Risk for tuberculosis in child contacts. Development and validation of a predictive score

Contact investigation of persons exposed to tuberculosis (TB) is resource intensive. To date, no clinical prediction rule for TB risk exists for use as a guide during contact investigation. We sought to develop and validate a simple and easy-to-use predictive score for TB risk, using data routinely available during contact investigation. The derivation cohort consisted of 9,411 children aged 0 to 12 years from 2008 to 2009 national contacts cohort. We used a multivariate Cox proportional hazards model to predict the risk of developing active TB. The validation cohort consisted of 2,405 children from the 2005 national contacts cohort. We calculated area under the receiver operating characteristic curves of the model as well as the predicted risk of TB for contacts with different scores. An 8-point scoring system was developed, including reaction to tuberculin skin test of the contacts, as well as smear-positivity, residence in high-incidence areas, and sex of the index cases. Area under the receiver operating characteristic curve was 0.872 (95% confidence interval, 0.810-0.935) for the derivation cohort and 0.900 (95% confidence interval, 0.830-0.969) for the validation cohort. The risk of developing active TB within 3 years is 100, 7.8, 4.3, 1.0, 0.7, and 0.2% for contacts with risk scores of 7, 6, 5, 4, 3, and 2, respectively. A risk predictive score was developed and validated to identify child contacts aged 0 to 12 years at increased risk for active TB. This predictive score can help to prioritize active case finding or isoniazid preventive therapy among children exposed to TB.