Impact of COVID-19 on tuberculosis mortality in Japan: a retrospective analysis of national surveillance data from 2017 to 2022 Authors: Lisa Kawatsu, Kazuhiro Uchimura Published: 2026 ABSTRACT BackgroundThe coronavirus disease 2019 (COVID‑19) pandemic has disrupted tuberculosis (TB) services globally, but its impact on TB mortality in Japan—where the TB burden is concentrated among older adults—remains unclear. This study assessed trends in TB mortality before and during the COVID‑19 pandemic and examined whether being diagnosed during the pandemic was associated with TB mortality.MethodsWe analyzed anonymized, individual‑level cohort data for all patients notified with active TB to the national surveillance system between 1 January 2017 and 31 December 2022. TB‑specific mortality within 365 days of notification was evaluated across three periods: pre‑COVID‑19, pre‑Omicron, and Omicron. Mortality rates per 100 person‑years were calculated, and Cox proportional hazards models, allowing for a time‑varying effect of age, were used to estimate adjusted hazard ratios (aHRs) for TB death in the overall cohort and in age‑stratified analyses. Covariates included demographic, clinical, and health‑system factors recorded at baseline.ResultsTB mortality increased from 10.1 (95% confidence interval [CI] 9.7–10.4) to 12.7 per 100 person-years (95% CI 12.0-13.5) between the pre-COVID-19 and Omicron periods. This increase was driven primarily by adults aged ≥ 65 years, whose mortality rose from 15.2 (95% CI 14.7–15.8) to 18.5 per 100 person‑years (95% CI 17.4–19.7). In the adjusted models, being diagnosed during the Omicron period was associated with higher TB mortality among adults aged ≥ 65 years (aHR 1.10, 95% CI 1.03–1.18). Among those aged 0–64 years, estimates were imprecise due to the small number of TB deaths (aHR 1.09, 95% CI 0.78–1.53), but the point estimates were similar across age groups. Thus, the difference in statistical significance should not be interpreted as evidence of a true subgroup difference.ConclusionsTB mortality in Japan increased during the COVID‑19 pandemic, particularly during the Omicron period, with the greatest impact observed among older adults. Although statistical significance varied by age group, similar point estimates suggest a broadly consistent effect across ages. These findings highlight the need to strengthen TB care pathways for older adults and ensure continuity of TB services during future public health emergencies. Background The coronavirus disease 2019 (COVID-19) pandemic and the wide-ranging public health, economic, and social measures implemented in response have affected the epidemiology of many diseases, including tuberculosis (TB). According to the WHO Global TB report, the world has experienced an 18% reduction in the number of TB cases notified in 2020, compared with 2019, 15% reduction in the number of people treated for drug-resistant TB, and 21% decrease in the number of people receiving preventive therapy for latent tuberculosis infection [1]. Treatment outcomes also deteriorated, with an estimated 1.32 million TB deaths in 2020 – the first year-over-year increase in TB mortality since 2005 [1]. Several modeling studies have similarly indicated the potential impact of COVID-19 on TB mortality. For example, Cilloni et al. predicted that globally, TB mortality could return to the level reported in 2013 [2], whereas Hogan et al. estimated that TB mortality could increase by up to 20% between 2020 and 2025 [3]. Japan achieved low-burden TB status during the pandemic, with a notification rate of 9.2 per 100,000 population in 2021 [4]. Its TB epidemiology is unique compared with that of other high-income countries, where the majority of the TB burden is now borne by the young migrant population while in Japan, approximately 70% of newly reported patients are Japan-born elderly individuals aged ≥ 65 years [4]. Naturally, older adults experience markedly poorer TB treatment outcomes; among patients notified in 2020, the proportion of deaths was 2.7% among those aged 0–64 years but 38.6% among those aged ≥ 65 years [4]. COVID-19 has also disproportionately affected older adults in Japan. As of May 2023, when COVID-19 was downgraded from Class 2 to Class 5 under Japan’s Infectious Diseases Control Law, approximately 34 million cases and 74,000 deaths had been reported, with individuals aged ≥ 65 years accounting for 95.7% of COVID-19 deaths [5]. Hence, Japan’s TB epidemiology is characterized by an exceptionally high proportion of older adults, who already experience markedly elevated TB mortality. Because COVID‑19 also disproportionately affects older adults, the intersection of these two conditions may uniquely amplify mortality risk in Japan. Despite this plausible synergy, no national‑level study has examined whether the pandemic—particularly the high‑transmission Omicron period—altered TB mortality patterns. This study therefore aimed to examine trends in TB mortality by different age groups before and during the COVID-19 pandemic and explore the effects, if any, of being diagnosed during the COVID-19 pandemic on TB mortality using national TB surveillance data. Methods Study design, study population and data sources We conducted a retrospective observational study using anonymized, individual-level data from the Japan Tuberculosis Surveillance System (JTBS). Details of the system have been described elsewhere [6], however, briefly, TB is classified as a Class 2 infectious disease in Japan, requiring all physicians, regardless of practice setting, to report diagnosed cases of active TB immediately to the local health authority. Upon notification, the local health authority staff enter patient information into the JTBS. Records remain in the system for up to two years, as mandated by Japan’s Infectious Diseases Control Law, which requires follow-up of TB patients for a maximum of two years after treatment completion. Patients are deregistered earlier if they die, transfer out, or are deemed not to require further follow-up (e.g., misdiagnosis). For this study, we utilized cohort-module data for all patients notified with active TB between 1 January 2017 and 31 December 2022. Of 81,332 newly notified patients, 81,003 (99.6%) had cohort‑module data available for analysis. Small discrepancies of this magnitude occasionally occur in the JTBS due to routine operational processes and do not reflect systematic exclusion of specific patient groups. The data were accessed from March 30, 2025, for 6 months. Operational definitions The primary outcome was TB‑specific mortality within 365 days of TB notification. TB‑specific mortality was identified using the mandatory “reason for deregistration” field, which classifies deaths as “TB death” or “non‑TB death.” All non‑TB deaths and other deregistration reasons were treated as censored observations. Time zero (i.e. baseline) was defined as the date of notification to the JTBS, and follow‑up continued for up to 365 days after notification. Follow‑up ended at the earliest of one of the following; TB death, non‑TB death, deregistration for reasons other than death, or upon passing of 365 days. TB deaths occurring after treatment completion but within the 365‑day follow‑up window were included in the outcome definition. Patients who did not experience TB death during follow‑up were censored at the date of non‑TB death, deregistration, or at 365 days, whichever came first. The time to TB death was defined as the number of days from the date of notification to the date of death. If the date of death was missing, the date of deregistration was used as a proxy. Exposure variable The main exposure was the COVID-19 period at the time of TB notification, categorized into: Pre-COVID-19 period (January 2017–June 2019):Patients were assumed to have completed treatment before COVID-19 emerged in Japan.Pre-Omicron period (July 2019–November 2021):Community exposure to COVID-19 was limited; fewer than 5% of the population had been infected [7]. It should be noted that the early part of this period (i.e. up to January 2020) had no COVID-19 transmission. However, considering the standard 6-months treatment for TB, it was assumed that some patients diagnosed during this period may have experienced the pandemic, albeit low-exposure, towards the end of their treatment.Omicron variant-dominant period (December 2021–December 2022):COVID-19 transmission was widespread, with approximately 27% of the population infected [7] (see also Fig. 1). Pre-COVID-19 period (January 2017–June 2019): Pre-COVID-19 period (January 2017–June 2019): Patients were assumed to have completed treatment before COVID-19 emerged in Japan. Patients were assumed to have completed treatment before COVID-19 emerged in Japan. Pre-Omicron period (July 2019–November 2021): Pre-Omicron period (July 2019–November 2021): Community exposure to COVID-19 was limited; fewer than 5% of the population had been infected [7]. It should be noted that the early part of this period (i.e. up to January 2020) had no COVID-19 transmission. However, considering the standard 6-months treatment for TB, it was assumed that some patients diagnosed during this period may have experienced the pandemic, albeit low-exposure, towards the end of their treatment. Community exposure to COVID-19 was limited; fewer than 5% of the population had been infected [7]. It should be noted that the early part of this period (i.e. up to January 2020) had no COVID-19 transmission. However, considering the standard 6-months treatment for TB, it was assumed that some patients diagnosed during this period may have experienced the pandemic, albeit low-exposure, towards the end of their treatment. Omicron variant-dominant period (December 2021–December 2022): Omicron variant-dominant period (December 2021–December 2022): COVID-19 transmission was widespread, with approximately 27% of the population infected [7] (see also Fig. 1). COVID-19 transmission was widespread, with approximately 27% of the population infected [7] (see also Fig. 1). Fig. 1Newly notified and cumulative number of COVID-19 cases in Japan, 2020–2022. Newly reported and cumulative number of COVID-19 cases in Japan in the “pre-Omicron” and “Omicron” variant-dominant periods between 2020 and 2022 (those notified in the “pre-COVID” era are not shown) Newly notified and cumulative number of COVID-19 cases in Japan, 2020–2022. Newly reported and cumulative number of COVID-19 cases in Japan in the “pre-Omicron” and “Omicron” variant-dominant periods between 2020 and 2022 (those notified in the “pre-COVID” era are not shown) Other covariates included age in years, sex, country of birth, TB diagnosis category as recorded in the JTBS, diabetes status, mode of detection, delay, and insurance status. TB diagnosis category served as a proxy for baseline severity. Diabetes status reflected patient‑ or physician‑reported comorbidity at baseline. Delay was defined as the interval between self‑reported symptom onset and the first clinical visit. For insurance status, the various health insurance types under Japan’s national insurance system were re-grouped as “insured.” Individuals unable to pay insurance premiums but protected under the social welfare scheme were categorized as “receiving social welfare.” All others—those without health insurance or social welfare protection, or whose insurance status was unknown—were grouped as “uninsured/other/unknown”. These categories are mutually exclusive. Although public health centers may update these fields during follow‑up, the JTBS does not indicate whether or when such updates occur; therefore, all covariates reflected the information at the time of notification. Statistical analysis R version 4.4.1 [8] was used for all statistical analyses. Descriptive statistics were summarized as means with standard deviations or frequencies with percentages. Differences in the patient characteristics across the three COVID‑19 periods were assessed using chi‑square tests for categorical variables. TB mortality rates were calculated using total person-years as the denominator for each COVID-19 period. Person‑years were calculated for each patient from the date of TB notification to the earliest of TB death, non‑TB death, deregistration from the JTBS, or 365 days of follow‑up. Temporal trends in the proportion of TB deaths across the three periods were assessed using the Cochran-Armitage trend test. Pairwise comparisons of the proportion of TB deaths between COVID-19 periods were performed using Fisher’s exact test with Hochberg correction for multiple comparisons, using the RVAideMemoire package in R. Because non‑TB death was treated as a censoring event, the Cox models estimated cause‑specific hazards for TB death. Accordingly, the Kaplan–Meier curves represent cause‑specific survival, not the cumulative incidence of TB death in the presence of competing non‑TB mortality. We considered a competing‑risk framework; however, because the primary objective was to evaluate the association between COVID‑19 period and the instantaneous risk of TB death, cause‑specific hazard models were used as the main analytical approach. Cox proportional hazards regression models were fitted to estimate crude and adjusted hazard ratios (aHRs) with 95% confidence intervals (CIs). Multicollinearity among covariates was assessed via the variance inflation factor (VIF). Proportional hazards assumptions were evaluated via Schoenfeld residuals. Age, treated as a continuous variable, violated the proportional hazards assumption and was therefore modeled with a time‑varying effect using the tt() function in the Cox model. Categorical covariates that violated the proportional hazards assumption were adjusted for through stratification via the strata() option. Statistical significance was defined as p < 0.05. Analyses were conducted for the entire cohort and separately for patients aged 0–64 years and those aged ≥ 65 years to assess potential age‑specific differences in the association between the COVID‑19 period and TB mortality. To formally assess whether the association between COVID‑19 period and TB mortality differed by age group, a likelihood ratio test was performed by comparing a Cox model including a Covid period x age group interaction term against a nested model without the interaction term, with age group entered as a binary covariate (0–64 vs. ≥ 65 years). The time‑varying specification of age used in the primary models was not applied in this supplementary interaction test. Results Sociodemographic characteristics and changes over the study period Characteristics of 81,003 patients are summarized by COVID-19 in Table 1. Across the three periods, the age distribution shifted toward older adults: the proportion of patients aged ≥ 65 years increased from 66.8% in the pre-COVID-19 period to 70.2% in the Omicron period (p < 0.001), consistent with Japan’s aging TB population. The proportion of those with extrapulmonary TB (EPTB) also increased (22.6% to 27.0%, p < 0.001). Regarding the mode of detection, the proportion of patients identified during outpatient visits or hospitalization increased from 81.8% to 84.0% (p < 0.001), suggesting increased reliance on clinical encounters rather than screening or contact investigations during the pandemic.Table 1Characteristics of the study participants by COVID-19 periodPre-CovidJan 1, 2017- Jun 30, 2019Covid: pre-OmicronJul 1, 2019- Nov 30, 2021Covid: OmicronDec 1, 2021-Dec 31, 2022n%n%n%Total39,37410030,46010011,169100.0Mean age in years (SD)68.2 (± 21.7)69.1 (± 21.7)70.0 (± 21.5)Age groups 0 to 64 years old13,08433.29,67031.73,32729.8 65 years old or above26,29066.820,79068.37,84270.2Sex Male23,52259.717,87658.76,68159.8 Female15,85240.312,58441.34,48840.2Country of birth Japan-born34,26687.026,28086.39,54685.5 Foreign-born3,95010.03,44211.31,34212.0 Unknown1,1582.97382.42812.5Comorbidity - diabetes Yes5,67814.44,58815.11,79716.1 No29,66775.322,12972.67,87770.5 Unknown4,02910.23,74312.31,49513.4Diagnosis Sputum smear positive PTB14,77437.511,07936.44,08636.6 Other bacteriologically positive PTB11,62929.58,77128.83,11327.9 Clinical PTB4,06910.32,8869.59548.5 EPTB8,90222.67,72425.43,01627.0Mode of detection Screening5,02212.83,62611.91,30711.7 Contact investigation1,4753.78742.92752.5 Upon hospital visit/during hospitalization☨32,21981.825,44283.59,38584.0 Others or unknown6581.75181.72021.8Delay < 2 weeks10,89727.78,25727.12,89425.9 ≧ 2 weeks, < 1 month3,2808.32,3167.68147.3 ≧ 1 month, < 3 months4,30710.92,9399.61,0599.5 > 3 months2,2035.61,5475.16035.4 Unknown8,98822.87,94726.13,21428.8 No symptoms9,69924.67,45424.52,58523.1Insurance status Insured35,67290.627,62990.79,97589.3 Receiving social welfare2,5806.61,9696.57396.6 Not insured, others or unknown1,1222.88622.84554.1SD Standard deviation, PTB Pulmonary TB, EPTB Extrapulmonary TB, ☨ Either at outpatient visit or during hospitalization for diseases other than TB Table 1Characteristics of the study participants by COVID-19 periodPre-CovidJan 1, 2017- Jun 30, 2019Covid: pre-OmicronJul 1, 2019- Nov 30, 2021Covid: OmicronDec 1, 2021-Dec 31, 2022n%n%n%Total39,37410030,46010011,169100.0Mean age in years (SD)68.2 (± 21.7)69.1 (± 21.7)70.0 (± 21.5)Age groups 0 to 64 years old13,08433.29,67031.73,32729.8 65 years old or above26,29066.820,79068.37,84270.2Sex Male23,52259.717,87658.76,68159.8 Female15,85240.312,58441.34,48840.2Country of birth Japan-born34,26687.026,28086.39,54685.5 Foreign-born3,95010.03,44211.31,34212.0 Unknown1,1582.97382.42812.5Comorbidity - diabetes Yes5,67814.44,58815.11,79716.1 No29,66775.322,12972.67,87770.5 Unknown4,02910.23,74312.31,49513.4Diagnosis Sputum smear positive PTB14,77437.511,07936.44,08636.6 Other bacteriologically positive PTB11,62929.58,77128.83,11327.9 Clinical PTB4,06910.32,8869.59548.5 EPTB8,90222.67,72425.43,01627.0Mode of detection Screening5,02212.83,62611.91,30711.7 Contact investigation1,4753.78742.92752.5 Upon hospital visit/during hospitalization☨32,21981.825,44283.59,38584.0 Others or unknown6581.75181.72021.8Delay < 2 weeks10,89727.78,25727.12,89425.9 ≧ 2 weeks, < 1 month3,2808.32,3167.68147.3 ≧ 1 month, < 3 months4,30710.92,9399.61,0599.5 > 3 months2,2035.61,5475.16035.4 Unknown8,98822.87,94726.13,21428.8 No symptoms9,69924.67,45424.52,58523.1Insurance status Insured35,67290.627,62990.79,97589.3 Receiving social welfare2,5806.61,9696.57396.6 Not insured, others or unknown1,1222.88622.84554.1SD Standard deviation, PTB Pulmonary TB, EPTB Extrapulmonary TB, ☨ Either at outpatient visit or during hospitalization for diseases other than TB Characteristics of the study participants by COVID-19 period SD Standard deviation, PTB Pulmonary TB, EPTB Extrapulmonary TB, ☨ Either at outpatient visit or during hospitalization for diseases other than TB Delay patterns also changed. While the proportion of those with shorter delays (< 2 weeks) decreased from 27.7% to 25.9% (p < 0.001), this decrease was not as much accompanied by an increase in the proportion of those with longer delays but rather an increase in those whose delay information was unknown (22.8% to 28.8%, p < 0.001). The proportion of patients who were uninsured, categorized as “other,” or had unknown insurance status also increased modestly (2.8% to 4.1%, p < 0.001). Trends in TB mortality and Kaplan‒Meier estimates of TB mortality for different COVID-19 periods Table 2 presents TB mortality by age group and COVID-19 period. Overall, the proportion of those deregistered from the JTBS due to TB death increased from 8.1% in the pre-COVID-19 period to 9.9% in the Omicron period. Consistent with this, the proportion of TB deaths showed a significant increasing trend across the three periods (Cochran–Armitage trend test, p < 0.001). TB mortality rates also rose across the three period: among all ages, mortality rose from 10.1 (95% confidence interval [CI] 9.7–10.4) to 12.7 per 100 person-years (95% CI 12.0–13.5), representing a relative increase of approximately 26%. This rise was driven largely by older adults: mortality among those aged ≥ 65 years increased from 15.2 (95% CI 14.7–15.8) to 18.5 per 100 person‑years (95% CI 17.4–19.7), a relative increase of 22%. In contrast, mortality among individuals aged 0–64 years remained low and stable (1.5 to 1.6 per 100 person‑years) (see also Fig. 2). Fig. 2Trends in TB mortality from the pre-COVID-19 period to the COVID-19 period by age group. The mortality rates for the overall population, those aged 0–64 years and those aged 65 years and above, for the pre-COVID-19, pre-Omicron and Omicron variant-dominant periods. Error bars represent 95% confidence intervals. ** indicates p < 0.01 by pairwise Fisher’s exact test with Hochberg correction for multiple comparisons, applied to the proportion of TB deaths in each period Trends in TB mortality from the pre-COVID-19 period to the COVID-19 period by age group. The mortality rates for the overall population, those aged 0–64 years and those aged 65 years and above, for the pre-COVID-19, pre-Omicron and Omicron variant-dominant periods. Error bars represent 95% confidence intervals. ** indicates p < 0.01 by pairwise Fisher’s exact test with Hochberg correction for multiple comparisons, applied to the proportion of TB deaths in each period Table 2TB mortality among different age groups by COVID-19 periodPre-CovidJan 1, 2017- Jun 30, 2019Covid: pre-OmicronJul 1, 2019- Nov 30, 2021Covid: OmicronDec 1, 2021-Dec 31, 2022 n % n % n %All ages Total39,374100.030,460100.011,169100.0 Of whom, deleted from JTBS13,96235.511,08136.44,44039.8 Of whom, deleted due to TB death3,1998.12,7228.91,1049.9 Person years31,80224,3828,681 TB mortality/100 person years (95% CI)10.1 (9.7–10.4)11.2 (10.8–11.6)12.7 (12.0–13.5)0–64 years old Total13,084100.09,670100.03,327100.0 Of whom, deleted from JTBS2,81121.52,02320.978923.7 Of whom, deleted due to TB death1731.31121.2481.4 Person years11,9268,8612,983 TB mortality/100 person years (95% CI)1.5 (1.2–1.7)1.3 (1.0–1.5)1.6 (1.2–2.1)65 years old and above Total26,290100.020,790100.07,842100.0 Of whom, deleted from JTBS11,15142.49,05843.63,65146.6 Of whom, deleted due to TB death3,02611.52,61012.61,05613.5 Person years19,87615,5225,698 TB mortality/100 person years (95% CI)15.2 (14.7–15.8)16.8 (16.2–17.5)18.5 (17.4–19.7)JTBS Japan Tuberculosis Surveillance System, TB Tuberculosis, CI Confidence interval Table 2TB mortality among different age groups by COVID-19 periodPre-CovidJan 1, 2017- Jun 30, 2019Covid: pre-OmicronJul 1, 2019- Nov 30, 2021Covid: OmicronDec 1, 2021-Dec 31, 2022 n % n % n %All ages Total39,374100.030,460100.011,169100.0 Of whom, deleted from JTBS13,96235.511,08136.44,44039.8 Of whom, deleted due to TB death3,1998.12,7228.91,1049.9 Person years31,80224,3828,681 TB mortality/100 person years (95% CI)10.1 (9.7–10.4)11.2 (10.8–11.6)12.7 (12.0–13.5)0–64 years old Total13,084100.09,670100.03,327100.0 Of whom, deleted from JTBS2,81121.52,02320.978923.7 Of whom, deleted due to TB death1731.31121.2481.4 Person years11,9268,8612,983 TB mortality/100 person years (95% CI)1.5 (1.2–1.7)1.3 (1.0–1.5)1.6 (1.2–2.1)65 years old and above Total26,290100.020,790100.07,842100.0 Of whom, deleted from JTBS11,15142.49,05843.63,65146.6 Of whom, deleted due to TB death3,02611.52,61012.61,05613.5 Person years19,87615,5225,698 TB mortality/100 person years (95% CI)15.2 (14.7–15.8)16.8 (16.2–17.5)18.5 (17.4–19.7)JTBS Japan Tuberculosis Surveillance System, TB Tuberculosis, CI Confidence interval TB mortality among different age groups by COVID-19 period JTBS Japan Tuberculosis Surveillance System, TB Tuberculosis, CI Confidence interval Kaplan–Meier curves showed that most TB deaths occurred within the first 100 days after notification, particularly among patients aged ≥ 65 years. Because non‑TB death was treated as a censoring event, these curves represent cause‑specific survival rather than cumulative incidence. In this framework, the curves depict the probability of surviving without TB death under the hypothetical condition that competing non‑TB deaths do not occur. This distinction is especially relevant for older adults, among whom competing non‑TB mortality is substantial. The steep early decline, particularly among older adults, reflects the concentration of TB deaths within the first 100 days after notification, consistent with the known clinical vulnerability of older TB patients. Median survival time was not reached in any group. Mean survival times decreased slightly across periods for all age groups, with the largest decline observed among older adults (328 days in the pre-COVID-19 period to 321 days in the Omicron period). Log-rank tests indicated significant differences in survival across COVID-19 periods for the overall cohort and for those aged ≥ 65 years, but not for those aged 0–64 years (Fig. 3a and c). Figs. 3Cause-specific survival estimates for TB death by COVID-19 period (Kaplan–Meier method, non-TB deaths censored). a Estimates for the whole cohort, (b) those aged 0–64 years, and (c) those aged 65 years and above Cause-specific survival estimates for TB death by COVID-19 period (Kaplan–Meier method, non-TB deaths censored). a Estimates for the whole cohort, (b) those aged 0–64 years, and (c) those aged 65 years and above Baseline characteristics of patients who died due to TB versus those who did not die from TB (including survivors and non‑TB deaths) Baseline characteristics of patients who died due to TB and those who did not die due to TB, namely, those survived the follow‑up period as well as those who died due to reasons other than TB, and are shown in Table 3. Compared with the latter, patients who died due to TB were more likely to be older, Japan‑born, diagnosed with sputum smear‑positive pulmonary TB, have diabetes, and be detected during hospital visits or hospitalization. These patterns were consistent across age‑stratified analyses. Table 3Baseline characteristics of patients who died due to TB versus no TB death*All ages0–64 years old65 years old and aboveNo TB deathTB deathNo TB deathTB deathNo TB deathTB deathAge, in yearsNANANANANANASex Female30,1292,85810,4015019,7282,808 Male43,8494,16715,34728328,5023,884Country of Birth Foreign-born8,671638,3322733936 Japan-born63,3936,69916,94029346,4536,406 Unknown1,914263476131,438250Diagnosis EPTB18,2051,4375,1934213,0121,395 Sputum smear positive PTB25,6374,3027,72124617,9164,056 Other bacteriologically positive PTB22,3631,1508,2553314,1081,117 Clinical PTB7,7731364,579123,194124Comorbidity - diabetes No54,9654,70821,11218233,8534,526 Yes10,8761,1871,979498,8971,138 Unknown8,1371,1302,6571025,4801,028Mode of detection Contact investigation2,599251,776082325 Upon hospital visit/during hospitalization✟60,2826,76416,19329144,0896,473 Screening9,883727,28682,59764 Others or unknown1,21416449334721130Delay No symptoms19,3443949,46469,880388 Less than 2 weeks19,5632,4854,9076614,6562,419 More than 2 weeks and less than 1 month5,8205901,905223,915568 More than 1 month and less than 3 months7,5367692,708474,828722 More than 3 months3,9893641,768512,221313 Unknown17,7262,4234,99614112,7302,282Insurance status With insurance67,0476,22923,28317643,7646,053 On social welfare2,1842551,20497980158 No insurance or unknown4,7475411,261603,486481COVID period Pre-COVID36,1753,19912,91117323,2643,026 COVID pre-Omicron27,7382,7229,55811218,1802,610 COVID Omicron10,0651,1043,279486,7861,056PTB Pulmonary TB, EPTB Extrapulmonary TB, * ‘No TB death’ includes patients who survived the follow‑up period as well as those who died from non‑TB causes, consistent with the JTBS classification of deregistration reasons.✟ Either at outpatient visit or during hospitalization Table 3Baseline characteristics of patients who died due to TB versus no TB death*All ages0–64 years old65 years old and aboveNo TB deathTB deathNo TB deathTB deathNo TB deathTB deathAge, in yearsNANANANANANASex Female30,1292,85810,4015019,7282,808 Male43,8494,16715,34728328,5023,884Country of Birth Foreign-born8,671638,3322733936 Japan-born63,3936,69916,94029346,4536,406 Unknown1,914263476131,438250Diagnosis EPTB18,2051,4375,1934213,0121,395 Sputum smear positive PTB25,6374,3027,72124617,9164,056 Other bacteriologically positive PTB22,3631,1508,2553314,1081,117 Clinical PTB7,7731364,579123,194124Comorbidity - diabetes No54,9654,70821,11218233,8534,526 Yes10,8761,1871,979498,8971,138 Unknown8,1371,1302,6571025,4801,028Mode of detection Contact investigation2,599251,776082325 Upon hospital visit/during hospitalization✟60,2826,76416,19329144,0896,473 Screening9,883727,28682,59764 Others or unknown1,21416449334721130Delay No symptoms19,3443949,46469,880388 Less than 2 weeks19,5632,4854,9076614,6562,419 More than 2 weeks and less than 1 month5,8205901,905223,915568 More than 1 month and less than 3 months7,5367692,708474,828722 More than 3 months3,9893641,768512,221313 Unknown17,7262,4234,99614112,7302,282Insurance status With insurance67,0476,22923,28317643,7646,053 On social welfare2,1842551,20497980158 No insurance or unknown4,7475411,261603,486481COVID period Pre-COVID36,1753,19912,91117323,2643,026 COVID pre-Omicron27,7382,7229,55811218,1802,610 COVID Omicron10,0651,1043,279486,7861,056PTB Pulmonary TB, EPTB Extrapulmonary TB, * ‘No TB death’ includes patients who survived the follow‑up period as well as those who died from non‑TB causes, consistent with the JTBS classification of deregistration reasons.✟ Either at outpatient visit or during hospitalization Baseline characteristics of patients who died due to TB versus no TB death* PTB Pulmonary TB, EPTB Extrapulmonary TB, * ‘No TB death’ includes patients who survived the follow‑up period as well as those who died from non‑TB causes, consistent with the JTBS classification of deregistration reasons.✟ Either at outpatient visit or during hospitalization Effect of being diagnosed during the COVID-19 pandemic on TB mortality Results of the Cox proportional hazards models are presented in Table 4. Among all ages, being diagnosed during the Omicron period was associated with a significantly higher risk of TB death compared with the pre‑COVID‑19 period (aHR 1.10, 95% CI 1.02–1.18, p = 0.010), whereas no significant association was observed for the pre‑Omicron period (aHR 1.04, 95% CI 0.99–1.10, p = 0.147). Model diagnostics indicated no major violations of proportional hazards assumptions after incorporating the time‑varying effect of age (global Schoenfeld test p > 0.05), and the models demonstrated moderate discrimination (concordance statistic 0.69).Table 4Hazard ratios for being diagnosed during the COVID-19 pandemic for TB mortalityNon-TB deathTB deathcHRaHR95% CIp valueAll ages Age (baseline)NA1.071.041.03–1.04< 0.001 Age (baseline), modeled with time-varying effectNANA1.011.01–1.01< 0.001 Pre-COVID36,1753,199Reference COVID: pre-Omicron27,7382,7221.101.040.99–1.100.147 COVID: Omicron10,0651,1041.241.101.02–1.180.0100–64 years old Age (baseline)NA1.091.051.03–1.09< 0.001 Age (baseline), modeled with time-varying effectNANA1.000.99–1.010.365 Pre-COVID12,911173Reference COVID: pre-Omicron9,5581120.880.890.69–1.140.357 COVID: Omicron3,279481.101.090.78–1.530.60665 years old and above Age (baseline)NA1.081.041.03–1.05< 0.001 Age (baseline), modeled with time-varying effectNANA1.011.01–1.01< 0.001 Pre-COVID23,2643,026Reference COVID: pre-Omicron18,1802,6101.091.050.99–1.100.103 COVID: Omicron6,7861,0561.191.101.03–1.180.007Age (baseline) and age modeled with a time‑varying effect (TVC) represent alternative specifications of the age term in separate Cox models. The COVID‑19 period estimates are taken from the model using the time‑varying effect of age, which satisfied the proportional hazards assumptionTB Tuberculosis, cHR Crude hazard ratio, aHR Adjusted hazard ratio, CI Confidence interval, TVC Time-varying covariate Table 4Hazard ratios for being diagnosed during the COVID-19 pandemic for TB mortalityNon-TB deathTB deathcHRaHR95% CIp valueAll ages Age (baseline)NA1.071.041.03–1.04< 0.001 Age (baseline), modeled with time-varying effectNANA1.011.01–1.01< 0.001 Pre-COVID36,1753,199Reference COVID: pre-Omicron27,7382,7221.101.040.99–1.100.147 COVID: Omicron10,0651,1041.241.101.02–1.180.0100–64 years old Age (baseline)NA1.091.051.03–1.09< 0.001 Age (baseline), modeled with time-varying effectNANA1.000.99–1.010.365 Pre-COVID12,911173Reference COVID: pre-Omicron9,5581120.880.890.69–1.140.357 COVID: Omicron3,279481.101.090.78–1.530.60665 years old and above Age (baseline)NA1.081.041.03–1.05< 0.001 Age (baseline), modeled with time-varying effectNANA1.011.01–1.01< 0.001 Pre-COVID23,2643,026Reference COVID: pre-Omicron18,1802,6101.091.050.99–1.100.103 COVID: Omicron6,7861,0561.191.101.03–1.180.007Age (baseline) and age modeled with a time‑varying effect (TVC) represent alternative specifications of the age term in separate Cox models. The COVID‑19 period estimates are taken from the model using the time‑varying effect of age, which satisfied the proportional hazards assumptionTB Tuberculosis, cHR Crude hazard ratio, aHR Adjusted hazard ratio, CI Confidence interval, TVC Time-varying covariate Hazard ratios for being diagnosed during the COVID-19 pandemic for TB mortality Age (baseline) and age modeled with a time‑varying effect (TVC) represent alternative specifications of the age term in separate Cox models. The COVID‑19 period estimates are taken from the model using the time‑varying effect of age, which satisfied the proportional hazards assumption TB Tuberculosis, cHR Crude hazard ratio, aHR Adjusted hazard ratio, CI Confidence interval, TVC Time-varying covariate Age was a strong predictor of TB mortality in all models. When modeled with a time‑varying effect, the influence of age increased over the follow‑up period (aHR 1.01, 95% CI 1.01–1.01, p < 0.001), reflecting the concentration of early deaths among older adults. In the age‑stratified analyses, the point estimates for the association between COVID‑19 period and TB mortality were similar across age groups. Statistical significance was reached only among those aged ≥ 65 years (Omicron, aHR 1.10, 95% CI 1.03–1.18), whereas estimates for individuals aged 0–64 years were imprecise due to the small number of TB deaths. These differences in statistical significance should therefore not be interpreted as evidence of a true subgroup difference. A formal likelihood ratio test for the interaction between COVID‑19 period and age group was non‑significant (χ²=2.05, df = 2, p = 0.359), providing no statistical evidence of effect modification by age. Discussion This nationwide retrospective study examined trends in TB mortality in Japan before and during the COVID‑19 pandemic and assessed whether being diagnosed during the pandemic was independently associated with TB mortality. Using six years of national surveillance data, we found that TB mortality increased progressively across the three periods, with the most pronounced rise occurring during the Omicron variant‑dominant period. This increase was driven almost entirely by older adults, who constitute the majority of TB patients in Japan. Among individuals aged ≥ 65 years, being diagnosed during the Omicron period was independently associated with a higher risk of TB death after adjustment (aHR 1.10, 95% CI 1.03–1.18, p = 0.007). Several shifts in patient characteristics occurred progressively across all three periods and may help explain the observed increase in TB mortality. The proportion of patients aged ≥ 65 years increased from 66.8% in the pre‑COVID period to 68.3% in the pre‑Omicron period and 70.2% in the Omicron period, and older patients consistently exhibited substantially higher TB mortality than younger patients. The prevalence of extrapulmonary TB rose from 22.6% to 25.4% to 27.0%, and diabetes prevalence increased from 14.4% to 15.1% to 16.1% across the three periods. A greater proportion of cases were also detected during hospitalization across all periods, which may indicate more advanced disease at presentation. These progressive changes suggest that the patient population became increasingly vulnerable over time, and that this underlying shift — rather than the pandemic alone — contributed to the gradual rise in TB mortality observed even before the Omicron period. At the onset of the pandemic, Japan classified COVID‑19 as a Class 2 infectious disease, the same category as TB, requiring mandatory reporting and intensive public health follow‑up [9, 10]. This reallocation of public health resources raised concerns that TB services might be disrupted, with early reports suggesting evidence of reductions in TB case detection and delays in care [11], which were consistent with global observations. However, several studies have eventually noted that Japan’s public health and medical systems remained relatively resilient during the initial pandemic waves, and TB services largely remained functional [12–15]. In the adjusted Cox model, being diagnosed during the pre-Omicron period was not independently associated with TB mortality (aHR 1.04, 95% CI 0.99–1.10, p = 0.147), despite a modest increase in crude mortality rates during this period (from 10.1 to 11.2 per 100 person-years overall, and from 15.2 to 16.8 per 100 person-years among those aged ≥ 65 years). The attenuation of this association after adjustment suggests that the gradual rise in crude mortality during the pre-Omicron period is largely accounted for by the progressive shifts in patient characteristics described above, rather than by an independent effect of the pandemic itself during this earlier phase. In contrast, the Omicron period was characterized by unprecedented numbers of COVID‑19 cases, despite the variant’s lower intrinsic virulence. The sheer volume of infections likely strained healthcare capacity and increased the probability of co‑infection with COVID‑19 among TB patients. International cohort studies have shown that COVID‑19 co‑infection has been associated with worse TB outcomes, including higher mortality, particularly among older adults and those with comorbidities [16, 17]. Although we could not assess co‑infection directly due to the absence of COVID‑19 clinical data in the JTBS, the elevated and independently significant mortality observed during the Omicron period - beyond what is explained by the progressive shift in patient characteristics - suggests an additional burden specific to this period, potentially reflecting the combined effects of increased exposure to COVID-19, healthcare system strain, and the heightened vulnerability of Japan’s predominantly older TB population. The rising proportion of patients with diabetes during the Omicron period also warrants attention. Although a formal likelihood ratio test for the interaction between diabetes and COVID‑19 period did not reveal strong statistical evidence of effect modification, diabetes is a well‑established risk factor for adverse TB outcomes. Its increasing prevalence across all three periods may have contributed to the heightened vulnerability of the patient population, and this trend may have been especially consequential during the Omicron period when it coincided with peak COVID‑19 transmission and healthcare strain. Beyond direct co‑infection, indirect effects of the pandemic may also have contributed to increased TB mortality among older adults. Prolonged social restrictions, reduced physical activity [18, 19], and worsening mental health [20–22], well‑documented consequences of the pandemic in Japan, may have disproportionately affected older individuals, potentially exacerbating frailty and impairing their ability to adhere to TB treatment. Poor activities of daily living and depression are known risk factors for adverse TB outcomes [23–26], including mortality. These broader societal impacts may therefore have played a role in the observed increase in TB deaths during the later stages of the pandemic. Regarding the age-stratified findings, TB mortality rates among individuals aged 0–64 years remained low and stable across all three periods (1.5, 1.3, and 1.6 per 100 person-years in the pre-COVID, pre-Omicron, and Omicron periods, respectively). In the Cox model, the adjusted hazard ratio for the Omicron period in this age group was 1.09 (95% CI 0.78–1.53, p = 0.606), a point estimate virtually identical to that observed among adults aged ≥ 65 years (aHR 1.10, 95% CI 1.03–1.18). The wide confidence interval around the younger group’s hazard ratio reflects the small absolute number of TB deaths across all periods in this group (173, 112, and 48 in the pre-COVID, pre-Omicron, and Omicron periods, respectively), which substantially limits statistical power. A formal likelihood ratio test for the interaction between COVID-19 period and age group was non-significant (χ²=2.05, df = 2, p = 0.359), providing no statistical evidence that the effect of the pandemic on TB mortality truly differed between age groups. The difference in statistical significance between age groups therefore reflects differences in statistical power rather than a fundamentally different epidemiological pattern. It is also important to interpret these findings in the context of competing mortality risks. Non‑TB deaths were substantial in this predominantly older cohort, and because non‑TB death was treated as a censoring event, our analyses estimated cause‑specific hazards rather than the cumulative incidence of TB death. Similarly, the Kaplan–Meier curves reflect cause‑specific survival, representing the probability of surviving without TB death under the hypothetical condition that competing non‑TB deaths do not occur. In older adults—who experience high background mortality—cumulative incidence functions may differ from cause‑specific survival curves. Nonetheless, cause‑specific models are appropriate for evaluating etiologic associations and for understanding how the instantaneous risk of TB death varied across COVID‑19 periods. Our findings underscore the unique vulnerability of older adults with TB in Japan. Even before the pandemic, TB mortality among older adults was substantially higher than among younger individuals. The additional pressures introduced by the pandemic appear to have amplified this disparity. As Japan continues to experience population aging, strengthening TB care for older adults—including early detection, comprehensive clinical management, and social support—will be essential, particularly during public health emergencies. This study has several limitations. First, because the analysis relied solely on national TB surveillance data, we lacked clinical information on COVID‑19 infection and vaccination status. Although several cohort studies from other countries have examined the impact of COVID‑19 co‑infection and vaccination on TB outcomes, such data were not available in Japan, and we were unable to assess these associations in our dataset. Other important confounders known to influence TB outcomes—such as malnutrition, smoking, and chronic obstructive pulmonary disease—were also unavailable and therefore unmeasured. HIV status is available from the JTBS, however, the data is missing for the vast majority of patients and was therefore excluded from the analysis. Given the overall low HIV prevalence rate [27] among the Japanese population and HIV-TB co-infection rate [4], the impact on our results is likely to be minimal. Second, the classification of “TB death” and “non-TB death” depends entirely on physicians’ reporting to public health centers. Given Japan’s extremely low autopsy rate compared with other high-income countries, both under- and over-reporting of TB deaths, as well as misclassification, may have occurred [28]. These issues may have been exacerbated during periods of heightened healthcare strain, such as the COVID‑19 pandemic. Moreover, the JTBS does not record the causes of “non‑TB death,” meaning that some TB‑related deaths may have been misclassified as non‑TB deaths. Third, because all patients were followed for up to 365 days from the date of notification, classical survivor bias due to variable follow‑up does not apply. However, a different form of bias may still arise: individuals who died before TB was diagnosed would not enter the JTBS at all. If pre‑diagnosis deaths increased during periods of healthcare strain, such as the Omicron wave, our estimates may underestimate the true mortality burden. Despite these limitations, the use of comprehensive national surveillance data covering over 81,000 patients across six years provides a robust foundation for examining TB mortality trends during the COVID‑19 pandemic in Japan. Conclusion This nationwide analysis demonstrates that TB mortality in Japan increased during the COVID‑19 pandemic, with the most pronounced rise occurring during the Omicron variant‑dominant period. Importantly, this increase was driven almost entirely by older adults, who already represent the majority of TB patients in Japan and who remain particularly vulnerable to adverse outcomes. Being diagnosed during the Omicron period was independently associated with higher TB mortality among individuals aged ≥ 65 years, whereas estimates among those aged 0–64 years were imprecise and should not be interpreted as evidence of absence of association. These findings highlight the need for strengthened TB care and support systems for older populations, especially during periods of widespread infectious disease transmission or healthcare system strain. Ensuring continuity of TB services, improving early detection, and addressing the broader health and social vulnerabilities of older adults will be essential for mitigating excess TB mortality during future public health emergencies.