Air pollution is one of the most important public health problems around the world (1). Sulfur dioxide (SO₂), mainly comes from various industrial processes, transportation, and vehicles, power plants, and fuel (coal) combustion, is one of the most common and irritant air pollutants in developing countries and industrialized areas (2–5). Several epidemiological studies have revealed that SO₂ exposure is related to human respiratory health (6, 7), including stimulating the respiratory tract (8), leading to the decline of pulmonary function (9, 10), and the increased mortality due to respiratory diseases (11, 12). However, there are few studies on the relationship between SO₂ and respiratory morbidity in developing countries. Inpatient visit is an important indicator of morbidity and has been widely used to assess the adverse effects associated with atmospheric pollutants (13). Therefore, it will be helpful to understand the impact of SO₂ on the respiratory system by evaluating the relationship between ambient air SO₂ and the number of hospitalized cases of the respiratory system.

Ganzhou is located in the southern part of Jiangxi Province. The ambient temperature and lifestyle of Ganzhou are typical of southern China. In 2019, the mean concentrations of PM₁₀, SO₂, NO₂, O₃, and CO in Ganzhou were 51, 12, 22, 74, and 1.2 mg/m³, respectively. Except for SO₂, the concentrations of other air pollutants in Ganzhou were lower than the mean concentration of 337 Chinese cities (63 μg/m³ for PM₁₀, 11 μg/m³ for SO₂, 27 μg/m³ for NO₂, 148 μg/m³ for O₃, and 1.4 mg/m³ for CO, respectively) (14). Compared with lower levels of other air pollutants, the concentration of SO₂ in Ganzhou is higher than that of the national average in China, drawing more attention to the potential respiratory health damage caused by SO₂.

In this study, the daily average concentrations of air pollutants, including SO₂ from 2017 to 2019, were collected from the local Environmental Protection Bureau. The respiratory inpatient visits were recorded during the same period in the biggest hospital in Ganzhou. We analyzed the association between the level of SO2 exposure and the daily respiratory inpatient visit, investigated populations more sensitive to SO₂ exposure, assessed the potential impact of seasonal changes on SO₂ lagging effects, and explored the exposure-response relationship between SO₂ concentrations and different population subgroups.

Table 1 presents descriptive results for ambient air pollutant concentrations, meteorological parameters, and daily respiratory inpatient visits. During the 1,095 days from January 1, 2017, to December 31, 2019, the average daily concentrations (standard deviation) of atmospheric pollutants were 62.7 (34.6) μg/m³ for PM₁₀, 18.5 (11) μg/m³ for SO₂, 24.2 (13) μg/m³ for NO₂, 71.7 (34.5) μg/m³ for O₃, and 1.3 (0.3) mg/m³ for CO, respectively. Moreover, the daily mean relative humidity and temperature were 74% and 19.6°C, respectively. A total of 9,668 respiratory inpatient visits were recorded from 2017 to 2019; the daily mean count of inpatient visits was 9. Women accounted for ∽34.1% of all the cases, and younger people (<65 years) accounted for ∽48.1% of all the cases.

The correlations between meteorological parameters and atmospheric pollutants are shown in Table 2. All the correlation coefficients between the meteorological factors and the atmospheric pollutants were statistically significant. The daily concentration of SO₂ was positively correlated with other air pollutants (PM₁₀, NO₂, O₃, and CO) and temperature (correlation coefficient = 0.07–0.71) and negatively correlated with relative humidity (correlation coefficient = −0.41). Moreover, PM₁₀ and NO₂ had positive associations with other air pollutants (correlation coefficient = 0.12–0.71), while there was a negative correlation between CO and O₃ (correlation coefficient = −0.16). The temperature was also negatively associated with relative humidity (correlation coefficient = −0.31).

Table 3 shows the percent changes in daily respiratory inpatient visits associated with a 10 μg/m³ increase in SO₂ on different lag days in different gender and age subgroups. The delayed effects of SO₂ were significant at lag3, lag4, lag0–3, lag0–4, lag0–5, lag0–6, and lag0–7, with the maximum effect observed at lag4 (PC: 3.4%; 95% CI: 0.4–6.4%) in single-lag days. The maximum effect of multi-lag days was at lag 0:5 with 6.9% (95% CI: 1.6–12.5%). For men, the association between SO₂ exposure and respiratory admission was not statistically significant at any lag days. The SO₂ concentration was significantly correlated with women at lag1, lag4, and all the multi-lag days, with the largest effect observed at lag1 (PC: 7.1%; 95% CI: 1.9–12.6%) in the single-lag day model. For younger people (<65 years), there is no statistically significant relationship between SO₂ and inpatient visits for respiratory diseases at any lag days. For the elderly (≥65 years), daily respiratory inpatient visits were significantly associated with SO₂ concentration at lag1, lag4, and all the multi-lag days except lag0:7; the strongest association was observed at lag1 (PC: 5.5%; 95% CI: 1–10.2%) in single-lag day model.

Figure 1 presents a positive exposure-response relationship between the SO₂ concentration and the relative risk of respiratory inpatient visits. The exposure-response curves of the total population, men, women, and the elderly (≥65 years) were raised relatively faster in the range of 0–20 μg/m³, which meant that the risk of respiratory hospitalization increases rapidly under relatively low concentrations of SO₂ exposure; the curves in the range of 20–50 μg/m³ were relatively flat, indicating that the risk of hospitalization remains at a relatively high level with the increase of SO₂ concentration. The curve for younger people (<65 years) decreased slightly in the range of 0–10 μg/m³ and increased at higher concentration (20–50 μg/m³), which showed the risk of hospitalization for younger people's respiratory system does not change much under relatively low concentration of SO₂ exposure.

As shown in Figure 2, the association between the SO₂ concentration and the daily respiratory inpatient visits was stronger in the warm season (May–Oct) rather than in the cold season (Nov–Apr). For the elderly (≥65 years), the positive association between the SO₂ concentration and the respiratory inpatient visits was statistically significant only in warm seasons. Table 4 summarizes the percent changes for daily respiratory inpatient visits associated with each 10 μg/m³ increase of SO₂ concentration in multiple pollutant models. We used the data from lag3 and lag4 because the single-day lag effect of SO₂ was statistically significant in the above lag days. The results from the multi-pollutant model indicated that the relationship between SO₂ and respiratory inpatient visits was meaningless at lag3 after PM₁₀ was controlled. When adjusted for all atmospheric pollutants, including PM₁₀, NO₂, O₃, and CO, the association between SO₂ and inpatient visits was still statistically significant at lag4. The effects of SO₂ on respiratory inpatient visits were slightly decreased after adjusting for PM₁₀, O₃, and CO (2.6–3.3%) at lag4 and the percent change of inpatient visits became 3.8% after adjusting for NO₂.

Our research is a quantitative evaluation of the relevance between SO₂ and daily respiratory inpatient visits in Ganzhou from 2017 to 2019 using the GAM and the DLNM. The results indicated that the elevation in SO₂ concentration was significantly associated with an increase in respiratory inpatient visits, especially in the warm season (May–Oct), women, and elderly (≥65 years) subgroups. This association was stable after adjusting for other atmospheric pollutants. The exposure-response curves of the SO₂ concentrations and the relative risk of respiratory inpatient visits were nearly non-linear with no obvious thresholds.

This study showed significant cumulative effects of SO₂ concentration in a single pollutant model, with the peak at lag0–5. At this point, per 10 μg/m³ elevation of SO₂ concentration was associated with a 6.9% (95% CI: 1.6–12.5%) increment in daily respiratory inpatient visits. The results of the previous studies were consistent with ours, indicating that the SO₂ concentration was positively related to respiratory diseases (21). Early in vivo studies have observed that SO₂ exposure causes bronchoconstriction and induces respiratory diseases (22). A study conducted in Thailand indicated that a 10 μg/m³ increase in the SO₂ concentration was associated with a 1.83% (95% CI: 1.2–2.4%) enhancement in total respiratory hospital admissions at lag0–4 (23). Similarly, the positive relationship between respiratory admissions and SO₂ was observed at lag 4 [relative risk (RR): 1.12; 95% CI: 1.05–1.21] in Malaysia (24). A systematic review also suggested that every 10 μg/m³ increase of the SO₂ levels corresponded to 0.7% (95% CI: 0.1–1.4%) increment respiratory morbidity at lag0 (25). An ecological study reported that in Shenyang, a typical heavily polluted city in China, per 10 μg/m³ increase in SO₂ was related to 0.7% (95% CI: 0–1.4%) increase in respiratory admissions at lag0 (26). The earliest observed positive relation between SO₂ exposure and respiratory inpatient visits in our study was at lag3 in a single-day lag model. After comparing the average SO₂ concentrations in Ganzhou (18.5 μg/m³) and Shenyang (55 μg/m³) during the study period, we speculated that the exposure of higher concentrations of SO₂ had intraday effects on the respiratory health, while the exposure of relatively low concentrations of SO₂ had a delayed effect.

In this study, the earliest observed association between the SO₂ concentration and the daily inpatient visits with respiratory diseases was at lag0–3 in the multi-day lag model. Conversely, some earlier studies indicated that the strongest effects were observed on the current day (25) or lag0–1 (23). A plausible explanation for the longer-than-normal hysteresis effect observed in Ganzhou may be that long-term exposure to SO₂ leads to a decrease in the sensitivity of residents to SO₂. In addition, the differences in the estimated effects and lagged patterns of SO₂ may be related to different outcome indicators of the study, local economic development, and the gender structure of the total population in different regions. A retrospective study from Switzerland suggested that the relationship between the SO₂ exposure and the hospital admissions for respiratory diseases was not statistically significant (27). This might reflect the difference in population sensitivity caused by different cultural backgrounds and dietary structures.

The stratified analysis suggested that women and elderly (≥65 years) are more sensitive to SO₂ exposure, which is consistent with previous studies (28, 29). An observational study conducted in Lanzhou, China, found that the estimated effect size of SO₂ on respiratory hospital admissions was slightly larger in women than in men (30). Physiological factors, such as lower red blood cell count and higher airway response, may be important reasons that cause women to be more sensitive to atmospheric pollutant exposure (31, 32). Moreover, the health differences according to sex may be affected by smoking, drinking, and other unhealthy living habits and occupational environment. Interestingly, in an earlier multi-city time series analysis, gender differences in the impact of SO₂ were not observed, and this study pointed out that the insignificant gender differences may be due to factors, such as study design, sample size, and modeling strategy (33).

Consistent results have been reported on the effects of ambient air pollution on specific age groups. Qiu et al. found that each 10 μg/m³ increase in SO₂ corresponds to a 3.4% (95% CI: 2.3–4.5%) increase in overall respiratory hospital admissions among the elderly (>65 years) in the Sichuan Basin (33). Similarly, earlier studies observed that SO₂ was more associated with hospital admissions for respiratory diseases in older adults (>65) than in other age subgroups (26). The elderly are more susceptible to pathogen exposure due to the weaker immune defenses and respiratory functions (34). This may be an important reason for their increased sensitivity to SO₂ exposure.

The health effects of SO₂ in different seasons have long been debated. The results showed a positive relation between SO₂ and daily respiratory inpatient visits during the warm season. Two study indicated that the relationship between SO₂ and respiratory disease morbidity could be observed only during the cold season (heating season) (30, 35). It was worth mentioning that the research site is located in northern China, where temperatures are well below freezing in winter, and fossil fuels need to be burned for heating. The SO₂ level in the cold season in northeast China was much higher than that in the warm season. However, there was no statistical difference in the SO₂ concentration in Ganzhou during the cold and warm seasons (Supplementary Table 1). A recent study, also from southern China, was consistent with our findings, confirming a statistically significant association between SO₂ and respiratory disease mortality during warm seasons (36). People in warm seasons tend to spend longer times outdoors. Even indoors, residents often open windows during warm seasons, and the indoor SO₂ level is closer to the ambient SO₂ level. Moreover, the different meteorological conditions (especially extreme temperatures) in different regions may be one of the important reasons for the variability (37–39).

Understanding the exposure-response curves of atmospheric pollutants is very important for the making of environmental and public health policies. As shown in Figure 1, we found that even at very low exposure levels, a positive correlation between SO₂ and daily respiratory inpatient visits could be observed in the elderly subgroups. Similarly, several studies demonstrated adverse effects of SO₂ at relatively low concentrations (40, 41). The health effect of SO₂ as a single pollutant or with other atmospheric pollutants has long been controversial. In this study, the association between the SO₂ concentration and the respiratory inpatient visits was stable after adjusting for other air pollutants at lag4. A study in Japan by Yorifuji et al. (42) also suggested that the relationship between SO₂ exposure and respiratory system mortality can still be observed after adjusting for atmospheric pollutants, such as nitrogen dioxide and particulate matter. Yang et al. (43) proposed that multi-pollutant models increase the standard error, so the effects of multi-pollutant models tend to be slightly lower than those of single-pollutant models, which was consistent with the results of this study. However, a systematic review suggests that after adjusting for PM₁₀ and NO₂, no association between respiratory morbidity and SO₂ had been observed (25). It is worth noting that there is a strong correlation between atmospheric pollutants, and it is difficult to accurately assess the adverse effects of a single pollutant even if multi-pollutant models are used (44).

In this study, the generalized additive model was adopted to adjust for the influence of confounding factors, such as long-term trends of the time, meteorological conditions, weekend, and holiday effects. Our study revealed a positive correlation between the SO₂ concentration and the morbidity of respiratory diseases in China. Nevertheless, the study has several limitations. Firstly, the average concentration at fixed air monitoring stations was used as a proxy for individual exposure, which may underestimate the adverse effects of SO₂ (45). Secondly, even if multi-pollutant models were used, the independent effects of SO₂ could not be fully explored because there are no data on other particulate pollutants except PM₁₀ in this study. Thirdly, the main model of this study did not include several confounding factors, such as daily activities and socioeconomic status, which may not fully reflect the association between SO₂ and respiratory inpatient visits.

The relationship between the elevated SO₂ concentrations and the daily respiratory inpatient visits was observed in Ganzhou, a subtropical city in southern China, even though the average daily SO₂ concentrations were lower than the minimum allowable exposure concentration set by the WHO (46) and China (47). We call on scholars in different regions to research on the health damage caused by common air pollutants to provide a theoretical basis for local health promotion and environmental health policy formulation. Future investigations should require more rigorous experimental designs (e.g., a combination of animal experiments and population epidemiological investigations) to identify the subgroups susceptible to respiratory damage caused by the SO₂ exposure.

Our study indicated that SO₂ concentration is positively associated with daily inpatient visits for respiratory disease. The association is closer in women, elderly people (≥65 years), and warmer seasons (May–Oct) subgroups. These results provide further evidence to support the potential health effects of exposure to SO₂. We hope that our study can remind researchers and managers to pay more attention to the adverse effects of SO₂ in developing countries.

The data analyzed in this study is subject to the following licenses/restrictions: the disclosure and use of inpatient information requires the consent of the relevant departments of the hospital. Requests to access these datasets should be directed to https://www.gyyfy.com/.

XZho carried out literature retrieval, determined the research direction and data analysis according to the existing literature, and wrote the article. YG participated in the data collection of articles and assisted in statistical analysis. DW and WC explained the data results, discussed them in combination with existing articles, and assisted in writing the first draft of the article. XZha was mainly responsible for contacting the hospital to provide the data required for this study and putting forward modification opinions on the first draft of the article. All authors read and approved the final manuscript.

This work was funded by Gannan Medical University (Fund No. QD201901) and Department of Education of Jiangxi Province (Fund No. GJJ190786).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.