Conference Abstract | Volume 8, Abstract NACNDC/19JASH043 (Oral) | Published: 24 Nov 2025
Louis Ocen1,&, Solome Najjingo1, Justus Muhangi1, Ronald Tamale1, Henry Suubi1, George Harris Kyambadde1
1Uganda Episcopal Conference – Uganda Catholic Medical Bureau, Kampala, Uganda
&Corresponding author: Louis Ocen, Uganda Catholic Medical Bureau, Kampala, Uganda. Email: louisocen@gmail.com/locen@ucmb.co.ug, ORCID: https://orcid.org/0009-0000-9170-6874
Received: 13 Sept 2025, Accepted: 20 Oct 2025, Published: 24 Nov 2025
Domain: Infectious Disease Epidemiology
Keywords: tuberculosis, CAD4TB, digital radiography, screening thresholds, Uganda
©Louis Ocen et al. Journal of Interventional Epidemiology and Public Health (ISSN: 2664-2824). This is an Open Access article distributed under the terms of the Creative Commons Attribution International 4.0 License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Cite this article: Louis Ocen et al., CAD4TB’s precision play: Threshold 45 optimizes TB screening in Westnile’s high burden setting of Uganda. Journal of Interventional Epidemiology and Public Health. 2025;8(ConfProc6):00043. https://doi.org/10.37432/JIEPH-CONFPRO6-00043
Uganda uses artificial intelligence-assisted chest x-ray screening with computer-aided diagnosis for tuberculosis (CAD4TB), applying a national threshold of 50 to identify presumptive TB cases. During the 2024 community awareness screening and testing (CAST+) campaign, a lower threshold of 30 was piloted in Westnile. This study evaluated CAD4TB version 7 performance in West Nile to identify an optimal threshold that improves efficiency without compromising case detection.
A retrospective cross-sectional study was conducted among individuals screened for TB using digital chest X-ray between 2021 and 2024. A total of 6,000 records were reviewed, including 188 GeneXpert-confirmed TB-positive and 4,690 TB-negative cases. Records with unavailable GeneXpert results (n = 1,122) were excluded. CXRs were analyzed using CAD4TB version 7. Diagnostic performance was assessed using Receiver Operating Characteristic analysis, evaluating sensitivity, specificity, referral burden, number needed to test (NNT), and TB positivity yield at thresholds of 30% (CAST+), 45%, and 50% (national standard).
CAD4TB demonstrated excellent discrimination with an area under the curve of 0.975. At 45%, sensitivity was 96.8% (95% CI: 93.1–98.9%) and specificity 94.2% (95% CI: 93.5–94.9%), detecting 182 of 188 TB cases with 453 referrals (NNT = 2; yield = 40%). At the national 50% threshold, sensitivity was 96.3% (95% CI: 92.5–98.6%) and specificity 98.0% (95% CI: 97.6–98.4%), identifying 181 cases with 274 referrals (NNT = 2; yield = 66%). Compared with CAST+ 30% (97.3% sensitivity, 70.3% specificity, 1,576 referrals identifying 183 TB cases, yield = 12%), the 45% threshold reduced referrals by 71% while detecting a similar number of cases.
CAD4TB v7 demonstrated food diagnostic accuracy. While the 50% threshold maintains high specificity, a locally optimized 45% threshold offered modest sensitivity gains with manageable referral. These findings support context-specific CAD4TB threshold adjustments in national TB screening programs to optimize case detection and operational efficiency.
Menu