Statistical analysis: Across-shift change in pulmonary function was calculated by taking the difference between preshift measurement and the last measurement measured on the same day. There was at least a 2 h difference between the two readings. In addition to the absolute difference, the shift change was also expressed as the percentage change from the preshift measurement. Annual rate change in pulmonary function was obtained by dividing the difference in the 1989/90 and 1994/95 pulmonary function values by the time between the two values.Means and standard deviations were calculated for continuous variables. Categorical variables were described by frequencies and percentages. Pearson correlation coefficients were used to quantify the association between the annual rate change and the shift change in pulmonary function measurements. Multiple regression analyses were used to examine the relationship between the annual rate change and across-shift change in pulmonary function measurements after adjusting for confounding variables.