A predictive risk estimation model developed through data linkage identified more cases of screen-detectable pulmonary tuberculosis while requiring fewer visa applicants to undergo diagnostic evaluation compared to standard immigration screening procedures in Australia, according to a population-based cohort study. Researchers evaluated records spanning more than 13 million individuals who applied for Australian visas between 2014 and 2019, linking administrative immigration files directly with nationwide tuberculosis notification registries and immigration medical examination results. By quantifying individual tuberculosis risk from routinely captured demographic and epidemiological characteristics, the objective data-driven system demonstrated superior sensitivity and efficiency over traditional screening rules that rely largely on broad country-level incidence thresholds and intended length of stay. The study highlights the feasibility of utilizing localized administrative data to establish transparent, adjustable screening policies in low-incidence settings.
Australia maintains a low domestic tuberculosis incidence of approximately six cases per 100,000 population, a rate that has remained largely stable since the 1980s despite limited ongoing local transmission within the country. However, national progress toward complete tuberculosis elimination remains constrained because approximately 90 percent of all tuberculosis diagnoses recorded in Australia each year occur among overseas-born individuals, reflecting the nation’s reliance on migration from regions with substantially higher disease incidence. Under the Migration Act of 1958, the Australian Department of Home Affairs is legally required to ensure that incoming migrants are free from active tuberculosis disease before granting long-term entry. Historically, tuberculosis screening policies across many low-incidence host nations have relied heavily on published international reports, expert consensus, and subjective risk perceptions, which can lead to inefficient allocation of diagnostic resources and variable health protection outcomes.
Under the conventional Australian screening framework, the immigration medical examination is a statutory requirement applied primarily to foreign nationals who declare an intention to stay in Australia for more than six months and who originate from countries where tuberculosis incidence is estimated at or above 40 cases per 100,000 population. Applicants intending to work in sensitive environments, such as healthcare, aged care, or disability services, are also mandated to undergo medical assessments, which encompass clinical symptom questionnaires, physical examinations, chest radiography, and laboratory sputum testing. Any applicant confirmed to have active tuberculosis disease must complete a full course of directly supervised antimycobacterial therapy and demonstrate negative sputum cultures before a visa can be approved. Despite these safeguards, the existing policy typically excludes short-term visitors from mandatory pre-arrival health assessments, regardless of their personal country of origin or underlying clinical vulnerability.
To build an empirical alternative to this framework, investigators extracted 18,175,532 electronic visa application records from the Department of Home Affairs covering the period between January 1, 2014, and December 31, 2019. The analysis excluded individuals who never entered Australia or whose physical arrival occurred more than 180 days after visa issuance, yielding a final analytical cohort of 13,691,772 visa applicants. The cohort comprised 51.6 percent female applicants (7,059,465 individuals) and 48.4 percent male applicants (6,631,429 individuals), with 46.2 percent aged 15 to 34 years and 25.1 percent aged 35 to 54 years. Citizens of countries with World Health Organization-estimated tuberculosis incidence rates of 40 or more cases per 100,000 population accounted for approximately 50 percent of the total applicant population. Visitor visas represented the largest single administrative visa category, accounting for 66.9 percent of the total cohort, followed by international students at 10.9 percent and temporary residents at 7.8 percent.
Probabilistic data linkage was conducted by the Australian Institute of Health and Welfare as an authorized third party, connecting the visa applicant database to both the collated Australian national tuberculosis register and the Department of Home Affairs immigration medical examination database using personal identifying fields including full name, date of birth, sex assigned at birth, and country of birth. Following linkage, all personal identifiers were removed to provide fully anonymized records. The study defined screen-detectable pulmonary tuberculosis as cases diagnosed during pre-departure immigration medical examinations or identified within state and territory tuberculosis registers within six months of visa issuance. In total, 1,842 tuberculosis diagnoses were documented across the linked databases over the study period, of which 863 cases met the specific definition of screen-detectable pulmonary tuberculosis, with 851 of those cases (98.6 percent) being confirmed bacteriologically through positive sputum smear microscopy, mycobacterial culture, or automated nucleic acid amplification testing.
Statistical modeling using multivariate logistic regression assessed associations between applicant baseline characteristics and the odds of being diagnosed with screen-detectable pulmonary tuberculosis. The strongest independent predictor of disease was the estimated tuberculosis incidence in the applicant’s country of citizenship. Compared with citizens of countries where baseline incidence was 10 or fewer cases per 100,000 population, applicants from nations with incidence rates of 200 or more per 100,000 exhibited an adjusted odds ratio of 44.12 for all pulmonary tuberculosis and 45.39 for bacteriologically confirmed disease. By contrast, age-stratified analyses revealed that individuals aged 15 to 34 years experienced the highest absolute risk outside of elderly applicants aged 75 years and older, who had wider confidence intervals. Across visa classifications, temporary resident unskilled visa applicants showed the highest adjusted odds of screen-detectable disease relative to standard visitor applicants, recording an odds ratio of 6.21 for all pulmonary tuberculosis.
The mathematical prediction model utilized regression coefficients and the model covariance matrix to derive logit values and estimated probabilities of pulmonary tuberculosis detection within a six-month window, adjusting standard error estimates for overdispersion using the ratio of model deviance to degrees of freedom. The full cohort was divided chronologically into a model development set of 10,551,971 applicants whose visas were granted between 2014 and 2018, and an independent validation cohort of 3,139,801 applicants granted visas throughout 2019. The area under the receiver operating characteristic curve reached 0.89 in the development cohort and 0.91 in the prospective 2019 validation cohort, demonstrating robust discrimination. The derived model was subsequently formatted as an accessible spreadsheet tool that accepts an applicant’s date of birth, sex, visa stream, and country of citizenship to generate an individualized six-month tuberculosis risk estimate per 100,000 individuals alongside 95 percent confidence intervals.
When applied to the 2019 validation cohort, which contained 134 individuals with verified screen-detectable pulmonary tuberculosis (equating to a cohort rate of 4.3 per 100,000 persons), the predictive tool exhibited marked performance improvements over the historical standard. Under the existing immigration screening protocol, Australian authorities screened 1,035,589 applicants, representing 33 percent of all applicants in 2019, successfully identifying 133 out of an estimated 248 total screen-detectable cases, an effective sensitivity of 54 percent that left 115 cases undetected. In contrast, applying the data-driven model at a conservative decision threshold of 7 cases per 100,000 population required screening 575,006 applicants, representing just 19 percent of the applicant pool, while detecting 87 percent of all screen-detectable pulmonary cases and missing only 18 cases. At a higher decision threshold of 17 per 100,000, the tool screened 298,026 individuals (10 percent of applicants) and detected 59 percent of screen-detectable cases, missing 55 cases while halving the screening volume required under the conventional policy.
The study highlighted key operational advantages in evaluating short-term travelers who are routinely exempted from pre-arrival health assessments under present immigration policy. The authors observed that airborne infectious pathogens operate independently of administrative visa classifications, noting that individuals arriving on short-term visas carry tuberculosis risks determined by underlying demographic and epidemiological factors rather than intended duration of stay. Furthermore, young adult migrants aged 15 to 34 years accounted for the majority of identified cases in the study cohort. Although clinical comorbidities are generally uncommon in this demographic, young migrants frequently originate from high-incidence settings to pursue higher education or temporary employment, facing substantial resettlement stressors, academic obligations, and financial pressures that may influence disease manifestation and healthcare-seeking behaviors.
The authors acknowledged several methodologic constraints, including the relatively small set of demographic variables available within administrative visa datasets. Factors such as documented household crowding, prior tuberculosis treatment history, specific occupation, and known exposure to infectious contacts could not be integrated into the predictive algorithm. Additionally, probabilistic matching without unique national health identifiers introduced a minor potential for matching error among individuals sharing identical names and birthdates, though this occurred infrequently. Classifying all tuberculosis diagnoses recorded within six months of visa issuance as screen-detectable may also have captured some cases that developed post-arrival, although the authors noted that this window conservatively reflects diagnostic delays inherent to passive surveillance and the recognized existence of subclinical disease. Because global migration dynamics and regional disease incidence rates change over time, the researchers emphasized that the risk estimation tool must undergo regular recalibration every two to five years to sustain predictive accuracy.
Subject of Research: Data-driven predictive risk modeling for migrant pulmonary tuberculosis screening in Australia
Article Title: A data-driven approach to tuberculosis screening of migrants entering Australia
Article References: Scandurra, G. M., Ampon, M. R., Martin, B., Marais, B. J., & Marks, G. B. (2026). A data-driven approach to tuberculosis screening of migrants entering Australia. The Lancet Regional Health – Western Pacific, 76, Article 102017. https://doi.org/10.1016/j.lanwpc.2026.102017
Image Credits: AI Generated
DOI: 10.1016/j.lanwpc.2026.102017
Keywords: tuberculosis, pulmonary tuberculosis, migrant screening, immigration health, Australia, predictive modeling, risk estimation, data linkage, Department of Home Affairs, public health, infectious disease surveillance, sensitivity and specificity, logistic regression, epidemiology, visa applicants, preventive medicine
News Source: Phoebe Ingram. (November 12, 2026). Data Linkage Model Enhances Screening for Tuberculosis Among Migrants Entering Australia. Scienmag.



