Health-related behaviour in context: a multilevel modelling approach

Do the determinants of health differ between people living in the community and in institutions?

Determinants of health studies have mainly dealt with samples of community-dwelling subjects. We utilized the 1996/97 Canadian National Population Health Survey community and institutional surveys to identify factors associated with the selection of individuals to institutions. We also assessed whether the same determinants of health variables explain variations in health in those living in the community and those living in institutions. Logistic regression analysis was used to investigate factors associated with selection to institutions. Results showed that health status, age, chronic conditions, education, race, marital status, smoking and alcohol were associated with the probability of being in an institution. Multiple linear regression analyses (dependent variable: Health Utilities Index Mark 3 [HUI3] score) with individual characteristics, socio-economic status, and health risk factors were also estimated. Results showed that advanced age, higher number of chronic conditions, lower education, smoking and being an alcohol non-drinker were negatively and significantly associated with HUI3 scores for the community sample. Except for age, chronic conditions and being an alcohol non-drinker, none of these factors were significantly associated with HUI3 for the institutional sample. Moreover, the association between age and HUI3 was weaker for the institutional sample in comparison to the community sample. Implications are that, for those who are institutionalized, the usual determinants of health factors are less important. In conclusion, there appears to be important heterogeneity in determinants of health between persons living in the institution and community.

Transferability of economic evaluations across jurisdictions: ISPOR Good Research Practices Task Force report

ABSTRACT A growing number of jurisdictions now request economic data in support of their decision-making procedures for the pricing and/or reimbursement of health technologies. Because more jurisdictions request economic data, the burden on study sponsors and researchers increases. There are many reasons why the cost-effectiveness of health technologies might vary from place to place. Therefore, this report of an ISPOR Good Practices Task Force reviews what national guidelines for economic evaluation say about transferability, discusses which elements of data could potentially vary from place to place, and recommends good research practices for dealing with aspects of transferability, including strategies based on the analysis of individual patient data and based on decision-analytic modeling.

Neighbourhood deprivation and small-for-gestational-age term births in the United States

Residential context has received increased attention as a possible contributing factor to race/ethnic and socio-economic disparities in birth outcomes in the United States. Utilising vital statistics birth record data, this study examined the association between neighbourhood deprivation and the risk of a term small-for-gestational-age (SGA) birth among non-Hispanic whites and non-Hispanic blacks in eight geographical areas. An SGA birth was defined as a newborn weighing <10th percentile of the sex- and parity-specific birthweight distribution for a given gestational week. Multi-level random intercept logistic regression models were employed and statistical tests were performed to examine whether the association between neighbourhood deprivation and SGA varied by race/ethnicity and study site. The risk of term SGA was higher among non-Hispanic blacks (range 10.8-17.5%) than non-Hispanic whites (range 5.1-9.2%) in all areas and it was higher in cities than in suburban locations. In all areas, non-Hispanic blacks lived in more deprived neighbourhoods than non-Hispanic whites. However, the adjusted associations between neighbourhood deprivation and term SGA did not vary significantly by race/ethnicity or study site. The summary fully adjusted pooled odds ratios, indicating the effect of one standard deviation increase in the deprivation score, were 1.15 [95% CI 1.08, 1.22] for non-Hispanic whites and 1.09 [95% CI 1.05, 1.14] for non-Hispanic blacks. Thus, neighbourhood deprivation was weakly associated with term SGA among both non-Hispanic whites and non-Hispanic blacks.

Determinants of HIV prevalence among female sex workers in four south Indian states: analysis of cross-sectional surveys in twenty-three districts

OBJECTIVE

In four states in southern India we explored the determinants of HIV prevalence among female sex workers (FSW), as well as factors associated with district-level variations in HIV prevalence among FSW.

METHODS

Data from cross-sectional surveys in 23 districts were analysed, with HIV prevalence as the outcome variable, and sociodemographic and sex work characteristics as predictor variables. Multilevel logistic regression was applied to identify factors that could explain variations in HIV prevalence among districts.

RESULTS

HIV prevalence among the 10 096 FSW surveyed was 14.5% (95% confidence interval 14.0-15.4), with a large interdistrict variation, ranging from 2% to 38%. Current marital status and the usual place of solicitation emerged as important factors that determine individual probability of being HIV positive, as well as the HIV prevalence within districts. In multivariate analysis, compared with home-based FSW, the odds of being HIV positive was greater for brothel-based FSW [adjusted odds ratio (AOR) 2.17, P <or= 0.001] and for public place-based FSW (AOR 1.32, P = 0.005). Unmarried FSW and those who were widowed/divorced/separated, or from the devadasi tradition, had higher odds of being HIV positive (AOR 1.79, P <or= 0.001 and 1.98, P < 0.001, respectively), than those currently married. The estimated district level variance in HIV prevalence was lowest (0.152) for brothel-based unmarried FSW, followed by brothel-based widowed/divorced/separated or devadasi FSW (0.192).

CONCLUSION

Heterogeneity in the organization and structure of sex work is an important determinant of variations in HIV prevalence among FSW across districts in India, much more so than the districts themselves. This understanding should help to improve the design of HIV preventive interventions.

Neighborhood effects on birthweight: an exploration of psychosocial and behavioral pathways in Baltimore, 1995--1996

Neighborhood characteristics have been proposed to influence birth outcomes through psychosocial and behavioral pathways, yet empirical evidence is lacking. Using data from an urban, low-income sample, this study examined the impact of the neighborhood environment on birthweight and evaluated mediation by psychosocial and behavioral factors. The sample included 726 women who delivered a live birth at Johns Hopkins Hospital in Baltimore, Maryland, USA between 1995 and 1996. Census-tract data were used to create a principal component index of neighborhood risk based on racial and economic stratification (% Black, % poverty), social disorder (violent crime rate), and physical deterioration (% boarded-up housing) (alpha=0.82). Information on sociodemographic, psychosocial, and behavioral factors was gathered from a postpartum interview and medical records. Random intercept multilevel models were used to estimate neighborhood effects and assess potential mediation. Controlling for sociodemographic characteristics, a standard deviation increase in neighborhood risk conferred a 76g birthweight decrement. This represents an approximate 300g difference between the best and worst neighborhoods. Although stress (daily hassles), perceived locus-of-control, and social support were related to birthweight, their adjustment reduced the neighborhood coefficient by only 12%. In contrast, the neighborhood effect was reduced by an additional 30% and was no longer statistically significant after adjustment for the behavioral factors of smoking, drug use, and delayed prenatal care. These findings suggest that neighborhood factors may influence birthweight by shaping maternal behavioral risks. Thus, neighborhood level interventions should be considered to address multiple maternal and infant health risks. Future studies should examine more direct measures of neighborhood stress, such as perceived neighborhood disorder, and evaluate alternative mechanisms by which neighborhood factors influence behavior (e.g., social norms and access to goods and services).

Heterogeneity in mammography use across the nation: separating evidence of disparities from the disproportionate effects of geography

Background

Mammography is essential for early detection of breast cancer and both reduced morbidity and increased survival among breast cancer victims. Utilization is lower than national guidelines, and evidence of a recent decline in mammography use has sparked concern. We demonstrate that regression models estimated over pooled samples of heterogeneous states may provide misleading information regarding predictors of health care utilization and that comprehensive cancer control efforts should focus on understanding these differences and underlying causal factors. Our study population includes all women over age 64 with breast cancer in the Surveillance Epidemiology and End Results (SEER) cancer registries, linked to a nationally representative 5% reference sample of Medicare-eligible women located in 11 states that span all census regions and are heterogeneous in racial and ethnic mix. Combining women with and without cancer in the sample allows assessment of previous cancer diagnosis on propensity to use mammography. Our conceptual model recognizes the interplay between individual, social, cultural, and physical environments along the pathways to health care utilization, while delineating local and more distant levels of influence among contextual variables. In regression modeling, we assess individual-level effects, direct effects of contextual factors, and interaction effects between individual and contextual factors.

Results

Pooling all women across states leads to quite different conclusions than state-specific models. Commuter intensity, community acculturation, and community elderly impoverishment have significant direct impacts on mammography use which vary across states. Minorities living in isolated enclaves with others of the same race/ethnicity may be either advantaged or disadvantaged, depending upon the place studied.

Conclusion

Careful analysis of place-specific context is essential for understanding differences across communities stemming from different causal factors. Optimal policy interventions to change behavior (improve screening rates) will be as heterogeneous as local community characteristics, so no "one size fits all" policy can improve population health. Probability modeling with correction for clustering of individuals within multilevel contexts can reveal important differences from place to place and identify key factors to inform targeting of specific communities for further study.

The social epidemiologic concept of fundamental cause

The goal of research in social epidemiology is not simply conceptual clarification or theoretical understanding, but more importantly it is to contribute to, and enhance the health of populations (and so, too, the people who constitute those populations). Undoubtedly, understanding how various individual risk factors such as smoking and obesity affect the health of people does contribute to this goal. However, what is distinctive of much on-going work in social epidemiology is the view that analyses making use of individual-level variables is not enough. In the spirit of Durkheim and Weber, S. Leonard Syme makes this point by writing that just "as bad water and food may be harmful to our health, unhealthful forces in our society may be detrimental to our capacity to make choices and to form opinions" conducive to health and well-being. Advocates of upstream (distal) causes of adverse health outcomes propose to identify the most important of these "unhealthful forces" as the fundamental causes of adverse health outcomes. However, without a clear, theoretically precise and well-grounded understanding of the characteristics of fundamental causes, there is little hope in applying the statistical tools of the health sciences to hypotheses about fundamental causes, their outcomes, and policies intended to enhance the health of populations. This paper begins the process of characterizing the social epidemiological concept of fundamental cause in a theoretically respectable and robust way.

Changes in health risk behaviors of elementary school students in northern Taiwan from 2001 to 2003: results from the child and adolescent behaviors in long-term evolution study

Background

Previous research has indicated that children's behaviors have long-term effects on later life. Hence it is important to monitor the development of health risk behaviors in childhood. This study examined the changes in health risk behaviors in fourth- to sixth-grade students in northern Taiwan from 2001 to 2003.

Methods

The Child and Adolescent Behaviors in Long-Term Evolution (CABLE) study collected data from 1,820 students from 2001 to 2003 (students were 9 or 10 years old in 2001). Exploratory factor analysis was used to determine the aggregation of health risk behaviors. A linear growth curve model was used to determine whether health risk behaviors changed over time.

Results

Of the 13 behaviors, staying up late and eating snacks late at night were the most prevalent (82.3% of subjects in 2001, 81.8% in 2002, 88.5% in 2003) and second most prevalent (68.7%, 67.4%, 71.6%) behaviors, respectively, from 2001 to 2003. The three least prevalent health risk behaviors were chewing betel nut (1.0%, 0.4%, 0.2%), smoking (1.4%, 1.0%, 0.8%), and drinking alcohol (8.5%, 6.0%, 5.2%). The frequencies of swearing and staying up late showed the greatest significant increases with time. On the other hand, suppressing urination and drinking alcohol decreased over time. Using exploratory factor analysis, we aggregated the health risk behaviors into three categories: unhealthy habits, aggressive behaviors, and substance use. Although students did not display high levels of aggressive behavior or experimentation with substances, the development of these behaviors in a small proportion of students should not be ignored. The results of the linear growth curve model indicated that unhealthy habits and aggressive behaviors increased over time. However, substance use slightly decreased over time.

Conclusion

We found that some health risk behaviors increased with time while others did not. Unhealthy habits and aggressive behaviors increased, whereas substance use slightly decreased during this period. Educational professionals should pay attention to the different patterns of change in these behaviors in elementary school students.

Neighborhood income and income distribution and the use of cigarettes, alcohol, and marijuana

Evidence about the relationship between contextual variables and substance use is conflicting. Relationships between neighborhood income and income distribution and the prevalence and frequency of substance use in 59 New York City (NYC) neighborhoods were assessed while accounting for individual income and other socio-demographic variables. Measures of current substance use (in the 30 days prior to the survey) were obtained from a random-digit-dial phone survey of adult residents of NYC and data from the 2000 U.S. Census to calculate median neighborhood income and income distribution (assessed using the Gini coefficient). Among 1355 respondents analyzed (female=56.2%, mean age=40.4), 23.9% reported cigarette, 40.0% alcohol, and 5.4% marijuana use in the previous 30 days. In ecologic assessment, neighborhoods with both the highest income and the highest income maldistribution had the highest prevalence of drinking alcohol (69.0%) and of smoking marijuana (10.5%) but not of cigarette use; there was no clear ecologic association between neighborhood income, income distribution, and cigarette use. In multilevel multivariable models adjusting for individual income, age, race, sex, and education, high neighborhood median income and maldistributed neighborhood income were both significantly associated with a greater likelihood of alcohol and marijuana use but not of cigarette use. Both high neighborhood income and maldistributed income also were associated with greater frequency of alcohol use among current alcohol drinkers. These observations suggest that neighborhood income and income distribution may play more important roles in determining population use of alcohol and marijuana than individual income, and that determinants of substance use may vary by potential for drug dependence. Further research should investigate specific pathways that may explain the relationship between neighborhood characteristics and use of different substances.

Understanding the relative influence of neighborhood, family, and youth on adolescent drug use

In the United States, a variety of programs have been developed to prevent substance use among youth. These programs often target youth directly, and may also have components that address the relational influence of families, schools, and communities. We discuss clustering of youth marijuana use within and between households and neighborhoods. As often discussed in the literature, we consider analyzing "components of variance" in a hierarchical sample design with two or more levels. With a continuous outcome variable, the estimated relative size of variance components at each level can be interpreted as its relative "importance." We estimate variance components when the outcome is dichotomous, and find that for the use of marijuana in the past year, the role of the individual (individual adolescent vs. role of household vs. role of neighborhood) is quite prominent (79% of variation). A similar result is observed for the continuous scale variable of individual positive attitudes toward drug use (83%). For continuous constructs related to either household (parental monitoring) or neighborhood (neighborhood disorganization) the majority of variation still occurs at the individual level (67% and 51%, respectively), although they reveal significant percent variation (about 30%) at the corresponding family or neighborhood levels as well. We discuss the use of variance component methodology and the relevance for prevention programs.

Geography: research and teaching in nurse education

This paper outlines how geography might be integrated into nurse education. At one level, researching nurse education geographically could add to the current academic understanding of the many transitional places that make educational experiences and influence outcomes. At another level, as part of a nursing curriculum, teaching geographical concepts and issues to students might provide them with unique insights into core subjects.

Mental health, places and people: a multilevel analysis of economic inactivity and social deprivation

Using data on 24,975 respondents to the Welsh Health Survey 1998 aged 17-74 years, we investigated associations between individual mental health status measured using the SF-36 instrument, social class, economic inactivity and the electoral division Townsend deprivation score. In a multilevel modelling analysis, we found mental health was significantly associated with the Townsend score after adjusting for composition, and this effect was strongest in respondents who were economically inactive. Further contextual effects were shown by significant random variability in the slopes of the relation between mental health and economic inactivity at the electoral division level. Our results suggest that the places in which people live affect their mental health, supporting NHS policy that multi-agency planning to reduce inequalities in mental health status should address the wider determinants of health, as well as services for individual patients.

Measuring and modeling the social and geographic context of trauma: a multilevel modeling approach

Increasingly it is recognized that health and illness are products of individual level risk and protective factors, as well as forces operating at contextual levels. In this article, we present the motivation and rationale for understanding trauma within its context. We use the example of the concept of social capital to illustrate the relevance of the contextual approach for trauma research and outline a multilevel modeling approach to examining contextual influences on trauma outcomes.

Health disparities in Canada today: some evidence and a theoretical framework

This paper documents contemporary evidence on patterns of health disparities in Canada and suggests theoretical mechanisms that give rise to these patterns. The overall health of Canadians, as measured by life expectancy or mortality, has improved dramatically over the past 30 years and some disparities have diminished slightly (e.g., life expectancy by income group for men), while others have increased (e.g., diabetes for Aboriginal peoples). Arguably the most egregious health disparities in Canada are those existing between Aboriginals and the rest of the Canadian population. This paper focuses specifically on three social determinants and their effects on disparities in health; Aboriginal status, income, and place. Overall we take the approach that disparities in health could be alleviated by reducing inequities in the distribution of these determinants. We further argue that these social determinants are proxies for opportunities, resources and constraints; all of which influence health outcomes. We suggest that policies focus on reducing the social inequities that lead to health disparities in Canada, rather than focusing on the disparities in health alone. Since the social determinants described here have been found to influence an array of disease outcomes, tackling them, rather than their outcomes, may have a greater overall influence on the health of the population.

Effect of Age on the Relationship of Occupational Social Class with Prevalence of Modifiable Cardiovascular Risk Factors and Cardiovascular Diseases

BACKGROUND

Previous studies on cardiovascular risk profile in different socioeconomic status were focused on younger populations and many of them have not been able to take into account age and sex differences.

OBJECTIVES

To investigate the relationship of occupational social class with the prevalence of cardiovascular disease risk factors and cardiovascular diseases in younger (<65 years) and older (>or=65 years) men and women.

METHODS

A population-based-cross sectional study was conducted in a general community in Norfolk, United Kingdom. Participants were 23,085 men and women aged 40-79 years, recruited from general practice age-sex registers as part of European Prospective Investigation into Cancer-Norfolk (EPIC-Norfolk). The prevalence of cardiovascular risk factors and cardiovascular diseases were examined.

RESULTS

The prevalence of smoking was significantly higher in those in manual social classes particularly in the younger (<65) age group. Younger women in manual social classes were more likely to be smokers compared to older women in the same social class. Being in manual social classes was associated with higher cholesterol levels in women but lower cholesterol levels in men. Manual social class was associated with higher physical activity in those younger than 65 years but this association was reversed in those 65 years or older.

CONCLUSION

Occupational social class is differently related to cardiovascular risk factors in individuals depending on their age and sex. This may reflect differences in behavior at work and leisure, which vary by sex and pre- and postretirement. Interventions to promote health and reduce social inequalities need to take age and gender into account.

Geographic aspects of poverty and health in Tanzania: does living in a poor area matter?

Previous studies have consistently found an inverse relationship between household-level poverty and health status. However, what is not well understood is whether and how the average economic status at the community level plays a role in the poverty-health relationship. The purpose of this study is to investigate the concentration of poverty at the community level in Tanzania and its association with the availability and quality of primary health care services, the utilization of services, and health outcomes among household categories defined by wealth scores. A principal component method has been applied to rank households separately by urban/rural location using reported levels of asset ownership and living conditions. The household wealth scores were also used to classify communities into three cluster-types based on the proportion of households belonging to the poorest wealth tercile. On average, all the wealth terciles living in low poverty concentration areas were found to have better health outcomes and service utilization rates than their counterparts living in high poverty concentration clusters. Consistent with the finding is that high poverty concentration areas were further away from facilities offering primary health care than low poverty concentration areas. Moreover, the facilities closest to the high poverty concentration areas had fewer doctors, medical equipment and drugs. Among the high poverty concentration clusters, the 10 communities with the best women's body mass index (BMI) measures were found to have access to facilities with a greater availability of equipment and drugs than the 10 communities with the worst BMI measures. Although this study does not directly measure quality, the characteristics that differentiate high poverty concentration clusters from low poverty concentration clusters point to quality as more important than physical access among the study population.

Individual-Level and Neighborhood-Level Income Measures

BACKGROUND

Census-based measures of income often are used as proxies for individual-level income. Yet, the validity of such area-based measures relative to 'true' individual-level income has not been fully characterized.

OBJECTIVES

The objectives of this study were (1) to determine whether area-based measures of household income are a suitable proxy for self-reported household income and (2) to assess whether these measures are associated with outcomes in a cardiac disease cohort.

RESEARCH DESIGN

We used a prospective cohort from the Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease (APPROACH) cardiac catheterization registry.

SUBJECTS

A total of 4372 patients having undergone cardiac catheterization and who also completed a 1-year follow-up questionnaire on self-reported income level were studied.

MEASURES

Our measurements were survival to 2.5 years after catheterization and health-related quality of life (EuroQoL).

RESULTS

Agreement between the 2 income measures generally was poor (unweighted Kappa = 0.07), particularly for the low-income patients. Despite this poor agreement, both income measures were positively associated with survival and EuroQoL scores. An outcome analysis that simultaneously considered individual level income and area-based income revealed that low-income individuals have poorer survival and lower quality of life scores if they live in low income neighborhoods, but not if they live in high income neighborhoods.

CONCLUSIONS

The area-based estimates of household income in these data demonstrate poor agreement with self-reported household income at the level of individual patients, particularly for low-income patients. Despite this, both income measures appear to be prognostically relevant, perhaps because individual and neighborhood income measure different constructs.

Race differentials in obesity: the impact of place

This article reveals race differentials in obesity as both an individual- and neighborhood-level phenomena. Using neighborhood-level data from the 1990-1994 National Health Interview Survey, we find that neighborhoods characterized by high proportions of black residents have a greater prevalence of obesity than areas in which the majority of the residents are white. Using individual-level data, we also find that residents of neighborhoods in which at least one-quarter of the residents are black face a 13 percent increase in the odds of being obese compared to residents of other communities. The association between neighborhood racial composition and obesity is completely attenuated after including statistical controls for the poverty rate and obesity prevalence of respondents' neighborhoods. These findings support the underlying assumptions of both institutional and social models of neighborhood effects.

Space, Place, and the Evidence Base: Part II-Rereading Nursing Environment Through Geographical Research

This, the second and final article in the short health geography series, articulates how, moving beyond the models and assumptions associated with the metaparadigm of Nursing Environment, as a focused subdisciplinary approach, health geography might provide unique insights into nursing. A case study of a fictional yet somewhat typical children's hospital is presented and demonstrates some wide-ranging geographical issues and research questions (and hence potential geographical data) pertaining to nursing and the allied health professions. Indeed, this broad-brush approach is purposeful to make as many connections as possible to readers with varied theoretical, methodological, empirical, and practice expertise. In addition to the case study, to indicate further how geographical inquiry might locate quite comfortably in nursing research, the article also makes some initial and tentative connections between geography and an established nursing framework for the uptake of research evidence for practice. Although it is acknowledged that geographical inquiry should certainly never have the first call on researching the relationships between nurses and their environments, it is argued that its conceptual focus on space and place provides dedicated and detailed attention and a sound basis for a reformed, "spatialized" route to a more comprehensive understanding. Moreover, it is argued that it also demonstrates great versatility in terms of the scales and the subject matter with which it might engage. Some important issues certainly remain with respect to what might be the correct form of engagement between geographical and nursing research, but arguably, as a reformed disciplinary approach, health geography has the potential to provide a wealth of focused evidence for nursing practice.

‘Airplanes are flying nursing homes’: geographies in the concepts and locales of gerontological nursing practice

With the geographical literatures and issues outlined in the first paper very much in mind, this paper focuses specifically on the places in the concepts and locales of gerontological nursing practice. Particular attention is paid to the role of nurses in the making of long-term care institutions and to some fundamental priorities of care. Consideration is also given to some broader impacts of neighbourhood and community as an important spatial context to nursing practice based both in institutions and the community. In the final section, the paper makes some observations on how place-sensitive practice and research might contribute to the uptake of evidence for practice. These discussions set the scene for the final paper in this special section focused on the future research agenda for place and older people nursing.

Socioeconomic inequalities in early childhood malnutrition and morbidity: modification of the household-level effects by the community SES

This paper examines variations among communities in childhood malnutrition and diarrhea morbidity, explores the influences of socioeconomic status (SES) on child health, and investigates how the SES of families and that of communities interact in this process. Using multilevel modelling and data from Demographic and Health Surveys of five African countries, it shows evidence of contextual effects and a strong patterning in childhood malnutrition and morbidity along SES lines, with community SES having an independent effect in some instances. It also reveals that living in poorest conditions increases the odds of suffering from both malnutrition and diarrhea, as opposed to experiencing only one of the two outcomes. Importantly, community SES significantly modifies the effects of the household SES, suggesting that measures to improve access of mothers and children to basic community resources may be necessary preconditions for higher levels of familial socioeconomic situation to contribute to improved child health.

Mortality and long-term exposure to ambient air pollution: ongoing analyses based on the American Cancer Society cohort

This article provides an overview of previous analysis and reanalysis of the American Cancer Society (ACS) cohort, along with an indication of current ongoing analyses of the cohort with additional follow-up information through to 2000. Results of the first analysis conducted by Pope et al. (1995) showed that higher average sulfate levels were associated with increased mortality, particularly from cardiopulmonary disease. A reanalysis of the ACS cohort, undertaken by Krewski et al. (2000), found the original risk estimates for fine-particle and sulfate air pollution to be highly robust against alternative statistical techniques and spatial modeling approaches. A detailed investigation of covariate effects found a significant modifying effect of education with risk of mortality associated with fine particles declining with increasing educational attainment. Pope et al. (2002) subsequently reported results of a subsequent study using an additional 10 yr of follow-up of the ACS cohort. This updated analysis included gaseous copollutant and new fine-particle measurements, more comprehensive information on occupational exposures, dietary variables, and the most recent developments in statistical modeling integrating random effects and nonparametric spatial smoothing into the Cox proportional hazards model. Robust associations between ambient fine particulate air pollution and elevated risks of cardiopulmonary and lung cancer mortality were clearly evident, providing the strongest evidence to date that long-term exposure to fine particles is an important health risk. Current ongoing analysis using the extended follow-up information will explore the role of ecologic, economic, and, demographic covariates in the particulate air pollution and mortality association. This analysis will also provide insight into the role of spatial autocorrelation at multiple geographic scales, and whether critical instances in time of exposure to fine particles influence the risk of mortality from cardiopulmonary and lung cancer. Information on the influence of covariates at multiple scales and of critical exposure time windows can assist policymakers in establishing timelines for regulatory interventions that maximize population health benefits.

Decomposing the total variation in a nested random effects model of neighborhood, household, and individual components when the dependent variable is dichotomous: implications for adolescent marijuana use

Multilevel modeling techniques have become a useful tool that enables substance abuse researchers to more accurately identify the contribution of multiple levels of influence on drug-related attitudes and behaviors. However, it is difficult to determine the relative importance of the different hierarchical levels because, in the case of dichotomous outcomes, the variance components estimation involves calculations using a log-odds metric at the lowest level of estimation. We present methods introduced by Goldstein and Rasbash [Goldstein, H., Rasbash, J., 1996. Improved approximations for multilevel models with binary responses. J. Roy. Stat. Soc. A 159, 505-513.] to convert the variance components from the log-odds to the probability metric. This method provides a more logical and interpretable way to examine variation for nonlinear outcomes, which tend to be heavily utilized in substance use research. Using data from the National Household Survey on Drug Abuse [Substance Abuse and Mental Health Services Administration (SAMHSA), 2001. 1999 National Household Survey on Drug Abuse. Data Collection Final Report. Office of Applied Studies (OAS), Rockville, MD. Available at . Accessed on July 1, 2003.], we partition variation among individual, household, and neighborhood levels for the binary outcome of past year marijuana use to illustrate this approach. We also conduct a stability analysis to examine the robustness across different estimation procedures commonly available in commercial multilevel software packages. Finally, we partition the variance components using a conventional continuously distributed outcome and compare the relative magnitudes across binary and continuous outcomes.

Evaluation of the school-based smoking-prevention program "BE smokeFREE"

This article examines the impact of the school-based smoking-prevention program "BE smokeFREE" on adolescent smoking. A national representative sample of 99 schools (195 classes, 4,441 students) was used when the intervention started in November 1994. Schools were allocated to one of four groups: a comparison group (A) and three intervention groups (B, C, and D). Group B received the most comprehensive intervention. A baseline (autumn 1994) and three follow-up data collections (1995, 1996, and 1997) were conducted. There were no significant differences in smoking habits among the four groups at baseline. The smoking habits in the group that was involved in the most comprehensive intervention (group B) changed more favourably than those of students in the comparison schools over the three follow-up data collections. At the third follow-up, the proportion of students smoking weekly or more in the comparison group was 29.2%, compared with 19.6% in the model intervention group. The two less comprehensive interventions (no teacher in-service courses in group C, and no involvement of parents in group D) appeared to be less effective than the model intervention. Multilevel multiple logistic regression analyses, comparing changes in smoking habits between students in group B with those among students in the comparison schools, confirm the conclusion that the comprehensive intervention was the most effective. This school-based intervention, based on a social influence approach, proved to be effective at reducing smoking rates among participants.

Healthy eating and exercising to reduce diabetes: exploring the potential of social determinants of health frameworks within the context of community-based participatory diabetes prevention

OBJECTIVES

We examined a community-based participatory diabetes intervention to identify facilitators of and barriers to sustained community efforts to address social factors that contribute to health.

METHODS

We conducted a case study description and analysis of the Healthy Eating and Exercising to Reduce Diabetes project in the theoretical context of a conceptual model of social determinants of health.

RESULTS

We identified several barriers to and facilitators of analysis of social determinants of a community-identified disease priority (in this case, diabetes). Barriers included prevailing conceptual models, which emphasize health behavioral and biomedical paradigms that exclude social determinants of health. Facilitating factors included (1) opportunities to link individual health concerns to social contexts and (2) availability of support from diverse partners with a range of complementary resources.

CONCLUSIONS

Partnerships that offer community members tangible resources with which to manage existing health concerns and that integrate an analysis of social determinants of health can facilitate sustained engagement of community members and health professionals in multilevel efforts to address health disparities.

Moving beyond the mother-child dyad: women's education, child immunization, and the importance of context in rural India

The argument that maternal education is critical for child health is commonplace in academic and policy discourse, although significant facets of the relationship remain empirically and theoretically challenged. While individual-level analyses consistently suggest that maternal education enhances child health outcomes, another body of literature argues that the observed causality at the individual-level may, in fact, be spurious. This study contributes to the debate by examining the contextual effects of women's education on children's immunization in rural districts of India. Multilevel analyses of data from the 1994 Human Development Profile Index and the 1991 district-level Indian Census demonstrate that a positive and significant relationship exists between the proportion of literate females in a district and a child's complete immunization status within that district, above and beyond the child's own mother's education as well as district-level socioeconomic development and healthcare amenities. However, results also indicate that the effect of maternal education cannot be downplayed. Thus, increasing women's literacy at the community level, in addition to mother's access to higher education-such as matriculation and beyond-at the individual-level, emerge as effective developmental tools.

Ecological complexity and West Nile virus: perspectives on improving public health response

The emergence of West Nile Virus, as well as other emerging diseases, is linked to complex ecosystem processes such as climate change and constitutes an important threat to population health. Traditional public health intervention activities related to vector surveillance and control tend to be reactive and limited in their ability to deal with multiple epidemics and in their consideration of population health determinants. This paper reviews the current status of West Nile Virus in Canada and describes how complex systems and geographical perspectives help to acknowledge the influence of ecosystem processes on population health. It also provides examples of how these perspectives can be integrated into population-based intervention strategies.

Neighbourhoods and self rated health: a comparison of public sector employees in London and Helsinki

STUDY OBJECTIVE

Mortality and morbidity vary across neighbourhoods and larger residential areas. Effects of area deprivation on health may vary across countries, because of greater spatial separation of people occupying high and low socioeconomic positions and differences in the provision of local services and facilities. Neighbourhood variations in health and the contribution of residents' characteristics and neighbourhood indicators were compared in London and Helsinki, two settings where inequality and welfare policies differ.

DESIGN

Data from two cohorts were used to investigate associations between self rated health and neighbourhood indicators using a multilevel approach.

SETTING

London and Helsinki.

PARTICIPANTS

From the Whitehall II study (London, aged 39-63) and the Helsinki health study (aged 40-60).

MAIN RESULTS

Socioeconomic segregation was higher in London than in Helsinki. Age and sex adjusted differences in self rated health between neighbourhoods were also greater in London. Independent of individual socioeconomic position, neighbourhood unemployment, proportion of residents in manual occupations, and proportion of single households were associated with health. In pooled data, residence in a neighbourhood with highest unemployment was associated with an odds ratio of less than good self rated health of 1.51 (95% CI 1.30 to 1.75). High rates of single parenthood were associated with health in London but not in Helsinki.

CONCLUSIONS

Neighbourhood socioeconomic context was associated with health in both countries, with some evidence of greater neighbourhood effects in London. Greater socioeconomic segregation in London may have emergent effects at the neighbourhood level. Local and national social policies may reduce, or restrict, inequality and segregation between areas.

Urban sprawl and risk for being overweight or obese

OBJECTIVES

I examined the association between urban sprawl and the risk for being overweight or obese among US adults.

METHODS

A measure of urban sprawl in metropolitan areas was derived from the 2000 US Census; individual-level data were obtained from the Behavioral Risk Factor Surveillance System. I used multilevel analysis to assess the association between urban sprawl and obesity.

RESULTS

After I controlled for gender, age, race/ethnicity, income, and education, for each 1-point rise in the urban sprawl index (0-100 scale), the risk for being overweight increased by 0.2% and the risk for being obese increased by 0.5%.

CONCLUSIONS

The current obesity epidemic has many causes, but there is an association between urban sprawl and obesity.

A multilevel analysis of tobacco use and tobacco consumption levels in France: are there any combination risk groups?

BACKGROUND

Both the predictors for tobacco use and the determinants of the amounts of tobacco consumed by smokers should be taken into account when designing prevention programmes.

METHODS

Using a sample of 12,948 individuals representative of the French population in 1999, multilevel models were used to carry out a comparative investigation for the predictors of tobacco use and the determinants of the amount of tobacco consumed by smokers.

RESULTS

At the individual level, a combination of risks (higher risk of smoking and larger amounts of tobacco consumed by smokers) was found for males, for individuals with a low level of education and for divorcees. At the level of the area of residence, both the risk of smoking (odds ratio 1.07, 95% confidence interval: 1.01-1.12 for an increase by one standard deviation) and the amount of tobacco consumed among smokers (percentage variation +4%, 95% confidence interval: 0% - +8%) increased with the gross domestic product per capita.

CONCLUSION

This study justifies the combined use, in such analyses, of consumption levels for smokers in addition to the risk of smoking, in order to identify the profiles with the highest risk. It was possible to identify various groups with both a high risk of tobacco use and a high level of consumption among smokers, on the basis of individual (male, divorced or less educated) and environmental (living in a high GDP area) factors. The prevention efforts should thus be focused on such groups.

Stress and physical health: the role of neighborhoods as mediating and moderating mechanisms

Using data from the 1995 Detroit Area Study (N = 1106) in conjunction with tract-level data from the 1990 census, this paper evaluates the relationship between residential stability and physical health among black and white adults. Results suggest that neighborhood-level variation in health is primarily mediated by key sociodemographic characteristics of individuals (e.g., age, race, and socioeconomic status). However, a significant portion of health differentials across neighborhoods is due to disparate stress levels across neighborhoods. Further, high levels of neighborhood stability provide an important buffer to the otherwise deleterious effects of increased stress levels on adults' overall health.

Spatial Analysis of Body Mass Index and Smoking Behavior among WISEWOMAN Participants

BACKGROUND

The WISEWOMAN program focuses on reducing cardiovascular disease (CVD) risk factors by providing screening and lifestyle interventions for many low-income and uninsured women. To provide the most effective interventions possible, it is important to understand the characteristics of WISEWOMAN participants and their communities.

METHODS

We used baseline data collected for WISEWOMAN participants from five states (Connecticut, Michigan, Nebraska, North Carolina, and South Dakota) who had enrolled in WISEWOMAN between January 2001 and December 2002 in order to examine body mass index (BMI) and smoking behavior for evidence of spatial clustering. We then examined whether neighborhood characteristics in clusters of high-risk factors differed from neighborhood characteristics in other locations.

RESULTS

Six percent of the WISEWOMAN participants lived in ZIP codes with high-BMI clusters, and 4% lived in ZIP codes with high-smoking clusters. High-BMI and high-smoking clusters occurred, however, in different locations from each other. The high-BMI-clustered ZIP codes were, on average, located in more disadvantaged areas. Most of the differences between the high-smoking-clustered ZIP codes and the remaining ZIP codes were not statistically significant.

CONCLUSIONS

Our analysis revealed spatial clustering in CVD risk factors among WISE-WOMAN participants. We also found evidence of a correlation between high-BMI clusters and low socioeconomic status of the surrounding community. A more in-depth analysis of the relationship between risk factors (e.g., BMI) and community characteristics in clustered locations will provide further information concerning the role of the community in affecting individual behavior and should allow for tailoring interventions to reduce these risk factors more effectively.

Country of birth, socioeconomic position, and healthcare expenditure: a multilevel analysis of Malmö, Sweden

STUDY OBJECTIVE

The principle of equity aims to guarantee allocation of healthcare resources on the basis of need. Therefore, people with a low income and persons living alone are expected to have higher healthcare expenditures. Besides these individual characteristics healthcare expenditure may be influenced by country of birth. This study therefore aimed to investigate the role of country of birth in explaining individual healthcare expenditure.

DESIGN

Multilevel regression model based on individuals (first level) and their country of birth (second level).

SETTING

The city of Malmö, Sweden.

PARTICIPANTS

All the 52 419 men aged 40-80 years from 130 different countries of birth, who were living in Malmö, Sweden, during 1999.

MAIN RESULTS

At the individual level, persons with a low income and persons living alone showed a higher healthcare expenditure, with regression coefficients (and 95% confidence intervals) being 0.358 (0.325 to 0.392) and 0.197 (0.165 to 0.230), respectively. Country of birth explained a considerable part (18% and 13%) of the individual differences in the probability of having a low income and living alone, respectively. However, this figure was only 3% for having some health expenditure, and barely 0.7% with regard to costs in the 74% of the population with some health expenditure.

CONCLUSIONS

Malmö is a socioeconomically segregated city, in which the country of birth seems to play only a minor part in explaining individual differences in total healthcare expenditure. These differences seem instead to be determined by individual low income and living alone.

Income, Income Inequality, and Cardiovascular Disease Mortality: Relations Among County Populations of the United States, 1985 to 1994

OBJECTIVES

Despite the major contribution of cardiovascular disease (CVD) to total mortality, and reports demonstrating strong relations between income and CVD, the joint relations of population-level income and income inequality with CVD mortality are not well described. This study was undertaken to describe relations among population-level income, income equality, and mortality due to cardiovascular disease, coronary heart disease, and stroke.

METHODS

County income distributions were determined from 1990 census data, and CVD mortality rates were obtained from the Compressed Mortality File. Relations among income, income inequality, and CVD mortality were examined in stratified and Poisson regression analyses.

RESULTS

County income was inversely related and income inequality was directly related to CVD, coronary heart disease, and stroke mortality. Relations were strongest for stroke. Relations of stroke mortality with income inequality were strongest in low-income populations.

CONCLUSIONS

The CVD mortality experiences of county populations are related to both income and income distribution in a complex, disease-dependent manner. The authors' findings are especially relevant to the Southeast, a region of high income inequality, low income, and high stroke mortality.

Statistical and Substantive Inferences in Public Health: Issues in the Application of Multilevel Models

Multilevel statistical models have become increasingly popular among public health researchers over the past decade. Yet the enthusiasm with which these models are being adopted may obscure rather than solve some problems of statistical and substantive inference. We discuss the three most common applications of multilevel models in public health: (a) cluster-randomized trials, (b) observational studies of the multilevel etiology of health and disease, and (c) assessments of health care provider performance. In each area of investigation, we describe how multilevel models are being applied, comment on the validity of the statistical and substantive inferences being drawn, and suggest ways in which the strengths of multilevel models might be more fully exploited. We conclude with a call for more careful thinking about multilevel causal inference.

Places and health

This glossary aims to provide readers with some key conceptual tools with which to address the issue of place and health; it is hoped that it will provoke thought and debate on the range of ways that places are connected to health.

Mortalidade hospitalar e tempo de permanência: comparação entre hospitais públicos e privados na região de Ribeirão Preto, São Paulo, Brasil

A avaliação de desempenho dos serviços de saúde é essencial. A comparação de indicadores de desempenho requer o uso de estratégias de ajuste de risco. O objetivo deste artigo é avaliar variações no desempenho clínico, mensurado pela mortalidade e pelo tempo de permanência, entre hospitais públicos e privados, levando em conta diferenças nas características dos pacientes tratados. Este estudo é limitado à região de Ribeirão Preto, São Paulo, Brasil. Entre os anos de 1996 e 1998, 32.906 pacientes admitidos com diagnósticos cardiovasculares e respiratórios foram estudados. As variáveis usadas para o ajuste de risco dos indicadores de desempenho foram: sexo, idade, diagnóstico principal e medidas de gravidade baseada em comorbidade. Os resultados mostraram que o desempenho clínico dos hospitais públicos, mensurado pela mortalidade hospitalar ajustada (razão de chance = 0,41), é superior ao dos privados. Os hospitais públicos e privados não foram estatisticamente diferentes com relação ao tempo de permanência dos pacientes. Ainda que problemas conceituais e metodológicos devam ser resolvidos, taxa de mortalidade e outros indicadores de desempenho ajustados devem ser considerados como instrumentos úteis para identificar problemas de desempenho dos serviços de saúde.

Selection of ecologic covariates in the American Cancer Society study

The American Cancer Society (ACS) Study of the effects of long-term exposure to ambient air pollution on mortality used metropolitan areas to assign exposures to individual cohort members (Pope et al., 1995); these authors did not, however, control for any other place-specific variables in their analysis. Consequently, the study has been criticized on the basis that the association observed between air pollution and mortality may be confounded by other unmeasured ecologic covariates. To address this criticism, the reanalysis team selected a set of place-specific variables that measured determinants of health ranging from the biophysical environment to the social environment and the healthcare system. This article outlines the process by which place-specific ecologic covariates were selected; data measuring these variables were obtained and geographic boundaries for places were delineated. Issues involved in obtaining and using geographically based ecological data are examined within the context of the reanalysis of the ACS study. Both the ecological fallacy and the atomistic fallacy are addressed and an argument is made for the importance of studying the effects of place-specific variables that are integral or contextual in nature. Issues relating to the Modifiable Areal Unit Problem (MAUP) are explored with reference to using ZIP codes and data from a variety of sources. It is argued that differences in the geographical scale of variability for various pollutants may prove to be the key to distinguishing between their relative impacts on health and that multilevel analyses are essential for understanding the impact of social and environmental determinants of health. A number of determinants of health are then briefly examined in terms of their association with mortality, the appropriateness of their being measured at the metropolitan scale, and the availability of data for the 1980s from U.S. sources. Finally, the article presents the database of place-specific ecologic covariates that was incorporated into the ACS models during the reanalysis in order to account for the influence that place may have above and beyond ambient air pollution.

An examination of social capital and social disorganisation in neighbourhoods in the British household panel study

Recent developments in social science research suggest that social environmental factors may be important for explaining community variations in health. We investigate the structural sources of two mechanisms that produce community variations in health. Using survey data collected from a representative cross-section of British households we examine variations in neighbourhood social capital and neighbourhood social disorganisation across a sample of British neighbourhoods. Adjusting for respondent's attributes, we assess the effects of neighbourhood characteristics measured by the 1991 census in Britain. The results show that concentrated affluence, residential instability and ethnic heterogeneity predict social capital for women. Population density is the only neighbourhood characteristic to predict social capital for men. For both men and women concentrated disadvantage and population density are associated with social disorganisation. Residential instability is additionally associated with social disorganisation for women. For women it was found that neighbourhood characteristics interact with individual social class in accounting for variations in social capital, the effects of neighbourhood characteristics being larger for those in professional and managerial and skilled non-manual occupations. The results show that neighbourhood structural characteristics influence social organisation processes. This helps establish a link between the structural characteristics of neighbourhoods and individual health outcomes.

Neighborhood differences in social capital: a compositional artifact or a contextual construct?

Assessment of social capital at the neighborhood level is often based on aggregating individual perceptions of trust and reciprocity. Individual perceptions, meanwhile, are influenced through a range of individual attributes. This paper examines the socioeconomic and demographic attributes that systematically correlate with individual perception of social capital and determines the extent to which such attributes account for neighborhood differences in social capital. Using improved multilevel modeling procedures, we ascertain the extent to which differences in social capital perception can be ascribed to true neighborhood-level variations. The analysis is based on the 1994-95 Community Survey of the Project on Human Development in Chicago Neighborhoods (PHDCN). The response measure is based on survey respondent's perceptions of whether people in their neighborhood can be trusted. The results suggest that even after accounting for individual demographic (age, sex, race, marital status) and socioeconomic characteristics (income, education), significant neighborhood differences remain in individual perceptions of trust, substantiating the notion of social capital as a true contextual construct.

Can the school make a difference? A multilevel analysis of adolescent risk and health behaviour

The main aim of this article is to assess the relationship between the structural and (health) policy variables of the school and characteristics of the individual on the risk and health behaviour of adolescents. Individual and school level effects on seven health-related behaviours are simultaneously estimated, using multilevel modelling. The data are from the Flemish health behaviour in school-aged children study in Belgium. Data are used from 29 Flemish schools in which students (N=3225), school administrators (N=29) and teachers (N=1132) were surveyed with anonymous written questions. The analysis confirms previous findings concerning individual level effects. Although differences between schools in risk and health behaviour were found to originate mainly from differences in pupil characteristics, substantial variation between schools remained with regard to regular smoking, drinking habits and tooth brushing after controlling for individual effects. A wide range of school structure and policy variables were taken into account, but only few of them were found to influence the health and risk behaviour of young people. Moreover, the study could not detect an effect of health promotion policy at school. The analysis therefore only partially confirms the hypothesis that the school has an impact on the health behaviour of young people. The findings demonstrate the need for a more thorough examination of the paths by which schools can influence the health behaviour of their pupils.

Towards a more place-sensitive nursing research: an invitation to medical and health geography

During recent years, nursing research has adopted and integrated perspectives and theoretical frameworks from a range of social science disciplines. I argue however, that a lack of attention has been paid in past research to the subdiscipline of medical geography. Although this may, in part, be attributed to a divergence between research priorities and foci, traditional 'scientific' geographical approaches may still be relevant to a wide range of nursing research. Furthermore, a recasting, redirecting and broadening of medical geography in the 1990s, towards what is termed health geography, has enhanced the discipline and provided a more cultural and expansive recognition of health, and a more comprehensive understanding of the dynamic relationship between people, health and place. Given the increasing range of places where health-care is provided and received, and some recent linkages made between nursing and place by nurse-theorists, these newer perspectives and concepts may be particularly useful for interpreting nurses' and patients' relationships both within and with a variety of healthcare settings and living spaces. Indeed, although a more place-sensitive nursing research is potentially a trans-disciplinary academic endeavor, a range of geographical approaches would be central to such a project.

The working-class context of pregnancy smoking

The risk of smoking during pregnancy in the US is strongly associated with women's individual socioeconomic status (SES) but little is known about the influence of local area context. The aim of this study was to examine whether local-area characteristics increase the risk of smoking during pregnancy above and beyond individual SES. In a hospital-based cohort of 878 pregnant women in California, who delivered between 1980 and 1990, we compared risk of smoking during pregnancy based on individual and local-area factors. Adjusting for individual SES, neighborhood social class was related to smoking in early pregnancy. Living in a predominantly working-class area significantly increased the risk of pregnancy smoking for both working-class and non-working-class women. However, local-area economic and demographic indicators were not related to smoking early in pregnancy. Individual and family characteristics alone may be insufficient to explain smoking during pregnancy; the social class context of the places in which pregnant women live may also influence this behavior.

Youth smoking initiation: disentangling context from composition

This paper examines the relative contributions of compositional and contextual effects on the one hand, and behavioural and material factors on the other, to help explain smoking initiation. We present results from a cross-sectional, multi-level analysis linking data from pre-adolescents, their households and their neighbourhoods across Quebec, Canada. Results show a significant geographical area variation in youth smoking initiation. The variation is explained not only by individual characteristics (parents' smoking status, parents" education and pre-adolescents' age), but also by aspects of the social structure at the neighbourhood level. When both the individual level and territory level predictors are entered in the model, only 2.66% of the between-territory variance in smoking initiation remain unexplained.

Emerging measurement and statistical methods in physical activity research

Although many studies have attempted to identify mediators and moderators of changes in physical activity involvement, the literature is inconclusive regarding which variable(s) relate to physical activity behavior change. The Cooper 2001 Conference series dedicated a session to discussing measurement and statistical methods that could contribute to advancing this research agenda. This article focuses on four such methodologic approaches: qualitative; psychometric; latent-variable, structural equation modeling; and multilevel modeling. The article presents a brief overview of these methods and discusses potential advantages and limitations of using them.

The use of surveillance data and market research to promote physical activity

Using various types of data sources for assessing and monitoring physical activity behaviors on a population level adds to our ability to explain the relationships between individuals and their surrounding social and physical environments. This article presents the findings from part of a panel presentation on available data sets at the 2001 Cooper Conference on Innovative Approaches to Understanding and Influencing Physical Activity. First, an overview of large national epidemiologic and surveillance data sets is offered, followed by a discussion on the use of market segmentation data to complement more traditional sources of data by adding new dimensions to our understanding of target groups and potential intervention strategies. The relative advantages and disadvantages of using each type of data are also given, as well as recommendations for further use.

Multilevel analysis of situational drinking among Canadian undergraduates

Using a multi-level approach, we examined the contribution of drinking setting characteristics and of individual characteristics on the alcohol intake per drinking occasion. The data are drawn from the Canadian Campus Survey, a national mail survey conducted in 1998 with a random sample of 8,864 students in 18 universities. For each student, up to five drinking occasions were investigated, resulting in 26,348 drinking occasions among 6,850 drinkers. At the individual level this study focused on the university life experience. At the situational level, information about alcohol intake was recorded relative to why, when, where and with whom drinking occurred. Our results show that drinking setting is as important as the individual characteristics in explaining the alcohol intake per occasion. Policies aimed at reducing students alcohol intake may be more beneficial if they address both situational and individual factors.