Article Text
PDF
Abstract
Introduction and objectives The objective measurement of cough frequency has changed the standards by which we evaluate anti-tussive treatments and study the mechanisms driving cough. The distribution of individual cough events over time can vary substantially between patients with identical cough frequencies but this has received little attention. The aim of this study was to explore the differences in the temporal distribution of cough in 24 hour acoustic recordings across a range of respiratory conditions using the Weibull Distribution.
Methods We studied 24 hour acoustic recordings made using the VitaloJAK™ cough monitoring system from 62 participants [healthy volunteers (n=11), chronic cough (n=15), asthma (n=10) chronic obstructive pulmonary disease (COPD) (n=16), Idiopathic Pulmonary Fibrosis (IPF) n=10)]. Using the exact location of the explosive phase of each cough sound, the lengths of inter-cough intervals were calculated in each recording and the fit to Weibull and exponential distributions assessed; the shape parameter which describes the ‘clustering’ of events was examined.
Results The number of coughs per 24 recording ranged from 0 to 3133, median 158.0 coughs and was increased compared with healthy controls in all disease groups [healthy volunteers median 13.0(IQR 1.0–22.0) coughs/24 hour versus chronic cough 509.0 (296.0–916.0), asthma 35.0 (13.5–144.0) COPD 158.0 (41.0–791.5), Idiopathic Pulmonary Fibrosis 193.5 (101.3–542.8), all p<0.001]. Overall the Weibull distribution equally or better fitted the inter-cough intervals compared with exponential distribution for 56/62 (90.3%) of subjects, based on Akaike’s Information Criterion (AIC). Interestingly, there were significant differences in shape parameter values between diagnostic groups [healthy volunteers median 0.22(IQR 0.00–0.30) chronic cough 1.08 (0.40–1.44), asthma 0.26 (0.23–0.57) COPD 0.32 (0.27–0.93), IPF 0.31 (0.26–0.59); p<0.001], see figure 1. Post-hoc testing suggested that healthy controls had significantly smaller shape parameter values (all p<0.02) whereas chronic cough patients had significantly larger shape parameter values (all p<0.002) compared with all other groups.
Conclusions The intervals between cough events are not random and can be described by a Weibull distribution. Significant differences in the Weibull shape parameter suggest differences in the temporal clustering of coughs between different respiratory conditions and therefore may be an important novel parameter from objective cough monitoring.