Heart and Respiration Rate Variability Analysis to Estimate the Apnoea-Hypopnoea Index in People With Intellectual Disabilities

BACKGROUND: Obstructive sleep apnoea (OSA) is highly prevalent in people with intellectual disabilities, and when left untreated, negatively influences daily activities and social interactions. Polysomnography (PSG) remains the diagnostic gold standard but can be an obtrusive and strenuous endeavour in people with intellectual disabilities, related to factors such as communicative impairment, anxiety, challenging behaviour and sensory hypersensitivity. Alternative methods to assess OSA severity by estimating the apnoea-hypopnoea index (AHI) have been proposed, based on heart and respiration rate variability signals. These signals could potentially be obtained with less obtrusive monitoring devices. We investigated whether this approach is also suitable in people with intellectual disabilities.

METHODS: We analysed overnight PSG data from 73 participants with intellectual disabilities. AHI was predicted by an algorithm trained to use cardiorespiratory inputs (from electrocardiogram and respiratory induction plethysmography) to detect the occurrence of sleep-disordered breathing events and total sleep time. It was compared to the PSG-derived AHI by means of Spearman's correlation and intraclass correlation coefficients (ICC). The diagnostic capacity of the algorithm to differentiate between OSA severity groups was evaluated using Cohen's κ coefficient of agreement and accuracy, using near-boundary double labelling, with the following boundaries: 'no OR mild OSA' 2.4 ≤ AHI < 7.0, 'mild OR moderate OSA' 12.4 ≤ AHI < 17.4 and 'moderate OR severe OSA', 26.6 ≤ AHI < 35.2.

RESULTS: The algorithm achieved a strong Spearman's correlation between the predicted and PSG-derived AHI of 0.76 (p < 0.001) and a moderate ICC of 0.74 (p < 0.001). Differentiation in OSA severity classes was done with a κ of 0.58 and accuracy of 68.5%, indicating a moderate level of agreement.

CONCLUSIONS: We show the potential of determining the severity of OSA in people with intellectual disabilities by estimating AHI using an algorithm based on surrogate cardiorespiratory signals. This allows the development of less obtrusive diagnostic modalities focusing only on cardiorespiratory inputs to assess OSA severity.