Prediction of latent tuberculosis infection in Venezuelan immigrants: construction and validation of a surveillance model.

Abstract

Background: Latent tuberculosis infection (LTBI) is a significant concern among migrant populations, particularly Venezuelans, due to its adverse health and social conditions. This study aimed to construct and validate a predictive model of LTBI among Venezuelan migrants.

Methods: This cross-sectional study utilized data from the project "TB and migrants in BRICS countries: The case of Brazil", carried out in Boa Vista, Roraima, in 2020. The final sample included 427 participants. For the analysis, 22 variables were selected, and simple and multiple logistic regression analyses were applied. General measures (Nagalkerke's R2 and Brier's score), discriminative capacity (accuracy, receiver operating characteristic curve, and area under the curve [AUC]), and calibration measures (Hosmer-Lemeshow test and calibration graph) were used to evaluate the model. The model was internally validated using bootstrapping. Finally, a nomogram and a clinical decision curve were constructed.

Results: Six LTBI predictors (marital status, social benefit, documentation status, smoking status, presence of comorbidities, and fever) were included in the final model. The predictive model demonstrated moderate discriminatory capacity (AUC: 0.676), good calibration, and was also validated with an AUC of 0.678. Additionally, a clinical decision analysis revealed that the use of the model offers superior benefits compared with traditional treatment strategies.

Conclusions: The predictive model and nomogram proved to be useful tools for LTBI screening in migrants, potentially guiding border health surveillance actions in this population.