Prediction and model evaluation for space-time data.

Evaluation metrics for prediction error, model selection and model averaging on space-time data are understudied and poorly understood. The absence of independent replication makes prediction ambiguous as a concept and renders evaluation procedures developed for independent data inappropriate for most space-time prediction problems. Motivated by air pollution data collected during California wildfires in 2008, this manuscript attempts a formalization of the true prediction error associated with spatial interpolation. We investigate a variety of cross-validation (CV) procedures employing both simulations and case studies to provide insight into the nature of the estimand targeted by alternative data partition strategies. Consistent with recent best practice, we find that location-based cross-validation is appropriate for estimating spatial interpolation error as in our analysis of the California wildfire data. Interestingly, commonly held notions of bias-variance trade-off of CV fold size do not trivially apply to dependent data, and we recommend leave-one-location-out (LOLO) CV as the preferred prediction error metric for spatial interpolation.

Effect modification of the association between fine particulate air pollution during a wildfire event and respiratory health by area-level measures of socio-economic status, race/ethnicity, and smoking prevalence.

Fine particulate air pollution (PM2.5) is decreasing in most areas of the United States, except for areas most affected by wildfires, where increasing trends in PM2.5 can be attributed to wildfire smoke. The frequency and duration of large wildfires and the length of the wildfire season have all increased in recent decades, partially due to climate change, and wildfire risk is projected to increase further in many regions including the western United States. Increasingly, empirical evidence suggests differential health effects from air pollution by class and race; however, few studies have investigated such differential health impacts from air pollution during a wildfire event. We investigated differential risk of respiratory health impacts during the 2008 northern California wildfires by a comprehensive list of socio-economic status (SES), race/ethnicity, and smoking prevalence variables. Regardless of SES level across nine measures of SES, we found significant associations between PM2.5 and asthma hospitalizations and emergency department (ED) visits during these wildfires. Differential respiratory health risk was found by SES for ED visits for chronic obstructive pulmonary disease where the highest risks were in ZIP codes with the lowest SES levels. Findings for differential effects by race/ethnicity were less consistent across health outcomes. We found that ZIP codes with higher prevalence of smokers had greater risk of ED visits for asthma and pneumonia. Our study suggests that public health efforts to decrease exposures to high levels of air pollution during wildfires should focus on lower SES communities.

Improvements in Air Quality and Health Outcomes Among California Medicaid Enrollees Due to Goods Movement Actions.

INTRODUCTION: In 2006, the California Air Resources Board (CARB) and local air quality management districts implemented an Emission Reduction Plan for Ports and Goods Movement program (referred to hereinafter as GM policy actions) (CARB 2006). The GM policy actions comprise approximately 200 actions with an estimated investment value of $6 to $10 billion. These actions targeted the major sources and polluters related to goods movements, such as highways; ports and railyard trucks; ship fuel and shore power; cargo equipment; and locomotives. These actions aimed to reduce total statewide domestic GM emissions to 2001 levels or lower by the year 2010; to reduce the statewide diesel particulate matter (DPM) health risk from GM by 85% by the year 2020; and to reduce the nitrogen oxides (NOx) emissions from international GM in the South Coast Air Basin by 30% from projected 2015 levels and 50% from projected 2020 levels. The years 2006 and 2007 marked an important milestone in starting to regulate GM polluters and adopting stricter standards for traffic-related air pollution.This project aimed to examine the impact of the GM policy actions on reductions in ambient air pollution and subsequent improvements in health outcomes of Medi-Cal fee-for-service (FFS) beneficiaries with chronic conditions in 10 counties in California. Specifically, we examined whether the GM policy actions reduced air pollution near GMC corridors more than in control areas. We subsequently assessed whether there were greater decreases in emergency room (ER) visits and hospitalizations for enrollees with chronic conditions who lived in the GM corridors (GMCs) than for those who lived in other areas. METHODS: The study used a quasi-experimental design. We defined areas within 500 m of truck-permitted freeways and ports as GMCs. We further defined non-goods movement corridors (NGMCs) as locations within 500 m of truck-prohibited freeways or 300 m of a connecting roadway, and areas out of GMCs and NGMCs as controls (CTRLs). We defined years 2004-2007 as the pre-policy period and years 2008-2010 as the post-policy period. We developed linear mixed-effects land use regression models and created annual air pollution surfaces for nitrogen dioxide (NO2), fine particulate matter (PM2.5), and ozone (O3) across California for years 2004-2010 at a spatial resolution of 30 m, then assigned them to enrollees' home addresses.We used a retrospective cohort of 23,000 California Medicaid (Medi-Cal) FFS adult beneficiaries living in 10 California counties with six years of data (September 1, 2004, to August 31, 2010). Cohort beneficiaries had at least one of four chronic conditions, including asthma, chronic obstructive pulmonary disease (COPD), diabetes, and heart disease.We used a difference-in-differences (DiD) model to assess whether air pollutant concentration and health care utilization (ER visits and hospitalizations) for cohort beneficiaries declined more for those living in intervention corridors (GMCs, NGMCs) than those living in CTRLs. All the models controlled for age, sex, language spoken, race/ethnicity, number of comorbidities in baseline years, county, time-varying health indicator variables, and several neighborhood variables.To facilitate interpretation, we calculated the DiD estimates in each of the three years after the policy intervention. The DiD was used to assess the causal impact of regulatory policy on reductions of air pollution, as well as for the improvements in health outcomes.We explored whether improvements in health outcomes were due to the air pollution reduction by using a multi- level mediation model, in which the effect of GM actions on health outcomes was mediated through the effect of actual air pollution reductions in the post-policy years. We used the Generalized Structural Equation Models for the estimation and combined the effects of NO2 and PM2.5 in the model. To further verify the causal inferences of the GM actions on reductions of exposures and improvements in health outcomes, we performed sensitivity analyses with propensity score weighting. RESULTS: We observed statistically significant reductions in pollutant NO2 and PM2.5 concentrations for enrollees in all 10 counties. The enrollees in GMCs experienced greater reductions in NO2 and PM2.5 from the pre- to the post-policy periods than those in CTRLs. Greater reductions were also observed among beneficiaries living in NGMCs versus those in CTRLs, but those reductions were smaller than among beneficiaries living in GMCs. For O3 concentrations, an opposite trend was observed.Furthermore, we observed significantly greater reductions in ER visits for patients with asthma and COPD living in GMCs than those in CTRLS in the post-policy years. For example, we saw in the DiD modeling results there were 170 fewer ER visits for 1,000 beneficiaries with asthma per year in GMCs if the regionwide trend in the CTRL group was considered not related to the GM policy. Similarly, among the beneficiaries with COPD, there were 180 fewer ER visits per 1,000 patients estimated in the GMCs for the third year after the implementation of the policy.We also observed greater reductions in ER visits among those with asthma, when comparing NGMCs with CTRLs, but reductions were smaller than comparisons between GMCs and CTRLs. The ER visits for those with COPD, diabetes, and the total sample in NGMCs also had downward trends in the post-policy year in comparison with those in CTRLs but the differences were not statistically significant; similar phenomena were also observed for the ER visits among those with diabetes and heart diseases and in the total sample when GMCs versus CTRLs and GMCs versus NGMCs were compared. Although hospitalizations also decreased more in GMCs than in NGMCs and more in NGMCs than in CTRLs in the post-policy period, results were not statistically significant.Using the mediation models, we observed 0.129 more reductions in the expected number of ER visits among individuals with asthma for a composite reduction in one unit NO2 and one unit PM2.5 (DiD = -0.129, P < 0.05) from the pre-policy years to the post-policy years. The reductions in NO2 and PM2.5 due to policy change estimated by the mediation model are essentially the same as shown in the respective DiD models. Mediation analyses suggested that the effects of GM policy interventions on health improvements were largely due to exposure reductions. Finally, sensitivity analyses with propensity scores produced similar DiD results. CONCLUSIONS: This project has produced empirical evidence that air pollution control actions reduced pollution exposures among disadvantaged and susceptible populations. More importantly, our findings suggest that the reductions in air pollution led to health outcome improvements among low-income people with chronic conditions. Our investigation also contributed to scientific methods for assessing the health effects of long-term, large-scale, and complex regulatory actions with routinely collected pollutants and medical claims data. Therefore, the results strongly support both short-term and long-term efforts to improve air quality for all members of society and future studies on the impact of air pollution control policies.

Private and public modes of bicycle commuting: a perspective on attitude and perception.

BACKGROUND: Public bicycle-sharing initiatives can act as health enhancement strategies among urban populations. The aim of the study was to determine which attitudes and perceptions of behavioural control toward cycling and a bicycle-sharing system distinguish commuters with a different adherence to bicycle commuting. METHODS: The recruitment process was conducted in 40 random points in Barcelona from 2011 to 2012. Subjects completed a telephone-based questionnaire including 27 attitude and perception statements. Based on their most common one-way commute trip and willingness to commute by bicycle, subjects were classified into Private Bicycle (PB), public bicycle or Bicing Bicycle (BB), Willing Non-bicycle (WN) and Non-willing Non-bicycle (NN) commuters. After reducing the survey statements through principal component analysis, a multinomial logistic regression model was obtained to evaluate associations between attitudinal and commuter sub-groups. RESULTS: We included 814 adults in the analysis [51.6% female, mean (SD): age 36.6 (10.3) years]. BB commuters were 2.0 times [95% confidence interval (CI) = 1.1-3.7] less likely to perceive bicycle as a quick, flexible and enjoyable mode compared to PB. BB, WN and NN were 2.5 (95% CI = 1.46-4.24), 2.6 (95% CI = 1.53-4.41) and 2.3 times (95% CI = 1.30-4.10) more likely to perceive benefits of using public bicycles (bicycle maintenance and parking avoidance, low cost and no worries about theft and vandalism) than did PB. CONCLUSION: Willing non-bicycle and public-bicycle commuters had more favourable perception toward public-shared bicycles compared to private cyclists. Hence, public bicycles may be the impetus for those willing to start bicycle commuting, thereby increasing physical activity levels.

Spatial analysis of exposure to traffic-related air pollution at birth and childhood atopic asthma in Toronto, Ontario.

Findings from the Toronto Child Health Evaluation Questionnaire (TCHEQ) study indicate that early childhood exposure to traffic-related air pollution (TRAP) is related to the onset of atopic childhood asthma. To test this hypothesis further, we investigated whether spatial patterns in the birth neighbourhood of TCHEQ subjects with atopic asthma (136 of 909 schoolchildren in grades 1-2) could be explained by TRAP and other risk factors. If a causal relationship exists between early childhood residential exposure to TRAP and the development of atopic asthma, we hypothesise that (1) clusters of current asthma should exist around the place of residence at birth, and (2) accounting for residential concentrations of TRAP at birth should explain some of the autocorrelation. Several high asthma clusters were observed. Adjusting for TRAP completely explained one cluster; elsewhere, clusters were only partially explained by TRAP. Findings suggest that exposure during early childhood to TRAP in Toronto is an important contributor to the development of the atopic asthma phenotype and reveal the likely importance of other risk factors not measured in the fixed effects of the model.

Determinants of manganese levels in house dust samples from the CHAMACOS cohort.

INTRODUCTION: Manganese (Mn) is an essential nutrient, but at high exposure levels Mn is a neurotoxicant. The fungicides maneb and mancozeb are approximately 21% Mn by weight and more than 150,000 kg are applied each year to crops in the Salinas Valley, California. It is not clear, however, whether agricultural use of these fungicides increases Mn levels in homes. MATERIALS AND METHODS: We collected house dust samples from 378 residences enrolled in the Center for the Health Assessment of Mothers and Children of Salinas (CHAMACOS) study with a second sample collected approximately nine months later from 90 of the residences. House dust samples were analyzed for Mn using inductively coupled plasma optical emission spectroscopy. Information from interviews, home inspections, and pesticide use reports was used to identify potential predictors of Mn dust concentrations and loadings. RESULTS: Mn was detectable in all dust samples. The median Mn concentration was 171 µg/g and median Mn loading was 1,910 µg/m(2) at first visit. In multivariable models, Mn dust concentrations and loadings increased with the number of farmworkers in the home and the amount of agricultural Mn fungicides applied within three kilometers of the residence during the month prior to dust sample collection. Dust concentrations of Mn and other metals (lead, cadmium and chromium) were higher in residences located in the southern Salinas Valley compared those located in other areas of the Salinas Valley. Dust loadings of Mn and other metals were also higher in residences located on Antioch Loam soil than other soil types, and in homes with poor or average housekeeping practices. CONCLUSIONS: Agricultural use of Mn containing fungicides was associated with Mn dust concentrations and loadings in nearby residences and farmworker homes. Housekeeping practices and soil type at residence were also important factors related to dust metal concentrations and loadings.

Parental stress increases body mass index trajectory in pre-adolescents.

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT: Rates of childhood obesity have increased since the mid-1970s. Research into behavioural determinants has focused on physical inactivity and unhealthy diets. Cross-sectional studies indicate an association between psychological stress experienced by parents and obesity in pre-adolescents. WHAT THIS STUDY ADDS: We provide evidence of a prospective association between parental psychological stress and increased weight gain in pre-adolescents. Family-level support for those experiencing chronic stress might help promote healthy diet and exercise behaviours in children. OBJECTIVE: We examined the impact of parental psychological stress on body mass index (BMI) in pre-adolescent children over 4 years of follow-up. METHODS: We included 4078 children aged 5-10 years (90% were between 5.5 and 7.5 years) at study entry (2002-2003) in the Children's Health Study, a prospective cohort study in southern California. A multi-level linear model simultaneously examined the effect of parental stress at study entry on the attained BMI at age 10 and the slope of change across annual measures of BMI during follow-up, controlled for the child's age and sex. BMI was calculated based on objective measurements of height and weight by trained technicians following a standardized procedure. RESULTS: A two standard deviation increase in parental stress at study entry was associated with an increase in predicted BMI attained by age 10 of 0.287 kg m(-2) (95% confidence interval 0.016-0.558; a 2% increase at this age for a participant of average attained BMI). The same increase in parental stress was also associated with an increased trajectory of weight gain over follow-up, with the slope of change in BMI increased by 0.054 kg m(-2) (95% confidence interval 0.007-0.100; a 7% increase in the slope of change for a participant of average BMI trajectory). CONCLUSIONS: We prospectively demonstrated a small effect of parental stress on BMI at age 10 and weight gain earlier in life than reported previously. Interventions to address the burden of childhood obesity should address the role of parental stress in children.

Physical environmental correlates of childhood obesity: a systematic review.

Increasing rates of childhood obesity in the USA and other Western countries are a cause for serious public health concern. Neighborhood and community environments are thought to play a contributing role in the development of obesity among youth, but it is not well understood which types of physical environmental characteristics have the most potential to influence obesity outcomes. This paper reports the results of a systematic review of quantitative research examining built and biophysical environmental variables associated with obesity in children and adolescents through physical activity. Literature searches in PubMed, PsychInfo and Geobase were conducted. Fifteen quantitative studies met the inclusion criteria for this systematic review. The majority of studies were cross-sectional and published after 2005. Overall, few consistent findings emerged. For children, associations between physical environmental variables and obesity differed by gender, age, socioeconomic status, population density and whether reports were made by the parent or child. Access to equipment and facilities, neighborhood pattern (e.g. rural, exurban, suburban) and urban sprawl were associated with obesity outcomes in adolescents. For most environmental variables considered, strong empirical evidence is not yet available. Conceptual gaps, methodological limitations and future research directions are discussed.

A distance-decay variable selection strategy for land use regression modeling of ambient air pollution exposures.

Land use regression (LUR) has emerged as an effective and economical means of estimating air pollution exposures for epidemiological studies. To date, no systematic method has been developed for optimizing the variable selection process. Traditionally, a limited number of buffer distances assumed having the highest correlations with measured pollutant concentrations are used in the manual stepwise selection process or a model transferred from another urban area. In this paper we propose a novel and systematic way of modeling long-term average air pollutant concentrations through "A Distance Decay REgression Selection Strategy" (ADDRESS). The selection process includes multiple steps and, at each step, a full spectrum of correlation coefficients and buffer distance decay curves are used to select a spatial covariate of the highest correlation (compared to other variables) at its optimized buffer distance. At the first step, the series of distance decay curves is constructed using the measured concentrations against the chosen spatial covariates. A variable with the highest correlation to pollutant levels at its optimized buffer distance is chosen as the first predictor of the LUR model from all the distance decay curves. Starting from the second step, the prediction residuals are used to construct new series of distance decay curves and the variable of the highest correlation at its optimized buffer distance is chosen to be added to the model. This process continues until a variable being added does not contribute significantly (p>0.10) to the model performance. The distance decay curve yields a visualization of change and trend of correlation between the spatial covariates and air pollution concentrations or their prediction residuals, providing a transparent and efficient means of selecting optimized buffer distances. Empirical comparisons suggested that the ADDRESS method produced better results than a manual stepwise selection process of limited buffer distances. The method also enables researchers to understand the likely scale of variables that influence pollution levels, which has potentially important ramifications for planning and epidemiological studies.

Modeling the intraurban variability of ambient traffic pollution in Toronto, Canada.

The objective of this paper is to model determinants of intraurban variation in ambient concentrations of nitrogen dioxide (NO2) in Toronto, Canada, with a land use regression (LUR) model. Although researchers have conducted similar studies in Europe, this work represents the first attempt in a North American setting to characterize variation in traffic pollution through the LUR method. NO2 samples were collected over 2 wk using duplicate two-sided Ogawa passive diffusion samplers at 95 locations across Toronto. Independent variables employed in subsequent regression models as predictors of NO2 were derived by the Arc 8 geographic information system (GIS). Some 85 indicators of land use, traffic, population density, and physical geography were tested. The final regression model yielded a coefficient of determination (R2) of .69. For the traffic variables, density of 24-h traffic counts and road measures display positive associations. For the land use variables, industrial land use and counts of dwellings within 2000 m of the monitoring location were positively associated with NO2. Locations up to 1500 m downwind of major expressways had elevated NO2 levels. The results suggest that a good predictive surface can be derived for North American cities with the LUR method. The predictive maps from the LUR appear to capture small-area variation in NO2 concentrations. These small-area variations in traffic pollution are probably important to the exposure experience of the population and may detect health effects that would have gone unnoticed with other exposure estimates.

A land use regression model for predicting ambient fine particulate matter across Los Angeles, CA.

Land use regression (LUR) models have been used successfully for predicting local variation in traffic pollution, but few studies have explored this method for deriving fine particle exposure surfaces. The primary purpose of this method is to develop a LUR model for predicting fine particle or PM(2.5) mass over the five county metropolitan statistical area (MSA) of Los Angeles. PM(2.5) includes all particles with diameter less than or equal to 2.5 microns. In the Los Angeles MSA, 23 monitors of PM(2.5) were available in the year 2000. This study uses GIS to integrate data regarding land use, transportation and physical geography to derive a PM(2.5) dataset covering Los Angeles. Multiple linear regression was used to create the model for predicting the PM(2.5) surface. Our parsimonious model explained 69% of the variance in PM(2.5) with three predictors: (1) traffic density within 300 m, (2) industrial land area within 5000 m, and (3) government land area within 5000 m of the monitoring site. These results suggest the LUR method can refine exposure models for epidemiologic studies in a North American context.

The health relevance of ambient particulate matter characteristics: coherence of toxicological and epidemiological inferences.

The aim of this article is to review progress toward integration of toxicological and epidemiological research results concerning the role of specific physicochemical properties, and associated sources, in the adverse impact of ambient particulate matter (PM) on public health. Contemporary knowledge about atmospheric aerosols indicates their complex and variable nature. This knowledge has influenced toxicological assessments, pointing to several possible properties of concern, including particle size and specific inorganic and organic chemical constituents. However, results from controlled exposure laboratory studies are difficult to relate to actual community health results because of ambiguities in simulated PM mixtures, inconsistent concentration measurements, and the wide range of different biological endpoints. The use of concentrated ambient particulates (CAPs) coupled with factor analysis has provided an improved understanding of biological effects from more realistic laboratory-based exposure studies. Epidemiological studies have provided information concerning sources of potentially toxic particles or components, adding insight into the significance of exposure to secondary particles, such as sulfate, compared with primary emissions, such as elemental and organic carbon from transportation sources. Recent epidemiological approaches incorporate experimental designs that take advantage of broadened speciation monitoring, multiple monitoring stations, source proximity designs, and emission intervention. However, there continue to be major gaps in knowledge about the relative toxicity of particles from various sources, and the relationship between toxicity and particle physicochemical properties. Advancing knowledge could be facilitated with cooperative toxicological and epidemiological study designs, with the support of findings from atmospheric chemistry.

Linking perceptions of neighbourhood to health in Hamilton, Canada.

STUDY OBJECTIVE: To investigate the association between perceptions of neighbourhood physical and social characteristics and three health outcomes (self assessed health status, chronic conditions, and emotional distress). DESIGN: Cross sectional survey data analysed in small neighbourhoods. SETTING: Hamilton, Ontario, Canada, a medium sized industrial city, located at the western end of Lake Ontario (population at the time of the study about 380 000). PARTICIPANTS: Random sample of 1504 adults aged 18 years and older residing in four contrasting neighbourhoods. MAIN RESULTS: Significant differences across the four neighbourhoods are apparent in self assessed health status and emotional distress, but not in chronic conditions. Neighbourhoods with lower SES reported poorer health and more emotional distress. Perceptions of the physical environment dominated social concerns in all neighbourhoods. For all three health outcomes, individual risk factors followed expectations, with measures of poverty, age, and lifestyle all significantly associated with poor health outcomes. Physical environmental problems were positively and significantly associated with poor physical and emotional health. Specifically, people reporting they dislike aspects of their neighbourhood's physical environment are 1.5 times more likely to report chronic health conditions (OR 1.56, 95% CI 1.19 to 2.05), while those reporting physical likes with their neighbourhood are less likely to report fair/poor health (OR 0.50, 95% CI 0.28 to 0.90) or emotional distress (OR 0.45, 95% CI 0.26 to 0.80). CONCLUSIONS: These results demonstrate the importance of neighbourhood perceptions as a determinant of health, as well as conventional factors such as low income, lifestyle, and age. The dominance of physical environmental concerns may have arisen from the industrial nature of Hamilton, but this result merits further investigation.

Do socioeconomic characteristics modify the short term association between air pollution and mortality? Evidence from a zonal time series in Hamilton, Canada.

STUDY OBJECTIVE: To assess the short term association between air pollution and mortality in different zones of an industrial city. An intra-urban study design is used to test the hypothesis that socioeconomic characteristics modify the acute health effects of ambient air pollution exposure. DESIGN: The City of Hamilton, Canada, was divided into five zones based on proximity to fixed site air pollution monitors. Within each zone, daily counts of non-trauma mortality and air pollution estimates were combined. Generalised linear models (GLMs) were used to test mortality associations with sulphur dioxide (SO(2)) and with particulate air pollution measured by the coefficient of haze (CoH). MAIN RESULTS: Increased mortality was associated with air pollution exposure in a citywide model and in intra-urban zones with lower socioeconomic characteristics. Low educational attainment and high manufacturing employment in the zones significantly and positively modified the acute mortality effects of air pollution exposure. DISCUSSION: Three possible explanations are proposed for the observed effect modification by education and manufacturing: (1) those in manufacturing receive higher workplace exposures that combine with ambient exposures to produce larger health effects; (2) persons with lower education are less mobile and experience less exposure measurement error, which reduces bias toward the null; or (3) manufacturing and education proxy for many social variables representing material deprivation, and poor material conditions increase susceptibility to health risks from air pollution.

Environmental influences on healthcare expenditures: an exploratory analysis from Ontario, Canada.

STUDY OBJECTIVE: This paper explores the relation between healthcare expenditures (HCEs) and environmental variables in Ontario, Canada. DESIGN: The authors used a sequential two stage regression model to control for variables that may influence HCEs and for the possibility of endogenous relations. The analysis relies on cross sectional ecological data from the 49 counties of Ontario. MAIN RESULTS: The results show that, after control for other variables that may influence health expenditures, both total toxic pollution output and per capita municipal environmental expenditures have significant associations with health expenditures. Counties with higher pollution output tend to have higher per capita HCEs, while those that spend more on defending environmental quality have lower expenditures on health care. CONCLUSIONS: The implications of our findings are twofold. Firstly, sound investments in public health and environmental protection have external benefits in the form of reduced HCEs. Combined with the other benefits such as recreational values, investments in environmental protection probably yield net social benefits. Secondly, health policy that excludes consideration of environmental quality may eventually result in increased expenditures. These results suggest a need to broaden the cost containment debate to ensure environmental determinants of health receive attention as potential complements to conventional cost control policies.

The spatial association between community air pollution and mortality: a new method of analyzing correlated geographic cohort data.

We present a new statistical model for linking spatial variation in ambient air pollution to mortality. The model incorporates risk factors measured at the individual level, such as smoking, and at the spatial level, such as air pollution. We demonstrate that the spatial autocorrelation in community mortality rates, an indication of not fully characterizing potentially confounding risk factors to the air pollution-mortality association, can be accounted for through the inclusion of location in the model assessing the effects of air pollution on mortality. Our methods are illustrated with an analysis of the American Cancer Society cohort to determine whether all cause mortality is associated with concentrations of sulfate particles. The relative risk associated with a 4.2 microg/m(3) interquartile range of sulfate distribution for all causes of death was 1.051 (95% confidence interval 1.036-1.066) based on the Cox proportional hazards survival model, assuming subjects were statistically independent. Inclusion of community-based random effects yielded a relative risk of 1.055 (1.033, 1.077), which represented a doubling in the residual variance compared to that estimated by the Cox model. Residuals from the random-effects model displayed strong evidence of spatial autocorrelation (p = 0.0052). Further inclusion of a location surface reduced the sulfate relative risk and the evidence for autocorrelation as the complexity of the location surface increased, with a range in relative risks of 1.055-1.035. We conclude that these data display both extravariation and spatial autocorrelation, characteristics not captured by the Cox survival model. Failure to account for extravariation and spatial autocorrelation can lead to an understatement of the uncertainty of the air pollution association with mortality.

Testing relationships among determinants of health, health policy, and self-assessed health status in Quebec.

By removing financial barriers, the Canada Health Act (1984) equalized access to health care services in Canada. Yet class, educational, and geographical disparities in individual and population health status persist. Recent health reform policies in Quebec assert that health and well-being are a function of income, educational level, housing conditions, employment, and other socioeconomic factors. They suggest that health policy should encompass social policies that influence individual and community socioeconomic factors which in turn affect health. Against the backdrop of these reforms, this study tests the importance of socioeconomic factors as a determinant of health--while controlling for other known determinants through a logistic regression model--with data from the Santé Quebec health surveys 1987 and 1992-93. The results confirm the importance of economic security as a determinant of individual health. This effect appears to operate through an individual income variable and through the community-level variable of regional unemployment. The importance of the income effect declined between 1987 and 1992-93. This may indicate that an increased focus on the socioeconomic determinants of health has reduced inequalities in health. It may also mean that health inequalities appear inevitable until health care policy merges completely with broader health and social policies. But such integration may well conflict with economic (and political) imperatives of the post-Fordist capitalist system.

Heterogeneity in the determinants of health and illness: the example of socioeconomic status and smoking.

Systematic variations in health and illness among social groups have persisted and, in some cases increased, in many countries in spite of improvements in the availability of, and access to, health care services. Health policy makers have responded by showing increasing interest in non-clinical determinants of health as a way of explaining the observed systematic variations in health and illness. Yet health care and non-health care "factors" are often seen as competing for society's scarce resources in the production of health. The purpose of this paper is to augment this traditional approach to understanding the determinants of health in populations by exploring heterogeneity in the smoking-health relationship between social groups. Logistic regression analyses were performed using data from the 1992-93 Santé Quebéc survey, a weighted random sample of the non-institutionalized population of the province of Québec, Canada (N = 23,564). The findings suggest that the likelihood of reporting health as poor or fair was greater for smokers than non-smokers. However, the difference between smokers and non-smokers was significantly greater among groups of the population with low incomes and without employment, but significantly less among sub-groups with lower levels of education. These findings suggest that the identification, evaluation and implementation of health care programmes need to pay greater attention to the interface between social, behavioural and clinical determinants of health.

Socioeconomic and environmental covariates of premature mortality in Ontario.

This paper contributes to debates on the broad determinants of health and on the policy shift from curative to preventive and protective interventions. It addresses empirically the relative importance of influences on health with a multiple regression analysis of ecologic data from the 49 counties of Ontario. One model achieved high predictive power (that is, Adj R2 > 75%, p < 0.0001). Educational levels were a strong predictor of population health, showing a consistent inverse relationship with premature mortality ratios for both sexes and it was the strongest predictor for females. A low income variable supplied the strongest prediction for male mortality. This variable displayed a positive association with male mortality. Municipal expenditures on environmental protection exerted a negative effect on male mortality. These findings raise questions about the current directions of health policy in Ontario where the provincial government has reduced funding to social and environmental programs, while promising to maintain health care funding.