Air pollution is one of the greatest environmental risks for public health. Reducing ambient air pollution has become an important priority for the United Kingdom (UK) government, as there is increased awareness of its impact on health. In 2020, the UK made history by ruling that failure to reduce pollution levels to legal limits (in line with levels recommended by the WHO) contributed to the death of a 9-year old girl in London (Laville, 2020).
While most of the research on ambient air pollution has focused on its impact on physical health through respiratory and cardiovascular diseases, more recent studies have found compelling evidence that air pollution can also increase the risk of serious mental illnesses, such as schizophrenia and depression. However, we still know relatively little about the extent of long-term impacts of air pollution on mental health, partly because most studies to date have generally used cross-sectional datasets and aggregated values for both air pollution and mental illnesses.
Therefore, in this paper (Newbury et al., 2021), the authors used a longitudinal survey to estimate the impact of ambient air pollution and mental health service use for psychotic and mood disorders, specifically focusing on the “duration and frequency of mental health service use”. This analysis provides important insights into the short term, and longer term implications of air pollution on mental health service utilisation.
The study (Newbury et al., 2021) used data from the Clinical Record Interactive Search (CRIS) system for over 13,000 individuals aged > 14 years old living in South London boroughs, who had accessed mental health services for schizophrenia and mood disorders between 2008-2012. Service use included both community mental health services (CMHS) and inpatient services for a one year and seven year follow-up period.
This data on mental health service use was then linked with estimates for air pollution including nitrogen dioxide (NO2), nitrogen oxides (NOx) and particulate matter (PM2.5 and PM10) concentrations. These data on air pollution were estimated through KCLurban, which uses emission estimates recorded in the London Atmospheric Emissions Inventory. Air pollution estimates were then linked to the residence of participants. The CRIS dataset used for service use was de-identified, so information on residence was collected using a de-identified dataset.
The authors also used information on various confounding variables in their model and the sensitivity analysis, which also accounted for specificity of diagnosis (i.e. psychotic disorder diagnoses versus mood disorder diagnoses), potential confounding effects of the pollutants, and time spent at residence. The authors also used methods to account for how causality between an exposure and an outcome might be affected by unmeasured confounding. To do this, they calculated E-values, which show how strong an association with the exposure and outcome an unmeasured confounder needs to be, to explain away the main associations.
Results showed that all pollutants were associated with increased use of mental health services, including both Community Mental Health Services (CMHS) and inpatient days. Furthermore, the authors showed that this association persisted after 7 years.
The strongest associations between exposure to pollutants and increased use of mental health services were observed for NO2 and NOx pollutants, in both the 1-year and 7-year follow-up periods.
Sensitivity analyses reported large E-values which increases confidence in the authors results. Additionally, an important finding shows that using a two-pollutant model could change both the direction and significance of the outcome, which suggests that future studies could adopt multi-pollutant approaches to avoid under estimating the strength of associations.
The authors concluded that:
Individuals with higher residential air pollution exposure used mental healthcare services more frequently in the months and years following their initial presentation to secondary mental healthcare services.
Between the two categories of utilisation, the authors found a more robust and positive association with CMHS use, which the authors suggest might be because this type of service is more commonly used.
The authors also discussed plausible mechanisms that might link pollution with psychotic and mood disorders. They note that there is evidence to suggest that pollutants can affect the brain through systemic inflammation or direct contact with the nervous system. The effects appear to be most directly linked to NO2 and NOx emissions typically caused by heavy traffic and diesel vehicle emissions.
Strengths and limitations
The research has several key strengths. First it provides novel evidence on the long-term impact of air pollution exposure on psychotic and mood disorders. Additionally, they used high-resolution (20 × 20 m) air pollution models to enhance the accuracy of the findings. Causality can be difficult to establish for environmental risk indicators, so the researchers have taken extra precautions through a range of sensitivity analysis and statistical techniques to ensure the robustness of the analysis. The methods and types of data used are well described and the findings are clearly presented. Additional methodological detail is also given in the supplementary materials.
There are also some caveats or limitations to consider.
One potential limitation is that the data on air pollution exposure has only been mapped at the time of initial service use. This would presume that there were no changes in ambient air pollution or emissions during the 1 year and 7 year follow up periods. There is however evidence to suggest that emissions, for example for NO2, have been steadily decreasing across the UK during the study period (Department for Environment Food & Rural Affairs, 2021). The change in air pollution should have implications for the magnitude of the health impact for the 1 year and 7-year follow-up period.
The analysis uses participants residential address to estimate exposure, but there are two main limitations with this approach. The first limitation was that the source of this information was not clear. The authors mention that this data was not available in CRIS, and that they have used information from a pre-CRIS database, but it was not clear whether this pre-CRIS database followed the participants residential address throughout the follow-up period. The second limitation is that many of the study participants will likely commute for work, education, or recreational reasons, and be exposed to emissions outside their residence. While the authors correctly point out that this can be difficult to capture, future research could build on other studies that have used modelling approaches to capture such potential variation in exposure (Beckx et al., 2009; Yann et al., 2014).
Lastly, the paper uses WHO guidelines to estimate health gains. The WHO has recently updated their guidelines on the recommended limit for pollutants, so any future analysis could update the estimates using limits from the new guidelines (World Health Organization, 2021).
Implications for practice
This is a timely analysis with important policy implications for reducing air pollution.
Mental illnesses are not included in the Institute for Health Metrics and Evaluation model, which estimates the burden of disease linked to air pollution (Brauer et al., 2016). Other than dementia, mental illness is also not included in the economic impact of air pollution used by the UK government (Public Health England, 2018). Based on the findings of this study, the UK and many other governments may be underestimating the impact of air pollution, and by extension, underfunding efforts to reduce air pollution.
More specifically, this paper makes an important case for policies to reduce vehicle emission. The analysis finds that NO2 and NOx have the largest impact on mental health service use, and vehicles are a major source for both these pollutants. Policies that may reduce vehicle emissions include promoting active travel by increasing road taxes, congestion charges, increasing bike lanes, as well as long-term investment in public transportation.
Statement of interests
Wajeeha Raza is PhD student funded by the NIHR Yorkshire and Humber Applied Research Collaboration (ARC). Peter Coventry is part funded by the NIHR Yorkshire and Humber ARC. The views expressed are those of the authors, and not necessarily those of the NIHR or the Department of Health and Social Care.
Newbury, J.B. et al. (2021) “Association between air pollution exposure and mental health service use among individuals with first presentations of psychotic and mood disorders: retrospective cohort study,” The British Journal of Psychiatry [Preprint]. doi:10.1192/bjp.2021.119.
Beckx, C. et al. (2009) “A dynamic activity-based population modelling approach to evaluate exposure to air pollution: Methods and application to a Dutch urban area,” Environmental Impact Assessment Review, 29(3), pp. 179–185. doi:10.1016/J.EIAR.2008.10.001.
Brauer, M. et al. (2016) “Ambient Air Pollution Exposure Estimation for the Global Burden of Disease 2013,” Environmental Science & Technology, 50(1). doi:10.1021/acs.est.5b03709.
Department for Environment Food & Rural Affairs (2021) Concentrations of nitrogen dioxide, https://www.gov.uk/government/statistics/air-quality-statistics/ntrogen-dioxide.
Laville, S. (2020) “Air pollution a cause in London girl’s death, coroner rules in landmark case,” The Irish Times, 17 December.
Public Health England (2018) Estimation of costs to the NHS and social care due to the health impacts of air pollution. London.
World Health Organization (2021) Ambient (outdoor) air pollution, https://www.who.int/news-room/fact-sheets/detail/ambient-(outdoor)-air-quality-and-health.
Yann, S. et al. (2014) “Health effects of ambient air pollution: Do different methods for estimating exposure lead to different results?,” Environment International, 66, pp. 165–173. doi:10.1016/J.ENVINT.2014.02.001.