Predicting risk of tuberculosis disease in people migrating to a low-TB incidence country: development and validation of a multivariable dynamic risk prediction model using health administrative data

Background: Tuberculosis (TB) incidence remains disproportionately high in people migrating to Canada and other low TB incidence countries, but systematic TB screening and prevention in migrants is often cost-prohibitive for TB programs. We aimed to develop and validate a TB risk prediction model to inform TB screening decisions in foreign-born permanent residents of Canada.

Methods: We developed and validated a proportional baselines landmark supermodel for TB risk prediction using health administrative data from British Columbia and Ontario, two distinct provincial healthcare systems in Canada. Demographic (age, sex, refugee status, year of entry, TB incidence in country of origin), TB exposure, and medical (HIV, kidney disease, diabetes, solid organ transplantation, cancer) covariates were used to derive and test models in British Columbia; one model was chosen for external validation in the Ontario cohort. The model's ability to predict 2- and 5-year TB risk in the Ontario cohort was assessed using discrimination and calibration statistics.

Results: The study included 715,423 individuals (including 1,407 people with TB disease) in the British Columbia derivation cohort, and 958,131 individuals (including 1,361 people with TB disease) in the Ontario validation cohort. The 2- and 5-year concordance statistic in the validation cohort was 0.77 (95%CI: 0.75-0.78) and 0.77 (95%CI: 0.76-0.78), respectively. Calibration-in-the-large values were 0.14 (95% CI: 0.08-0.21) and -0.05 (95% CI: -0.12-0.02) in 2- and 5-year prediction windows.

Conclusions: This prediction model, available online at https://tb-migrate.com, may improve TB risk stratification in people migrating to low incidence countries and may help inform TB screening policy and guidelines.