Document resource
Introduction Telemonitoring of home non-invasive ventilation (H-NIV) is gaining interest as a tool for detecting deteriorating patients in the community (Khirani, Patout and Arnal 2024). Although clinical alerts can be configured within telemonitoring software, evidence supporting its routine use remains limited.Methods This single-centre retrospective study analysed NIV telemonitoring data (NIV-TD) collected from AirView® for 70 H-NIV patients between 2020–2025. All patients were compliant with NIV (≥4 hours usage on ≥70% of days) and had experienced at least one acute hospital admission since NIV initiation (admission group). NIV-TD was collected for two 10-day periods: prior to a routine clinical review (stable period) and prior to hospital admission (pre-admission period). Stable-period data from a control group (n=70) were used for logistic regression analysis only.Results A Kruskal–Wallis test demonstrated significantly higher hourly NIV usage during the stable period in patients with neuromuscular disease (NMD) compared with obesity hypoventilation syndrome (OHS) (9.50 hours [IQR 8.13–14.44] vs 6.82 hours [3.88–7.72], H(4)=12.99, p=0.011). Wilcoxon signed-rank tests showed that during the pre-admission period, respiratory rate (p≤0.001, r=0.414) and spontaneous triggered breaths (p≤0.001, r=0.345) significantly increased with moderate effect sizes, while tidal volume significantly decreased (p=0.03, r=0.183) with a weak effect size ( figure 1A). Multivariate logistic regression identified increases in respiratory rate (OR 1.211, 95% CI 1.010–1.450, p=0.038) and hourly usage (OR 1.211, 95% CI 1.026–1.431, p=0.024), alongside decreases in spontaneous cycled breaths (inverse OR 1.032, 95% CI 1.009–1.056, p=0.006) and compliance days (inverse OR 2.13, 95% CI 1.38–3.29, p<0.001), as significantly associated with hospital admission (RUSC model). ROC curve analysis of the RUSC model (figure 1B) showed fair predictive performance (AUC 0.784), improving to good performance with the addition of stable-period PaCO2 (AUC 0.850).Conclusion NIV usage differed significantly between NMD and OHS groups, helping address gaps in evidence on normal NIV-TD variation (Jeganathan et al., 2021). Changes in respiratory rate, spontaneous triggered breaths, and tidal volume were evident prior to hospital admission. The RUSC model could inform telemonitoring alerts for unwell H-NIV patients, though multi-centre studies are required to improve generalisability and predictive accuracy.Abstract P2 Figure 1ANIV telemonitoring variables: analysis between stable and pre-admission periods boxplots shown for the telemonitoring data calculated from AirView® across the 10-day stable period (blue) and pre-admission period (green) in the admission group, for respiratory rate (A), spontaneous triggered breaths (B) and tidal volume (C). Line: median, Box: interquartile range (IQR) (Q1-Q3), Whiskers: 1.5x IQR, Dots: outliers (1.5-3x IQR)Abstract P2 Figure 1BReceiver operating characteristic (ROC) curve for the multivariate rusc model & univariate single variable analysis area under the curve (AUC) displayed with associated significance level (p value <0.05). Higher AUC values indicate better predictive performance. RUSC, respiratory rate, hourly usage, spontaneous cycle & compliance; PaCO2, partial pressure of carbon dioxide in arterial blood