Emergency Department (ED)-Directed Interventions to Improve Asthma Outcomes

This prospective, randomized, open label study will include asthmatics 17-55 years of age with no evidence of chronic obstructive pulmonary disease. Patients will be eligible if they visit one of four Emergency Departments (EDs) in Edmonton or Calgary and receive treatment for acute asthma resulting in discharge home. All patients should have a primary care provider (PCP: Family Physician, Internist or Nurse Practitioner) with whom to follow-up or one will be found for them. At discharge patients will be randomized into three groups: A: Usual care; PCPs will receive a faxed copy of the ED chart and patients will receive a discharge plan and a paper-based educational pamphlet (treatment in the ED and at discharge will be left to ED physicians' discretion); B: Usual care + personalized fax to the patients' PCP including a copy of the ED chart and a opinion-leader (OL) letter encouraging follow-up within two weeks and providing management suggestions; or C: Usual care + personalized fax to the patients' PCP including the OL letter as described above + involvement of a case manager who will encourage patients' to pursue follow-up, provide management review and offer brief education within the next week. Outcomes will be ascertained blinded to treatment allocation through telephone follow-up at 30 and 90 days. A sample of 366 patients (122 per group) is required based on the proportion with relapse at 90 days (40%) and a chi-square test of association and post-hoc tests (groups A vs. B, groups B vs. C). This sample will allow for the detection of an moderate effect size of at least 0.171, and a difference between groups A and B of 50% (i.e., 40% vs. 20%) and between groups B and C of 75% (i.e., 20% vs. 5%) using two-sided z-tests, 80% power, α=0.025. Intention to treat analyses will be conducted. Relapse rates by group and associated 95% confidence intervals (CI) will be calculated. Proportions among groups will be compared using chi-square tests and if statistically significant, separate proportion tests will compare pairs of groups adjusting for multiple testing. A multivariable logistic regression model will adjust effect estimates for potential baseline imbalances and site-specific differences in effectiveness.