Using air quality monitoring data for public health action: Health Impact Assessment Studies

1. Using air quality monitoring data for public health action:Health Impact Assessment Studies Marco Martuzzi World Health Organisation, Regional Office for Europe European Centre for Environment and Health, Rome Division

2. Rationale • Clinical and epidemiological evidence on health effects of outdoor air pollution is abundant • Risks have been identified, and dose-response relationships have been characterised for several pollutants and health endpoints • Compared to other risk factors risks are small • Exposure is ubiquitous, majority of people exposed • Need to assess overall public health relevance of air pollution • Air quality monitoring data increasingly available

3. Assessing the impact • Policy makers under growing pressure • Underlying cost-benefit type question, “what would we gain if we could reduce concentrations to X?” • Recent research has addressed this issue • Metrics for health impact: attributable risks (risk assessment studies), years of life lost (YLL), and economic evaluations • Air pollution impact studies have been published, e.g., France-Switzerland-Austria (Kuenzli 2000); UK (Hurley 2000), Italy (submitted), US, …

4. Methods for air pollution HIA • Use PM as a summary indicator of all pollutants (cannot evaluate separate roles); recent studies include ozone • Risk functions for selected outcomes • Exposure estimates, usually average concentrations for large population • Observed rates or prevalence • “Prudent” estimates, i.e., identify part of the health effects effectively attributable to AP

5. 8 Italian Cities • PM10 data from monitoring stations • Mortality, morbidity, hospital admission • Average concentration 52.6 g/m3 • Estimate rates or prevalences predicted at lower concentrations • Compare with observed rates • Reference PM10 levels: 20, 30, 40 g/m3 • “Conservative” risk coefficients, e.g. for long term mortality: 1.026 / 10 g/m3 (95% CI 1.009 – 1.043)

6. Methods E=A*B*C*P A = Attributable proportion [(RR-1)/RR] B = occurrence of health endpoint C = change in concentration (from reference value) P = exposed population

8. Mortality • Long term effects (from cohort studies), age 30+, excluding accidental causes: Austria 5,600; France 31,700; Switzerland 3,300 • PM10 reference level: 7.5 g/m3 • Dose response coefficient for mortality: 1.04 / g/m3 • But who dies? When? (Important especially for economic evaluation) • UK study: estimate YLL

10. Health Impact studies • “First generation” studies • Rough approximations involved, generally thought to be “conservative” • Work is needed, two levels: • Methods for risk assessment of air pollution • Interpretation and use of results in public health and risk management

11. Methods for AP risk assessment • Validity of average concentrations (consistent with epidemiological studies) • Extrapolation across populations • Naïve estimates of uncertainty • Different pollutants (ozone probably needs separate treatment) • Better dose-response models • More health endpoints, esp. short- vs. long-term effects

12. Interpretation • Attributable vs preventable • Susceptible subgroups (the elderly being a group of special interest) • Consider realistic scenarios of reduction of concentrations • Risk assessment vs HIA assessment (e.g., of transport policies)

13. Conclusions • HIA studies useful to fill the knowledge gap between laboratory, clinical, epidemiological evidence and public health policy • Make public health action on AP more compelling • Need to improve methodology and evaluate implications more thoroughly (possibly better communication) • Quality and completeness of AP data has been improving • Still need: higher spatial resolution, partitioning of sources • HIA transport: Noise, accidents, cycling and walking, psychosocial effects