K. Demissie et al., BAYESIAN-ESTIMATION OF ASTHMA PREVALENCE, AND COMPARISON OF EXERCISE AND QUESTIONNAIRE DIAGNOSTICS IN THE ABSENCE OF A GOLD STANDARD, Annals of epidemiology, 8(3), 1998, pp. 201-208
PURPOSE: To estimate the sensitivities, specificities, and predictive
values of exercise challenge and questionnaire, when these tests are u
sed to diagnose asthma in children. METHODS: Participants were childre
n, predominantly aged 6 to 12 years, selected from three primary schoo
l grades among 18 different schools in Montreal. Of 1111 participants,
989 successfully completed a six-minute free running test at school a
nd returned a respiratory questionnaire. A total of 952 children had c
omplete information that could be used for analysis. A history of whee
zing in the past year in conjunction with a past diagnosis of asthma d
efined current asthma by questionnaire. Exercise responsiveness was de
fined as a greater than or equal to 10% fall in FEV1 after a six-minut
e free run. As there is no perfectly accurate diagnostic test for asth
ma, we analyzed the data using a previously published Bayesian method
that allows for the estimation of test properties when no gold standar
d test is available. RESULTS: Current asthma by questionnaire was foun
d to have significantly higher specificity (94.9%, 95% credible interv
al (CI): 93.2-96.5 versus 82.6%, 95% CI; 79.9-85.1) and positive predi
ctive value (53.8%, 95% CI: 41.0-66.7 versus 19.2%, 95% CI: 12.3-27.8)
in comparison to exercise challenge. While there was no statistically
significant difference between the two tests with respect to sensitiv
ity and negative predictive values, the estimates were higher for curr
ent asthma (64.4%, 95% CI: 50.9-76.6 and 96.7%, 95% CI: 94.6-98.1, res
pectively) in comparison to exercise challenge (51.3%, 95% CI: 37.8-64
.5 and 95.4%, 95% CI: 93.2-97.1, respectively). Agreement between the
two diagnostic methods was poor and the combined use of the two tests
did not significantly improve the likelihood of correctly identifying
children with asthma. CONCLUSIONS: Our findings support the view that
exercise testing adds little to a well designed questionnaire in ident
ifying subjects with asthma in community based studies.