Journal Article

0557 Sleep apnea detection using photoplethysmography using wearable electronic devices : A systematic review and meta-analysis

Mirza, Wasique, Nguyen-Luu, Tristan, Saraza, Mohammad, Sharma, Rohit, Singh, Prashant Kumar, Sardar, Sundus, Abbas, Sayeda Z

2023

Overview

Introduction Sleep apnea, a hyper common disorder is estimated to affect almost 1 billion people with the prevalence of more than 50% in some countries. It is almost impossible to do gold standard polysomnography for all the population for screening. Multiple studies have reported the use of cost effective wearable devices using photoplethysmography (PPG) for the screening. To our knowledge, we presented the first meta-analysis of all studies looking at PPG and screening of sleep apnea. Methods We conducted a systematic review and meta-analysis under the PRISMA guidelines. Study Eligibility criteria included patients aged > 18 without a prior diagnosis of sleep disorder. The intervention included screening using a wearable device using PPG with comparison using standard of care sleep study (e.g polysomnography). The databases used were Cochrane, PubMed, Embase and Google scholar. Search strategy with Boolean logic included Sleep apnea, wearable device, smart, photoplethysmography, PPG, smart and smartwatch. Outcomes were pooled diagnostic odds ratio, meta-analyzed area under the curve (AUC),, and forest plots of sensitivity & specificity. Statistical analysis was done using the programming language R with R package mada. Fixed and random effects models were used to derive diagnostic odds ratio (DOR) and paired forest plots for sensitivity & specificity. Summary Receiver operation curve was made using proportional hazard model and Reitsma et al model with a bivariate approach. Results The search strategy revealed following number of studies: PubMed (270), Cochrane (0), Embase (41), and Google Scholar(728). Total of 12 studies were included. We found high DOR with the fixed effect model which was 28.979. Mean DOR with the random effects model was 27.6. Bivariate diagnostic random-effects meta-analyzed AUC was 0.902. Studies used various electronic devices some of which are commercially available such as E4 Wristband, Galaxy watch 4, Smartwatch GT2, Belun ring etc. 4 studies used PPG solely while others used multimodal data. 7 studies used conventional machine learning algorithm for analysis. Conclusion Screening of sleep apnea could be done in a cost-effective manner from commercially available devices using photoplethysmography owing to possible high sensitivity, AUC and DOR. More studies for data replicability and further standardization is needed. Support (if any)

• Publisher: Oxford University Press
• Subject: Meta-analysis / Sleep apnea / Smartwatches / Systematic review / Wearable computers