Our study aimed to comprehensively explore the DALYs of LRIs, URIs, and otitis media attributable to PMP from 1990 to 2021 via GBD 2021 data. The data on URIs and otitis media attributable to PMP from 1990 to 2021 are available only for infants (<1 year) in the GBD 2021 database. DALYs were evaluated by combining years of life lost due to premature death (YLLs) and years lived with disability (YLDs) by location, age, sex, year, and cause. Furthermore, the socio-demographic index (SDI) is employed to assess the sociodemographic development of a country or territory via three components: the fertility rate in females under 25 years, the schooling attainment of individuals over 15 years, and lag-distributed income per capita. SDI values are categorized into five quintiles: low, low-middle, middle, high-middle, and high (16).

The GBD study defines LRIs as pneumonia or bronchiolitis. The diagnostic codes for LRIs include 079.82, 466–469, 470.0, 480–481.9, 482.0–482.89, 483.0–483.9, 484.1–484.2, 484.6–484.7, 487–490.9, 510–511.9, and 513.0–513.9 in the ICD-9 diagnostics criteria, and A48.1, A70, B96.0–96.1, B97.21, B97.4–B97.6, J09–J11.89, J12–J13.9, J14–J14.0, J15–J15.8, J20–J21.9, J85.1, J91.0, P23.0–P23.4, and U04–U04.9 in the ICD-10 diagnostic criteria, excluding tuberculosis or COVID-19.¹ The GBD study defines URIs as conditions encompassing cough, acute nasopharyngitis, sinusitis, pharyngitis, tonsillitis, laryngitis/tracheitis, epiglottitis, rhinitis, rhinosinusitis, rhinopharyngitis, supraglottitis, and the common cold. For the classification of URIs, the relevant ICD-10 codes are J00-J02, J02.8-J03, J03.8-J06.9, and J36-J36.0, while ICD-9 codes include 460–465.9, 476.9, and 475–475.9.² Otitis media refers to an infection of the middle ear, including acute otitis media, and chronic otitis media. The relevant ICD-10 codes are H65-H75.83, and ICD-9 codes are 381–384.9.³ PMP encompasses both ambient particulate matter pollution and household air pollution. Ambient particulate matter pollution, is calculated from the population-weighted annual mean concentration of airborne PM₂.₅ (μg/m³). Household air pollution is defined based on the proportion of individuals utilizing solid fuel for cooking and the corresponding PM₂.₅ exposure levels.⁴

Frontier analysis delineates the nonlinear relationship between the SDI and disease burden across countries or territories by quantifying the disparity between the current burden and its theoretical minimum. LOESS (Locally Estimated Regression) is a nonparametric smoothing technique that fits local polynomial regressions to subsets of data in order to model variable relationships (17, 18). We applied this method to explore the nonlinear association between SDI and age-standardized DALY rate (ASDR) (19). To ensure robustness, 500 bootstrap samples were performed to estimate the average age-standardized rate (ASR) for each SDI value. The absolute distance between each country’s or territory’s 2021 ASDR and the frontier line was calculated to assess the potential for improvement (20).

We utilized the decomposition method developed by Das Gupta to systematically examine the contributions of population growth, aging, and epidemiological changes (epidemiologic changes is age-standardized DALY rates) to the PMP-attributable burden. This approach utilizes mathematical techniques to disaggregate overall changes into distinct components, enabling the quantification of each factor’s specific contribution. We report shares (%) as each component’s ΔDALYs divided by the net change in DALYs. Consequently, values may exceed 100% when one component overcompensates the others; negative values indicate opposition to the net change.

In this study, we employed absolute and relative inequality—the slope index of inequality (SII) and the concentration index—to assess the distributional inequality in the PMP-attributable burden of LRIs, and of URIs and otitis media among infants across countries or territories (21). The SII, representing absolute inequality, was computed by regressing national DALY rates for the total population on a relative position scale associated with sociodemographic development. The concentration index (CI), reflecting relative inequality, was derived through numerical integration of the area under the Lorenz curve, which was generated using the cumulative DALY and the corresponding cumulative population distribution ranked by the SDI (22). A negative value of the SII or concentration index indicates that a higher SDI is associated with a lower ASDR, and vice versa.

The autoregressive integrated moving average (ARIMA) model, a time series analysis technique, was employed to forecast the future burden in the next decades. This model integrates autoregression (AR), differencing (I), and moving average (MA) components to forecast from historical data (23), which is less prone to overfitting than higher-complexity alternatives. In the ARIMA (p, d, q) model, “p” represents the number of autoregressive terms, “d” indicates the order of differencing, and “q” defines moving average terms. The optimal ARIMA model parameters (p, d, q) were selected using the auto.arima() function from the forecast package in R. Model fit was evaluated by assessing the consistency between the observed and predicted values. Residual diagnostics in Supplementary Figures S1–S3, including Q-Q plots, as well as autocorrelation function (ACF) and partial autocorrelation function (PACF) plots, were used to assess whether the residuals followed a normal distribution. The Ljung-Box test was conducted to evaluate the robustness of the residuals, confirming that the model was with randomness (white noise). The ACF and the PACF were used to guide identification of appropriate parameters (p and q). The parameter d involves the difference calculation between adjacent observations to achieve stationarity. We selected the best ARIMA (p, d, q) models with the lowest Akaike information criterion (AIC)/Bayesian information criterion (BIC) (Supplementary Table S4) and satisfactory residual diagnostics to predict the disease burden (24).

The Bayesian age-period-cohort (BAPC) model, incorporating age, period, and cohort effects, is used to project trends in LRIs attributable to PMP from 2022 to 2046. It applies generalized linear models to explore associations, and employs a Bayesian framework integrating prior sample data and information for parameter estimation (25).

The age-standardized rate (ASR) was weighted by the corresponding proportions of the GBD standard population, and the weighted rates were then summed to obtain the ASR per 100,000 population (26). The estimates and their 95% uncertainty intervals (UIs) for DALYs were obtained from GBD 2021. Estimated annual percentage change (EAPC) is a more precise parameter for evaluating trends in age-standardized DALY rate (ASDR) over time. The EAPC was estimated via the formula 100 * (e^β − 1) and reported with a 95% confidence interval (95% CI), where β denotes the slope derived from the log-linear regression model. The ASDR was considered to have increased if both the EAPC estimates and the lower bound of the 95% CI were greater than 0. Conversely, a decrease in ASDR was considered if both the EAPC estimates and the upper bound of the 95% CI were less than 0. Otherwise, the ASDR was considered stable over time (27). R software (version 4.4.0) was used for database organization, and analysis.

Between 1990 and 2021, PMP-attributable ASDRs of LRIs, infant URIs, and infant otitis media decreased globally. For LRIs, the number of PMP-attributable DALYs declined from 87,315,640 (95% UI: 20,780,746–139,682,971) to 29,098,331 (95% UI: 6,988,265–48,127,683), and the ASDR decreased from 1462.2 (95% UI: 340.59–2338.31) per 100,000 to 420.09 (95% UI: 106.35–693.12) per 100,000 globally (Supplementary Table S1). For infant URIs, the number of PMP-attributable DALYs saw a decrease from 23781.75 (95% UI: 6477.74–48147.78) to 11708.46 (95% UI: 2800.5–28081.79), and ASDR decreased from 0.37 (95% UI: 0.1–0.75) per 100,000 to 0.19 (95% UI: 0.05–0.45) per 100,000 globally (Supplementary Table S2). For infant otitis media, the number of PMP-attributable DALYs declined from 1363.14 (95% UI: 512.18–2946.84) to 327.40 (95% UI: 156.27–626.36), and ASDR decreased from 0.02 (95% UI: 0.01–0.05) per 100,000 to 0.01 (95% UI: 0–0.01) per 100,000 globally (Supplementary Table S3). The EAPCs for the ASDRs from 1990 to 2021 were −3.76 (95% CI: −3.96 to 3.55) for PMP-attributable LRIs, −2.32 (95% CI: −2.39 to 2.25) for PMP-attributable URIs in infants, and −4.54 (95% CI: −4.8 to 4.28) for PMP-attributable otitis media in infants (Supplementary Tables S1–S3). Overall, global trends in the PMP-attributable burden of LRIs, infant URIs, and infant otitis media are similar, showing declines in both numbers of DALYs and ASDRs.

The age patterns of global DALYs for LRIs attributable to PMP are displayed in Table 1, and Figure 1A. The age group under 5 years and above 95 years presented the highest ASDR of PMP-attributable LRIs. From 1990 to 2021, all age groups displayed reductions in the ASDRs of LRIs attributable to PMP, while infants showed declining ASDRs for URIs and otitis media attributable to PMP (Figures 1A–C). From 1990 to 2021, both males and females presented a decreasing trend in the PMP-attributable ASDRs of LRIs, infant URIs, and infant otitis media (Figures 1D–F).

From 1990 to 2021, the PMP-attributable ASDRs for LRIs, infant URIs, and infant otitis media decreased across different SDI quintiles, and the low-SDI region presented the highest ASDR over time (Figures 1G–I). Regionally, Western Sub-Saharan Africa, Eastern Sub-Saharan Africa, and Central Sub-Saharan Africa had higher ASDRs in 2021 for LRIs, infant URIs, and infant otitis media, all attributable to PMP (Supplementary Tables S1–S3). The largest decrease in the PMP-attributable burden of LRIs, and of infant URIs occurred in East Asia (EAPC of LRIs = −9.13 (95% CI: −9.45 to 8.81), EAPC of URIs = −10.99 (−11.53 to 10.44), Supplementary Tables S1, S2). The largest decrease in the PMP-attributable burden of infant otitis media occurred in Central Europe [EAPC = −11.27 (−13.62 to 8.85), Supplementary Table S3].

With respect to countries or territories, Central African Republic, Chad, and South Sudan showed the highest ASDRs for LRIs attributable to PMP in 2021 (Figure 2A). Tajikistan, Somalia, Oman, and Central African Republic showed the highest ASDRs for infant URIs attributable to PMP in 2021 (Figure 2B). South Sudan, Somalia, and Madagascar showed the highest ASDRs for infant otitis media attributable to PMP in 2021 (Figure 2C). From 1990 to 2021, the largest decrease in the burden of LRIs attributable to PMP occurred in Finland, China, and Ireland, whereas the largest increase occurred in Argentina, Kuwait, and Lesotho (Figure 2D). From 1990 to 2021, the largest decrease in the DALYs of infant URIs attributable to PMP occurred in China, Latvia, and Estonia, while the largest increase occurred in Afghanistan, Qatar, and Kuwait (Figure 2E). From 1990 to 2021, the largest decrease in the DALYs of infant otitis media attributable to PMP occurred in Poland, Albania, and Bosnia and Herzegovina, while the largest increase occurred in Saudi Arabia, and Kuwait (Figure 2F).

A frontier analysis was performed to evaluate the potential for improvement in the PMP-attributable ASDRs for LRIs, infant URIs, and infant otitis media, accounting for national and regional SDI levels. The PMP-attributable ASDRs for LRIs, infant URIs, and infant otitis media tended to decrease with increasing SDI (Figure 3). The countries or territories with the largest effective distance in ASDR for PMP-attributable LRIs, characterized by disproportionately high ASDRs relative to their sociodemographic resources, included the Central African Republic, Zimbabwe, Lesotho, and Eritrea (Figures 3A,B). For PMP-attributable URIs in infants, countries or territories such as Tajikistan, Oman, Central African Republic, and Guinea demonstrated the largest effective distances (Figures 3C,D). Similarly, for PMP-attributable otitis media in infants, the countries or territories with the largest effective distances included South Sudan, Madagascar, Malawi, and Burundi (Figures 3E,F).

Significant absolute and relative inequalities in the burden of LRIs attributable to PMP were observed across countries or territories with varying SDIs, with lower SDI-associated countries and territories disproportionately shouldering a greater burden. The slope index of inequality (SII) revealed that absolute inequalities in ASDRs between the highest and lowest SDI-associated countries or territories. The SII decreased from −1679.45 (95% CI: −2701.13 to −657.77) in 1990 to −780.85 (95% CI: −1037.21 to −524.48) in 2021 (Figure 4A). Similarly, the concentration index was 0.06 (95% CI: 0.00 to 0.11) in 1990 and decreased to 0.01 (95% CI: −0.06 to 0.08) by 2021. These findings revealed that a decrease in both absolute and relative inequalities from 1990 to 2021 (Figure 4B).

For PMP-attributable URIs in infants, the SII decreased from −0.33 (95% CI: −0.66 to 0.00) in 1990 to −0.28 (95% CI: −0.46 to −0.1, p < 0.05) in 2021 (Figure 4C). Similarly, the concentration index declined from 0.08 (95% CI: 0.01 to 0.15) in 1990 to 0.02 (95% CI: −0.08 to 0.12) in 2021, reflecting a reduction in both absolute and relative inequalities (Figure 4D). For PMP-attributable otitis media in infants, the SII declined from −0.02 (95% CI: −0.04 to 0.00) in 1990 to −0.01 (95% CI: −0.01 to −0.00, p < 0.05) in 2021, indicating a decline in absolute inequalities (Figure 4E). However, the concentration index increased from 0.00 (95% CI: −0.01 to 0.10) in 1990 to 0.04 (95% CI: −0.06 to 0.13) in 2021 (Figure 4F), suggesting an increase in relative inequalities over time.

Our study employed decomposition analysis to evaluate the PMP-attributable DALYs of LRIs, infant URIs, and infant otitis media, from 1990 to 2021. This study focused on quantifying the contributions of factors such as aging, population growth, and epidemiological changes to the trends observed in DALYs. Globally, the DALYs of LRIs attributable to PMP demonstrated a declining trend across all SDI regions, with the most pronounced reduction observed in low-middle-SDI regions. Aging and epidemiological changes accounted for 22.53 and 119.37%, respectively, of the global decline in the burden of LRIs attributable to PMP. The impact of population growth exhibited considerable variation across the SDI region, contributing to burden increases of 140.96% in low-SDI regions, 51.89% in low-middle-SDI regions, 27.37% in middle-SDI regions, 15.45% in high-middle SDI regions, and 46.32% in high-SDI regions. The contribution of aging was more substantial in high SDI regions, with specific contributions of 75.48% in high-SDI regions, compared with −28.23% in low-SDI regions, −28.46% in low-middle SDI regions, −18.43% in middle-SDI regions, and −10.84% in high-middle SDI regions. Epidemiological changes significantly reduce the disease burden of PMP-related LRIs globally, particularly in high-SDI regions (−221.8%) (Figure 5A; Table 1).

The PMP-attributable DALYs for URIs in infants also showed a global decline across all SDI regions, with the largest reduction observed in middle-SDI regions. Population growth contributed to a 8.89% increase in the burden, whereas epidemiological changes led to a 108.89% decrease. The impact of population growth was especially prominent in low-SDI regions, contributing 594.78% to the increase in the burden. Epidemiological changes were the predominant factor in reducing the global burden of infant URIs due to PMP, with a particularly high contribution in low-SDI regions (694.78%) (Figure 5B; Table 1).

The PMP-attributable burden of otitis media in infants also largely driven by epidemiological changes, which contributed to 104.98% of the global reduction. However, population growth increased the disease burden, particularly in low-SDI regions (76.11%) and low-middle-SDI regions (6.8%). Notably, epidemiological changes played a decreasing role in the burden globally, especially in low-SDI regions (−176.11%) (Figure 5C; Table 1).

Overall, epidemiological changes emerged as significant contributors to the global reduction in the PMP-attributable burden of LRIs, infant URIs, and infant otitis media, while population growth generally increased the burden, particularly in low-SDI regions. Aging exhibited a variable impact on the burden of LRIs attributable to PMP, contributing to an increased burden in high-SDI regions while reducing the burden in lower-SDI regions.

We employed the ARIMA model to predict the PMP-attributable burden of LRIs, infant URIs, and infant otitis media for the period 2022 to 2031. The Ljung–Box test p-values, ARIMA model parameters (p, d, q), and corresponding AIC and BIC values are presented in Supplementary Table S4. Using the ARIMA prediction model, the ASDR of LRIs attributable to PMP in the total population is predicted to decrease from 386.47 per 100,000 in 2022 to 83.92 per 100,000 in 2031. Among males, the ASDR is projected to decline from 413.75 per 100,000 to 120.22 per 100,000, and among females, it is projected to decline from 361.72 per 100,000 to 47.92 per 100,000 over the same period (Figure 6A). Using the BAPC prediction model, we projected the global trends in LRIs attributable to PMP from 2022 to 2046. The ASDR is predicted to decline from 345.69 per 100,000 to 63.74 per 100,000, with consistent decreases across all age groups (Supplementary Figures S4, S5). The similarity of projections from both BAPC and ARIMA models supports the robustness of our predictions.

For PMP-attributable URIs in infants, the ASDR of the total population is forecast to decline slightly, from 0.19 per 100,000 in 2022 to 0.16 per 100,000 in 2031 based on the ARIMA model. Among males, the ASDR is predicted to increase marginally, from 0.17 per 100,000 to 0.18 per 100,000, while for females, it is anticipated to decrease slightly from 0.20 per 100,000 to 0.15 per 100,000 (Figure 6B). In the case of PMP-attributable otitis media in infants, the ASDR of the total population is predicted to increase slightly, from 0.005 per 100,000 in 2022 to 0.006 per 100,000 in 2031 based on the ARIMA model. For males, the ASDR is expected to decrease from 0.006 per 100,000 to 0.004 per 100,000, whereas for females, it is projected to increase slightly from 0.004 per 100,000 to 0.006 per 100,000 (Figure 6C).