Development of OASYS-2: a system for the analysis of serial measurement of peak expiratory flow in workers with suspected occupational asthma.

BACKGROUND: Serial peak expiratory flow (PEF) measurement is usually the most appropriate first step in the confirmation of occupational asthma. Visual assessment of the plotted record is more sensitive and specific than statistical methods so far reported. The use of visual analysis is limited by lack of widespread expertise in the methods. A computer assisted diagnostic aid (OASYS-2) has been developed which is based on a scoring system developed from visual analysis. This removes the requirement for an experienced interpreter and should lead to the more widespread use of the technique.
METHODS: PEF records were collected from workers attending an occupational lung disease clinic for investigation of suspected occupational asthma and from workers participating in a study of respiratory symptoms in a postal sorting office. PEF records were divided into two development sets and two gold standard sets. The latter consisted of records from workers in which a final diagnosis had been reached by a method other than PEF recording. An experienced observer scored individual work and rest periods for the two development set PEF records; linear discriminant analysis was used to compare measurements taken from development set 1 records with visual scores. Two equations were produced which allowed prediction of scores for individual work or rest periods. The development set 2 was used to determine how these scores should be used to produce a whole record score. The first gold standard set was used to determine the whole record score which best separated those with and without occupational asthma. The second set determined the sensitivity and specificity of the chosen score.
RESULTS: Two hundred and sixty eight PEF records were collected from 169 workers and divided into two development sets (81 and 60 records) and two gold standard sets (60 and 67 records). Linear discriminant analysis produced equations predicting the score for work periods incorporating five indices of PEF change and one for rest periods using seven indices. These equations correctly predicted the score for development set 1 work and rest periods on 61% of occasions (kappa = 0.47). The whole record score for development set 2 records, after weighting for definite or definitely no occupational effect, correlated with the visual score (correlation coefficient 0.86). Comparison with gold standard set 1 identified a cut off which proved to have a sensitivity of 75% and a specificity of 94% for an independent diagnosis of occupational asthma when applied to gold standard set 2.
CONCLUSIONS: These results suggest that the sensitivity and specificity of analysing PEF records for occupational asthma using OASYS-2 approaches that of visual analysis, but it should be absolutely reproducible. The performance of OASYS-2 is more specific and approaches the sensitivity of other statistical methods of analysis. The evaluation of a large number of PEF records from workers exposed to different sensitising agents suggests that these results should be robust and should be repeatable in clinical practice.