109; 95% CI 87.2%–97.4%)

  • Kappa (agreement with 12‑lead ECG): 0.86.

Updated software

  • Sensitivity: 100% (39/39; 95% CI 91.0%–100%)

  • Specificity: 95.7% (67/70; 95% CI 88.0%–99.1%)

  • Overall accuracy: 97.3% (106/109; 95% CI 92.2%–99.4%)

  • Kappa (agreement with 12‑lead ECG): 0.94.

Cardiologist A

  • Sensitivity: 100% (39/39; 95% CI 91.0%–100%)

  • Specificity: 90.0% (63/70; 95% CI 80.5%–95.9%)

  • Overall accuracy: 93.6% (102/109; 95% CI 87.2%–97.4%)

  • Kappa (agreement with 12‑lead ECG): 0.87.

Cardiologist B

  • Sensitivity: 94.9% (37/39; 95% CI 82.7%–99.4%)

  • Specificity: 94.3% (66/70; 95% CI 86.0%–98.4%)

  • Overall accuracy: 95.4% (104/109; 95% CI 89.6%–98.5%)

  • Kappa (agreement with 12‑lead ECG): 0.88.

Validation set (n=204)

Updated software

  • Sensitivity: 97.9% (47/48; 95% CI 80.9%–99.9%)

  • Specificity: 96.8% (151/156; 95% CI 92.7%–98.9%)

  • Overall accuracy: 97.1% (198/204; 95% CI 93.7%–98.9%)

  • Kappa (agreement with 12‑lead ECG): 0.92.

Conclusions

The authors concluded that the AliveCor Heart Monitor and the AliveECG app have high diagnostic accuracy and are useful for community screening for AF.

Abbreviations: AF, atrial fibrillation; CI, confidence interval; ECG, electrocardiogram; n, number of patients; NR, not reported.

Table 2 Overview of Lowres et al. (2014) study

Study component

Description

Objectives/hypotheses

Assess the feasibility and cost effectiveness of community‑based AF screening using the AliveCor Heart Monitor and the AliveECG app.

Study design

Observational study of 1000 people, mean age 76 years and cost‑utility analysis comparing the AliveCor System in a screening programme with no screening. The cost utility analysis used a hypothetical strategy not matched by the data collection method where people are given a provisional diagnosis by the AliveECG app. Those with suspected AF are referred on to a cardiologist for further testing. This analysis projected costs and benefits over a 10 year period.

Setting

Screening intervention to detect AF, based in 10 Australian pharmacies between June 2012 and January 2013: ECGs done under supervision of trained pharmacists.

Inclusion/exclusion criteria

All individuals aged 65 years and over entering participating pharmacies were eligible for inclusion.

Primary outcomes

Cost effectiveness of AF screening using the AliveCor system, sensitivity and specificity of the AliveCor Heart Monitor with the AliveECG app used to indicate AF and societal prevalence of AF.

Statistical methods

Summary statistics and decision tree economic model. Monte Carlo simulation adopted, selecting age and gender‑specific incidence rates from log‑normal distributions to generate the 95% CIs.

Patients included

Analysis conducted on 1000 pharmacy users. 1051 individuals recruited with 51 exclusions due to withdrawal of consent (n=2), duplicate recruitment (n=3), incorrect age (n=42) and incomplete screening (n=4).

Results

Cost effectiveness

Base‑case analysis reported an ICER of £2799 (95% CI £745–£6280). Two‑way sensitivity analyses altered treatment adherence rates, costs and QALY gain per stroke avoided. ICERs ranged from £1448 (95% CI £27–£3867) to £7742 (95% CI £4371–£13,471). (All costs were converted into pounds sterling from Australian dollars, using 2014 purchasing power parities [OECD 2015]).

Sensitivity and specificity of the AliveECG app with cardiologist's interpretation of the AliveCor reading as the reference test

Sensitivity: 98.5% (67/68; 95% CI 92.1%–100.0%)

Specificity: 91.4% (849/929; 95% CI 89.4%–93.1%)

Prevalence based upon cardiologist's interpretation of the AliveCor readings

Prevalence of AF identified by screening was 6.7% (67/1000; 95% CI 5.2%–8.4%). Newly identified AF was found in 15 people (1.5%; 95% CI 0.8%–2.5%). Of these, 10 people had no prior history of AF, and 5 had an unknown recurrence of AF ≥3 years after cardioversion.

Conclusions

The authors concluded that pharmacy‑based screening for AF using the AliveCor system in individuals over 65 years is feasible and cost effective.

Abbreviations: AF, atrial fibrillation; CI, confidence interval; ECG, electrocardiogram; ICER, incremental cost‑effectiveness ratio; OECD, Organisation for Economic Co‑operation and Development; QALY, quality‑adjusted life year.

Table 3 Overview of Haberman et al. (2015) study

Study component

Description

Objectives/hypotheses

To compare the accuracy of the AliveCor Heart Monitor with 12‑lead ECG in assessing ECG intervals, heart rate and heart rhythm.

Study design

Diagnostic accuracy evaluation and user satisfaction survey.

Setting

Individuals enrolled between August and December 2012 from a US college campus and cardiology treatment centre. 12‑lead ECG was done immediately after the completion of the AliveCor Heart Monitor reading. All ECGs were collected within a clinical setting supervised by the study investigators.

Inclusion/exclusion criteria

Assessment conducted within 3 separate populations: student athletes, healthy young adults and cardiology clinic patients. Cardiology patients were recruited from patients pre‑scheduled for a 12‑lead ECG.

Primary outcomes

Sensitivity and specificity of the AliveCor for AF or flutter.

Statistical methods

2×2 contingency tables were used to calculate sensitivity and specificity for each abnormality and Kappa plots were constructed. Post participation surveys were compared using a Chi‑square test.

Patients included

Student athletes, n=123

Healthy young adults, n=128

Cardiology clinic patients, n=130.

Results

Sensitivity and specificity of detection of AF or atrial flutter

Based upon comparison of the AliveCor Heart Monitor with 12‑lead ECG reading, which was considered fully accurate.

  • Athletes: sensitivity – N/A (0/0); specificity – 99.2% (122/123; 95% CI 95.6%–100.0%); Kappa – N/A.

  • Healthy young adults: sensitivity – N/A (0/0); specificity – 100% (128/128; 95% CI 97.2%–100.0%); Kappa – N/A.

  • Cardiology patients: sensitivity – 94.4% (17/18; 95% CI 72.7%–99.9%); specificity – 99.1% (111/112, 95.1%–10.0%); Kappa – 0.94.

  • Total: sensitivity – 94.4% (17/18; 95% CI 72.7%–99.9%; specificity – 99.4% (361/363; 95% CI 98.0%–99.9%); Kappa – 0.91.

User satisfaction (selected items)

Results presented in diagrammatic format only, so figures are rounded to nearest 5%. All statements were positively worded. The post participation survey was completed by student athletes and healthy young adults only.

Level of agreement with selected statements:

  • I had no problems with the smartphone ECG: 95%

  • The smartphone ECG was more comfortable: 70%

  • The smart phone ECG was faster: 75%

  • I prefer the smartphone ECG: 75%.

Conclusions

The authors concluded that the AliveCor Heart Monitor accurately detects atrial rate, conduction intervals and common arrhythmias such as AF, and can be used for large scale screening.

Abbreviations: AF, atrial fibrillation; AV block, atrioventricular block; ECG; electrocardiogram; n, number of patients; N/A, not applicable.

Table 4 Overview of Tarakji et al. (2015) study

Study component

Description

Objectives/hypotheses

Compare the accuracy and usability of the AliveCor Heart Monitor with traditional transtelephonic monitors (TTM), to monitor individuals after an ablation procedure for AF.

Study design

Diagnostic accuracy evaluation, comparison of quality of readings and user satisfaction survey. For a comparison of the quality of monitor readings, the electrophysiologists responsible for interpreting the device readings were asked to comment on their quality compared to the TTM readings.

Setting

55 people recruited from a single US treatment centre after cardiac ablation were monitored for 3 months using the AliveCor Heart Monitor and TTM in a non‑clinical setting. User satisfaction survey conducted 3–4 months after ablation procedure. Enrolment took place between July and November 2013.

Inclusion/exclusion criteria

All of the people in the study had had an ablation procedure for AF. Inclusion criteria were ownership of an iPhone, age 18 to 75 years, history of paroxysmal or persistent AF and willingness to use the AliveCor Heart Monitor. People were excluded if they were unable or unwilling to use an iPhone or lived outside the USA.

Primary outcomes

  • Sensitivity and specificity of the AliveCor Heart Monitor reading interpreted by electrophysiologists using TTMs as a reference test.

  • Level of agreement between the AliveCor Heart Monitor and TTMs.

  • Patient satisfaction with the AliveCor Heart Monitor.

  • Electrophysiologists' assessment of quality of the AliveCor Heart Monitor readings compared with TTMs.

Statistical methods

Power calculation carried out to determine a sample size to detect a sensitivity of at least 90% with the AliveCor Heart Monitor. The Kappa statistic of >0.8 was excellent agreement. TTM was assumed to have 100% sensitivity and specificity.

Patients included

Final analysis done on 384 simultaneous readings from 55 people. Although 60 people enrolled, 5 did not complete the study due to overseas travel (n=1), change of phone (n=1) and consent withdrawn (n=3). Five readings were excluded because of non‑interpretability, 4 from the AliveCor Heart Monitor and 1 from TTM.

Results

Diagnostic accuracy

AF and flutter treated as a single condition:

  • sensitivity: 100% (46/46; 95% CI 92.3%–100%)

  • specificity: 97.0% (328/338; 95% CI 94.6%–98.6%)

  • agreement between the AliveCor Heart Monitor and TTM: Kappa = 0.82.

User satisfaction(selected items)

  • How easy was it to use each device?

    • Very easy: TTM 47%; AliveCor 80%

    • Moderately easy: TTM 35%; AliveCor 18%

    • Difficult: TTM 18%; AliveCor 2%.

  • How often did you have access to each monitor when you had symptoms and needed to transmit a recording?

    • Every time: TTM 45%; AliveCor 84%

    • Sometimes: TTM 40%; AliveCor 11%

    • Not at all: TTM 15%; AliveCor: 5%.

  • Which device would you prefer to use in the future?

    • TTM 8%; AliveCor 92%.

Electrophysiologists' interpretations of quality of readings

  • 72.7% (40/55) were rated of equal quality

  • 14.6% (8/55) rated TTM readings of higher quality

  • 12.7% (7/55) rated the AliveCor Heart Monitor readings of higher quality.

Conclusions

The AliveCor Heart Monitor could be a useful monitoring tool, with high sensitivity, specificity and user satisfaction. Authors noted concerns about data security, resourcing, insurance coverage and incorporating readings into medical records.

Abbreviations: AF, atrial fibrillation; CI, confidence interval; ECG, electrocardiogram; TTM, traditional transtelephonic monitors