Analysis
We calculated the interrater reliability of chart abstractor diagnosis using a k statistic. We compared patient characteristics using the Student t test for continuous variables and the c2 test for categorical variables. Using chart abstractor diagnosis as the reference standard, we
calculated true-positive, true-negative, false-positive, and false-negative rates; the sensitivity and specificity of each search; and the Youden index (J = sensitivity + specificity
- 1). We also performed a discordance analysis (an in-depth chart review to identify reasons for misclassification in patients whose test results were falsely positive or falsely negative) in the algorithm with the highest Youden index. We calculated 95% CIs for test characteristics,
assuming a binomial distribution. To maximize sensitivity, patients who had both asthma and COPD
were categorized as asthma. We used SAS, version 9.3,for all analyses.