Tuberculosis (TB) was a significant public health burden in Taiwan during the last century. In 1978, TB was the eighth leading cause of death, with 2,809 deaths (16.55 per 100,000) and accounted for 3.56% of all deaths. By 2022, it had declined to the 24th leading cause, with 477 deaths (2.05 per 100,000) and 0.23% of total mortality. Notably, 83.65% (399 of 477) of TB-related deaths in 2022 occurred in individuals aged 65 years or older. As of May 2024, 18.7% of Taiwan's population was aged ≥65 years, and the country is projected to become a super-aged society—defined by over 21% of the population aged ≥65—by 2026. These demographics highlight that TB continues to pose a health challenge among older populations in Taiwan.
To analyze long-term trends, the study employed both Joinpoint regression and age-period-cohort (A–P–C) models. Joinpoint regression identifies statistically significant changes in trend over time and computes the annual percentage change (APC), which indicates the rate and direction of change. In contrast, the A–P–C model disaggregates the effects of age, time period, and birth cohort on mortality trends, offering deeper insight into underlying epidemiological dynamics. This study is the first to apply both models simultaneously to Taiwan’s TB mortality data from 1978 to 2022.
Mortality data were obtained from Health Statistics issued by Taiwan’s Ministry of Health and Welfare, using ICD-8 (1978–1980: codes 010–019), ICD-9 (1981–2007: codes 010–018), and ICD-10 (2008–2022: codes A15–A19). Corresponding population data were sourced from Demographic Statistics by the Department of Household Registration, Ministry of the Interior. Age-standardized mortality rates (ASMRs) were calculated using the WHO World Standard Population 2000.
All crude and standardized mortality rates from 1978 to 2022 were calculated using Microsoft Excel. Joinpoint regression analyses used annual crude and age-standardized TB mortality rates, with the rate calculation function implemented in the Joinpoint Regression Program (Version 5.0.2, May 2023). The maximum number of joinpoints allowed was seven, and model selection was based on the Weighted Bayesian Information Criterion (WBIC). Confidence intervals for APC and average APC (AAPC) were calculated using the Empirical Quantile Method.
For A–P–C analysis, TB mortality rates were calculated by sex and age group across nine 5-year calendar periods (1978–1982 through 2018–2022) and 17 age groups (0–4 to 80–84 years). The oldest age group (85+) was excluded to meet the model’s 5-year interval requirement. As a result, 25 birth cohorts (from 1898–1902 to 1998–2002) were included. The average annual mortality rate was calculated for each age group within each period. All computations were performed using Excel.
Ethical approval was not required, as the data used were aggregated and derived from official government publications. To compare TB mortality rates before and during the COVID-19 pandemic, Poisson regression was used—a standard log-linear model for count data, assuming the mean equals the variance. Model dispersion was checked, and no overdispersion was detected. These analyses were conducted using SAS software.
In the Joinpoint analysis, APCs were estimated for sex-specific crude and age-standardized mortality rates. For A–P–C models, log-linear regressions were fit to estimate overall net drift (analogous to the APC for age-standardized rates) and local drift (age-specific APCs over time). The longitudinal age curve illustrated TB mortality across 5-year age groups for the reference cohort (1958–1962). Period rate ratios (RRs) were estimated by comparing each period to the reference period (1998–2002), and cohort RRs were computed by comparing each birth cohort to the reference cohort (1958–1962).
The age-standardized mortality rate (ASMR) for tuberculosis (TB) in Taiwan decreased substantially from 1978 to 2022. Among males, it declined from 41.70 to 1.59 per 100,000, and among females, from 14.58 to 0.45 per 100,000—representing reductions of 96.19% and 96.91%, respectively. The ASMR rate ratios (1978 vs. 2022) were 26.23 for males and 32.40 for females, indicating a steeper proportional decline among females.
Crude mortality rates also decreased, though less sharply. In males, the crude rate dropped from 23.46 to 3.02 per 100,000, and in females, from 8.84 to 1.09 per 100,000—equivalent to reductions of 87.13% and 87.67%, respectively. The crude rate ratios were 7.77 in males and 8.11 in females. Compared to ASMR trends, crude mortality trends showed a more gradual decline, likely reflecting Taiwan’s aging population.
During the COVID-19 pandemic, TB crude mortality further declined in males compared to 2019 but rebounded in both sexes in 2022. For crude mortality, the average annual percentage change (AAPC) was −4.68% in males and −4.99% in females. In males, all Joinpoint segments except 1987–1990 showed statistically significant declines (p < 0.05), with the steepest drop in 2003–2006 (APC: −15.89%) and a slight increase in 1987–1990 (APC: +1.47%). In females, all segments showed a decline, but the reductions in 1987–2001 and 2007–2022 were not statistically significant. The sharpest decrease occurred during 2001–2007 (APC: −8.21%).
For ASMR, the AAPC was −7.17% in males and −7.64% in females. All segments significantly declined (p < 0.05) except 2020–2022 in females. Among males, the greatest drop was in 2000–2010 (APC: −12.01%), while the least was in 2010–2022 (APC: −5.35%). Among females, the sharpest decline occurred from 1999–2020 (APC: −9.81%), followed by an increase during 2020–2022 (APC: +11.44%).
Detailed segment analysis of crude rates also revealed that in males, the largest drop occurred in 2003–2006 (APC: −13.23%), and the only increase (though not significant) was seen in 1986–1989 (APC: +0.94%). In females, the sharpest drop occurred in 1982–1988 (APC: −8.16%), and a notable rise in 2020–2022 (APC: +7.33%). For ASMR, the most significant male decline occurred in 2003–2006 (APC: −16.60%), with the smallest decline in 1986–1989 (APC: −1.34%). In females, the largest drop occurred in 2000–2007 (APC: −11.52%), with an increase in 2020–2022 (APC: +7.36%).
Regarding age-specific trends, only older age groups (≥55 years) were analyzed due to sparse data in younger groups. The age-sex-specific APCs from Joinpoint models ranged from −3.55% to −10.53%. Differences between Joinpoint and age-period-cohort (A–P–C) model estimates ranged from −0.52% to +1.42%.
After adjusting for A–P–C effects, all estimated APCs and age-specific APCs remained negative, confirming a consistent decline in TB mortality. The average annual reduction was −8.83% for males and −9.77% for females. The most significant drop in males was observed in the 25–29 age group (EAPC: −10.19%, 95% CI: −11.53% to −8.82%), while in females, it was in the 0–4 age group (EAPC: −11.92%, 95% CI: −21.36% to −1.35%).
TB mortality trends between ages 45 and 74 showed significantly steeper declines in females than males. However, among those aged ≥65 years, mortality declined more slowly in both sexes. The highest longitudinal age-specific mortality rates were observed in the 0–4 age group: 6.31 per 100,000 in males (95% CI: 4.28–9.29) and 7.87 per 100,000 in females (95% CI: 5.14–12.05).
The longitudinal mortality curves were similar for both sexes under age 24 but diverged after age 30, with consistently higher rates in males. In males, mortality increased with age except for a peak in the 0–4 age group. In females, mortality slightly rose to a second peak at 20–24 years, declined to a nadir at 55–59 years, and then rose again slightly.
When comparing period rate ratios (RRs) to the reference period (1998–2002), the highest RRs occurred in 1978–1982 (RR: 7.20 in males; 10.26 in females), while the lowest were observed in 2018–2022 (RR: 0.19 in males, 95% CI: 0.16–0.23; RR: 0.17 in females, 95% CI: 0.13–0.21). These declines were steeper in females.
Cohort RRs also showed consistent declines across successive birth cohorts, with peaks in the 1898–1902 cohort for males and 1903–1907 cohort for females. Similar to period RRs, cohort-based declines were more pronounced in females than in males.
The difference in annual percentage change (APC) estimates between the Joinpoint regression and the age–period–cohort (A–P–C) model can be attributed to differences in adjustment factors and the number of age groups included in the analyses. In Joinpoint regression, crude rate APCs were unadjusted, while ASMR APCs were adjusted only for age. In contrast, the A–P–C model adjusted simultaneously for age, period, and cohort effects.
Another methodological distinction is the inclusion of age groups. For Joinpoint analysis, ASMRs were calculated using 18 age groups, including the oldest group (85+). However, in the A–P–C model, only 17 age-specific rates were used, excluding the 85+ group due to the required 5-year age interval structure. Since TB mortality is typically higher and decreases more gradually in the oldest age groups, excluding the 85+ group likely contributed to a slightly larger APC estimate in the A–P–C model.
TB mortality in Taiwan has shown a steady decline over the past 45 years, with the exception of a slight increase in 2020–2022 among females. Compared to previous studies, TB mortality in Taiwan remains higher than that reported in Japan. The overall trends of age-standardized mortality rates in this study resemble those observed in high-middle sociodemographic index (SDI) countries. The accelerated decline in mortality after 1999–2000 may reflect improved health interventions and policy implementation.
Sex differences in TB mortality may stem from a combination of biological factors (e.g., hormonal and genetic differences), behavioral risk exposures (e.g., smoking and alcohol use), disparities in treatment compliance and response, and changing gender-related health inequalities. Taiwan's public health advancements—such as the introduction of 9-year compulsory education in 1968, the national vaccination program (since 1949), the launch of National Health Insurance in 1995, and the implementation of directly observed treatment, short-course (DOTS)—have played important roles in reducing sex-based disparities in TB outcomes.
TB mortality declined less among older age groups. While Bacillus Calmette–Guérin (BCG) vaccination has been effective in preventing TB-related deaths, particularly among children under 15, the vaccine's protective effect wanes with age. Older adults face a combination of reduced immunity, greater exposure to risk factors, and immunosenescence, which may explain the smaller estimated APCs with increasing age.
The significant period and cohort effects observed in this study reflect Taiwan’s socioeconomic transformation. Between 1978 and 2022, Taiwan transitioned from an agricultural society to an industrialized, high-income nation, with GDP per capita increasing from USD 1,606 to USD 32,756. National nutrition status improved markedly, shifting from deficiency to sufficiency. The institutional response to TB evolved as well—from a single TB ward established in 1915 for treatment to a centralized public health system under the Taiwan CDC focusing on prevention. As outlined in Supplementary Table S1, anti-TB programs expanded from treatment-only to integrated prevention strategies.
The BCG vaccination program, introduced in 1949, has maintained a coverage rate of approximately 98% over the past decade. In addition, free TB treatment and widespread implementation of DOTS have ensured high treatment completion rates. The observed period and cohort effects in TB mortality are thus likely the result of Taiwan’s socioeconomic development, improved nutrition, strengthened healthcare infrastructure, universal BCG coverage, and access to free TB treatment.
Source: Chen, S.Y., 2024. Trends and annual percentage changes in tuberculosis mortality estimated by Joinpoint regression and age-period-cohort analyses, Taiwan, 1978–2022. Scientific Reports, 14(1), p.29101.
No comments:
Post a Comment