Mobile health technology for Atrial Fibrillation

Application screening using smart devices-based PPG

The HUAWEI Heart Study
（Pre-MAFA Study）

Yutao Guo, M.D. Ph.D, Yundai Chen, M.D. Ph.D and Gregory Y H
Lip, M.D

On behalf of the MAFA II Study Investigators

Sep 2,2019 .Paris
 Background
• Low detection and nonadherence in AF management
• Photoplethysmography (PPG) and mobile health
(mHealth) technology may enable a screening
approach combined with intervention.

Limit data on AF screening integration with clinical AF care

 Objective
To investigate the effectiveness of AF screening in a large population-based
cohort using smart device based photoplethysmography (PPG) technology,
combined with a clinical care AF management pathway using a mHealth
approach

Ø Primary outcome: the incidence of AF identified

Ø Secondary outcome: the anticoagulant use

Independent academic study
 PPG algorithm and smart devices have been validated before into 
HUAWEI HEART STUDY
Stage 1 Stage 2 Stage 3
Stage 4
2018.3 2018.6 2018.9 2018.10
STAGES
Pilot verified PPG Remodeling& optimization Validation in a real-world setting Pre-MAFA
HUAWEI HEART STUDY

115 Subjects 279 subjects 375 cases

SUBJECTS  • AF 52 cases • AF 113 cases • 373 cases with active measurement >100,000 
• SR 56 cases • SR 151 cases • 200 cases with continuous detection, Validation of a 
• Excluded 7 cases • Excluded 15 • taken 60s every 10 minutes for 14 day
cases
population screening 
approach and 
integrated AF care
PPG SIGNALS 614 signals 2,654 signals 26,217 signals

*Subjects were excluded if they had a pacemaker or implantable defibrillator, or they were unable to use smart wearable devices.

Fan Y, Guo Y, et al. JMIR Mhealth Uhealth. 2019

Guo Y, et al. Int J Clin Pract. 2019
Guo Y. m-Health in Arrhythmias . 2019 ESC congress
AF screening flow diagram

MAFA: mobile Atrial Fibrillation Application

 Individuals with “suspected” AF
by PPG algorithm

MAFA Telecare team 

Check and follow up at 24 hour AF confirmed by

clinical evaluation,
Can’t contact ECG, 24 h Holter
Follow up on First week
Individual
s who Can’t contact
refuse
the Follow up on Second week MAFA II cluster
follow-up
Can’t contact
MAFA doctors randomized clinical trial
Follow up for 12 months
Follow up on Fourth week

Can’t contact

Follow up on Sixth week

Can’t contact

Follow up on Eighth week

Can’t contact
The confirmation and follow-up of  
Loss of follow-up
individuals with “suspected” AF
Inclusion 246,541 downloaded Screening App Exclusion
Adult ≥18 years (between October 26, 2018 and May 20, 2019) 
Adult <18 years
Huawei phone (Android 5.0 or higher)
Inability to use smart phone or devices
Smart devices:
Huawei Watch GT (Version 1.0.3.52 or higher)
187,912 used smart devices
Honor Watch (Version 1.0.3.52 or higher)
Honor Band 4 (Version 1.0.0.86 or higher)

424 (0.2%) with “suspected” AF

262 (61.8%) “suspected” AF followed by 
MAFA Telecare team and MAFA doctors

227 (87.0%) “suspected” AF confirmed by doctors with medical 
history, physical examine, ECG, 24-H Holter, etc.

216 (95.1%) AF into MAFA with AF integrated care

AF screening, confirmation, and transference into MAFA 
 Baseline characteristics
  Overall Cohort Notification Individuals with  Identified AF
clinical evaluation
(n=187,912) (n=424) (n=227)
(n=262)

Suspected AF episodes, n 265,139 265,139 203,985 186,956

Female, n (%) 24938 (13.3) 55(13.0) 43(16.4) 42(18.5)

Age, mean (SD) 34.7(11.5) 54.1(14.3) 54.9(14.0) 56.1(13.7)

 

≥65, n (%) 3419 (1.8) 95(22.4) 62(23.7) 58(25.5)

 

55-64, n (%) 7491 (4.0) 112(26.4) 71(27.1) 69(30.4)

 

40-54, n (%) 44432 (23.6) 136(32.1) 82(31.3) 64(28.2)

 

20-39, n (%) 132570 (70.5) 81 (19.1) 47(17.9) 36(15.9)

Results: 71% of AF episodes were found within 14 days, but nearly 
one third of AF episodes were recorded after two weeks 
The rate of AF, %

70.0% 63.9%
61.1%
60.0% Suspected AF (n=424) Identified AF (n=227)
50.0%
40.0%
30.0%
About one third
20.0% 14.9%15.9% of AF episodes
9.7% 7.9% 8.0% 6.2%
10.0% 6.4% 6.2% were recorded
0.0% after two weeks
0-7 days 8-14 days 15-21 days 22-30 days >31 days

Time to first AF

Monitoring time to first detected AF episode

Results: The automatic periodical measurements were much likely to 
detect “suspected” AF episodes 
The rate of measure, 
60.0% 55.4%

50.0%
*
40.0% 37.0%
Automatic periodical
30.0%
%

measurements were
20.0% more likely to identify
7.5%
10.0% AF episodes
0.0%
Active measure (n=32) Automatically periodic  Active and periodic 
measure (n=157) measure (n=235)

Monitoring method for first “suspected” AF

*Compared to active measure, p<0.001, ** Compared to automatically periodic measure, p<0.001.
Results: The highest proportion of AF episodes was among the elderly aged over 65 
years, and there was a consistent increases  of incident ‘suspected AF’ and ‘identified’ AF over 
age

Positive PPG  Total PPG  PPV of PPG 

signals signals signals (95% 
CI)
186,956 203,985 91.6%
( 91.5%-91.8%)

Number of Number of Proportion of

cases with cases with positive AF
identified AF suspected (95% CI)
AF
227 262 86.6%
(82.0%-90.2%)

Incident “suspected” and “identified” AF among the screened population (n=187912 )
Results 80% of patients at high-risk were anticoagulated
Oral anticoagulant use in AF patients
Low risk Intermedi High risk
  ate risk 

N ( %) 91 (42.1) 71(32.9） 54 (25.0) 2007-2009 year

1034 Patients with AF
Anticoagulant use at baseline, n% 5 (5.49) 9 (12.68) 43 (79.63)
OACs: 14% 
（Guo Y, et al. Int Card J. 2013)
Anticoagulant use at 3 months, n% 3 (3.30) 29 (40.85) 42 (77.78)

p 0.372 <0.001 0.673

* Low risk: CHA2DS2-VASc of 0 in males, or 1 in females; Intermediate risk: CHA2DS2-VASc of 2 in female, 1 in male; High
risk: CHA2DS2-VASc ≥3 in females, ≥2 in males. McNemar's test was used for testing the difference.
The reasons for patients with or without oral anticoagulants (OACs) on baseline:

Low risk patients with OACs at baseline: 2 patients undergoing AF ablation, with OAC used after discharge, 2 patients with
current onset acute AF episodes, and 1 patient with rheumatic valvular heart disease.

High risk patients without OACs at baseline: six patients who were unwilling to accept anticoagulants, four patients with
antiplatelets (aspirin or clopidogrel), and one patient anticoagulated with traditional Chinese medicine.
 Limitations
• Although we had strict follow-up procedures for ‘suspected AF’, there
were 38% of individuals with ‘suspected AF’ who could not be effectively
followed up, which would decrease the proportion of identified AF.
• For the PPV calculation with PPG signals, we did not have real-time 12-lead ECG
data synchronized with PPG-based smart devices.
• Incident AF detection in present study might be impacted by the availability of
smart phones and devices.
 Comparison with Apple Heart study
Main Aspects Huawei Heart Study  Apple Heart Study
Algorithm 
PPG Detecting PPV 91.6% 71.3%
accuracy
Based on PPG，Periodic measurement 
Period  Based on PPG，Periodic measurement results 
results are given, the proportion of 
measurement  are not given
irregular rhythms is analyzed
Measurement
Period  Every 2h for baseline mode，an irregular 
measurement  Every 10 minutes tachogram initiates a cascade of more 
frequency frequent mode (every 16 minutes)
Follow-up by doctor, combined with a 
Follow-up mode Follow-up with video visit online
clinical care AF management
Confirmed with medical history, physical 
Confirmation mode Confirmed with ePatch
examination, ECG or 24-Holter by doctors
Follow up and 
confirmation 44% first study visit * effectively followed up 
Follow-up rate 62% with MAFA telecare and doctors
with ePatch returned and analyzed: 21% 
Proportion of 
confirmed AF by  87% 34%
doctors or ePatch
 Summary

Mobile Atrial Fibrillation App

AF  screening in 
186,956 population
0.2 % received  87% confirmed  95% with MAFA for
the notification of  AF by doctors AF  integrated care ABC
suspected  AF

80% of high-risk patients

were successfully
anticoagulated.
 Clinical application

Continuous home-monitoring with smart device-based PPG technology is

a feasible approach for screening and early detection of AF in a large
population.

This could help our efforts at screening and detection of AF, as well as
early interventions to reduce stroke and other AF-related complications.
 Acknowledgments
This research project was funded by the National Natural Science
Foundation of China (H2501) and was funded by the Health and Family
Planning Commission of Heilongjiang Province, China (2017-036), and
partly supported by the NIHR Global Health Research Group on Atrial
Fibrillation management at the University of Birmingham, UK.
Thanks to HUAWEI Heart Health Research Team for the development
and optimization of PPG algorithm, headed by Mr. Xiaoxiang He. Team
members include Jiabing Yan, Wenjuan Chen, Qin Chen, Jie Zhang, Xi
Huang, and Hongbao Li.
THANK YOU

JACC. In press
�User reported adverse events
  Overall cohort 
Total 186
  Any device connection and data  123
synchronization issues
  Login and experience issues (any) 63
 Skin irritation, anxiety, pressure 0