Introduction Sleep Disordered Breathing (SDB) is prevalent in patients with chronic heart failure (CHF) and is associated with increased mortality. The current gold standard for making this diagnosis in patients is polysomnography or polygraphy (PSG/PG). This is cumbersome and labour intensive. Repeat studies where there is diagnostic doubt are also difficult and challenging due to such resource constraints. The SleepMinder device is a novel non-contact bedside biomotion sensor that is uses an ultra low power radiofrequency transceiver able to detect movement and breathing. We sought to evaluate its accuracy in identifying and quantifying SDB in patients with CHF compared to full in hospital PSG.
Methods 30 adult patients with systolic heart failure had simultaneous overnight in-hospital PSG and SM recordings after consent in the HF clinic. An Apnoea Hypopnoea Index (AHI) algorithm was developed using intelligent feature extraction, vector transformation and pattern recognition techniques, from a training set of 20 patients and tested on the remainder of 10 patients. We aligned the PSG and SM signals using a cross correlation and sliding window system and the SM's Sleep/Wake analysis and integrated movement detector algorithm was employed to determine the SM calculated Total Sleep Time (TST) for the purpose of AHI estimation. The PSG signals were scored by an independent expert scorer using standard criteria.
Results The median age of studied patients was 70.5 years. The mean ejection fraction, and Body Mass Index were 33.6±6.05%, and 29±5 kg/m2 respectively. 37% of patients had SDB with an AHI >15 determined by expert PSG scoring. The correlation between SMAHI and PSGAHI was excellent with a co-efficient of 0.95 for the training and 0.84 for the test set. The Bland and Altman plot also showed good agreement across the range of AHI and did not vary by more than 10 events per hour in 70% of patients for all patients (figure 1). Our algorithm had good diagnostic accuracy across the board but performed best with a diagnostic threshold of AHI >15, which represents clinically significant SDB, with a sensitivity of 91% and a specificity of 90%. The area under receiver operator characteristic curve at this threshold was 0.93.
Conclusions The SleepMinder compares favourably with full in-hospital PSG, detecting every patient with an AHI >15 with a low false positive rate of 9% in a CHF clinic population. Its novel contactless feature makes it convenient to use and patient friendly. We plan to extend this work to involve a larger study cohort to confirm the diagnostic value of this new tool and are also continuing to explore the unique opportunity this device presents for the longitudinal assessment of nocturnal respiratory patterns and nightly variations of SDB in patients with chronic heart failure.