Performance evaluation of a ring-worn pulse oximeter for the identification and monitoring of obstructive sleep apnea

Obstructive sleep apnea (OSA) is a highly prevalent chronic disorder that is challenging to monitor clinically. While single-night laboratory-based polysomnography (PSG) is the current gold standard for OSA assessment, its utility is limited by cost and inaccessibility. Overnight pulse oximetry is a feasible approach for simplified at-home monitoring of OSA. In this study, we evaluate the performance of a modified finger-worn pulse oximetry device ("Ring") for OSA assessment.

In all, 25 patients with OSA [age: 55.5 ± 7.7 years (mean ± SD), body mass index (BMI): 31.8 ± 5.1 kg/m², 14M:11F, and Fitzpatrick scale score I-II: 15, III-IV: 6, and V-VI: 4] completed up to four in-laboratory PSG studies with simultaneous Ring oximetry measurements (90 studies in total). Correlation and agreement analyses compared Ring-derived measures of the oxygen desaturation index (ODI4_RING, desaturations ≥4%) against PSG measures (ODI4_PSG and AHI4_PSG). Likewise, Ring-derived hypoxic burden (HB_RING) was compared against its PSG counterpart (HB_PSG). Receiver operator characteristic (ROC) curve analysis quantified the ability of ODI4_RING to identify moderate-to-severe OSA (AHI4_PSG > 15 events/h).

Median [interquartile range (IQR)] of AHI4_PSG was 18.0 [9.6, 31.7] events/h. ODI4_RING was positively correlated with ODI4_PSG (Pearson r = 0.87, root mean square error [RMSE] = 6.6 events/h, intraclass correlation [ICC] = 0.85) and AHI4_PSG (r = 0.85, RMSE = 7.1 events/h, ICC = 0.84). The bias (mean difference) and limits of agreement (1.96 SD) between ODI4_PSG and ODI4_RING were 2.9 and 14.2 events/h, while for AHI4_PSG and ODI4_RING, the bias and limits of agreement were 1.4 and 16.3 events/h, respectively. HB_RING was positively correlated with HB_PSG (r = 0.75, RMSE = 24.6% min/h, ICC = 0.73), with a mean difference of 3.7% min/h and limits of agreement of 60.6% min/h. The receiver operator characteristic curve analysis of ODI4_RING to identify moderate-to-severe OSA produced an area under the curve of 0.92 (ODI4_PSG > 15 events/h, "excellent") and 0.84 (AHI4_PSG > 15 events/h, "excellent").

Our results show that a low-cost, convenient, and simple-to-use finger-worn pulse oximeter is a reliable tool for continuous monitoring of OSA severity and therapy responses. It also offers excellent discriminative value for screening moderate-to-severe OSA in this population.