Background Radiographic screening for pulmonary tuberculosis is used in high risk groups in many countries. There are no published reports of the sensitivity and specificity of this technique in an operational setting with high screening throughput. A Department of Health screening programme has used mobile digital chest radiography to identify pulmonary tuberculosis amongst homeless people, problem drug users and prisoners in London since April 2005. We aimed to use data from this programme and from national tuberculosis surveillance to establish the sensitivity and specificity of mobile digital X-ray screening for the identification of pulmonary tuberculosis.

Methods Data collected on those screened for the period March 2005 to March 2010 were collated. Abnormal radiographs were classified as: 1) suspected active tuberculosis; 2) signs of old inactive tuberculosis; 3) other suspicious findings that warrant further investigation (eg, possible cancer); 4) abnormal but no further investigation required; 5) normal. Demographic data from those screened were matched to the national tuberculosis surveillance database using probabilistic algorithms. Duplicates within the screening database (ie, persons having been screened more than once) were removed using the same algorithms. To calculate sensitivity and specificity all those who were referred with suspected active tuberculosis were classified as ‘screen positive’ with all others classified as ‘screen negative’. The gold standard comparator was a notification of pulmonary tuberculosis within the four months after screening. Non-pulmonary cases of tuberculosis were also excluded.

Results 39 225 individuals were screened during the study period, 75% were male and the median age was 36 (IQR 27–45). 430 of those screened were referred with suspected active tuberculosis for further investigation. Matching of the screening data to surveillance data identified 44 screened individuals who were notified with pulmonary tuberculosis within 4 months of screening, 36 of these were screen positive. Based on these results, sensitivity was 82% (95% CI 67.3 to 91.8%) and specificity 99.3% (95% CI 99.1 to 99.3%) (see Abstract S6 Table 1).

Conclusion High throughput mobile digital radiographic screening for pulmonary tuberculosis has high sensitivity and specificity in high risk groups and should be considered as a key tool for active case finding in these populations.