TB Risk-Prediction Model Using Health Administrative Data

A study set out to develop and validate a pre-tuberculosis (TB) screening risk-prediction tool using demographic and clinical data derived from health administrative records in two Canadian provinces: British Columbia and Ontario. This retrospective cohort study aimed to support earlier identification of individuals at high risk for developing TB, particularly within migrant populations.

Researchers analyzed data from 715,423 individuals in British Columbia, among whom 1,407 developed TB, and 958,131 individuals in Ontario, where 1,361 developed the disease. Using this large-scale, linked administrative dataset, the model demonstrated good discrimination ability, with concordance values around 0.75, indicating it could reliably differentiate between individuals at high and low risk of TB. The model also showed strong internal validity, evidenced by low optimism scores ranging from 0.005 to 0.012. Calibration-in-the-large values were close to zero, suggesting minimal overestimation of overall risk, though calibration slopes under 1 indicated a degree of overfitting.

Key predictors of TB risk within a two-year period included several strong risk factors: individuals aged 75 years or older (hazard ratio [HR] 15.80), originating from countries with TB incidence rates of 300 or more per 100,000 population (HR 21.34), living with HIV (HR 16.45), undergoing dialysis for chronic kidney disease (HR 16.09), and having had close (HR 6.57) or non-close (HR 3.99) contact with a TB case.

External validation, referred to as Model F, performed best when prescription variables were excluded and a 5-year landmark period was applied. However, the model tended to underestimate risk over a 2-year window and overestimate it over 5 years, as reflected in the expected/observed risk ratios. Subgroup analyses revealed a tendency to overestimate TB risk among older adults, refugees, individuals living with HIV, and patients on dialysis.

In conclusion, the study successfully created and externally validated a TB risk prediction model using existing health administrative data. The model shows significant potential for identifying individuals at elevated risk for TB, especially in migrant populations. Nonetheless, further calibration and cost-effectiveness assessments are necessary before widespread implementation in clinical or public health settings.

Source: Puyat, J.H., Brode, S.K., Shulha, H., Romanowski, K., Menzies, D., Benedetti, A., Duchen, R., Huang, A., Fang, J., Macdonald, L. and Marras, T.K., 2025. Predicting Risk of Tuberculosis (TB) Disease in People Who Migrate to a Low-TB Incidence Country: Development and Validation of a Multivariable, Dynamic Risk-Prediction Model Using Health Administrative Data. Clinical Infectious Diseases, 80(3), pp.644-652.