Ambient air pollution and cause-specific risk of hospital admission in China: A nationwide time-series study

Abstract

Background: The impacts of air pollution on circulatory and respiratory systems have been extensively studied. However, the associations between air pollution exposure and the risk of noncommunicable diseases of other organ systems, including diseases of the digestive, musculoskeletal, and genitourinary systems, remain unclear or inconclusive. We aimed to systematically assess the associations between short-term exposure to main air pollutants (fine particulate matter [PM2.5] and ozone) and cause-specific risk of hospital admission in China over a wide spectrum of human diseases.

Methods and findings: Daily data on hospital admissions for primary diagnosis of 14 major and 188 minor disease categories in 252 Chinese cities (107 cities in North China and 145 cities in South China) from January 1, 2013, to December 31, 2017, were obtained from the Hospital Quality Monitoring System of China (covering 387 hospitals in North China and 614 hospitals in South China). We applied a 2-stage analytic approach to assess the associations between air pollution and daily hospital admissions. City-specific associations were estimated with quasi-Poisson regression models and then pooled by random-effects meta-analyses. Each disease category was analyzed separately, and the P values were adjusted for multiple comparisons. A total of 117,338,867 hospital admissions were recorded in the study period. Overall, 51.7% of the hospitalized cases were male, and 71.3% were aged <65 years. Robust positive associations were found between short-term PM2.5 exposure and hospital admissions for 7 major disease categories: (1) endocrine, nutritional, and metabolic diseases; (2) nervous diseases; (3) circulatory diseases; (4) respiratory diseases; (5) digestive diseases; (6) musculoskeletal and connective tissue diseases; and (7) genitourinary diseases. For example, a 10-μg/m3 increase in PM2.5 was associated with a 0.21% (95% CI 0.15% to 0.27%; adjusted P < 0.001) increase in hospital admissions for diseases of the digestive system on the same day in 2-pollutant models (adjusting for ozone). There were 35 minor disease categories significantly positively associated with same-day PM2.5 in both single- and 2-pollutant models, including diabetes mellitus, anemia, intestinal infection, liver diseases, gastrointestinal hemorrhage, renal failure, urinary tract calculus, chronic ulcer of skin, and back problems. The association between short-term ozone exposure and respiratory diseases was robust. No safety threshold in the exposure-response relationships between PM2.5 and hospital admissions was observed. The main limitations of the present study included the unavailability of data on personal air pollution exposures.

Conclusions: In the Chinese population during 2013-2017, short-term exposure to air pollution, especially PM2.5, was associated with increased risk of hospitalization for diseases of multiple organ systems, including certain diseases of the digestive, musculoskeletal, and genitourinary systems; many of these associations are important but still not fully recognized. The effect estimates and exposure-response relationships can inform policy making aimed at protecting public health from air pollution in China.

Fig 1. Percent change in hospital admissions per 10-μg/m³ increase in PM_2.5 and O₃ by major disease category, on average across all cities.

Results are presented as point estimates and 95% CIs of the percentage increase in daily hospital admissions associated with a 10-μg/m³ increase in PM_2.5 or O₃. Major disease categories are based on the chapter division of the ICD-10 diagnostic coding system. Single-day exposure on the same day (lag 0) was used as the exposure metric of PM_2.5. Two-day moving average exposure (lag 0–1) was used as the exposure metric of O₃. In single-pollutant models, the effects of PM_2.5 and O₃ were estimated without adjustment for co-pollutants; in 2-pollutant models, the effects of PM_2.5 were estimated after adjustment for O₃, and the effects of O₃ were estimated after adjustment for PM_2.5. The Benjamini–Hochberg procedure was applied to adjust the P values across the 14 major disease categories. *Statistically significant estimate (P < 0.05).

Fig 2. Percent change in hospital admissions per 10-μg/m³ increase in PM_2.5 by major disease category for study subgroups, on average across all cities.

(A) Endocrine, nutritional, and metabolic diseases; (B) nervous diseases; (C) circulatory diseases; (D) respiratory diseases; (E) digestive diseases; (F) genitourinary diseases. Results are presented as point estimates and 95% CIs of the percentage increase in daily hospital admissions associated with a 10-μg/m³ increase in PM_2.5. Major disease categories are based on the chapter division of the ICD-10 diagnostic coding system. Single-day exposure on the same day (lag 0) was used as the exposure metric of PM_2.5. The effects of PM_2.5 were estimated after adjustment for O₃. The cool season is from October to March; the warm season is from April to September. The 2 regions of China (North and South) are divided by the Huai River–Qinling Mountain line. The P values were not adjusted for multiple comparisons. Results for all disease categories and for O₃ are shown in S6 and S7 Figs. *Statistically significant estimate (P < 0.05).

Fig 3. Exposure–response curves of the effects of PM_2.5 on hospital admissions by major disease category, on average across all cities.

(A) Endocrine, nutritional, and metabolic diseases; (B) nervous diseases; (C) circulatory diseases; (D) respiratory diseases; (E) digestive diseases; (F) genitourinary diseases. The vertical scale represents the relative change from the mean effect of PM_2.5 on daily hospital admissions, with solid lines indicating point estimates and dashed lines indicating 95% CIs. Major disease categories are based on the chapter division of the ICD-10 diagnostic coding system. Single-day exposure on the same day (lag 0) was used as the exposure metric of PM_2.5. The national average exposure–response curves were developed using multivariate meta-analysis approaches. Results for all disease categories and for O₃ are shown in S8 and S9 Figs.