Statistical analysis of physiological quality assurance (QA) data
The results suggest that weekly physiological QA is sufficient to highlight problems and using a range of statistical indices to analyse QA data should ensure more consistent internal quality.
Abstract
Background: QA is an key part of lab practice. Using physiological controls we compared differences between devices in the same lab. Methods: QA data was obtained for FEV 1 and FVC, TLC & RV (body box) and CO diffusion (KCO, VA) from nSpire devices, on 5 physiologists from 07/12 to 09/14. Each device had physiological QA weekly with physical QA each day. Each index was compared between devices for each physiologist using coefficient of variation ( cv ), ANOVA, kurtosis (g2: shape of distribution) and skewness (b1: symmetry of distribution). Data was analysed with GraphPad Prism v6. cv should be Results: Results for FEV 1, VA and KCO for 2 physiologists (P1, P2) are shown in Tables 1 & 2. Similar results were obtained for physiologists P3 to P5. Across the devices, device 4 had higher variability with a high cv. RV was the most variable index with every device having a high cv, regardless of physiologist. Statistical differences (p Conclusion: These results suggest that weekly physiological QA is sufficient to highlight problems and using a range of statistical indices to analyse QA data should ensure more consistent internal quality.