We defined the Medicare-eligible population as persons aged 65 years or older who were either U.S. citizens or eligible non–U.S.-born residents in 2022, excluding Medicare-eligible persons younger than 65 years (21, 22). Within this population, we defined a TB risk population based on 3 groups of risk factors that confer elevated risk for developing TB (Supplement Table 1, available at Annals.org). Group A included persons with 1 of 4 factors (non-U.S. birth, history of homelessness, history of incarceration, or residence in a long-term care facility) recommended by the U.S. Preventive Services Task Force (USPSTF) for LTBI testing and treatment in primary care settings (7). Group B included persons with 1 of 4 factors (recent contact with a person with TB disease, diagnosed HIV infection, use of immunosuppressive medications, or receipt of a solid organ transplant) recommended for LTBI testing and treatment for which services are typically provided outside primary care settings (8). Finally, group C included persons with 1 of 7 factors (diagnosed diabetes, end-stage renal disease, chronic kidney disease, cancer of the head or neck or blood, gastrectomy, body mass index <20 kg/m², or current smoking) that are not specifically part of the USPSTF screening recommendation but that produce elevated risks for TB disease among persons with LTBI (8, 9).

We extracted population data by age, sex, race and ethnicity, and nativity from the 2021 ACS (American Community Survey) (23) and estimated the numbers of Medicare-eligible persons aged 65 years or older in 2022. Data on non-U.S. birth and most risk factors in groups B and C were sourced from the ACS, NHIS (National Health Interview Survey), and NHANES. For risk factors not covered by these sources, we used national surveillance reports or published studies (Supplement Table 1). We applied mixed-effects generalized linear models using these data to predict the prevalence of each risk factor in the Medicare-eligible population. We summed these values to estimate the population with at least 1 risk factor, after adjusting for individuals with multiple risk factors (Supplement, available at Annals.org). We extracted data on interferon-gamma release assay (IGRA) test results for TB infection from the 2011-to-2012 NHANES cycle (24, 25) and fitted logistic regression models to estimate IGRA positivity for each analytic stratum. We estimated LTBI prevalence by adjusting IGRA positivity based on reported estimates of IGRA sensitivity and specificity (26) using a published approach (27).

We examined 2 LTBI testing and treatment scenarios. The first assumed expanded LTBI screening via a one-time offer of screening for all Medicare beneficiaries (aged ≥65 years) with at least 1 TB risk factor in 2022. The second scenario assumed a one-time offer of LTBI screening for persons aged 65 years who were newly eligible for Medicare in 2022 and had at least 1 TB risk factor. Under this second scenario, the intervention was assumed to be offered as part of the “Welcome to Medicare” visit for new Medicare enrollees. Under both scenarios, we assumed that LTBI testing would be via IGRA or tuberculin skin test. We assumed that 100% of the non–U.S.-born and 50% of the U.S.-born Medicare population would be tested using IGRA. Those with positive results would receive a chest radiograph to rule out TB disease before being offered treatment to prevent TB disease (either 3 months of self-administered weekly rifapentine and isoniazid or 4 months of self-administered daily rifampin) (28, 29).

We compared intervention scenarios versus a counterfactual scenario with no LTBI screening. For all scenarios, persons developing TB disease in the future were assumed to receive prompt diagnosis and treatment consistent with national performance indicators (Supplement Table 2, available at Annals.org) (30).

We constructed a Markov cohort model representing TB natural history and receipt of related health services (Supplement Figure 1, available at Annals. org), parameterized using published evidence (Supplement Table 2). For individuals in the TB risk population, the proportion initiating LTBI treatment was calculated as the product of 1) the fraction without a history of LTBI treatment (88.1% for those with LTBI and 99.4% for those without) (31), 2) the fraction that would accept testing if offered (76.2%) (32), 3) the fraction testing positive and diagnosed with LTBI (varied by risk factor), and 4) the fraction that would accept treatment if offered (76.2%) (30). We assumed that a fraction of treated individuals would discontinue the regimen before completion and a fraction would have treatment hepatotoxicity (0.3%) (28) with potential associated death (33, 34). For those with LTBI completing LTBI treatment, we assumed that most (93.0%) (28, 35) would clear LTBI and have no future risk for reactivation of TB. Individuals not completing the regimen were assumed to receive no benefit. Those incorrectly diagnosed with LTBI (that is, false-positive diagnoses) could receive LTBI treatment (Supplement Figure 1) and potentially have treatment toxicity but no other health effects. Individuals with LTBI not receiving LTBI treatment were assumed to have similar outcomes to those in the “no screening” counterfactual.

The model simulated future TB-related outcomes, including progression to TB disease, TB diagnosis and treatment initiation, and recovery through treatment. Individuals could die of TB disease, or deaths could be related to excess mortality from modeled risk factors or background mortality. Individuals surviving TB disease were assumed to face no further TB-related health consequences. Background mortality rates specific to age, sex, and race were extracted from U.S. life tables (36). Excess mortality rates associated with TB disease and coprevalent conditions were modeled by applying condition-specific mortality rate ratios (Supplement Table 2). We did not model future exposure to infection because the consequence of these infections would be unaffected by the screening intervention. We did not consider the consequences of secondary transmission from future TB cases in the main analysis but considered this in sensitivity analysis.

We estimated costs from a TB service perspective, a health care sector perspective, and a societal perspective, with the societal perspective used in the main analysis (Supplement Table 3, available at Annals.org) (37). Costs (Supplement Table 2) were inflated to 2021 values using relevant price indices (38) and reported in 2021 U.S. dollars. To calculate cost-effectiveness results, we applied a 3% annual discount rate to future costs and benefits.

We compared results of intervention versus no intervention scenarios to calculate the numbers of TB cases averted, TB deaths averted, and QALYs gained. We aggregated results across demographic strata and reported outcomes as absolute values as well as per 100 000 persons in the TB risk population. We estimated incremental costs and ICERs from each perspective, for each risk factor, and for each risk factor group, and we present these ICERs rounded to the nearest thousand.

We specified probability distributions representing the uncertainty associated with each model parameter (Supplement Table 2) and estimated outcomes using a second-order Monte Carlo simulation (39). To do so, we sampled 1000 parameter sets using a Latin hypercube sample (40) and simulated results for each of these parameter sets. We calculated point estimates for each outcome as the mean of these 1000 values and calculated 95% uncertainty intervals as the 2.5th and 97.5th percentiles. We calculated partial-rank correlation coefficients (PRCCs) to quantify the influence of individual parameters on study outcomes (41) and cost-effectiveness acceptability curves to report uncertainty in results. We did several sensitivity analyses and reestimated results under the following scenarios: 1) assuming lower (1.7%) (42) and 2) higher (7.0%) discount rates, 3) using an alternative source for the base rate of reactivation of TB for persons with LTBI (0.084 [43] vs. 0.051 in the main analysis [27]), 4) assuming 100% acceptance and completion of LTBI testing and treatment, 5) using LTBI test characteristics from a published USPSTF review (7), and 6) considering the consequences of averted secondary TB cases (parameter values shown in Supplement Table 2). Analyses were done using R, version 4.2.1 (R Foundation) (44).

Employees of the Centers for Disease Control and Prevention participated as study investigators.

Compared with no testing or treatment, a one-time intervention of LTBI testing and treatment provided to the Medicare-eligible population aged 65 years or older was estimated to result in 2.6 million persons initiating LTBI treatment, preventing 10 946 future TB cases and 2579 TB deaths, with 13 234 lifetime QALYs gained (Figure 1; Supplement Table 5, available at Annals.org). Among persons with USPSTF-recommended risk factors (group A), an estimated 7975 TB cases and 1870 TB deaths would be averted, with 10235 QALYs gained (Figure 1; Supplement Table 5). The health benefits estimated for the intervention varied across risk factor groups, reflecting differences in population size, LTBI prevalence, and LTBI reactivation rates (Figure 1; Supplement Table 4). For 65-year-olds newly eligible for Medicare in 2022, we estimated that the intervention would result in 0.2 million LTBI treatments initiated, 1168 TB cases and 237 TB deaths averted, and 1632 QALYs gained compared with no testing and treatment (Figure 2; Supplement Table 5). Most of these benefits were estimated to accrue among persons with USPSTF-recommended risk factors (group A) (Figure 2).

Compared with no intervention, LTBI testing and treatment of the Medicare-eligible population aged 65 years or older was estimated to have incremental costs of $1474 million from a TB service perspective ($2040 million of intervention vs. $566 million of no intervention), $1715 million from a health care perspective, and $1802 million from a societal perspective, which includes all future health care and non–health care expenditures as well as productivity losses from early death (Supplement Table 6, available at Annals.org). Most incremental costs were estimated for persons with USPSTF-recommended risk factors (group A), with estimated societal incremental costs of $1126 million (Figure 1). For 65-year-olds newly eligible for Medicare in 2022, we estimated that the incremental costs of the intervention would be $91 million ($153 million of intervention vs. $62 million of no intervention) from a TB service perspective, $120 million from a health care perspective, and $119 million from a societal perspective (Figure 2; Supplement Table 6).

Compared with no LTBI testing and treatment, the intervention for the Medicare-eligible population aged 65 years or older was estimated to have an ICER of $192 000 per QALY gained from a societal perspective. For persons with USPSTF-recommended (group A) risk factors, the ICER was estimated as $155 000 per QALY gained. For those with group B and group C risk factors, the ICERs were estimated as $220 000 and $324 000 per QALY gained, respectively (Table 2). The magnitude of estimated intervention health benefits, costs, and ICERs varied widely across the 15 risk factor populations. The greatest health benefits were estimated for the non–U.S.-born, Medicare-eligible population aged 65 years or older, which had an ICER of $116 000 per QALY gained (Figure 3 and Table 2).

For 65-year-olds newly eligible for Medicare in 2022, the ICER for the intervention (compared with no LTBI testing and treatment) was estimated as $109 000 per QALY gained in the overall TB risk population. The ICER was $89 000 per QALY gained for the subset of this population with USPSTF-recommended risk factors (group A) (Table 3). The greatest health benefits were estimated for non–U.S.-born individuals newly eligible for Medicare (aged 65 years), who had an ICER of $72 000 per QALY gained (Figure 3).

Supplement Table 7 (available at Annals.org) reports PRCCs identifying parameters with the greatest influence on cost-effectiveness results. In this analysis, the base rate of reactivation of TB for individuals with LTBI in the overall population (PRCC, −0.94) and the cost of IGRA testing (PRCC, 0.62) were found to have the strongest relationship with the ICER for the intervention provided to Medicare-eligible persons aged 65 years or older with at least 1 TB risk factor.

Supplement Figure 2 (available at Annals.org) shows cost-effectiveness acceptability curves for the intervention offered to the overall TB risk population aged 65 years or older and for risk factor groups A, B, and C. In these results, there was a 23.6% probability that the ICER would fall below $150 000 per QALY for the intervention provided to persons with at least 1 TB risk factor. This probability was 42.9% for those with group A risk factors. Supplement Figure 3 (available at Annals.org) shows cost-effectiveness acceptability curves for the intervention offered to 65-year-olds newly eligible for Medicare.

Among Medicare-eligible persons aged 65 years or older with at least 1 TB risk factor, the estimated ICER was $166 000 per QALY gained at a discount rate of 1.7% and $291 000 per QALY gained at a discount rate of 7% (Supplement Tables 8 and 9, available at Annals.org). When we reestimated results with a higher base LTBI reactivation rate (0.084 [43] vs. 0.051 [27] per 100 person-years), estimated health benefits were approximately 64% higher and cost-effectiveness ratios more favorable (39% lower) (Supplement Table 10, available at Annals.org). When we assumed 100% acceptance and completion of LTBI testing and treatment, estimated health benefits increased by approximately 87% and ICERs decreased by 24% (Supplement Table 11, available at Annals.org). Similarly, using an alternative source of LTBI test characteristics (a recent USPSTF review [45]) and considering the consequences of averted secondary TB cases both produced greater health benefits (QALYs gained, 5% and 12% higher, respectively) and more favorable cost-effectiveness ratios (incremental cost per QALY gained, 14% and 16% lower, respectively) (Supplement Tables 12 and 13, available at Annals.org).