Prospective validation of a wearable device for automated paediatric OSA diagnosis
File(s)
Accepted version
Author(s)
Pramono, Renard
Imtiaz, Syed Anas
Evans, Hazel
Elphick, Heather
Rodriguez Villegas, Esther
Type
Journal Article
Abstract
Background: Obstructive sleep apnoea (OSA) is common in children and associated with significant childhood morbidity. The current reference standards for paediatric OSA diagnosis are attended polysomnography (PSG) and cardio-respiratory polygraphy (CR-PG) - both resource-intensive and access-limited. No fully-automated diagnostic device has previously been prospectively validated against PSG or CR-PG.
Methods: Prospective, blinded, multicentre diagnostic-accuracy study at two UK tertiary paediatric centres. Consecutive children referred for suspected OSA were studied overnight by attended in-hospital paediatric sleep study (reference standard PSG or CR-PG, scored
per AASM paediatric rules) and by AcuPebble SA100, a wearable sensor producing an automated diagnosis, set up by carers who also completed a usability questionnaire.
Results: 141 children had valid data (median age 7.1 years, range 1.2–17.3 years; 88.7% with comorbidities; 36.2% with learning difficulties). For moderate-or-severe OSA (OAHI ≥5/h), AcuPebble achieved sensitivity 100% (95%-CI 80.49–100), specificity 96.77%, accuracy 97.16%, positive predictive value 80.95%, negative predictive value 100%, and positive likelihood ratio 31.0. Performance was similar on the broader AHI (additionally capturing central events): sensitivity 94.12%, specificity 98.13%, positive likelihood ratio
50.4. Among surveyed carers, 97.0% completed setup unassisted; 81.9% rated AcuPebble as more comfortable than conventional sensors; only 10.5% would prefer PSG/CR-PG.
Conclusions: Automated diagnosis of paediatric OSA with AcuPebble met recommended diagnostic-accuracy thresholds. Carer-led setup was feasible and acceptable, supporting the device's use as an alternative to attended in-hospital sleep studies. Performance was maintained in a complex tertiary-referral cohort, where sleep studies are most difficult to
deliver.
Methods: Prospective, blinded, multicentre diagnostic-accuracy study at two UK tertiary paediatric centres. Consecutive children referred for suspected OSA were studied overnight by attended in-hospital paediatric sleep study (reference standard PSG or CR-PG, scored
per AASM paediatric rules) and by AcuPebble SA100, a wearable sensor producing an automated diagnosis, set up by carers who also completed a usability questionnaire.
Results: 141 children had valid data (median age 7.1 years, range 1.2–17.3 years; 88.7% with comorbidities; 36.2% with learning difficulties). For moderate-or-severe OSA (OAHI ≥5/h), AcuPebble achieved sensitivity 100% (95%-CI 80.49–100), specificity 96.77%, accuracy 97.16%, positive predictive value 80.95%, negative predictive value 100%, and positive likelihood ratio 31.0. Performance was similar on the broader AHI (additionally capturing central events): sensitivity 94.12%, specificity 98.13%, positive likelihood ratio
50.4. Among surveyed carers, 97.0% completed setup unassisted; 81.9% rated AcuPebble as more comfortable than conventional sensors; only 10.5% would prefer PSG/CR-PG.
Conclusions: Automated diagnosis of paediatric OSA with AcuPebble met recommended diagnostic-accuracy thresholds. Carer-led setup was feasible and acceptable, supporting the device's use as an alternative to attended in-hospital sleep studies. Performance was maintained in a complex tertiary-referral cohort, where sleep studies are most difficult to
deliver.
Date Acceptance
2026-09-14
Citation
ERJ Open Research
ISSN
2312-0541
Publisher
European Respiratory Society
Journal / Book Title
ERJ Open Research
Copyright Statement
Copyright This paper is embargoed until publication. Once published the Version of Record (VoR) will be available on immediate open access.
License URL
Publication Status
Accepted