Performance of the Alice PDx Device With the Somnolyzer Automated Scoring Algorithm for the Diagnosis of Obstructive Sleep Apnea.

ABSTRACT

Objective Automated scoring of respiratory events could allow a swifter obstructive sleep apnea (OSA) identification. We assessed the accuracy of the Alice PDx device with the Somnolyzer automated scoring algorithm, compared to the manually reviewed scoring by a trained sleep technician, for the diagnosis of OSA. Methods A prospective study was conducted between March 2021 and March 2022 in Centro Hospitalar do Baixo Vouga, a level 2 hospital in Aveiro, Portugal. Patients with high pre-test probability for OSA performed a type III home sleep apnea testing with the Alice PDx device. Data were scored automatedly by the Sleepware G3 with the Somnolyzer digital system and manually by a trained sleep technician. Correlation and dependent t-tests were used. Sensitivity, specificity, positive predictive values (PPVs), negative predictive values (NPVs), and area under the receiver operating characteristic curve (AUROC) of automated scoring were calculated. Data were analyzed using the Stata Statistical Software (Release 17, StataCorp., 2023, College Station, TX StataCorp LLC). Results In 150 participants (mean age 57.8 ± 13.9 years), the mean apnea-hypopnea index (AHI) was 21.9 ± 21.8 events/hour by manual scoring and 25.4 ± 21.6 events/hour by automated scoring. The mean difference was 3.4 ± 4.4 events/hour, and a strong, positive, linear correlation was found between the two scores (r = 0.98). At the altered AHI (AHI ≥ 5 events/hour), mild, moderate, and severe OSA, the automated scoring sensitivity/specificity values were 91.2%/100.0%, 80.0%/68.6%, 91.6%/41.9%, and 98.1%/80.9%, respectively. The PPVs/NPVs for the same categories were 100.0%/69.4%, 89.3%/51.1%, 79.7%/66.7%, and 91.8%/95.0%, respectively. Finally, the AUROC was 0.85, 0.70, 0.73, and 0.93, respectively. Conclusion The automated scoring obtained from the Alice PDx portable device, using Sleepware G3 with the Somnolyzer digital system, seems accurate enough to diagnose OSA and validate the initiation of PAP therapy in the correct clinical setting. Nevertheless, it does not replace manual reviewing by a trained sleep technician in the case of mild and moderate OSA, to obtain a correct severity classification. With this valuable time-saving tool, we expect to hasten OSA diagnosis and treatment and thus tackle the underdiagnosis problem.