Are peak flow and symptom measures good predictors of asthma hospitalizations and unscheduled visits?

Epidemiologic studies of pediatric respiratory health often include objecti ve measures such as peak expiratory flow (PEF), and subjective measures suc h as symptom reports. These measures, however, are poorly correlated with e ach other, and there is little evidence that PEF is useful in predicting im portant health outcomes. Within a cohort of 791 inner-city children with as thma, we examined correlations between a series of five peak flow measures and five symptom scores obtained from 2-week diaries. The strongest correlations were found between "total peak flaw lability" de fined as: [(diary maximum - diary minimum)/diary mean] and "% of days with chest tightness" (r = 0.31). Logistic models evaluated peak flow and sympto ms as predictors of an important health outcome: hospitalization or emergen cy department or unscheduled clinic visit for asthma within 30 days of star ting the diary. Each of the peak flow and symptom measures was significantl y related to utilization. However, the predictive power of each measure was low (range of area under ROC curve, 0.54-0.67). Models including only peak flow or symptoms had greater prediction than models with risk factors such as atopy, asthma persistence, and age. The prediction from a model with th e risk factors and symptoms was not improved by adding a peak flow measure to the model (increase in area under ROC, 0.67-0.68). Stratified analyses s uggest that prediction was similar in the fall vs. winter, spring, and summ er months. Greater prediction of health outcomes was found among more persi stent asthmatics and children who were nonatopic. These findings suggest that in a research setting, peak flow monitoring in children did not add prediction beyond that obtained from symptom reports.