Comparing the Sensitivity and Specificity of Various TB Screening Mechanisms
We explored whether routine TB screening for coughing of two weeks or longer yields greater sensitivity and specificity than relying on the discretion of health workers or a random approach.
Our analysis is based on a prevalence survey of 31 007 adults in eight high TB prevalence communities in the Western Cape (2010). Culture confirmed pulmonary TB was defined as M.tuberculosis (M.tb) isolated from a respiratory sample after inoculation into manual mycobacterial growth indicator (MGIT BD) tubes. Individuals were considered to have culture positive TB if the growth in the MGIT tube was confirmed to be M.tb by 16SrRNA sequencing. Participants were asked about demographic data, typical TB symptoms and health seeking behaviour.
Sensitivity and specificity of the following TB screening procedures were evaluated using hypothetical models: (1) the recommended protocol of testing those who report a cough for two weeks or more; (2) three enhanced protocols suggested by the literature; (3) random allocation of coughing clients to testing and (4) screening procedures subject to health worker discretion (as reported by participants). With the three enhanced protocols, screening rules were simulated to include: i) patients with known HIV-positive status ii) patients with known HIV-positive status who tested after a HIV campaign (uptake 34%) and iii) patients with at least 3 of the 5 TB symptoms.
The hypothetical protocol models were based on information about symptoms and health seeking behaviour reported by participants who were currently coughing. We estimated the specificity and sensitivity of actual screening procedures used in clinics based on self-reported information regarding a visit to a clinic for a cough.
The recommended protocol outperformed current practice in clinics in terms of sensitivity but not specificity. Enhanced protocols tended to have lower specificity but increased sensitivity. The proportion of coughing patients offered a test ranged from 52.6% to 89.5% for the eight facilities.
Given that the benefits associated with increased sensitivity are likely to be larger than those associated with an increased specificity, the results provide support for adherence to the recommended protocol. The level of variation across clinics for offering a TB test to symptomatic individuals, is concerning. These findings need to be validated and explored further under operational conditions.
This work was presented as a poster at the 44th Union World Conference on Lung Health in Paris in November 2013 and is currently being converted into a journal article.
Collaborators: Mareli Claassens, Helen Ayles, Peter Godfrey-Fausset, Nulda Beyers