Resumen

PEREZ BEJARANO, Domingo et al. Diagnostic performance and reproducibility of a pulse oximeter with artificial intelligence algorithms in patients with obstructive sleep apnea at the Luque General Hospital. Mem. Inst. Investig. Cienc. Salud [online]. 2026, vol.24, n.1, e24122601.  Epub 30-Mar-2026. ISSN 1812-9528.  https://doi.org/10.18004/mem.iics/1812-9528/2026.e24122601.

Obstructive sleep apnea (OSA) is a common yet underdiagnosed disorder, particularly in resource-limited settings such as Paraguay, where polysomnography—the diagnostic gold standard—is unavailable in the public health system. To address this gap, we evaluated the diagnostic performance and reproducibility of a wrist-worn pulse oximeter equipped with artificial intelligence algorithms (BM2000A Wrist Pulse Oximeter®) against a home respiratory polygraph (SleepFairy®). A prospective study was conducted between 2018 and 2023 at Luque General Hospital, enrolling 84 adults selected by convenience sampling, excluding those with incomplete recordings (<4 hours). Variables including the apnea-hypopnea index (AHI), mean oxygen saturation (SpO2), oxygen desaturation index (ODI), time with SpO2 <90%, and heart rate were recorded over two consecutive nights. The oximeter demonstrated excellent performance in detecting moderate-to-severe OSA (AHI ≥15/h): sensitivity of 100%, specificity of 96.4%, and an AUC of 0.98. Reproducibility was high (kappa = 0.77 for AHI ≥15/h; r = 0.94 between nights). However, its accuracy was limited for mild OSA (AHI ≥5/h), with an AUC of 0.69. These findings suggest that the device is a viable, cost-effective tool for screening and diagnosing moderate-to-severe OSA in low-resource environments, though it should not replace more comprehensive testing in mild cases or patients with complex comorbidities.

Palabras clave : sleep apnea, obstructive; oximetry; artificial intelligence; reproducibility of results; polysomnography.

        · resumen en Español     · texto en Español     · Español ( pdf )