Comparative modeling of tuberculosis epidemiology and TB control efforts in California

The United States reported its lowest tuberculosis (TB) incidence ― or the number of new TB cases ― in 2018: 2.8 cases per 100,000 persons.¹ However, the decline in TB incidence has slowed in recent years^1,2 and the current pace of this decline is estimated³ to be too slow to reach the national goal of TB elimination (defined as an annual incidence of less than one case per 1 million population)^4,5 within this century. In recent years, TB causes approximately 500 deaths annually in the United States ⁶ and results in over U.S. $500 million in TB disease treatment and societal costs of premature death,⁶ necessitating efforts to explore innovative approaches for TB prevention and control. Recent modeling efforts⁴ suggested that substantial progress can be made toward reducing the U.S. TB caseload and the associated mortality and costs by scaling up existing programmatic efforts.

The set of recommendations for how to further strengthen TB control efforts could be informed by comparing three mathematical models of TB transmission and epidemiology and by identifying consensus evidence to support policies and strategies to accelerate existing programmatic interventions to enhance TB prevention. The models were developed separately by academic institutions at Harvard University, Johns Hopkins University, and the University of California at San Francisco, in partnership with CDC scientists. All three models used the same data inputs from California,⁴ which accounted for 23% of U.S. TB cases in 2018.¹

The models included in this comparison varied in terms of their key structural features, such as:

• modelling approach: deterministic models (the same inputs always produce the same outputs) versus stochastic models (the same inputs produce a range of outputs due to randomness)
• age structure: all ages versus age 15 years and older; stratified by single year of age or stratified into several age bands
• stratification of non–U.S.-born residents: years in the United States and/or country of origin
• risk factors for TB exposure and progression: list of included risk factors; number of risk factors
• other population stratifications: prior LTBI/TB treatment alone or in combination with TB drug resistance

Each of the models projected TB incidence, TB deaths, and incident (i.e., newly occurring) and prevalent (i.e., existing) latent TB infection (LTBI) from 2020 to 2050 under several standardized scenarios. The base-case scenario estimated the effect of continued current population coverage and treatment services for TB and LTBI, including routine contact investigation of persons exposed to infectious TB patients. Under the base-case, each model corroborated findings of an earlier analysis using national data³ and confirmed that without major scale-up of TB interventions or changes in major TB determinants, TB incidence and deaths will continue to decline slowly and remain well above the elimination threshold through 2050.

The remaining two scenarios modeled the effect of:

1. Ensuring no TB or LTBI among new California residents from 2018 onwards
2. Targeted one-time testing and treatment for LTBI for 25% of the non–U.S.-born population, conducted in 2018.

Despite their differences, all models agreed on the scenarios having potential to further TB elimination. Eliminating TB and LTBI among new California residents (scenario 1) consistently had the highest predicted impact, leading to an estimated reduction between 21.9% and 38.1% in cumulative TB cases and highlighting the importance of strengthening and expanding testing and treatment of LTBI programs. For the scenario that scaled up one-time LTBI testing and treatment for 25% of all non–U.S.-born residents of California (scenario 2), models predicted a relatively stable 4.6%―13.4% reduction in TB cases, demonstrating that even short-term interventions to identify and treat LTBI might generate ongoing benefits for reducing TB cases and deaths. Across examined scenarios, model agreement was strongest when there were rigorous empirical data available to parameterize and calibrate models (e.g., TB case notifications and deaths), highlighting the importance of the high-quality, up-to-date data for mathematical modelling.

These findings highlight that strategies to achieve TB elimination goals will need to incentivize providers to test and treat LTBI among non-U.S.-born populations and to provide greater access to LTBI diagnosis and treatment for non-U.S.-born individuals, such as access and linkage to healthcare services upon arrival in the United States.^5,7 Although the models used in this analysis were based on TB epidemiology in California, the conclusions about programmatic interventions with the greatest potential impact were similar to those previously analyzed on a national level,² and may be applicable to other parts of the United States. Similar modeling at the jurisdiction level can account for the area-specific nuances of TB epidemiology and dynamics critical for adjusting and prioritizing interventions and estimating local resources needed to achieve TB elimination goals.²