Study population
Reviewers assessed 460 charts, of which 41 (8.9%) had no available data, 21 (4.6%) were duplicates, and 398 (86.5%) were fully reviewed. These 398 patients consisted of 112 (28.1%) with asthma, 81 (20.4%) with COPD, 104 (26.1%) with other respiratory conditions (neither asthma nor COPD), and 101 (25.4%) with nonrespiratory conditions.Nine of 112 (8.0%) asthma patients had coexisting COPD.Concordance between the 2 reviewers in chart abstraction diagnosis was high (k = 0.89, 95% CI 0.80 to 0.97) (Table 2).
Search algorithms
Results from each unique search query and the 5 algorithms with the best test characteristics (based on the Youden index) are presented in Table 3. All tested search algorithms as well as search characteristics by clinic are available upon request. True-positive and false-negative rates for these searches and for the most sensitive and most specific individual algorithms are represented in Figure 2.
Discordance analysis
The algorithm combining asthma in the cumulative patient profile (CPP) or use of billing code 493 (asthma or allergic bronchitis) had the highest Youden index (Table 3) and was used for the discordance analysis. There were 11 false-negative results and 46 false-positive results. Among
the 11 false-negative results, 4 charts (36.4%) simply did not have asthma in the CPP and had not been billed for asthma despite clear chart documentation of asthma. Of the remaining 7 charts, 6 (85.7%) had been diagnosed with asthma by an outside specialist. In these cases, we suspect that clinicians might have been less likely to update the CPP because the diagnosis was made elsewhere. Also, 6 of these 7 had COPD in addition to asthma, and clinicians appeared to default to using COPD rather than asthma billing codes for respiratory-related visits in these patients. Among the 46 false-positive results, 14 (30.4%) had COPD as opposed to asthma, and clinicians might have confused this with asthma when completing the CPP or when billing.
Another 13 (28.3%) were initially suspected of having had asthma, but later had negative objective test results for asthma. Five of these patients also saw other specialists and received the following alternate diagnoses: eosinophilic bronchitis (n = 2); bronchiectasis (n = 2); and gastroesophageal reflux disease (n = 1). Some of these charts had falsely positive results because the asthma billing code was used at the time of the initial diagnostic suspicion, and others because the CPP had not been updated in light of objective testing and non–family physician specialist results. Another 16 (34.8%) had an upper or lower respiratory
tract infection that resulted in (usually) isolated use of an asthma billing code (with no other evidence of asthma in the chart). Finally, 3 (6.5%) had a single asthma billing code with no plausible explanation and no other evidence of asthma in the chart.