An algorithm for systematic TB screening should combine one or several screening tests and a separate diagnostic evaluation for TB disease, as recommended by WHO (12). A negative diagnostic test result may be followed up by further clinical evaluation if clinical suspicion of TB is still high. This could include re-testing with the same or another diagnostic method and/or close follow-up of clinical symptoms with or without chest imaging. A positive diagnostic test result might have to be re-confirmed with further testing and clinical evaluation if the positive predictive value of the test result is low.
Different configurations of screening tests have different implications for the sensitivity, specificity and costs of the algorithm. Single screening algorithms comprise one screening test; people who screen positive require diagnostic evaluation for TB. Examples of single screening algorithms are screening all clinic attendees for any cough or a mobile van screening campaign in which everyone in the community is screened by CXR.
Parallel screening algorithms comprise an initial screening step with two screening tests (e.g. screening for symptoms and CXR simultaneously). A positive or abnormal result in either (or both) screening test is an indication for referral onwards towards a diagnostic evaluation. Parallel screening algorithms are more sensitive, as they capture a broader population of people to be evaluated for TB with a diagnostic test. This approach is ideal if the goals of screening are to maximize case detection or to measure the prevalence of TB in the population being screened. (A parallel screening approach is used in prevalence surveys, in which screening for symptoms is combined with CXR) (15). Parallel screening algorithms are, however, typically less specific and therefore have higher cost implications because of the larger number of people referred for diagnostic evaluation and a higher risk of falsepositive screening results.
Serial screening algorithms comprise two screening tests conducted successively, with referral for a second screening test according to the results of the first test. A sequential positive serial screening algorithm is one in which a positive or abnormal result on the first test requires referral to a second screening test, followed by diagnostic evaluation of those who screen positive on both screening tests. An example of this approach is screening for any TB symptom, followed by screening by CXR for those with symptoms. This screening approach increases the pre-test probability of TB in the population being screened before referral for diagnostic evaluation, thereby increasing the efficiency of the screening programme and reducing the risk of false-positive diagnoses. This approach is, however, less sensitive.
A sequential negative serial screening algorithm is one in which a positive or abnormal result on the first screening test results in referral to diagnostic evaluation, while a negative or normal result on the first screening test results in referral for a second screening test and then subsequent referral for diagnostic evaluation for those who screen positive or abnormal in the second screening test. A sequential negative serial screening algorithm has the same sensitivity and specificity as a parallel screening algorithm with the same tests (the same number of people will be referred for diagnostic evaluation) but reduces the cost, because the second screening test is limited to individuals who test negative in the first. For example, an algorithm that begins with screening for symptoms and then CXR for those who do not present with symptoms will result in fewer CXRs being conducted with the same case detection as CXR plus symptom screening for all. This may, however, introduce delays, given that the tests are not run simultaneously. The specificity of a negative sequential screening approach will be lower than that of a positive sequential algorithm because of the larger number of people referred for diagnostic evaluation and the higher risk of false-positive screening results.