Dados do Trabalho

Título

Accuracy of an ultra-portable X-ray system with an automated interpretation for tuberculosis active case finding in prisons

Introdução

Despite the growing evidence suggesting that persons deprived of liberty (PDL) have markedly increased risk of tuberculosis (TB) disease, most institutional correctional facilities have limited resources to support TB screening/detection efforts. We evaluated the diagnostic performance of an ultra-portable X-ray system paired with an automated interpretation used in a TB active case finding in prisons.

Objetivo (s)

The World Health Organization recommends systematic active case finding especially among high-risk population including incarcerated individuals, who have increased risk of TB. In this study, we estimated the diagnostic accuracy of an ultra portable digital chest radiography system paired with LunitTB, a machine learning-based algorithm to detect TB disease.

Material e Métodos

We performed a prospective active case finding study for tuberculosis in two prisons located in the state of Mato Grosso do Sul, Brazil, from February 2023. Eligible individuals included adults (≥18years) without a TB history. Fujifilm FDR Xair-LunitTB system was used to screen consented individuals for TB disease irrespective of their TB symptoms. Fujifilm FDR Xair is a lightweight (~3.5kg) handheld digital radiography system. TB case was defined by either a) a positive sputum culture or b) two positive Xpert or c) one positive Xpert and one trace Xpert results. Area under ROC curve (AUC) and 95% confidence intervals (CI) were estimated to determine the accuracy of FDR Xair-LunitTB system as a quantitative TB diagnostic test. 

Resultados e Conclusão

We screened a total of 3203 individuals, and 2652 (82.8%) met our eligibility criteria. The median age of participants was 30 years (interquartile range [IQR] 26 – 37). The prevalence of TB was 3.5% (93/2652, 95%CI 2.9 – 4.3). Any TB symptoms were reported among 15.2% of participants. The AUC for FDR Xair-LunitTB system in our cohort was 0.88 (0.84 – 0.93). The accuracy of FDR Xair-LunitTB system among those with any TB symptoms was significantly higher (AUC=0.96, 95%CI 0.92 – 1.00) compared to those without TB symptoms (AUC=0.84, 95%CI 0.78 – 0.91) (DeLong p=<0.01). The FDR Xair-LunitTB system enabled us to screen PDL rapidly, with a high diagnostic accuracy especially among those reported any TB symptoms. Further studies to assess the performance of FDR Xair-LunitTB system in other settings are still warranted.

Palavras Chave

Tuberculosis; automated x-ray interpretation; active case finding