DISCUSSION
In this article, we present algorithms that will enable clinicians to accurately identify their patients with asthma by searching data routinely recorded in Canadian EMR systems. These will enable clinicians to create asthma registries that can be used for practice audits and to target and evaluate quality improvement measures.
A simple individual search for use of billing code 493 provided a reasonable balance of sensitivity (78.6%) and specificity (89.2%). However, combining individual searches into algorithms further improved their diagnostic yield, with the best overall accuracy achieved by the combination of asthma in the CPP or use of billing code 493. As expected, further excluding patients for whom a COPD code had been billed resulted in an algorithm with a higher specificity but a lower sensitivity owing to the fact that patients with a combination of asthma and
COPD were no longer counted (Table 3).
Observed sensitivities and specificities of individual search strategies and findings of the discordance analysis offer insight into both care and charting patterns. Prescription of asthma medications was neither particularly sensitive (78.6%) nor specific (63.6%). Sensitivity was limited because medications are not always required in mild asthma (14.3% of asthma subjects had not been prescribed a medication within the past year) (Table 2). Specificity was reduced both
by formulations of these drugs used for non-asthma conditions (eg, nasal steroid sprays) and because asthma medications are also used for COPD. Indeed, exclusion of patients prescribed medications used predominantly for COPD increased the specificity of this algorithm to 73.4%.