Article Text

Download PDFPDF

Original article
Ambient concentrations of NO2 and hospital admissions for schizophrenia
  1. Lijun Bai1,2,
  2. Xulai Zhang3,
  3. Yanwu Zhang1,2,
  4. Qiang Cheng1,2,
  5. Jun Duan1,2,
  6. Jiaojiao Gao1,2,
  7. Zihan Xu1,2,
  8. Heng Zhang1,2,
  9. Shusi Wang1,2,
  10. Hong Su1,2
  1. 1 Department of Epidemiology and Health Statistics, School of Public Health, Anhui Medical University, Hefei, China
  2. 2 Department of Epidemiology and Health Statistics, Anhui Province Key Laboratory of Major Autoimmune Diseases, Hefei, China
  3. 3 Department of geriatric psychology, Anhui Mental Health Center, Hefei, China
  1. Correspondence to Professor Hong Su, Department of Epidemiology and Health Statistics, School of Public Health, Anhui Medical University, Hefei 230032, China; 271244914{at}


Objectives Schizophrenia is a chronic and severe mental disorder affecting more than 21 million people worldwide. Short-term exposure to nitrogen dioxide (NO2) has been associated with hospital admissions (HAs) for mental disorders, but no study has evaluated the specific association of NO2 and schizophrenia. Additionally, the shape of the concentration–response (C–R) curve has not yet been assessed at present. This study aims to investigate the relationship between short-term exposure to NO2 and HAs for schizophrenia in Hefei, from 2014 to 2016. We also attempt to explore the C–R and the underlying effect modifiers of the association.

Methods Daily number of HAs for schizophrenia was derived from the computerised medical record system of Anhui Mental Health Center. We used a time-series Poisson generalised linear regression combined with distributed lag non-linear models to model the NO2–schizophrenia relationship.

Results A total of 11 373 HAs were identified during the study period. An increase in levels of NO2 was significantly associated with elevated schizophrenia HAs. The estimated relative risk per IQR increase in NO2 at lag 01 was 1.10 (95% CI 1.01 to 1.18). Greater association was observed in young patients (relative risk: 1.11, 95% CI 1.02 to 1.19). The modelled C–R curves of the NO2–schizophrenia relationship suggested possible threshold effects of NO2 for all ages combined, young patients, men and both seasons.

Conclusions Short-term exposure to NO2 may be associated with increased schizophrenia HAs. Findings indicated potential threshold effects of NO2, which has important implications for health-based risk assessments.

  • nitrogen dioxide
  • schizophrenia
  • concentration-response
  • time series study

Statistics from

Key messages

What is already known about this subject?

  • Air pollution has been linked to mental disorders, but evidence for schizophrenia is lacking.

  • Research about the association between short-term exposure to nitrogen dioxide (NO2) and schizophrenia is scarce.

  • The shape of concentration–response curve for the specific association is poorly understood.

What are the new findings?

  • Elevated NO2 concentration is associated with increases in schizophrenia hospital admissions (HAs).

  • Possible threshold effects of NO2, above which the risk of schizophrenia HAs begins to increase, are observed.

  • Age may be an effect modifier of this association.

How might this impact on policy or clinical practice in the foreseeable future?

  • This research can provide basis for quantitative risk assessments of public health.

  • The present study also suggests that holding NO2 below the possible threshold level may reduce the risk of schizophrenia hospitalisations.


Schizophrenia is a serious mental disorder affecting more than 21 million people worldwide, and people with schizophrenia are two to three times more likely to die in their early years than the general population. According to the Global Burden of Disease study in 2016, schizophrenia is a leading cause (ranked ninth) of years lived with disability in China, and the incidence of schizophrenia (23.6 new cases per 100 000) ranked first in the world. A full understanding about the risk factors of schizophrenia is of great importance to public health. Existing evidence indicates that genetic,1 socioeconomic2 and behavioural risk factors3 are primary contributors to schizophrenia. Environmental factors are also important contributors.4 5 Air pollution, as one of the world’s largest health challenges,6 has already been linked to the incidence of human psychological health.7 For example, exposure to air pollution has been found to increase the risk of suicide,8 depression,9 maternal stress,10 autism11 and so on.

Existing epidemiological studies mainly focused on the mental health effect of particulate matter (PM). The role of ambient nitrogen dioxide (NO2) in affecting mental health has received less attention currently, although NO2 is deemed to be a more relevant health-based exposure indicator than PM in several studies.12 13 Some research has reported increased hospital admissions (HAs) of mental disorders following days of elevated NO2 concentrations, implicating NO2 may be a risk factor for mental health.14 15 However, these studies primarily investigated depression, suicide and total mental disorders, and none has studied schizophrenia. Furthermore, a number of epidemiological studies only quantified the strength of the relationship. Very few explicitly explored the shape of concentration–response (C–R) function, which has strong implications for quantitative risk assessment, especially in low-income and middle-income countries with severe air pollution problems, such as China. Chen et al reported the C–R relationship curves between NO2 and daily hospitalisations for mental disorders, but they only investigated total mental disorders, with the C–R for schizophrenia unknown.16 Therefore, it is needed to determine whether NO2 is a risk factor for schizophrenia HAs. In this study, we aimed to evaluate the effects of short-term exposure to ambient NO2 on daily HAs for schizophrenia in Hefei, China, and to investigate the shape of the C–R curve. Further, stratified analyses were performed by season, sex and age group.

Hefei, the capital city of Anhui Province, is situated in the eastern region of China (31°52 N, 117°17 E), with a total population of 717 7200 and an area of 11 445.1 km2 (National Bureau of Statistics of China, 2015). This city has a subtropical monsoon climate and four distinct seasons. As an important transportation hub in China, this city is confronted with increasingly serious traffic air pollution, which is considered to be one of the principal pollutants.


Data collection

Daily number of HAs for schizophrenia from 1 January 2014 to 31 December 2016 was collected from the computerised medical record system of Anhui Mental Health Center. The available information of each case included date of admission, age, gender and residential address. Schizophrenia cases were defined according to the 10th version of the International Classification of Diseases  (F20–F29). We excluded cases whose residential address is not in Hefei.

Daily air quality data from 1 January 2014 to 31 December 2016 were obtained from Hefei Environmental Monitoring Center. Concentration of each air pollutant was measured at 10 fixed-site monitoring stations. We evaluated 24-hour average concentrations of ambient NO2, particulate matter with aerodynamic diameter less than 10 mm (PM10) and carbon monoxide (CO) by averaging the daily concentrations of air pollutants from the 10 monitoring stations. Daily meteorological data, including daily mean temperature (MT, °C) and relative humidity (RH, %), were collected from the Hefei Bureau of Meteorology. The missing value was interpolated by using the mean of the previous and the following days’ values.17

Statistical analysis

Daily admissions for schizophrenia are a small probability event, which is approximately a Poisson distribution. Thus, a Poisson generalised linear regression18 was applied to investigate the effects of short-term exposure of NO2 on daily schizophrenia HAs. To control for the potential effects of weather conditions on schizophrenia admissions, we used a distributed lag non-linear model (DLNM) for daily MT and RH. The basic model includes the following: (1) a natural cubic spline function of calendar time with 8 df per year to exclude long-term and seasonal trends18; (2) a natural cubic spline of 4 df for RH (%) and MT (°C)19; (3) a categorical variable for day of the week to control the variation of emergency admissions within a week; and (4) an indicator variable for public holidays adjusting for holiday effects.20 We used a 27-day lag for MT21 and a 14-day lag for RH according to Akaike information criterion (AIC). The core model in the present study was described as follows:

Embedded Image

Embedded Image

t is the day of the observation, Y t indicates the observed schizophrenia visit count on day t, X t indicates the pollutant concentrations on day t, α is the model intercept, β represents the vector of coefficients for X t,l , and l is the lag days; X t,l , MT t,l and RH t,l are the cross-basis matrix produced by DLNM; ns() denotes a natural cubic spline; Time t refers to the time variable to adjust for time trends; Dow t is a categorical variable for day of the week, and ŋ is the coefficient; Holiday t is the binary variable indicating public holidays, and γ is the coefficient.

Accounting for the lowest AIC value, 1 day of lag was used to capture any immediate and cumulative effects. To explore the lag patterns in the effects of NO2, we further introduced both single-day lags from 0 to 6 and a 7-day overall cumulative effect of the current and previous 6 days (lag 06). To check the possible modifiers of the association, analyses were stratified by gender (male and female), age (<40 years and ≥40 years) and season (cool season (from November to April) and warm season (from May to October)).22 The statistical significance for the differences in estimates across strata was tested by calculating the 95% CI as Embedded Image , where Q1 and Q2 were the estimates for two categories (such as male and female), and SE1 and SE2 represented the corresponding SEs.23

Several sensitivity analyses were conducted to check the stability of our findings. First, since traffic is an important source of NO2, in order to explore whether the association was sensitive to the effects of other traffic-related pollutants, such as PM10 and CO,24 two-pollutant and three-pollutant models were fitted. According to the observed shape of C–R (online supplementary figure S1), PM10 and CO were included as non-linear terms in models, respectively. Two-pollutant models included PM10 or CO, and three-pollutant models included both PM10 and CO. Second, we changed the df for time t from 6 to 8 per year. Third, different dfs were chosen for MT and RH to evaluate the robustness of the results.

Supplemental material

A natural cubic spline with 3 df for NO2 (lag 0–1 day) was also applied in single-pollutant models to check whether the association between NO2 and schizophrenic HAs was linear or non-linear.19 Effect estimates were described as relative risk (RR) and 95% CIs in daily HAs for schizophrenia per IQR (IQR=75th percentile–25th percentile of air pollutants) increase in NO2.

All statistical analyses were conducted using ‘dlnm’ package in R V.3.4.1 software to derive the risk estimates.25 The statistical tests were two-sided, and associations of p<0.05 were considered statistically significant.


Table 1 summarises the descriptive statistics of this study. During the study period, a total of 11 373 HAs were recorded. All admissions were for acute episodes of schizophrenia including both first onset and repeated cases. People aged under 40 years accounted for 70.2%, and patients’ sexes were distributed evenly (50.9% and 49.1%). The number of HAs was slightly higher in the cool season than that in the warm season. The daily mean concentration of NO2 was 35.5 µg/m³, and the values at 25th, 50th and 75th percentiles were 24.0 µg/m³, 33.0 µg/m³ and 42.0 µg/m³. Daily MT and RH were 16.8°C and 75.8%. Correlations among air pollutants and weather factors ranged from 0.019 to 0.66 (online supplementary figure S2).

Supplemental material

Table 1

Summary statistics of daily environment and health data from 2014 to 2016 in Hefei, China

Figure 1 describes the RRs in single-pollutant models using different lags (0–6 days). Except lag 6, positive associations between NO2 and schizophrenia HAs appeared in all lag days, but the positive association maintained statistically significant only on the first 2 days, with the largest estimates presented on lag 01. Thus, use of a lag structure from the same day up to 1 day before the outcome was sufficient to capture the short-term effects of NO2 on schizophrenia admissions. Interestingly, negative association between NO2 exposure and schizophrenia HAs was observed at lag 6, which could be interpreted as a harvesting effect, possibly caused by the fact that the pool of vulnerable people was depleted faster than usual due to the air pollution event and consequently resulting in low-than-expected number of schizophrenia cases a few days after.26 Figure 2 and figure 3 show the C–R curves for the associations between NO2 exposure at lag 01 and daily HAs for schizophrenia, implying possible threshold effects (a minimum HAs concentration of the air pollutant)27 for total admissions, men, patients under 40 years old, the cool season and the warm season, respectively. The threshold concentrations of NO2 for different subgroups are listed in online supplementary table S1, ranging from 19.1 µg/m³ to 26.1 µg/m³. However, the relationships were approximately linear among women and patients aged <40 years. Thus we evaluated the effect size in the context of non-linear and linear C–R relationship separately. Based on non-linear analysis, we used a lag of 0–1 day in the linear C–R analyses.

Supplemental material

Figure 1

Relative risk (RR) and 95% CI of schizophrenia hospital admissions per IQR increase in concentrations of nitrogen dioxide for single-day lags from 0 to 6 and cumulative lags of lag 01 (the current and previous 1 day) and lag 06 (the current and previous 6 days).

Figure 2

The concentration–response relationship curves between NO2 (lag 01 day) and daily total hospital admissions for schizophrenia. NO2, nitrogen dioxide; RR, relative risk.

Figure 3

Associations between short-term NO2 exposure and schizophrenia hospital admissions stratified by different subgroups (A for younger patients aged <40, B for older adults ≥40, C for men, D for women, E for cool season, D for warm season). NO2, nitrogen dioxide; RR, relative risk.

Table 2 summarises the effect estimates in single-pollutant models, which showed increases in schizophrenia HAs were significantly associated with an IQR increase in NO2 concentration at different lag days. The associations of NO2 with daily total schizophrenia hospitalisations were positive, but remained statistically significant only at lag 1 and lag 01, with the respective RRs of 1.07 (95% CI 1.01 to 1.13) and 1.10 (95% CI 1.03 to 1.18). The results for the possible effect of modification by age, sex and season are also shown in table 2. The risk estimates for men and women were similar. At lag 01, RRs for men and women were 1.10 (95% CI 1.00 to 1.20) and 1.11 (95% CI 1.01 to 1.22), respectively. The RR for people aged <40 years was 1.11 (95% CI 1.02 to 1.19), but the association became non-significant among the people aged ≥40 years. Positive relationship was presented in both seasons, but non-significant in the cool season. No significant difference was presented for season-specific associations, although the effect size in the warm season (1.13, 95% CI 1.00 to 1.27) seemed larger than that in the cool season (1.09, 95% CI 0.99 to 1.20).

Table 2

Relative risk and 95% CI for schizophrenia hospital admissions per IQR increase in concentrations of nitrogen dioxide stratified by gender, age and season in Hefei, China, 2014–2016

Our sensitivity analyses indicated robustness of the associations between ambient NO2 and schizophrenia HAs. Table 3 provides the risk estimates in the two-pollutant and three-pollutant models, with statistically significant RRs after adjusting for other pollutants (PM10 and CO). Estimated effects changed slightly under varying df for the time, and the RRs were similar when the df for temperature and RH ranged from 3 to 5 (online supplementary table S2).

Table 3

Relative risk and 95% CI for schizophrenia hospital admissions per IQR increase in concentrations of nitrogen dioxide in two-pollutant and three-pollutant models


To the best of our knowledge, this was one of the few studies investigating the short-term effect of NO2 on schizophrenia HAs. The present study showed that elevated short-term ambient NO2 level was associated with statistically significant increases in schizophrenia HAs. According to our sensitivity analyses, the effect of NO2 was robust. Effect modification of sex and season was not observed, but the associations were stronger among people aged <40 years than that among people aged ≥40 years. Our study also showed a possible threshold effect of NO2 on daily schizophrenia HAs in Hefei. The associations were approximately linear for women and adults aged ≥40 years.

Although the exact mechanisms are unclear, the relationship between NO2 and increased schizophrenia HAs seems biologically plausible. Exposure to NO2 may increase the risk of schizophrenia admissions through increased neuroinflammation and oxidative stress,28 29 as supported by the following findings from animal studies: associations between diesel exhaust inhalation and neuroinflammation in brain regions,30 relationship between NO2 inhalation and brain pathology, lower antioxidant defences and deregulation of apoptosis-related genes expression,31 links between diesel exhaust and microglial activation and upregulation of oxidative stress,28 and associations between vehicle emissions and anxiety-like behaviour, depression-like behaviour, neurobehavioural as well as cognitive deficits.24 Epidemiological research also indicated that traffic-related air pollution was associated with autism and decreased cognitive function.11 32 33 After entering the brain, NO2 can activate the microglia, and then microglial-mediated inflammatory processes have been proposed to cause secondary neuronal degeneration, decreased neurogenesis, synaptic dysfunction and structural brain changes, and thus onset or relapse of schizophrenia.34 35 In addition, activated microglia36 and impaired antioxidant defences37 due to NO2 exposure may induce failure to protect the cell membranes against free-radical generation, with resulting dysfunction that might impact on neurotransmission, and ultimately acute exacerbations of schizophrenia.38

Currently, few studies have attempted to examine the shape of the C–R curve for short-term NO2 exposure and schizophrenia HAs. A threshold concentration of air pollutant is generally expected to protect the public health by holding the pollutant below this level. Our findings suggested there might be a threshold effect of NO2 for daily schizophrenia HAs. Chen et al reported the exposure–response relationship curves between NO2 and daily HAs for mental disorders in Shanghai, China, and showed a steep slope at concentrations >60 µg/m3.16 This may be explained by different population susceptibility to NO2, since the level of NO2 in Shanghai was higher than that in Hefei, as the statistics of the environment data showed in their article. Findings from other regions are lacking; thus, it is essential to explore the C–R relationship between NO2 and schizophrenia admissions in the future.

The threshold concentration of NO2 in our study is 21.5 µg/m3, lower than the class I level of the Chinese national standards (annual mean: 40 µg/m3 for NO2) (GB3095-2012). However, the concentration of air pollution level below the current standards does not represent that there is no adverse health effects. Di et al reported significant risk effect of short-term exposure to air pollutants lower than the current national air quality standards.39 Meanwhile, an Australian study conducted by Li et al 27 observed that increased risks of preterm birth were associated with increase in the mean concentrations of NO2 in 24 hours before birth, with a threshold of 7.6 parts per billion (ppb), far below the standard of WHO (40 µg/m3). Although C–R relationships may depend on a number of factors, such as region, susceptibility of population, climatic features and air pollution composition, our present findings still have public health implications that standards for NO2 concentration may need to be re-evaluated to reduce schizophrenia HAs.

The present study suggested that stronger effect was observed among those aged <40 years than those aged ≥40 years, although the age difference was not statistically significant. Similarly, a recent study in Tokyo found the effect of NO2 on suicide among those aged under 30 years was significantly greater than those aged 30–60 years.15 In another study in Beijing, positive and statistically significant associations between air pollution and schizophrenia admissions were only found among patients younger than 45 years old.5 This could be explained by the following reasons. First, people aged <40 years generally have more opportunities of exposure to outdoor NO2 on the way to school, to work or home every day. Additionally, people in this age group in China are often confronted with more burdensome mental pressure, including academic pressure, work-related stress, family pressures and role change, which may increase the risk of schizophrenia hospitalisation.40 In contrast to our results, findings from many epidemiological studies demonstrated that older adults were more vulnerable to air pollutants due to their weaker immune system.8 14 What underlie the differences was uncertain. But given the fact that higher age-specific incidence of schizophrenia in China is prevalent in younger adults aged 15–34 years (Global Burden of Disease (GBD), 2016), our study may bring greater public health benefit and protect more vulnerable people.

Although gender difference of the association between air pollution and mental disorders was discovered in many studies,8 14 their conclusions were conflicting, and the clear reasons for gender differences were not well established. In line with some studies in China,5 14 gender difference was not observed in our study. As for season-specific analysis, non-significantly higher effect sizes were found in the warm season than in the cool season, which was consistent with several studies.17 This may be related to more natural ventilation in warmer seasons. Additionally, exposure to extremely high temperature (or heatwave) during the warm season is also likely to increase HAs for mental disorders,41 although weather variables have been modified in our analyses. However, there were studies discovering a stronger association in cool seasons than in warm seasons.14 42 The exact season pattern deserved further investigation.

The findings of this study may have implications for local environmental and social policies. We found adverse and acute effects of NO2, suggesting timely and effective prevention measures might play an important role in protecting the population in affected regions. Potential beneficial measures should be taken, for example, establishing widespread air pollution network to keep the citizens up to date with the NO2 concentration; getting the mass media involved in helping the public, clinical workers and caregivers to know the harmful effects of NO2; and reducing outdoor time or even temporally close some institutions such as schools and factories during heavily polluted days. We also found young adults were a vulnerable population, implying that protecting them against high exposure to NO2 will bring about greater public health benefits. Generally, a high level of NO2 occurs during rush hour when students are on their way to school or home and workers on their way to work or home. Therefore, cutting down traffic exhaust emission is the primary measure in both the warm season and the cool season.

Interestingly, no significant effect of PM10 and CO was observed in this study, although HAs were seemingly increased following elevated air pollution concentration (online supplementary figure S1). Conversely, Gao et al found PM10 was significantly associated with schizophrenia admissions in Beijing.5 It may be driven by differences in population, such as susceptibility, or by differences in the pollution level between cities. For example, concentration of PM10 in Beijing is higher than in Hefei; thus, the link between PM10 and schizophrenia HAs may be stronger in Beijing.

There were several limitations in our study. First, concentrations of air pollutants were collected from fixed-site air quality monitors, which may lead to an inevitable exposure misclassification. Second, the data of daily HAs were collected from one hospital, resulting in the possible selection bias. However, the selected hospital is the largest hospital in Hefei specially targeting mental disease, so generally patients with schizophrenia are more inclined to visit this hospital. Third, a longer study period can increase the robustness of the results, but we only obtained 3-year data due to data availability issue. Finally, although we have considered some other possible confounding effects of air pollutants (PM10, CO) and weather factors (MT, RH), there may be other potential confounders, such as tetraethyl lead, which may play a role in the relationship between environmental pollution and psychotic disorders.43


This study found a positive association between short-term exposure to NO2 and HAs for schizophrenia. Also, the findings suggested potential threshold effects of NO2, which can be used to protect local people by keeping the NO2 below the identified threshold point.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.


  • Contributors LB and HS designed the study. LB, YZ and QC conducted the study and analysed the data. All authors helped interpret the results. LB wrote the first draft. All authors contributed to the manuscript revision. Then this was revised critically for important intellectual content by the other authors, before LB incorporated the amendments to produce the final draft.

  • Funding This work was supported by the National Natural Science Foundation of China (grant number: 81773518).

  • Competing interests None declared.

  • Patient consent Not required.

  • Ethics approval The study was approved by the Ethics Committee of Anhui Medical University.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.