B105-12 Evaluating the Accuracy of Ventilator Generated Apnea-Hypopnea Index Measures by a Home Bi-Level Non-Invasive Ventilator in Pediatric Patients With Neuromuscular Disease

Abstract Rationale Home Bi-level non-invasive ventilation (NIV) is increasingly used in children with neuromuscular disorders to provide long-term ventilatory support. Continuous ventilatory monitoring is crucial to ensure adequate ventilation as well as to detect residual sleep-disordered breathing. While polysomnography (PSG) remains the gold standard, it is resource-intensive and not always available. Modern ventilators provide automated apnea-hypopnea index (AHI) estimates from flow and pressure signals, but the accuracy of these values in pediatric patients remains uncertain. Methods We retrospectively analyzed pediatric patients with neuromuscular disease using home Bi-level NIV who underwent in-lab PSG at The Hospital for Sick Children between January 2018 and December 2024. Studies were included if total sleep time at the home settings was ≥ 60 minutes. The AHI obtained from ventilator downloads (“device AHI”) was compared with the PSG-derived AHI (sum of obstructive and central indices). Correlation and agreement were assessed using Wilcoxon signed-rank testing, Spearman’s correlation coefficients, Bland-Altman analysis, Deming regression, and the intraclass correlation coefficient (ICC). Results A total of 117 studies from 32 unique patients were analyzed (mean age 11.6 years, 65.6% male). Median (IQR) total sleep time and REM sleep under home conditions were 143.4 (71.8-241.5) minutes and 17.3 (0-42) minutes, respectively. Device-derived and PSG-derived AHI values differed significantly (p 0.001, Wilcoxon test). The correlation between the two methods was moderate (Spearman ρ = 0.50; p 0.001). Bland-Altman analysis demonstrated a mean bias of + 0.81 events/h, with 95% limits of agreement from −4.68 to + 6.31 events/h, indicating a slight but variable overestimation by the device. Deming regression (PSG AHI = 0.71 + 0.42 × Device AHI) showed that the device underestimated higher PSG AHI values. The ICC was 0.33 (95% CI 0.16-0.48), confirming low agreement. Conclusions Device-reported AHI values from home NIV downloads showed moderate correlation but poor absolute agreement with PSG-derived indices. The device tended to overestimate low AHI values and underestimate higher ones, with broad variability across the range. While these automated indices may assist with longitudinal trend monitoring or screening, they cannot substitute for a PSG in evaluating ventilation adequacy or guiding therapeutic adjustments in children receiving chronic NIV. This abstract is funded by: None