OBJECTIVES : Various measures of Short-Acting β-agonist (SABA) use, namely Asthma Medication Ratio (AMR) and SABA canister count, predict future asthma exacerbations. The goal of this study was to validate and compare these measures related to their predictive ability and clinical usefulness.

METHODS : 1,799 patients in the Observational Study of Asthma Control & Outcomes study were included (from Apr 2011 – Jun 2012). The primary outcome of interest is future asthma exacerbation (asthma-related hospitalizations, emergency room visits, and outpatient visits with oral steroid bursts). SABA measures of interest (AMR and SABA canister count) were measured in varying observation/follow-up windows. SABA measures were cross-validated using a generalized linear model framework with performance measures including: overall model performance, discrimination, calibration, and net benefit.

RESULTS : The optimal number of SABA canister threshold identified was greater than one SABA canister during an observation window of nine months and follow-up window of 12 months had a brier score of 0.804, AUROC of 0.536, calibration p-value of 0.995, and net benefit of 0.261. On the other hand, the optimal AMR threshold identified was AMR ≥ 0.1 during an observation window of six months and follow-up window of 12 months had a brier score of 0.805, AUROC of 0.525, calibration p-value of 0.999, and net benefit of 0.259. The probability threshold for the measures of SABA use that optimized prediction were 0.526. The optimal measures (SABA canister ≥ 1 and AMR ≥ 0.1) had a Pearson correlation coefficient of -0.33 with a p-value < 0.0001.

CONCLUSIONS : The accuracy and calibration of the SABA measures were high, however they also exhibited low discriminatory power (which aligns with the results of previous studies). SABA canister ≥ 1 had the greatest predictive ability and clinical usefulness with regards to predicting future exacerbation in members of Kaiser Permanente Colorado.