Characteristics of and Deaths among 333 Persons with Tuberculosis and COVID-19 in Cross-Sectional Sample from 25 Jurisdictions, United States

Little is known about co-occurring tuberculosis (TB) and COVID-19 in low TB incidence settings. We obtained a cross-section of 333 persons in the United States co-diagnosed with TB and COVID-19 within 180 days and compared them to 4,433 persons with TB only in 2020 and 18,898 persons with TB during 2017‒2019. Across both comparison groups, a higher proportion of persons with TB–COVID-19 were Hispanic, were long-term care facility residents, and had diabetes. When adjusted for age, underlying conditions, and TB severity, COVID-19 co-infection was not statistically associated with death compared with TB infection only in 2020 (adjusted prevalence ratio 1.0 [95% CI 0.8‒1.4]). Among TB–COVID-19 patients, death was associated with a shorter interval between TB and COVID-19 diagnoses, older age, and being immunocompromised (non-HIV). TB–COVID-19 deaths in the United States appear to be concentrated in subgroups sharing characteristics known to increase risk for death from either disease alone.

The COVID-19 pandemic also affected TB epidemiology and program management across epidemiologic contexts (11).In the United States, reported TB incidence declined ≈20% in 2020 compared with 2019 (12).Limited information suggests that some persons with TB in the United States may have had more clinically severe disease in 2020 than before the COVID-19 pandemic (13), and TB diagnoses may have been delayed (14).We aimed to describe demographic, social, and clinical characteristics of persons with TB and COVID-19 in the United States, including risk for death, and to identify populations who may benefit most from integrated interventions.

Design and Population
We established a voluntary collaboration of US health jurisdictions to obtain a cross-sectional sample of persons with TB and COVID-19 diagnosed within 180 days (hereafter TB-COVID-19).We used the population-based National Tuberculosis Surveillance System (NTSS) for cases reported during 2017-2021 for standardized demographic, social, underlying conditions, and TB-specific diagnosis and treatment variables (15).Each jurisdiction captured a subset of the standardized data elements from the National Notifiable Disease Surveillance System for COVID-19 cases (16) and contributed them to this project.The core set of COVID-19 and TB surveillance data elements were consistent across jurisdictions.We included all persons with COVID-19 meeting the public health surveillance case definition for confirmed or probable COVID-19 (17) reported during January 1, 2020-June 30, 2021, who were also persons with TB reported in 2020 (i.e., persons with TB-COVID- 19).Although the methods used by participating jurisdictions to identify persons with TB-COVID-19 varied (Appendix Table 1, https:// wwwnc.cdc.gov/EID/article/29/10/23-0286-App1.pdf), each jurisdiction systematically identified their residents with co-diagnoses of TB and COVID-19 using personal identifiers.Of 26 participating jurisdictions, 11 (42.3%)used a software algorithm that included name and date of birth to match persons (several also included various combinations of sex, race/ethnicity, and place of residence), 8 (30.8%) had integrated surveillance systems (i.e., a given individual's TB and COVID-19 diagnoses were already linked to a single record), and 1 (3.8%) with an integrated surveillance system also performed a name-based software match (Appendix Table 1).Directly identifiable information in the surveillance registries was retained by participating jurisdictions and not shared with investigators in other jurisdictions or with CDC.
Each jurisdiction securely transmitted data on persons with TB-COVID-19 to CDC.We then excluded persons with TB-COVID-19 with unknown TB or COVID-19 diagnosis dates and for which diagnoses occurred >180 days apart, regardless of which disease was diagnosed first (Figure 1).For analysis of TB treatment outcomes, we excluded jurisdictions with incomplete TB case outcome data.To identify characteristics of persons with TB-COVID-19 that differed from persons with TB only, we compared them to characteristics of persons with TB reported in 2020 without COVID-19 (i.e., 2020 TB-only) and TB reported during the 3 most recent pre-COVID-19 years, 2017-2019 (i.e., 2017-2019 TB-only).

Data Elements
NTSS data included demographic, social, and clinical characteristics, and TB diagnosis and treatment outcomes.We used a composite all-cause death outcome that included TB diagnosed after death, death occurring before or during TB treatment, and death recorded on the COVID-19 case report form.We defined the TB diagnosis date as the earliest among positive smear or tissue collection, positive nucleic acid amplification test result, first culture specimen collected for phenotypic drug-sensitivity testing, or TB treatment start date.For the COVID-19 diagnosis date, we used the date of specimen collection of the first positive nucleic acid amplification test or antigen test.We defined persons with disseminated TB as having meningeal or miliary disease, both pulmonary and extrapulmonary disease, or having a positive culture for Mycobacterium tuberculosis complex from blood.

Analytic Methods
We compared characteristics of persons with TB-COVID-19 with those of persons with 2020 TB-only and 2017-2019 TB-only, calculating statistically significant differences of bivariate frequencies by using the Mantel-Haenszel χ 2 test (or Fisher exact test for small cell counts) with Bonferroni correction for multiple comparisons.We also calculated Clopper-Pearson binomial 95% CIs for some frequencies.For continuous variables, we assessed differences in parametric means by using t-tests.We used the Wilcoxon rank-sum test to compare nonparametric continuous variables.We calculated prevalence ratios (PRs) and 95% CIs by using log-binomial RESEARCH multivariable regression employing backward selection in logistic regression models to identify statistically significant (α = 0.05) variables for inclusion in the log-binomial models.The final models included all variables reaching statistical significance and the COVID-19 co-diagnosis status as the exposure of interest.We did not assess interaction terms in multivariable models.Rather than exclude persons with missing covariate data, we classified missing values as unknown and retained them in the models.We stratified outcomes by the proximity in timing of TB and COVID-19 diagnoses (i.e., within 30, 90, and 180 days) and fit independent log-binomial models to each time interval.

Ethics Considerations
This activity was determined to meet the requirements of public health surveillance as defined in 45 CFR 46.102(l) (2).Informed consent was not required because the project was classified by CDC as nonresearch.Although most participating jurisdictions relied on the CDC project determination, some independently classified the activity as nonresearch.

TB-COVID-19 Analytic Population
The 26 participating jurisdictions accounted for 62.9% of US TB cases in 2020 and 67.0% of the 2020 US population (18).The number of all TB cases reported in 2020 per participating jurisdiction ranged from 10 to 1,703: 12 jurisdictions (46.2%) reported <75 cases, 8 (30.8%) reported 75-149 cases, and 6 (23.1%) reported >150 cases.Participating jurisdictions reported ≈64% of the ≈46,353,000 COVID-19 cases reported in US states and territories reported during the observation period (1).Jurisdictions using more robust methods (i.e., a software algorithm or an integrated surveillance system) to identify persons with TB-COVID-19 (Appendix Table 1) accounted for 91.7% of the TB cases among the 26 participating jurisdictions.
One of 26 jurisdictions (North Dakota) did not identify persons with TB-COVID-19 meeting our criteria, and so we excluded numerator and denominator data for this jurisdiction from statistical analyses.We did not find statistically significant (95% CI with Bonferroni correction) bivariate differences in persons with TB-COVID-19 relative to the 2020 TB-only and 2017-2019 TB-only comparison groups for sex, residence in a correctional facility, homelessness, or excessive alcohol use (Table 1).The TB-COVID-19 group had a higher proportion of Hispanic persons compared with both of the reference groups (Table 1).Higher proportions of persons with TB-COVID-19 also were residents of long-term care facilities at TB diagnosis compared with both reference groups.

TB-COVID-19 Clinical Characteristics Compared with 2020 TB-only and 2017-2019 TB-only
We did not find statistically significant differences, compared with either the 2020 TB-only or 2017-2019 TB-only reference group, for the proportion of persons with TB-COVID-19 by the status of a previous episode of TB, HIV infection, TB disease site (i.e., pulmonary-only, extrapulmonary, or both sites), or TB disease dissemination (Table 1).Compared with both reference groups, persons with TB-COVID-19 had a higher rate of diabetes and endstage renal disease.sputum smear positivity for acid-fast bacilli (aPR 1.4 [95% CI 1.1-1.8]).We observed similar associations when persons had TB and COVID-19 diagnosed within shorter intervals (i.e., 30 and 90 days) (Appendix Table 2).

Risk Factors for Death among TB-COVID-19 Patients
Among the 288 persons with  The baseline mortality rate for persons with TB was ≈9% annually in the United States in 2017 and 2018 (18).In our study, ≈17% of persons diagnosed with TB and COVID-19 within 180 days died.Nonetheless, in multivariable analysis corrected for age and underlying conditions, COVID-19 was not an independent predictor of death among persons with TB diagnosed within 180 days.Those findings suggest that poor outcomes for persons with TB-COVID-19 may be driven by the overlapping sociodemographic and medical risk factors common to each TB and COVID-19 that already place persons with TB at risk for death with TB, rather than the effect of COVID-19 coinfection alone.Compared with countries that have high TB prevalence, TB disease in the United States and other low TB incidence countries is more concentrated in older persons who have underlying conditions such as diabetes and renal disease (18).The timing of TB and COVID-19 co-diagnoses and its association with TB mortality warrants more investigation, given that our model demonstrated an association between a smaller diagnostic interval (90 days) and death among persons with TB-COVID-19.In addition to biological mechanisms to explain the association, persons with more severe COVID-19 may have been more likely to receive chest imaging and additional diagnostic testing to reveal concurrent TB.
Delayed TB diagnoses could have led to more severe TB disease at clinical evaluation in our analysis population.Another study from a low TB incidence setting showed a higher proportion of positive microscopic examinations during the COVID-19 pandemic compared with historical trends (20), similar to the observation in this US cohort of persons with TB-COVID-19.This finding suggests longer delays until TB diagnosis during the COVID-19 pandemic.The timing of TB diagnosis after COVID-19 (a substantial proportion had TB diagnosed within 14 days after COVID-19) could also reflect delayed TB diagnoses, suggesting that COVID-19 could have brought persons with undiagnosed TB into care.
TB program participation was nonrandom, which limits the representativeness of results to the entire United States, perhaps especially related to race and ethnicity.Nonetheless, the cohort represented a crosssection of US jurisdictions with varying TB prevalence.An important distinction in comparison with other studies is that we were unable to compare outcomes for persons with TB-COVID-19 with those for persons with COVID-19 only.Other limitations are that the completeness of COVID-19 case reporting may have differed by jurisdiction and phase of the epidemic.Missing data may have lessened the accuracy of some descriptive characteristics; missing death dates precluded hazards analyses of time to death.Longitudinal case management for persons on TB treatment probably captured most, but potentially not all, deaths among persons with TB.Our definition for disseminated TB is intended to capture most cases resulting from hematogenous spread that might be associated with delayed diagnoses or poor outcomes.It may not reflect all disseminated TB characterized by isolated extrapulmonary lymphadenitis or TB misclassified because of incomplete tissue sampling.The Bonferroni correction may have raised the risk for type II error in bivariate comparisons, and the small number of persons with TB-COVID-19 and having sociodemographic characteristics potentially influencing outcomes (e.g., experiencing homelessness) limited our ability to describe them.Strengths include the high completeness of sociodemographic data available in NTSS (15).Still, some underlying conditions strongly associated with poor COVID-19 outcomes (e.g., cardiovascular disease and obesity) were not available.
In conclusion, this analysis of a US cohort of persons with TB-COVID-19 suggests deaths among persons with TB-COVID-19 in the United States is concentrated in subgroups having shared characteristics known to increase risk for death with either disease alone.Timely consideration for TB disease among persons with COVID-19 and TB risk factors should be reinforced.Because death was associated with shorter intervals between co-diagnoses, prioritizing additional early medical interventions for persons with concurrent disease processes who are at highest risk for death might improve outcomes.COVID-19 patients with severe disease may be given immunomodulating treatments that could reactivate latent TB infection.Therefore, COVID-19 patients with risk factors for TB infection could be considered for screening and treatment of latent TB infection.Last, integration of screening for TB infection (risk factor review and serum interferon gamma release assays testing) with community COVID-19 prevention efforts among subpopulations with shared risk profiles, as has been done for persons at increased risk for COVID-19 and diabetes (21), may expand high-yield opportunities to prevent TB.

Figure 3 .
Figure 3. Frequency of 333 persons with TB and COVID-19 co-diagnosed (TB-COVID-19), by sequence and days between TB and COVID-19 diagnoses, 25 US jurisdictions, 2020.The percentage denominator accounts for all 333 persons.Individual percentages may not sum to 100% because of rounding.TB, tuberculosis.