All data used in this study were obtained from the Global Burden of Disease (GBD) 2021 database produced by the Institute for Health Metrics and Evaluation (IHME),¹ which details the data, statistical modeling, and methodology used for inclusion in the study. The GBD database, based on age and gender, includes 204 countries. The GBD database contains 371 diseases and injuries from 204 countries and territories, quantifying and estimating health losses from 1990 to the present by age and sex. The data are standardized to multiple correlates for each disease or injury separately. Censuses, household surveys, civil registration and vital statistics, disease registries, healthcare utilization, air pollution monitoring, satellite imagery, and disease notification from 86,249 sources were used as data sources (16). In addition, we used the Query Exchange tool to obtain data on the incidence, prevalence, and disability-adjusted life years (DALYs) of hearing loss from the GBD database between 1990 and 2021. DisMod-MR is a Bayesian regression algorithm used to assess overall incidence, prevalence, and other metrics in the GBD database in 2021 (17). Specifically, a dataset on hearing loss among people aged 60 years and older from 1990 to 2021 was obtained from the GBD 2021 database.

GBD 2021 uses standardized tools to model processed data for most diseases and injuries, disaggregated by age, sex, location, and year of study. It provides a range of interrelated indicators to assess the burden of disease, including incidence, prevalence, mortality, and years of life lost (YLLs), as well as composite health indicators such as years of life lived with disability (YLDs) and disability-adjusted life years (DALYs). DALYs are widely used as a composite measure of health status and are a useful tool for comparing health disparities across health statuses and populations. DALYs are the total number of years lost due to illness, disability, or premature death and are calculated as the sum of DALYs and DALYs (18).

The SDI we use is a precise indicator that serves as a composite indicator and a composite index that captures national social development from the geometric mean of the following: the total fertility rate for those under 25 years of age, the lagged distribution of per capita income, and the average years of schooling for the population 15 years of age and older, which is then materialized into individual scores on a scale of 0 to 1.1 We use the SDI as an indicator of social development in the countries and regions we include. We artificially categorized the included countries and regions according to five SDI thresholds: high SDI (>0.81), medium-high SDI (0.70–0.81), medium SDI (0.61–0.69), medium-low SDI (0.46–0.60), and low SDI (<0.46), and estimated all the values according to the uncertainty intervals (UIs), which are the values that we The 2nd-5th percentile and 97th-5th percentile of the sample determined the 95% uncertainty intervals of the estimates to derive the distribution values (19). All analyses were conducted using R (version 4.2.2) and adhered to the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) to ensure scientific rigor and methodological transparency. According to the established inclusion and exclusion criteria (Supplementary file 1), data on hearing loss among individuals aged 60 years or older were extracted from the GBD database. The detailed data extraction process is shown in Supplementary Figure 1.

We used descriptive analyses to explore spatial and temporal trends in disease and comprehensively analyze and compare the burden of HL across different dimensions, including global HL incidence from 1990 to 2021. Both metrics were assessed by the number of cases, age-standardized or age-adjusted rates (ASRs), and estimated annual percentage change (EAPC) for males, females, and all genders between 1990 and 2021. The Estimated Annual Percentage Change (EAPC) indicates trends in burden at different scales over a given period of time and is a practical way to quantify overall and localized trends in epidemiologic outcomes. In addition, we compared case counts, ASRs, and EAPCs on HL incidence, prevalence, and disability-adjusted life years (DALYs) at the global, regional, and national levels. At the global level, we examined trends in the incidence and prevalence of HL among populations aged 60 years and older, by sex and age. To control for age structure, we calculated age-standardized rates. At the regional level, we analyzed differences in the burden of HL among people aged 60 years or older in several different geographic regions. We explored the relationship between these differences and SDI. At the national level, we compared the burden of HL among people aged 60 years or older across 204 countries and regions, focusing on those in the high, medium-high, medium, medium-low, and low SDI quintiles, and assessed the burden of HL in these countries and regions.

We assessed global trends in the dynamics of HL deaths, disability-adjusted life years (DALYs), and annual deaths over age 60 between 1990 and 2021. We explored trends in the burden of HL disease from the overall level to different regional levels. We used the Estimated Annual Percentage Change (EAPC) to illustrate global or regional trends in burden over a given period of time. EAPC’s assessment of ASR trends is a more reliable indicator for monitoring changes in the burden of disease because it is calculated using a linear regression model (20). The linear regression model is expressed as y = α + βx, where y = ln(ASR) and x = calendar year. The EAPC is calculated as (exp(β) − 1)×100%, and its 95% confidence interval (CI) is also derived from the model (21). The natural logarithm of the regression model was fitted through the time variable, and the natural logarithm of each observation was fitted to a straight line. The slope of the straight line is used to estimate the annual rate of change for a given period. Specifically, positive and negative values of EAPC, along with their 95% confidence intervals, indicate an upward or downward trend in disease burden.

Joinpoint regression software version 4.9.1, developed by the National Cancer Institute, was used to analyze local trends in the burden of HL disease in people aged 60 years and older. Detailed methods for Joinpoint regression can be found on the software’s website² and in the Li et al. (22). We used the Joinpoint regression model to divide the temporal trends in the burden of disease from 1990 to 2021 into consecutive, meaningful time periods. The analysis included annual percentage change (APC) and average annual percentage change (AAPC), along with corresponding 95% confidence intervals (CIs) to represent year-to-year (APC) and average (AAPC) trends in disease burden over a given time period (23). Specifically, positive or negative APC and AAPC values (p < 0.05) and their corresponding 95% confidence intervals indicate an increasing or decreasing trend in the burden of disease over a given period of time, respectively, whereas values close to zero indicate stabilization.

The contribution of age, period, and cohort effects to outcomes is of great significance to epidemiologic conclusions, but traditional methods have failed to overcome several shortcomings, including the inability to eliminate covariance between factors. Therefore, eigenestimation (IE) was then used to assess the effects of age, period, and cohort separately. The age-period-cohort model is a linear statistical method that can be used to demonstrate and analyze disease burden information, bypassing the previous method, which involved a multiclass model. Specifically, it can be illustrated with the following sentence: In (Refg) = α + Ae + Pf + Cg, where “Refg” denotes the incidence or mortality of PF in g birth cohorts, “e” refers to the age group, “f” stands for the period, and “Ae,” “Pf,” and “Cg” stand for the effects of age, period, and cohort, respectively.

Age, period, and cohort refer to the outcomes of population aging, objective and temporal variations in disease prevalence, and variations in outcomes among participants in the same birth cohort, respectively. Age, period, and cohort relative risks (RR) represent the relative risk of each age/period/cohort relative to the reference age/period/cohort, controlling for the other two dimensions. The RR is statistically significant when compared with 1. The RR is not statistically significant when compared with 1. For precise estimation, the data were recoded into consecutive 5-year age cohorts spanning the 5-year periods from 1990 through 2021, along with the corresponding 5-year birth cohorts. To determine the relative risk (RR) of a particular age, period, or birth cohort relative to the average portfolio level, we calculated these coefficients and aggregated them. This approach allows us to visualize the impact of different time dimensions on disease risk and provides a scientific basis for targeted prevention strategies. The systematic and comprehensive analysis provided by the age-period-cohort model provides stronger theoretical support for epidemiological studies, enabling us to accurately identify and address public health challenges.

We used the decomposition analysis proposed by Das Gupta et al. (23), combined with the improved scheme developed by Cheng et al. (24), to examine the contributions to HL disability-adjusted life-years (DALYs) among individuals aged 60 years and older over the past three decades, considering population growth, population aging, and epidemiological changes. The contribution of changes in DALYs, py, ey = (ai,y * py * ei,y) for each region was calculated using the formula DALYs, py, ey = (ai,y * py * ei,y), following the method proposed by Xie et al. (25). The relative contribution of each variable to changes in DALYs in MAS populations was quantified by isolating the standardized effects of each multiplying factor. In addition, we decomposed DALY changes by gender and SDI groups to identify specific drivers of DALY changes in HL among people aged 60 years and older from 1990 to 2021.

We used the Slope of Inequality Index (SII) and the Concentration Index (CI), which are able to show cross-national inequalities in HL burdens among people aged 60 years or older and are standardized metrics to quantify inequalities in HL burdens in many regions, allowing for a better narrowing of differences in the distribution of health and for further improvements in related policies, programs, and practices, as well as a comprehensive assessment of health inequalities. In this study, we compared data from 204 countries and territories over the period 1990 to 2021.

SII is a measure that uses regression modeling to represent the absolute difference in burden projections between those with the highest and lowest levels of SDI. Differences in regression-based estimates of health outcomes allow for the overall distribution of socio-economic factors to be taken into account. Positive/negative values of SII indicate the burden concentration in countries with higher/lower SDI levels. Absolute inequalities in health indicators can be quantified; the higher the absolute value, the greater the inequality.

The CI is a relative measure of inequality that quantifies the extent to which disadvantaged groups with low SDI burdens or advantaged groups with high SDI burdens are affected, and it concentrates the SDI to show a gradient in health across multiple subgroups in a natural ordering. It is derived by numerical integration under the Lorenz concentration curve, which is fitted to the cumulative percentage of DALYs relative to the population’s cumulative distribution by SDI (26). The health burden in low SDI countries is concentrated on the equality line under the Lorenz curve with a positive CI (27).

In this study, a Bayesian age-period-cohort (BAPC) modeling approach was used to further predict the burden of HL in terms of ASR, prevalence, and YLDs for people aged 60 years or older in the next 20 years, aiming to provide a scientific and reliable basis for decision-making by health organizations and researchers in response to future public health challenges. By accurately calculating marginal posterior distributions using the INLA method, the approach avoids the mixing and convergence issues associated with Markov chain Monte Carlo (MCMC) sampling, thereby improving the stability and efficiency of predictions (28). The theoretical basis and specific applications of various statistical analysis methods are detailed in Supplementary Table 1.

Globally, between 1990 and 2021, the absolute number and age-standardized rates (ASRs) of disability among individuals aged 60 and above due to hearing loss (HL) have both significantly increased, with the estimated annual percentage change (EAPC) for disabilities caused by hearing loss slightly surpassing the prevalence (Tables 1, 2).

From a regional perspective, we observed significant differences in the burden indicators across social demographic index (SDI) quintiles and geographical areas in 2021 (Figure 1). In terms of the absolute number of cases, the burden is highest in the median SDI quintile and lowest in the low SDI quintile. In contrast, when age-standardized rates are considered, the prevalence in high SDI quintiles shows the highest age-standardized rate (ASR), while the prevalence in low SDI quintiles is the lowest. The incidence rate among the top fifth of the population with high SDI has shown the largest increase over time (measured by EAPC), while the incidence rate among the bottom fifth of the population with low SDI has increased the least.

Geographically, East Asia has consistently had the highest absolute numbers of cases and YLDs. In contrast, Western Europe has the lowest age-standardized incidence rate (ASR), while Eastern sub-Saharan Africa has the highest YLD ASR. South Asia (incidence) and East Asia (YLDs) show the most significant upward trends in ASRs, while Western Europe and southern Latin America have shown stable or slightly declining trends during the study period.

Nationally, the patterns reflected these regional disparities. China recorded the highest absolute number of prevalent cases and YLDs, while Kenya had the highest ASR for YLDs. At the other end of the spectrum, countries such as Sweden and the United Kingdom were among the lowest ASR countries. A comprehensive listing of national-level data for all 204 countries and territories is provided in Supplementary Tables S2, S3.

The results in Figure 2 show that the overall trend from 1990 to 2021 indicates a continuous increase in the burden of older adults HL across all indicators, with distinct demographic characteristics. The absolute number of cases has grown most significantly, primarily influenced by global population growth and aging. The relationship between national SDI and HL burden presents a complex non-linear pattern. Overall, as SDI increases, the ASR of YLD declines. However, many regions significantly deviate from this general pattern. For example, East Asia, South Asia, and East Sub-Saharan Africa lie above the smoothed curve, indicating that their burden is higher than what is expected based solely on their SDI levels. In contrast, Western Europe is well below the expected value, indicating that its burden is lower than anticipated. This underscores that, in addition to overall socio-economic development, specific factors such as environmental exposures or healthcare policies also play a crucial role in determining HL burden.

In addition, it is worth mentioning that gender differences are evident across all indicators. Throughout the study period, the crude and age-standardized incidence rates (ASRs) of men consistently exceeded those of females in terms of prevalence and disease burden (YLDs). However, due to the generally longer lifespan of females, the absolute number of cases is highest among the women aged 60 and above group, followed by males, and then the combined group of both sexes (‘bBoth’) (Supplementary Figure 2).

The Joinpoint regression analysis revealed specific temporal inflection points in the trend of HL burden, which are not evident in the overall EAPC (Figure 3A). In terms of the age-standardized prevalence rate (ASR), the entire period shows a steady increase (AAPC = 0.13). However, this trend is not consistent: there was a particularly rapid increase from 2014 to 2018 (APC = 0.28).

The trend of YLDs’ ASR is more dynamic, resembling a ‘W’-shaped curve. Significant declines occurred from 1990 to 1994 (APC = −0.24) and from 2005 to 2011 (APC = −0.04), while the fastest growth occurred from 2000 to 2005 (APC = 0.45). This non-linear trajectory indicates that changes in YLDs are influenced by the interaction of various complex factors, including changes over time in disability weights, diagnostic practices, and treatment availability (Figure 3B).

Age-period-cohort (APC) analysis shows that age, calendar period, and birth cohort have independent effects on the risk of hearing loss (HL) (Figure 4 and Table 3). Unsurprisingly, the impact of age is greatest, with relative risks (RRs) increasing monotonically with age. The rate of increase slows after age 85. The period effect shows that from 1990 to 2021, RRs have gradually but steadily increased, indicating that the influencing factors affect all age groups simultaneously, such as improvements in detection and diagnosis or changes in environmental risk factors. In contrast, a significant birth-cohort effect was observed; after controlling for age and period, the risk of HL decreased across successive birth cohorts born after the early 20th century. This suggests that improvements in childcare and nutrition, or a reduction in early life exposure to ototoxic factors, may have a protective effect on later-born populations. The age-stratified prevalence RRs for hearing loss by gender are shown in Supplementary Table 4. For details of the age-period-cohort analysis model, see Supplementary Figure 3.

Globally, over a 30-year research period, the total increase in YLDs caused by hearing loss was primarily driven by population growth (90.25%) and aging of the population (4.64%), while changes in epidemiological incidence accounted for only a small proportion (5.12%) (Figure 5 and Supplementary Table 5). However, this pattern varies significantly by SDI. In high- and upper-middle SDI regions, the epidemiological change component is negative, indicating that the underlying risk of HL has increased (decreased) over time, which is offset by demographic factors. In low- and lower-middle SDI regions, the impact of population growth is greatest. Notably, in upper-middle SDI quintile regions, the impact of epidemiological changes (despite being negative) is proportionally greater than in other quintile regions, suggesting that these areas are in a transitional phase in which public health interventions may already be positively affecting HL risk factors. The breakdown analysis by sex reflects these overall patterns. The changes in years lived with disability due to hearing loss among older adults are detailed in Supplementary Table 5.

Inequality analysis confirms that a significant transnational socioeconomic gradient in HL burden persisted throughout the study period, although there was a trend of narrowing (Figure 6). Between 1990 and 2021, the inequality slope index (SII) and concentration index (CI) of crude YLD rates and prevalence were both negative. This indicates that the burden of HL has been concentrated in countries with lower SDI. However, the degree of inequality has decreased over time. For instance, the SII of the crude YLD rate dropped from −562.06 per 100,000 (95% UI: −706.34, −417.79) in 1990 to −354.79 (95% UI: −496.07, −213.51) in 2021. This convergence suggests that, while the burden in low-income countries remains disproportionate, the gap between high- and low-SDI countries has been narrowing, which may be attributed to economic development and improvements in healthcare services across many regions of the world.

According to the Bayesian age-period-cohort model, the burden of older adults HL is expected to continue to rise until 2040 (Figure 7). This is primarily due to the anticipated global population growth and aging, with the absolute number of prevalent cases and YLD expected to surge dramatically. For instance, the number of YLD is projected to increase from 27.5 million in 2022 to 46.9 million in 2040. In contrast, age-standardized rates (ASRs) are expected to remain relatively stable, with only a slight increase. This disparity highlights that future challenges in managing HL will be driven more by changes in population structure than by worsening age-specific risks. Supplementary Table 6 provides detailed predictions for the data above.