Abstract

Background: The efficacy of decision making based on longitudinal spirometric measurements depends critically on the precision of the available data, which is determined by the magnitude of the within-person variation.

Aims: Firstly, to describe and investigate two statistical methods—a pairwise estimate of within-person standard deviation s_p and the reliability coefficient G—for use in the monitoring of precision of longitudinal measurements of forced expiratory volume in one second (FEV1). Secondly, to investigate the effect of longitudinal data precision on the detectable excess rate of decline in FEV1.

Methods: The authors “monitored” retrospectively on a yearly basis the magnitude of the within-person variation s_p and the coefficient G in 11 workplace based spirometric monitoring programmes conducted from 1987 to 2001 on 12 729 workers in various industrial plants.

Results: The plant-specific mean values s̄_p (range 122–166 ml) and Ḡ (range 0.88–0.95), averaged over all years of follow up, correlated well with the plant-specific within-person standard deviation s_r (range 130–177 ml) estimated from all longitudinal data. The correlations were 0.90 for s̄_p and 0.68 for Ḡ. The average precision of the longitudinal FEV1 measurements affected the duration of follow up needed to identify a “true” excess rate of decline in FEV1 in an individual.

Conclusions: The results show that monitoring of longitudinal spirometry data precision (1) allows that data precision can be improved or maintained at levels that allow individuals with a rapid decline to be identified at an earlier age; and (2) attaches a measure of precision to the data on which decision making is based.