Digital Chest X-Ray with Computer-aided Detection for Tuberculosis Screening within Correctional Facilities.

Rationale: Realizing the Global Plan to End Tuberculosis (TB) will require reaching at least 90% of people in key populations, such as inmates, through optimizing case-finding approaches.

Objectives

To evaluate the value of adding digital chest X-ray (d-CXR) with computer-aided detection (CAD) to symptom-based screening on TB yield among inmates.

Methods

Consecutive adult inmates from four correctional facilities in South Africa were screened for TB using symptoms and d-CXR. Any person with at least one symptom or CAD score of ⩾50 provided two sputa for liquid culture and GeneXpert MTB/RIF Ultra (Xpert Ultra) testing. In a sample of 800 symptom-negative inmates with CAD score <50, Xpert Ultra testing was also conducted. TB yield was defined as the proportion of new patients with bacteriologically confirmed TB who were identified.

Results

We enrolled 3,576 participants: 99.6% male, median age of 34 years (interquartile range, 28-41), and 584 (16.3%) with positive test results for human immunodeficiency virus. Of those screened, 867 (24.2%) participants required investigation (394 [11.2%] symptomatic, 685 [19.1%] with abnormal CAD results, and 867 [24.2%] with either). Sputum was taken in 747 (86.2%) participants, with 28 (7.8 per 1,000 population) new TB cases diagnosed. On the basis of hypothesized screening modalities, yield would have been 3.6 per 1,000 population on the basis of symptoms alone and 7.0 per 1,000 population on the basis of d-CXR alone. Among an additional 800 inmates tested who initially screened symptom negative and had a CAD score <50, five TB cases were diagnosed. There was no difference in TB yield when comparing Xpert Ultra against culture (5.6 vs. 4.8 per 1,000 population; P = 0.21).

Conclusions

The addition of d-CXR identified two times more patients with undiagnosed TB than did investigation of symptoms alone. Complementary use of d-CXR may potentially overcome the subjectivity inherent in symptom screening alone for identifying TB in this population.