Model overview
We constructed dynamic, age-specific and sex-specific life tables to estimate years of life lived, PALYs, and associated economic costs in China from 2021 to 2030 among the working-age population (men aged 20–63 years and women aged 20–58 years). The life tables were stratified into two subpopulations: individuals with COPD and those without the condition.
To reflect the dynamic nature of demographic change, the models incorporated: (1) projected population changes over the follow-up period, including age- and sex-specific mortality trends; (2) new COPD cases based on age- and sex-specific expected trends; and (3) cohort entry at age 20 years and exit at the retirement age. In 2024, China announced a phased plan to raise the statutory retirement age to 63 years for men (from 60) and to 55 or 58 years for women (from 50 to 55), depending on occupation. The model assumed retirement at age 63 for men and 58 for women to fully account for COPD-related productivity loss [14]. Thus, the baseline included males aged 10 to 63 years and females aged 10 to 58 years. Individuals aged 10–19 years in 2020 were also included, because they would progressively enter the working-age population (≥ 20 years) during the 10-year projection period. For example, individuals aged 10 in 2020 would reach 20 years of age by 2030, the final year of the projection.
This study employed two comparative models to estimate the potential benefits of COPD prevention: a counterfactual scenario in which COPD incidence was set to zero throughout the study period, and a reference scenario reflecting current epidemiological projections. Differences in outcomes (years of life, PALYs, and associated economic costs) between the two scenarios represented the potential gains achievable through effective COPD prevention. The key model inputs and their data sources are shown in Table S1. The World Health Organization (WHO) standard 3% annual discount rate [15] was applied to all years of life lived, PALYs lived and related costs.
Model population
The number and demographic profile of the baseline population was obtained from the 2020 China Census and stratified by sex and five-year age groups (shown in Table 1) [16]. Data on COPD prevalence in 2020 and incidence from 1990 to 2021 were sourced from the Global Burden of Disease Study 2021 (GBD 2021), reported in five-year age intervals and by sex [17].
To generate continuous age-specific estimates, values for each five-year age group were assumed to represent the midpoint of the interval (e.g., age 22 for the 20–24 group) [8, 9]. These were then interpolated to single-year ages using quadratic regression for incidence rate and exponential regression for prevalence, based on model fit assessed by R2.
To project the incidence of COPD from 2021 to 2030, we applied a Bayesian age-period-cohort (BAPC) model using integrated nested Laplace approximations. This approach has shown improved accuracy and coverage relative to alternative forecasting methods [18, 19]. Analyses were conducted with the BAPC package in R, following standard procedures established in previous studies [19, 20].
Mortality rates
Baseline mortality data were derived from the 2020 China Census [16]. Temporal trends in population mortality risk over the modelling period were projected using the EAPC for China, calculated from the United Nations World Population Prospects forecast [21].
Temporal trends in rates were assessed using a log-linear regression model, from which the EAPC was derived. The model was specified as Y = α + βX + ε, where Y denotes the natural logarithm of the rate, X is the calendar year, α is the intercept, β is the coefficient indicating the trend over time, and ε is the random error. The EAPC was computed as 100 × [exp(β) − 1], quantifying the average annual rate of change [18].
Mortality estimates for COPD and non-COPD subpopulations were derived by combining population-wide mortality data with COPD-specific excess mortality risks. First, we obtained a fixed-effect pooled hazard ratio (HR) of 1.40 (95% CI 1.28–1.53) from a meta-analysis assessing the association between mortality and various COPD diagnostic criteria [22]. The pooled hazard ratio was then applied to estimate mortality rates for the two subpopulations. Calculation of annual mortality rates in those with and without COPD in the Chinese population by age group and sex was based on the following formula [9, 11, 23]:
$$\begin{array}{c}Mort_{Non-COPD}\;=\;Mort_{Tot}\;/\;\lbrack HR\;\times\;Prev_{COPD}+\;(1-Prev_{COPD})\rbrack\\Mort_{COPD}\;=\;Mort_{Non-COPD}\;\times\;HR\end{array}$$
Where MortTot is the mortality rate for the total population; MortCOPD is the mortality rate for the sub-population with COPD; MortNon−COPD is the mortality rate for the non-COPD sub-population; PrevCOPD is the prevalence of COPD; and HR is the hazard ratio for mortality associated with COPD relative to non-COPD.
Labor force participation
To estimate productivity loss, we applied sex- and age-specific labor force participation rates in five-year intervals, covering males aged 20–63 years and females aged 20–58 years. Labor force participation in China was sourced from the International Labor Organization (ILO) estimates for 2020 [24, 25]. Labor force participation was lowest in men aged 60–64 years (55.9%) and in women aged 55–59 years (55.3%). The participation was highest in men aged 30–34 years (91.2%) and in women aged 40–44 years (84.4%).
Productivity indices
Consistent with prior studies [8,9,10], we assessed productivity loss attributable to COPD using two indices: the COPD-related labor force dropout rate, which quantifies the reduction in workforce participation among individuals with COPD compared to those without; and the productivity index, which represents an individual’s work output on a scale from 0 (no productivity) to 1 (full productivity), capturing performance impairments due to health limitations.
Although numerous studies have explored the impact of COPD on work productivity [26,27,28,29], a consolidated quantitative estimate has been lacking. To address this gap, we performed meta-analyses to derive two key measures: the COPD-related labor force dropout rate and composite productivity index. The pooled labor force dropout rate among individuals with COPD was 92.794%. The latter incorporated both absenteeism and presenteeism components.
Absenteeism referred to the number of workdays missed annually due to COPD and was calculated as a proportion of the total possible working days (245 days/year in China) [9], yielding an estimated productivity reduction of 0.846% based on an average of 2.073 lost days per year. Presenteeism captured self-reported reductions in on-the-job productivity and was estimated at 4.5%. Together, these figures indicated a total productivity loss of 5.346%, corresponding to a productivity index of 0.947.
In contrast, those without the condition were assumed to experience no productivity losses, maintaining a dropout rate of zero and a productivity index of 1. These relative measures were subsequently applied to the 2020 ILO data, stratified by sex and age, to estimate workforce participation in populations with and without COPD. More methodological details were provided in the Supplementary Appendix.
Estimation of PALYs and value of PALYs
To estimate the PALYs lived by the population, the number of years spent in the labor force was calculated by multiplying the years of life lived by the corresponding labor force participation rate [9, 11]. These labor force years were then adjusted by productivity indices to derive PALYs [8, 9].
Due to the absence of data distinguishing between full- and part-time employment, all individuals were assumed to be engaged in equivalent full-time (EFT) work. Consistent with previous research [9, 11, 23], we assumed that the economic value of each PALY was equivalent to the annual GDP per worker. Data on per-worker GDP in China from 2010 to 2023 were obtained from the Organization for Economic Co-operation and Development (OECD) Compendium of Productivity Levels [30]. Based on these data, we calculated the EAPC and used it to project GDP per worker from 2024 to 2030 (Shown in Table S2).
Sensitivity and scenario analyses
To evaluate the potential productivity gains from mitigating risk factors, we modelled two intervention scenarios. The first aligned with the targets of the “Healthy China 2030” initiative, which aims to reduce the national smoking rate to 20% by 2030 [31]. The second followed the WHO’s air quality guidelines [32], setting a target annual mean concentration for PM2.5 at 15 µg/m³. In the model, we assumed a linear decline in smoking prevalence and ambient PM2.5 exposure over the 10-year projection period. These reductions in exposure were expected to lead to decreases in both COPD incidence and all-cause mortality. The health benefits resulting from these reductions were quantified using potential impact fractions (PIFs), which were applied to age- and sex-specific incidence and mortality rates to estimate the avoidable disease burden attributable to each intervention [33]. Specifically, PIFs were used to adjust baseline rates by multiplying the original COPD incidence or all-cause mortality rates by (1 − PIF), yielding the expected rates under the reduced exposure scenarios. The difference between the baseline and adjusted rates reflects the proportion of cases or deaths potentially averted due to the intervention. More methodological details were provided in the Supplementary Appendix.
In addition, sensitivity analyses were conducted to explore the impact of key input data on years of life lived, PALYs, and associated costs. First, the annual discount rate was increased to 5%, consistent with the WHO’s standard recommendation [15]. Second, GDP per worker was assumed to remain constant at 2020 levels. Third and fourth, both COPD incidence and mortality rates were held fixed at their 2020 values. Lastly, prevalence estimates were replaced with figures derived from a national population-based survey [3] to evaluate the influence of alternative disease burden assumptions on the outcomes.
Probabilistic sensitivity analyses
To reflect uncertainty in model inputs, we performed 1,000 simulation iterations. In each run, key parameters were randomly sampled from normal distributions defined by the respective standard errors. For parameters where CIs were not available, we assumed a 10% standard error around the mean [34]. From the resulting distributions of estimated life years, PALYs, and costs, we reported the median values along with the 2.5th and 97.5th percentiles as the point estimates and corresponding 95% confidence intervals. All statistical analyses were performed using Stata and R software, with two-sided tests applied throughout.
link
