FeedbackThis operational handbook includes 10 screening algorithm options for screening the general population and groups at higher risk (not including people living with HIV or children), consisting of a combination of one or two screening tests and a diagnostic evaluation (Annex 1). Algorithms for screening people living with HIV are discussed in Chapter 5 and algorithms for screening children in Chapter 6.
The algorithms differ in sensitivity and specificity and, therefore, have different yields of detection of prevalent TB, predictive values and associated costs. The performance of the algorithms also depends on the prevalence of TB in the population being screened. Tables A2.1-A2.3 in Annex 2 contain modelled estimates of the performance of the algorithms described below, including the results of true- and false-positive diagnoses for the entire algorithm, consisting of the screening test(s) followed by diagnostic evaluation with an mWRD.
For all algorithms, the risk of a false-positive diagnosis increases as the prevalence declines; therefore, special attention must be paid to diagnostic accuracy of the screening algorithm, particularly when the prevalence of TB in the screened population is < 1%. At a TB prevalence of 0.5% in the screened population, all the algorithms have a positive predictive value of < 75% (i.e. 25% give a false-positive diagnosis). Efforts must therefore be made to ensure high-quality diagnostic procedures and clinical assessment, especially when the TB prevalence in the screened population is moderate to low.
In each given screening situation, it is critical to consider the proportions of false-positive and falsenegative results that are unacceptable. Ethical considerations such as unnecessary anxiety and inappropriate TB treatment due to a false-positive diagnosis and the adverse consequences of missing or delaying a TB diagnosis should guide the acceptable sensitivity and specificity of the algorithm. Considerations will depend on risk groups. For groups of individuals who are at high risk of dying or of other severe negative effects of a missed or delayed diagnosis and treatment, however, the algorithm used should have very high sensitivity, even at the expense of lower specificity.
The algorithms have different costs and requirements in terms of human resources and health systems. The choice of algorithm depends on the risk group, the prevalence of TB, the availability of resources and the feasibility of implementation.
Algorithms that begin with screening for cough
Fig. A.1.1 – Screening with cough (page 60)
Fig. A.1.2 – Parallel screening with cough and CXR (page 61)
Fig. A.1.3 – Sequential positive serial screening with cough and CXR (page 62)
Fig. A.1.4 – Sequential negative serial screening with cough and CXR (page 63)
Algorithms that begin with screening for any symptom compatible with TB
Fig. A.1.5 – Screening with any TB symptom (page 64)
Fig. A.1.6 – Parallel screening with any TB symptom and CXR (page 65)
Fig. A.1.7 – Sequential positive serial screening with any TB symptom and CXR (page 66)
Fig. A.1.8 – Sequential negative serial screening with any TB symptom and CXR (page 67)
Algorithm that begins with screening with CXR (page 68)
In addition to the parallel and sequential algorithms that include CXR above, the algorithm in Fig A.1.9 presents an option to screen only with CXR, followed by referral for diagnostic evaluation for people with an abnormal CXR.
Algorithm that begins with screening with mWRD (page 69)
The algorithm in Fig A.1.10 presents a screening approach that begins with an mWRD, followed by a thorough clinical evaluation (including physician assessment and further tests such as CXR or repeat mWRDs on additional sputum samples) for those with a positive test result.