Objective: To identify risk factors for defaulting from anti-tuberculosis treatment.
Setting: Directly observed treatment in Hong Kong Government chest clinics.
Design: Defaulters were recruited from a cohort of tuberculosis patients registered from 1 January to 31 March 1999. Three controls per case, matched for age and sex, were selected randomly from the cohort. Patient factors, initial tuberculosis characteristics and treatment-related variables were collected by review of medical records.
Results: On matching 102 defaulters and 306 controls, a logistic risk model of default that considered patient factors, initial disease characteristics and treatment-related factors identified seven risk factors: current smoking (OR 3.00, 95% CI 1.41-6.39), past TB with default (OR 6.23, 95% CI 1.95-19.91), poor initial adherence (OR 117.21, 95% CI 13.52-1015.92), fair initial adherence (OR 11.02, 95% CI 2.15-56.43), unknown initial adherence (OR 6.59, 95% CI 3.47-12.49), treatment side effects (OR 13.30, 95% CI 3.23-54.79), and subsequent hospitalisation (OR 0.27, 95% CI 0.11-0.67). Its predictive power was 85%. Another model that considered only factors on registration for treatment gave a lower predictive power of 70%.
Conclusions: Treatment default could be predicted fairly accurately by considering patient and treatment-related factors.