Introduction: Non-invasive ventilation (NIV) effectively treats chronic respiratory failure but requires individualisation of settings to ensure comfort and efficacy.  An evolution of ventilators has provided the capacity to review device data remotely.  These data have been used to guide adjustment to ventilator settings but provide less detail than polysomnography (PSG) or polygraphy (PG).  The aim of this study was to compare NIV settings obtained from 3 methods of ventilator setting evaluation: laboratory NIV PSG, ventilator data only (V) and ventilator data with augmented signals.  

Methods: 50 NIV PSG studies were retrospectively selected and analysed by two respiratory and sleep physicians.  Studies were triplicated and re-analysed in random order under the following signal conditions: NIV PSG, ventilator derived signals (V) and polygraphy (PG): ventilator signals, respiratory bands, transcutaneous carbon dioxide and oximetry. NIV setting prescription were recorded for each study. Analysis including Bland and Altman plots were used to compare the ventilator setting prescription under each signal condition.  

Results: The absolute change in pressure support (PS) prescription was 0.32cmH2O (standard deviation (SD) 0.96) for V, 0.90(SD 1.11) for PG and 0.98 (SD 1.25) for PSG. Bland and Altman plot analysis between study type PSG and V demonstrated mean difference (MD) 0.66cmH2O (limits of agreement (LOA) -1.92 to 3.25), and PSG and PG MD 0.08 (LOA -2.72 to 2.88).

Conclusion: Although mean PS prescription did not significantly differ between signal groups, the absolute change was lower in the V signal group.  The broad LOA suggest that a reliance upon ventilator derived signals alone may markedly alter PS prescription.