Contact predictors of tuberculosis disease in all household contacts (both multidrug-resistant tuberculosis contacts and drug-susceptible tuberculosis contacts).

1These variables were included in the multivariate regression as they were determined to be p 0.2 on univariate analysis or were known confounding variables identified a priori that have previously been associated with second cases of tuberculosis.2Contacts who had not been tested for HIV/diabetes were assumed to be negative.3These variables were identified as being time-varying covariates with increasing hazards over the length of the study; the corresponding HR should therefore be regarded as an average over the follow-up period.Contact predictors of tuberculosis disease in all household contacts (both multidrug-resistant tuberculosis contacts and drug-susceptible tuberculosis contacts).