Document resource
Background Almost 1/3 of incident pulmonary TB (PTB) cases in England start treatment >4 months after symptom onset. 1 Leicester, a high incidence TB area (>40 cases/100,000 population) provides a centralised rapid access tuberculosis (RATB) pathway for suspected TB cases from primary and secondary care, facilitating urgent assessment. To expedite diagnosis, radiologists can assign urgent referral codes, such as the TB code, to streamline referrals to appropriate multidisciplinary teams.Aim We investigated the accuracy of radiology reports for diagnosing PTB and the impact of reporting codes and keywords on time to PTB notification in patients referred to the RATB service.Methods We assessed CXR reports of 1124 patients referred to the RATB service at University Hospitals of Leicester (Jan 2019–Dec 2023) for inclusion of TB or other urgent referral codes - lung cancer code (RAL), GPCTFU (GP CT follow-up), XRCCT (urgent CT follow-up) and TB keywords (TB*, tubercul*, mycobac*), and the intervals between CXR request, report and TB diagnosis.Results There were 216 cases (19.2%) of PTB among the 1124 patients referred to RATB service. CXRs were performed promptly (median same-day), but median reporting times varied by requesting source - 2 days (A&E), 2 days (primary care), 6 days (inpatient), 11 days (outpatient) (p<0.001). Urgent codes or TB keywords were reported in 138 (63.9%) of the PTB cases. Interestingly, 22 PTB cases (10.2%) had reports containing TB keywords but no referral code assigned, and 41 cases (21.2%) had other abnormalities reported but no link to TB diagnosis was made. The diagnostic performance of different codes and TB keywords is noted in table 1. Inclusion of the RATB code or inclusion of TB keywords was associated with a significantly shorter median notification delay than non-coded cases (9 vs. 46.5 days (p<0.001)).Abstract M38 Table 1Effect of CXR coding and reporting of TB keywords on time to TB notification CXR CODING CATEGORIES IN PATIENT REFERRALS TO RATB Variable RATB coded RAL, GPCTFU or XRCCT urgent codes applied (RATB excluded) Non-urgent X-ray follow-up code (RATB and urgent codes excluded) Non-coded scans including TB keywords Non-coded scans with no TB keywords Number of CXRs in category (% total CXRs) 308(27.4) 103(9.2) 19(1.7) 79(7.0) 615(54.7) Number of PTB cases in the category (% of given category) 93(30.2) 23(22.3) 4(21.1) 22(27.8) 74(12.0) Sensitivity (%) [95% CI ] 43.1[36.4–49.9] 10.6[6.9–15.5] 1.9[0.5–4.7] 10.2[6.5–15.0] 34.3[28.0–41.0] Specificity (%) [95% CI ] 76.3[73.4–79.1] 91.2[89.2–93.0] 98.3[97.3–99.1] 93.7[91.9–95.2] 40.4[37.2–43.7] Median days from CXR report to PTB notification (IQR) 9(6–15) 29(11–45) 51.5(29–63) 10(5.5 – 24.5) 46.5(15–78) Conclusion Accurate reporting and inclusion of a TB code on CXR reports is a key driver of accelerated PTB diagnosis for RATB pathways. However, the code was applied to only a subset of cases, suggesting TB-related CXR changes lack specificity. These data provide a useful reference for evaluating AI-driven approaches to CXR reporting and coding for TB.Reference https://www.gov.uk/government/publications/tuberculosis-in-england-2023-report-data-up-to-end-of-2022/tb-treatment-and-outcomes-england-2022