The rise of artificial intelligence reading of chest X-rays for enhanced TB diagnosis and elimination.

We provide an overview of the latest evidence on computer-aided detection (CAD) software for automated interpretation of chest radiographs (CXRs) for TB detection. CAD is a useful tool that can assist in rapid and consistent CXR interpretation for TB. CAD can achieve high sensitivity TB detection among people seeking care with symptoms of TB and in population-based screening, has accuracy on-par with human readers. However, implementation challenges remain. Due to diagnostic heterogeneity between settings and sub-populations, users need to select threshold scores rather than use pre-specified ones, but some sites may lack the resources and data to do so. Efficient standardisation is further complicated by frequent updates and new CAD versions, which also challenges implementation and comparison. CAD has not been validated for TB diagnosis in children and its accuracy for identifying non-TB abnormalities remains to be evaluated. A number of economic and political issues also remain to be addressed through regulation for CAD to avoid furthering health inequities. Although CAD-based CXR analysis has proven remarkably accurate for TB detection in adults, the above issues need to be addressed to ensure that the technology meets the needs of high-burden settings and vulnerable sub-populations.

Resource implications of the latent tuberculosis cascade of care: a time and motion study in five countries.

BACKGROUND: The End TB Strategy calls for global scale-up of preventive treatment for latent tuberculosis infection (LTBI), but little information is available about the associated human resource requirements. Our study aimed to quantify the healthcare worker (HCW) time needed to perform the tasks associated with each step along the LTBI cascade of care for household contacts of TB patients. METHODS: We conducted a time and motion (TAM) study between January 2018 and March 2019, in which consenting HCWs were observed throughout a typical workday. The precise time spent was recorded in pre-specified categories of work activities for each step along the cascade. A linear mixed model was fit to estimate the time at each step. RESULTS: A total of 173 HCWs in Benin, Canada, Ghana, Indonesia, and Vietnam participated. The greatest amount of time was spent for the medical evaluation (median: 11 min; IQR: 6-16), while the least time was spent on reading a tuberculin skin test (TST) (median: 4 min; IQR: 2-9). The greatest variability was seen in the time spent for each medical evaluation, while TST placement and reading showed the least variability. The total time required to complete all steps along the LTBI cascade, from identification of household contacts (HHC) through to treatment initiation ranged from 1.8 h per index TB patient in Vietnam to 5.2 h in Ghana. CONCLUSIONS: Our findings suggest that the time requirements are very modest to perform each step in the latent TB cascade of care, but to achieve full identification and management of all household contacts will require additional human resources in many settings.

Computer-aided detection of pulmonary tuberculosis on digital chest radiographs: a systematic review.

OBJECTIVE: To systematically review the diagnostic accuracy of computer-aided detection (CAD) of pulmonary tuberculosis (PTB) on digital chest radiographs (CXR). DESIGN: We searched four databases for articles published between January 2010 and December 2015 comparing CAD of PTB on CXR to a microbiologic reference standard (smear, culture or polymerase chain reaction). We collected and summarised data on study design, CAD software and diagnostic accuracy (sensitivity, specificity, area under the curve [AUC]). RESULTS: We included 5 of 455 articles identified by searching databases. PTB prevalence ranged from 18% to 60%, and human immunodeficiency virus (HIV) prevalence from 33% to 68%. All articles evaluated CAD4TB, the only commercially available software. AUC ranged from 0.71 to 0.84. Software settings that increased sensitivity resulted in important reductions in specificity, and vice versa. Risk of bias was low in prospective studies (n = 2), and high in retrospective studies (n = 3). CONCLUSION: Evidence assessing CAD's diagnostic accuracy is limited by the small number of studies, most of which have important methodological limitations, the availability and evaluation of only one software programme, and limited generalisability to settings where PTB and HIV are less prevalent. Additional research is required.

Low rates of recurrence after successful treatment of multidrug-resistant tuberculosis in Tomsk, Russia.

SETTING: Tomsk, Russia, where multidrug-resistant tuberculosis (MDR-TB) is prevalent. OBJECTIVES: To report rates of recurrence following successful treatment of MDR-TB in a program providing individualized treatment regimens designed according to the current global standard of care. DESIGN: A retrospective cohort study of 408 adults successfully treated for pulmonary MDR-TB from 10 September 2000 to 1 November 2004, and followed for up to 6 years post-treatment. We used Poisson regression with generalized estimating equations to assess whether recurrence rates changed significantly with time. RESULTS: We analyzed 399 (97.5%) patients with at least one follow-up visit (15 850 person-months of observation [PMO]). Baseline resistance to second-line drugs was common (65.2%); 398 patients (99.7%) were human immunodeficiency virus (HIV) negative. In the first year of post-treatment follow-up, there were six episodes of recurrence (1.4/1000 PMO, 95%CI 0.5-3.0). After the first post-treatment year, there were 21 episodes of recurrence (1.8/1000 PMO, 95%CI 1.1-2.8). The rate did not change significantly with time. CONCLUSION: Individualized regimens designed according to the current global standard of care achieved low rates of MDR-TB recurrence among non-HIV-infected persons treated in a programmatic setting.