A Prospective Validation of a Clinical Algorithm to Detect Tuberculosis in Child Contacts

To the Editor: Over 60% of pediatric tuberculosis cases are undetected by healthcare services in low-income settings (1).Untreated children with tuberculosis have fatality rates of .20%,reaching above 40% in children ,5 years old (2).Specific, effective, and validated interventions to increase case detection in children are urgently needed.Household contact tracing has been widely recommended to increase detection of childhood tuberculosis cases.Although it can provide high yields (3, 4), child contact tracing has rarely been implemented in high-burden settings (5).Methods are urgently needed to incentivize contact tracing for national tuberculosis programs in low-income settings.A clinical risk score was recently proposed in Taiwan (a high-income, low-prevalence setting) to increase the effectiveness of child contact tracing (6).This algorithm uses a combination of contact, index, and environmental characteristics to allocate child contacts into high-and low-risk groups for secondary disease.External validation in high-burden settings has not been performed.To investigate whether this score ( 6) is useful for detecting tuberculosis in exposed children from Sub-Saharan Africa, we used data from a large Ugandan child contact cohort.This was a prospective cohort study of household contacts of tuberculosis cases; the study has been described previously (7-9).Briefly, we identified adult patients in Kampala, Uganda, with a new diagnosis of tuberculosis.Index cases were microbiologically confirmed through a positive culture test and evaluated through physical examination and medical history.Households were visited by field workers within 2 weeks of diagnosis, and contacts were defined as spending at least 7 consecutive days in the index's household 3 months before diagnosis.Coprevalent tuberculosis was defined as tuberculosis within 3 months of baseline.Contacts were evaluated for tuberculosis through a medical examination, posteroanterior chest radiographs, specimen microscopy, and mycobacterial culture.Contacts without coprevalent tuberculosis were followed for incident tuberculosis for 2 years.Contacts were grouped by age (,13 yr old) and disease status to categorize the data according to the scoring algorithm (6).Contacts were given a point score based on the algorithm (6): contact tuberculin skin test (TST) induration (2 points if 10-14 mm, 3 points if 15-20 mm, and 4 points if >20 mm), index smear result (1 point if smear positive), index sex (1 point if female), and burden of residence area (2 points if index lived in a high-incidence area).Because all index cases in the Ugandan cohort lived in a high-incidence area, all contacts were automatically given 2 points.The score's discrimination and classification accuracy was compared with results from the original derivation and internal † The Cochran-Armitage test was used to evaluate trends within groups.‡ Score cutoffs were chosen as shown in Reference 6 to provide proper comparison between studies.