Performance of algorithms for tuberculosis active case finding in underserved high-prevalence settings in Cambodia: a cross-sectional study

Background: Most studies evaluate active case findings (ACF) for bacteriologically confirmed TB. Adapted diagnostic approaches are needed to identify cases with lower bacillary loads. Objectives: To assess the likelihood of diagnosing all forms of TB, including clinically diagnosed pulmonary and extra-pulmonary TB, using different ACF algorithms in Cambodia. 

Methods: Clients were stratified into 'high-risk' (presumptive TB plus TB contact, or history of TB, or presumptive HIV infection; n = 12,337) and 'moderate-risk' groups (presumptive TB; n = 28,804). Sputum samples were examined by sputum smear microscopy (SSM) or Xpert MTB/RIF (Xpert). Initially, chest X-ray using a mobile radiography unit was a follow-up test after a negative sputum examination [algorithms A (Xpert/X-ray) and B (SSM/X-ray)]. Subsequently, all clients received an X-ray [algorithms C (X-ray+Xpert) and D (Xray+SSM/Xpert)]. X-rays were interpreted on the spot. 

Results: Between 25 August 2014 and 31 March 2016, 2217 (5.4%) cases with all forms of TB cases were diagnosed among 41,141 adults. The majority of TB cases (1488; 67.1%) were diagnosed using X-ray. When X-rays were taken and interpreted the same day the sputum was collected, same-day diagnosis more than doubled. Overall, the number needed to test (NNT) to diagnose one case was 18.6 (95%CI:17.9-19.2). In the high-risk group the NNT was lower [algorithm D: NNT = 17.3(15.9-18.9)] compared with the 'moderate-risk group' [algorithm D: NNT = 20.8(19.6-22.2)]. In the high-risk group the NNT was lower when using Xpert as an initial test [algorithm A: NNT = 12.2(10.8-13.9) or algorithm C: NNT = 11.2(9.6-13.0)] compared with Xpert as a follow-up test [algorithm D: NNT = 17.3(15.9-18.9)]. 

Conclusion: To diagnose all TB forms, X-ray should be part of the diagnostic algorithm. The combination of X-ray and Xpert testing for high-risk clients was the most effective ACF approach in this setting.