Health-related behaviour in context: a multilevel modelling approach

Place effects on health: how can we conceptualise, operationalise and measure them?

In this paper we highlight what we consider to be a lack of adequate conceptualisation. operationalisation and measurement of "place effects". We briefly review recent historical trends in the study of the effects of place on health in industrial countries, and argue that "place effects" often appear to have the status of a residual category, an unspecified black box of somewhat mystical influences on health which remain after investigators have controlled for a range of individual and place characteristics. We note that the distinction between "composition" and "context" may be more apparent than real, and that features of both material infrastructure and collective social functioning may influence health. We suggest using a framework of universal human needs as a basis for thinking about how places may influence health, and recommend the testing of hypotheses about specific chains of causation that might link place of residence with health outcomes.

A theoretical and empirical analysis of context: neighbourhoods, smoking and youth

Numerous studies are currently addressing the issue of contextual effects on health and disease outcomes. The majority of these studies fall short of providing a theoretical basis with which to explain what context is and how it affects individual disease outcomes. We propose a theoretical model, entitled collective lifestyles, which brings together three concepts from practice theory: social structure, social practices and agency. We do so in an attempt to move away from both behavioural and structural-functionalist explanations of the differential distribution of disease outcomes among areas by including a contextualisation of health behaviours that considers their meaning. We test the framework using the empirical example of smoking and pre-adolescents in 32 communities across Québec, Canada. Social structure is operationalised as characteristics and resources; characteristics are the socio-economic aggregate characteristics of individuals culled from the 1996 Canadian Census, and resources are what regulates and transforms smoking practices. Information about social practices was collected in focus groups with pre-adolescents from four of the participating communities. Using zero-order and partial correlations we find that a portrait of communities emerges. Where there is a high proportion of more socio-economically advantaged people, resources tend to be more smoking discouraging, with the opposite being true for disadvantaged communities. Upon analysis of the focus group material, however, we find that the social practices in communities do not necessarily reflect the "objectified" measures of social structure. We suggest that a different conceptualisation of accessibility and lifestyle in contextual studies may enable us to improve our grasp on how differential rates of disease come about in local areas.

Do places matter? Socioeconomic disadvantage and behavioral problems of children in Canada

This study evaluated the influence of neighborhoods and socioeconomic disadvantage on behavioral problems rated by parents and teachers in a nationally representative sample of children ages 4 to 11 years living in Canada. Between-neighborhood variation accounted for 7.6% and 6.6% of parent and teacher ratings, respectively. About 25.0% of this neighborhood variation could be explained by socioeconomic variables evenly divided between neighborhood and family-level measures. Family socioeconomic status, lone-parent family status, and percentage of lone parents in neighborhoods were strong, reliable predictors of behavioral problems. Ratings were contextualized: Fewer behavioral problems were assessed in children from well-off families living in disadvantaged neighborhoods, whereas more problems were assessed in children from poor families living in advantaged neighborhoods.

Predicting small area health-related behaviour: a comparison of multilevel synthetic estimation and local survey data

A recent paper in Social Science and Medicine (Twigg et al. 50 (2000) 1109) outlined an approach to the estimation of prevalences of small-area health-related behaviour using multilevel models. This paper compares results from the application of the multilevel approach with those derived using the more traditional strategy of the local 'lifestyle' survey. Estimations of smoking prevalence and high alcohol consumption are examined and critical assessments made of both estimation approaches. It is concluded that the alternative method is more suited towards the prediction of smoking rates as opposed to unsafe alcohol consumption.

Social environment effects on health and aging: integrating epidemiologic and demographic approaches and perspectives

This paper provides an overview of epidemiological and demographic research linking social characteristics of both individuals and communities to differences in both morbidity and mortality risks. Evidence is presented linking three broad aspects of the social environment to health--the network of personal social relationships within which most of us live our lives, individual socioeconomic status (SES), and community-level social characteristics. Large and consistent bodies of literature from both epidemiology and demography provide clear evidence for the generally health-promoting effects of personal social relationships and SES. The bulk of the evidence relates to mortality although both fields have begun to examine other health outcomes, including aspects of physical and cognitive functioning as well as disease outcomes. A smaller but growing body of community-level data, reflecting both the socioeconomic/resource characteristics of these broader communities and, more specifically, social features of these environments, also point to health impacts from these more macro level social environment characteristics. Much remains to be elucidated, however, concerning the actual mechanisms through which something as complex and multifaceted as SES "gets under the skin." This necessarily includes consideration of external characteristics of the environments (both physical and sociocultural) where people live and work, and individual characteristics, as well as possible interactions between these in producing the observed SES gradients in health and mortality. These questions concerning links between social environment conditions and health may be a particularly fruitful area of future collaboration, drawing on the shared interest of demographers and epidemiologists in understanding how different social conditions promote variation in distributions of better versus worse health outcomes within a population.

Ecologic versus individual-level sources of bias in ecologic estimates of contextual health effects

A number of authors have attempted to defend ecologic (aggregate) studies by claiming that the goal of those studies is estimation of ecologic (contextual or group-level) effects rather than individual-level effects. Critics of these attempts point out that ecologic effect estimates are inevitably used as estimates of individual effects, despite disclaimers. A more subtle problem is that ecologic variation in the distribution of individual effects can bias ecologic estimates of contextual effects. The conditions leading to this bias are plausible and perhaps even common in studies of ecosocial factors and health outcomes because social context is not randomized across typical analysis units (administrative regions). By definition, ecologic data contain only marginal observations on the joint distribution of individually defined confounders and outcomes, and so identify neither contextual nor individual-level effects. While ecologic studies can still be useful given appropriate caveats, their problems are better addressed by multilevel study designs, which obtain and use individual as well as group-level data. Nonetheless, such studies often share certain special problems with ecologic studies, including problems due to inappropriate aggregation and problems due to temporal changes in covariate distributions.

Socioeconomic differences in road traffic injuries during childhood and youth: a closer look at different kinds of road user

STUDY OBJECTIVE

To investigate if there are socioeconomic differences in road traffic injuries among Swedish children and adolescents, and if this applies to the same extent to all categories of road users. To assess the modification effect of gender of child.

DESIGN

A closed population-based cohort study based on the Swedish Population and Housing Census of 1985. Individual census records are linked to Sweden's National Hospital Discharge Register (1987-1994).

SETTING AND SUBJECTS

All children aged 0-15 years in 1985 (approximately 1.5 million subjects) were monitored for five categories of road traffic injuries over eight years, and divided into seven socioeconomic groups on the basis of parental socioeconomic status. Odds ratios and population attributable risks were computed using the children of intermediate and high level salaried employees as reference group.

MAIN RESULTS

The injury risks of pedestrians and bicyclists are 20% to 30% higher among the children of manual workers than those of intermediate and high level salaried employees. Socioeconomic differences are greatest for injuries involving motorised vehicles-that is, moped, motorcycle and car. If all children had the same rate as children in the reference group, the rate for all groups would be 25% lower for moped riders and 37% lower for car drivers.

CONCLUSIONS

Socioeconomic differences in road traffic injuries are substantial for both boys and girls. Socioeconomic injury-risk differentials increase when young people use motorised vehicles.

The relationship of neighborhood socioeconomic characteristics to birthweight among 5 ethnic groups in California

OBJECTIVES

This study sought to examine relationships between neighborhood socioeconomic characteristics and birthweight, accounting for individual socioeconomic characteristics, among 5 ethnic groups.

METHODS

Birth records were linked to census block-group data for 22 304 women delivering infants at 18 California hospitals during 1994-1995. Information on income and additional factors was obtained from a surveyed subset of 8457 women. Neighborhood levels of poverty, unemployment, and education were examined.

RESULTS

After adjustment for mothers' individual socioeconomic characteristics and other risk factors, less-favorable neighborhood socioeconomic characteristics were associated with lower birthweight among Blacks and Asians. No consistent relationship between neighborhood socioeconomic characteristics and birthweight was found among Whites, US-born Latinas, or foreign-born Latinas overall, but birthweight increased with less-favorable neighborhood socioeconomic characteristics among foreign-born Latinas in high-poverty or high-unemployment neighborhoods. These findings were not explained by measured behavioral or cultural factors.

CONCLUSIONS

In addition to individual socioeconomic characteristics, living in neighborhoods that are less socioeconomically advantaged may differentially influence birthweight, depending on women's ethnicity and nativity.

Does the state you live in make a difference? Multilevel analysis of self-rated health in the US

This paper investigates the different sources of variation between US states in self-rated health using multilevel statistical procedures. The different sources that are considered are based on individual- and state-level factors. Data for the analysis comes from the 1993-94 Behavioral Risk Factor Surveillance System and the 1986-90 General Social Surveys. Results show that individual-level factors (such as low income, being black, smoking) are strongly associated with self-rated poor health. Significant variation, however, remain between states after allowing for individual characteristics. Crucially, between-state variation in self-rated health is different for different income groups. State-level contextual effects are found for per-capita median-income and 'social capital'. While not strong, there seems to be a differential impact of state income-inequality on high-income groups, such that the affluent report better health from living in high inequality states. The paper substantiates the need to connect individual health to their macro socioeconomic context. Importantly, it is argued that without adopting an explicitly multilevel approach, the debate on linkages between individual health and income-inequality/social capital cannot be adequately addressed.

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.

Short term power output in relation to growth and maturation

OBJECTIVE

To examine short term power output during growth and maturation using a multilevel modelling approach.

METHODS

Body mass, stature, and triceps and subscapular skinfold thicknesses of boys and girls, aged 12.2 (0.4) years (mean (SD)) at the onset of the study, were measured at age 12, 13, and 17 years. Sexual maturation, classified according to Tanner's stage of pubic hair development, was assessed on the first two occasions and assumed to be stage 5 at 17 years. Peak power (PP) and mean power (MP) were assessed on each occasion using the Wingate anaerobic test.

RESULTS

Initial models, founded on 417 determinations of short term power output, identified body mass, stature, and age as significant explanatory variables of both PP and MP. The values for girls were significantly lower than those for boys, and a significant age by sex interaction described a progressive divergence in the MP of boys and girls. The introduction of sum of two skinfold thicknesses produced a model with an improvement in fit as indicated by a significant change in log likelihood. The stature term was negated and the body mass term increased. The age and sex terms were reduced but remained significant. The age by sex interaction term remained a significant explanatory variable for MP. Maturity effects were non-significant additional explanatory variables in all models of power output.

CONCLUSION

The values of PP and MP for boys are higher than those for girls, and, for MP, sex differences increase with age. Body mass and skinfold thicknesses are significant influences on both PP and MP, but age exerts a positive but non-linear effect on power output independent of body size and fatness.

Multilevel analysis in public health research

Over the past few years there has been growing interest in considering factors defined at multiple levels in public health research. Multilevel analysis has emerged as one analytical strategy that may partly address this need, by allowing the simultaneous examination of group-level and individual-level factors. This paper reviews the rationale for using multilevel analysis in public health research, summarizes the statistical methodology, and highlights some of the research questions that have been addressed using these methods. The advantages and disadvantages of multilevel analysis compared with standard methods are reviewed. The use of multilevel analysis raises theoretical and methodological issues related to the theoretical model being tested, the conceptual distinction between group- and individual-level variables, the ability to differentiate "independent" effects, the reciprocal relationships between factors at different levels, and the increased complexity that these models imply. The potentialities and limitations of multilevel analysis, within the broader context of understanding the role of factors defined at multiple levels in shaping health outcomes, are discussed.

Evaluating the impact of a locality based social policy intervention on mental health: conceptual and methodological issues

Urban regeneration initiatives provide an opportunity for examining the impact of changes in socio-economic circumstances on the mental health of different groups and individuals within localities. This paper sets out the conceptual and methodological bases for evaluating the impact of a population based social policy intervention on mental health. We suggest the need to integrate a range of disciplinary and methodological developments in research on health inequalities in exploring the impact of urban regeneration on mental health. A combination of multi-level modelling, subjective indicators and narrative accounts of individuals about mental health in the context of locality and personal changes are central for developing theories and methods appropriate for exploring the action and interaction of effects operating between structural and individual/agency levels.

Small area variations in health related behaviours; do these depend on the behaviour itself, its measurement, or on personal characteristics?

In this paper we examine the patterning, by small areas, of four health related behaviours (smoking, alcohol consumption, diet, and exercise) in the West of Scotland, after controlling for a range of individual/household characteristics, using multilevel models. Smoking and drinking were measured both as binary and as continuous variables, and diet and exercise were each measured in two ways: 'good' (health promoting) and 'bad' (health damaging). 'Area effects' (unattributed variation by post code sector) were found for 'bad' diet only. 'Good' and 'bad' diet, 'bad' exercise patterns and current smoking were associated with postcode sector deprivation. For 'bad' diet this effect was found only for individuals in more affluent households, and for 'good' exercise and current smoking the association with area deprivation differed between adolescents and adults. We conclude that the influence of area on health related behaviours varies according to the behaviour and the way it is measured, and that the influence of area deprivation and/or of area can vary by age and household deprivation.

A multilevel city health profile of Moscow

This report describes a multilevel city-wide profile of physical health in Moscow, examining individual and urban level factors. Objectives of the paper were to: (1) identify macro and micro risk factors for poor physical health in Moscow;(2) assess the effect of two dimensions of micro determinants--personal health habits and social connectivity, such as social cohesion, social support, and social networks; (3) examine the hypothesis that relative social inequality is a significant structural condition at the community level which influences the physical health of individuals, as a main and as a joint effect with psychosocial behaviors. A random sample of Moscow adults, with household telephones (N = 2000), was collected 17-19 September, 1991, and had a completed interview rate of 81.8%. The questionnaire replicated items from the California Alameda Study and the US Health Interview Survey. Respondents' urban area of residence was linked to macro measures of inequality derived from the Moscow census. This report describes the baseline survey of a prospective study design. Results of this study demonstrate that the social context in a community affects the health of people living there independently from the effects of individual health lifestyle or social connectivity. The structural conditions in Moscow which significantly increased the vulnerability of specific social groups for poor physical health were identified in a hierarchical linear regression: relative social inequality in the form of income inequality; urban area poverty risks; and mean level of alcohol consumption in urban areas. The psychosocial conditions included poor diet, lack of social cohesion and social support, involvement in formal social networks in the form of professional groups. A multilevel theoretical perspective is important for defining the targets of preventive health policy by identifying the structural conditions which increase the health disadvantage of some social groups. Further research is needed in refining the concept of relative social inequality, as well as investigating whether psychosocial factors, such as social cohesion, are mediating links between sick societies and their sick citizens.

Disclosure to the family of breast/ovarian cancer genetic test results: patient's willingness and associated factors

Informed probands are key actors for disclosing genetic information to their relatives when a mutation has been identified in the family. The objectives were to study women's attitudes towards the family disclosure of positive breast cancer genetic testing results and to determine the predictive factors of the diffusion patterns observed. A national multi-center cross-sectional survey was carried out at five French cancer genetic clinics during a 1-year period. Self-administered questionnaires were completed after the consultation by 84.5% (398/471) of women attending breast cancer genetic clinics for the first time. Among the 383 respondents who had at least one living first-degree relative to inform, 8.6% would inform none, 33.2% would inform at least one of them, and 58.2% would inform all of them. The sibship would be the most frequently informed blood relatives, sisters in 86.9% and brothers in 79% compared with mother in 71.4%, children in 70.4%, and father in 64.9%. Women of the family would be more frequently informed than men (P < 0.05). After multivariate adjustment, age, the fact to be affected by cancer, the number of daughters, and the emotional disturbance due to cancer in a close relationship were the main determinants (P < 0.05) of the diffusion patterns observed. The first step of the relatives' attendance to genetic counseling and the proband's willingness to disclose breast cancer genetic tests results was high in this study and was clearly dependent on the women's personal and emotional characteristics.

Contributions of acute bouts of vigorous physical activity to explaining diurnal variations in feeling states in active, middle-aged women

This study describes diurnal variations in feeling states in 84 women as a function of involvement in physical activity. Women completed feeling state checklists before and after bouts of vigorous physical activity that lasted a minimum of 20 min and on a stratified random-sampling protocol in response to a stimulus from an electronic pager. Participants self-reported the date, the time of day, the stimulus for responding, their current feeling states, and their ongoing activity. A total of 7,295 complete mood and activity reports were recorded in response to pager calls that were further coded into experience samplings that occurred on a nonexercise day, prior to exercise on an exercise day, and following exercise on an exercise day. Multilevel modeling analyses showed that in samplings recorded following exercise on exercise days, positive engagement, revitalization, and tranquility were elevated in comparison with predicted diurnal patterns.

Do patients matter? Contribution of patient and care provider characteristics to the adherence of general practitioners and midwives to the Dutch national guidelines on imminent miscarriage

OBJECTIVE

To assess the relative contribution of patient and care provider characteristics to the adherence of general practitioners (GPs) and midwives to two specific recommendations in the Dutch national guidelines on imminent miscarriage. The study focused on performing physical examinations at the first contact and making a follow up appointment after 10 days because these are essential recommendations and there was much variation in adherence between different groups of providers.

DESIGN

Prospective recording by GPs and midwives of care provided for patients with symptoms of imminent miscarriage.

SETTING

General practices and midwifery practices in the Netherlands.

SUBJECTS

73 GPs and 38 midwives who agreed to adhere to the guidelines; 391 patients were recorded during a period of 12 months.

MAIN MEASURES

Adherence to physical examinations and making a follow up appointment were measured as part of a larger prospective recording study on adherence to the guidelines on imminent miscarriage. Patient and care provider characteristics were obtained from case recordings and interviews, respectively. Multilevel analysis was performed to assess the contribution of several care provider and patient characteristics to adherence to two selected recommendations: the number of recommended physical examinations at the first contact and the number of days before a follow up appointment took place.

RESULTS

In the multilevel model explaining variance in adherence to physical examinations, the care provider's acceptance of the recommendations was the most important factor. Severity of symptoms and referral to an obstetrician were significant factors at the patient level. In the model for follow up appointments the characteristics of the care provider were less important. Referral to an obstetrician and probability diagnosis were significant factors at the patient level.

CONCLUSIONS

The study showed that characteristics of both the patient and care provider contribute to the variability in adherence. Furthermore, the contribution of the characteristics differed per recommendation. It is therefore advised that the contribution of both patient and care provider characteristics per recommendation should be carefully examined. If implementation is to be successful, strategies should be developed to address these specific contributions.

Do attitude and area influence health? A multilevel approach to health inequalities

This paper demonstrates an area effect on health and suggests improvements in research practice for work in this field. We use an area classification which can be related clearly to factors that influence the health of individuals and take account of people's differing propensities to draw influence from their area. Multilevel analysis demonstrates that the degree of deindustrialisation which an area experienced in Britain, in the 1980s, has an independent association with the health of resident individuals. A significant relationship between a person's attitude to their community and their health is shown to be independent of individual and area characteristics. We conclude that both individual and area characteristics influence health.

Distance and health care utilization among the rural elderly

This paper explores the relationship between distance and the utilization of health care by a group of elderly residents in rural Vermont. By drawing on recent work on the geography of health we frame the decision to visit a primary care physician in the context of the experience of place. The paper devises a test of this broader reading of the role of distance for utilization, and operationalizes this test using a custom designed survey. Using a randomized mail survey of elderly residents of Vermont's North East Kingdom we explore how grocery shopping, travel to work, home location relative to local services, access to private transportation, and living arrangements are associated with the number of doctor visits made to primary health care providers. Although the results confirm the idea that increased distance from provider does reduce utilization, they strongly suggest that distance to provider is a surrogate for location in a richer web of relations between residents and their local communities. We conclude by calling for further research that establishes links between place and the use of health facilities.

Ecological effects in multi-level studies

Multi-level research that attempts to describe ecological effects in themselves (for example, the effect on individual health from living in deprived communities), while also including individual level effects (for example, the effect of personal socioeconomic disadvantage), is now prominent in research on the socioeconomic determinants of health and disease. Such research often involves the application of advanced statistical multi-level methods. It is hypothesised that such research is at risk of reaching beyond an epidemiological understanding of what constitutes an ecological effect, and what sources of error may be influencing any observed ecological effect. This paper aims to present such an epidemiological understanding. Three basic types of ecological effect are described: a direct cross level effect (for example, living in a deprived community directly affects individual personal health), cross level effect modification (for example, living in a deprived community modifies the effect of individual socioeconomic status on individual health), and an indirect cross level effect (for example, living in a deprived community increases the risk of smoking, which in turn affects individual health). Sources of error and weaknesses in study design that may affect estimates of ecological effects include: a lack of variation in the ecological exposure (and health outcome) in the available data; not allowing for intraclass correlation; selection bias; confounding at both the ecological and individual level; misclassification of variables; misclassification of units of analysis and assignment of individuals to those units; model mis-specification; and multicollinearity. Identification of ecological effects requires the minimisation of these sources of error, and a study design that captures sufficient variation in the ecological exposure of interest.

Predicting small-area health-related behaviour: a comparison of smoking and drinking indicators

Health-related behaviours are of central importance to health promotion and to the promotion of enhanced population health. In the UK, localised knowledge of the quantitative dimensions of health-related behaviours is traditionally attained by conducting a costly sample survey. Such surveys seldom generate reliable data at scales more local than that of the health authority, they also need to be repeated regularly. This paper outlines an alternative framework for generating statistics on small-area health related behaviours using routinely available data from the annual Health Survey for England (N = 17,000) and the decennial Population Census. Using a multilevel modelling approach nesting individuals within postcode sectors within health authorities, and focusing on the prevalence of smoking and 'problem' drinking, the paper comprises four sections: a consideration of the modelling strategy, a comparison of the smoking and drinking models, an outline of the estimation strategy, and the presentation and discussion of ward-level estimates of smoking and drinking behaviour for England. The paper concludes that the method is better at estimating smoking than drinking but that it offers a feasible, cheap and more informative alternative to the survey approach to the generation of information on smoking and drinking behaviour.

Death and deprivation: an exploratory analysis of deaths in the health and lifestyle survey

An analysis is undertaken of deaths of respondents in the UK- representative Health and Lifestyle Survey. The sample was originally interviewed in 1984/5 and followed initially until May 1997. Using multilevel logistic and Cox-proportional hazards models, the relationships between death and a wide range of social circumstances and behaviours is explored. It is found that place deprivation interacts with individual social class in accounting for variations in mortality. This is the case even when account is taken of personal health-related behaviour. There appears to be some evidence of a threshold relationship such that the differential effects of social class are only found at high-levels of deprivation. No statistically significant interactions are found for social and behavioural variables, for behavioural and place deprivation variables, and for social and place deprivation variables with the exception of social class. The study is deliberately exploratory and a wide range of models have been fitted which will be subject to more rigorous evaluation as the HALS death study proceeds.

Collective lifestyles as the target for health promotion

The last five years have witnessed intense debate among health researchers in Canada regarding the overlap of the health promotion and population health discourses. Meanwhile, strong currents within health promotion have attempted to move the field beyond a focus on individual behaviour towards the influence of social environments on health, although the tendency is often to fall back on individual behaviour modification as the primary lever for change. The Population Health research agenda bypasses behavioural determinants of health and explores instead social determinants. This body of knowledge provides useful insight for addressing some of the tensions in the health promotion discourse. This paper explores two of these tensions: whether individuals at risk or general populations should be targeted for change; and whether lifestyle is an individual or a collective attribute. We propose the notion of collective lifestyles as a heuristic for understanding the interaction between social conditions and behaviour in shaping health.

Health, environmental assessments and population health: tools for a complex process

Place is more than physical and natural environment. The role of biophysical environment has still to be articulated in population health discourse and its relations with human health are fraught with scientific uncertainty and dissension. An environmental impact assessment (EA) evaluates the environmental effects of a proposal--a rational and technical process. Sometimes health assessments are included, usually by quantitative risk assessments which are subject to the limits of scientific knowledge and bedevilled by data limitations. The goal must be to add health to the process, yet the relevant features to include are complex. Impacts are non-specific and they interact and have spatial and temporal characteristics. To integrate environment into population health, there is a need for a physical environment-health database and inter-sectorial policy and action. There is also a need for different types of indicators to measure process, impact and effectiveness, and for new tools (stories, photography) to account for context and values.

Social determinants and inequalities in urban health

Social and environmental factors are health determinants, in association with behavior, biological factors, and health services. Whereas socioeconomic characteristics (age, gender, ethnicity, social status) describe individuals, social determinants work through broader policies that are influenced by governments. The relation between health and four social areas is discussed. Social capital, measured as social networks and social support, appears to be protective in developing some heart disease and mental illnesses; job control at work is also found to protect against heart disease; early life experiences affect both biological and social development; and the degree of income inequality within societies correlates with health status. The Independent Report on Inequalities in Health, published in the United Kingdom in 1998, is also reviewed. The report (a) briefly describes inequalities in health by social class, sex, and ethnicity; (b) reviews the literature on policy areas that affect health; (c) includes a section indicating relatively little inequality of access or provision in the National Health Service; and (d) makes 132 recommendations. Social determinants is a new area of research having the potential to link epidemiology and environmental sciences at small area level.

[Multilevel models or the importance of ranking]

Many researchers in Public Health have data bases with a hierarchical structure. The studied patients (level 1) can be nested in groups, i.e., district, doctor, hospital, etc. (level 2). It is possible that patients in the same group be similar, so traditional regression models can not be used because the hypothesis of independent observations is not satisfied. A Multilevel Analysis, using hierarchical models, can be a solution for this problem; these models take into account the distribution of the data at different levels to estimate two types of variability: one due to individuals in the study and another due to the groups in which patients are nested. These types of models were applied in education in the last decade, however they have been recently applied in Health Research. This paper is a review about multilevel analysis. A discussion about hierarchichal models versus traditional regression models is presented and some applications in Epidemiology and Health Research are showed.

The social environment and health: a discussion of the epidemiologic literature

The environment can be thought of in terms of physical and social dimensions. The social environment includes the groups to which we belong, the neighborhoods in which we live, the organization of our workplaces, and the policies we create to order our lives. There have been recent reports in the literature that the social environment is associated with disease and mortality risks, independent of individual risk factors. These findings suggest that the social environment influences disease pathways. Yet much remains to be learned about the social environment, including how to understand, define, and measure it. The research that needs to be done could benefit from a long tradition in sociology and sociological research that has examined the urban environment, social areas, social disorganization, and social control. We summarize this sociological literature and discuss its relevance to epidemiologic research.

Neighbourhood characteristics and use of benzodiazepines in The Netherlands

This paper analyses the relationship between individual and neighbourhood characteristics and the use of benzodiazepines within a Dutch city. It is hypothesized that the proportion of users is lower in more socially integrated and less deprived neighbourhoods. Hypotheses have been tested by using multi-level analysis to distinguish between composition and context effects. Age and gender have a clear relation to the use of benzodiazepines and neighbourhood differences in the proportion of users are partly the effect of population composition by age and gender. The proportion of users is higher in neighbourhoods with a higher percentage of one-parent families, with a lower percentage of social rented housing and with a larger number of rooms per person. The strength of the relation between age and use is influenced by neighbourhood characteristics. Neighbourhood variation in the amount used only depends on population composition.

Educating patient educators: enhancing instructional effectiveness in physical therapy for low back pain patients

The objective of this research project was to study the effectiveness of a training program for the enhancement of patient education skills in physical therapy. In this paper the improvement of five of these skills is tested. These skills are aimed at a better monitoring of adherence problems during the treatment and at enhancing self-efficacy of the patient after treatment. In order to test the effectiveness of the program, complete treatments of 19 physiotherapists have been assessed before (1142 sessions, 130 patients) and after (775 sessions, 88 patients) the training program. Information on the instructions and solutions given to the patients was obtained with a registration form, completed after each session by the physiotherapist. The patient's perception of the effectiveness and feasibility of instructions was obtained from questionnaires, completed by the patient on three occasions. After the training only a minority of the trained skills appeared to be improved. All in all, the training program was not very effective. More effort is needed to develop training programs aimed at promoting patients' self-efficacy as well as measurement instruments to assess the effects of such programs.

Mortality variations in England and Wales between types of place: an analysis of the ONS longitudinal study. Office of National Statistics

This study investigates the extent to which individuals, in England and Wales, in different types of place experience differential mortality once account is taken of personal (individual and household) social circumstances. Data comes from the Longitudinal Study of England and Wales of the Office of National Statistics, the respondents being a one percent national random sample of people aged between 25 and 74 at the 1971 census, followed until the end of 1985. For males and females separately, differences in mortality are found for the 36 types of Craig-Webber classification in models which include, at the individual level, a number of demographic and socio-economic variables (women being classified by their own occupation). In general, for both males and females, the same types of place have elevated or lowered mortality. For males a (cross-level) interaction exists between the proportion in the area in professional social classes and individual social class, the effects of individual social class being larger in areas containing a higher proportion of those in professional occupations. For females mortality is negatively related to the proportion of car-ownership in the area.

Socioeconomic characteristics and melanoma incidence

PURPOSE

This investigation examines the relationship between socioeconomic status (SES) and melanoma incidence in counties included in the Surveillance, Epidemiology, and End Results Registry (SEER) in the United States from 1973 to 1993.

METHODS

Cases included whites, aged at least 15 years, with a morphologic diagnosis of malignant melanoma, residing in one of 199 counties at the time of diagnosis. County level measures of SES including median household income, percentage of high school graduates, and percentage of families below poverty were abstracted from the 1950, 1960, 1970, 1980, and 1990 U.S. Census data. The relationship between SES factors and melanoma rates was examined by hierarchical Poisson regression.

RESULTS

The percentage of high school graduates was significantly and positively associated with the incidence of melanoma (relative risk (RR), 1.28; 95% confidence interval (CI), 1.21-1.35), after controlling for age at diagnosis, gender, time period, latitude, and percentage of Hispanics in the county. Percentage of families below poverty was significantly inversely associated with the incidence of melanoma (RR, 0.66; 95% CI, 0.55-0.78). When education and poverty were included in the same model, both the positive effects of education (RR, 1.23; 95% CI, 1.16-1.31) and the negative effects of poverty (RR, 0.85; 95% CI, 0.74-0.98) persisted. In contrast, median household income was not associated with melanoma incidence in a similar multivariable model (RR, 1.00; 95% CI, 0.99-1.00).

CONCLUSION

Whether the effect of education on incidence of melanoma reflects lifestyle behaviors that modify exposure to sunlight or some other factor remains unclear. Nonetheless, the findings of this study suggest that the determinant is primarily related to education, not income.

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.

The influence of households on drinking behaviour: a multilevel analysis

This paper examines the influence of household membership and area of residence on individual drinking behaviour using a multilevel modelling approach. The effects are investigated using data from the Health Survey for England (HSE) in which multiple interviews were conducted in the same household. With the use of postal address, the data were organised into a hierarchical structure of individuals within households within enumeration districts. After controlling for characteristics of individuals thought to influence or correlate with drinking behaviour, unexplained variation in alcohol consumption was attributed to individual, household and area effects. Household influences on drinking behaviour far outweigh the influences of place of residence. Policies aimed at reducing alcohol consumption, particularly by heavy drinkers, may be best targeted at the household level.

Bringing context back into epidemiology: variables and fallacies in multilevel analysis

A large portion of current epidemiologic research is based on methodologic individualism: the notion that the distribution of health and disease in populations can be explained exclusively in terms of the characteristics of individuals. The present paper discusses the need to include group- or macro-level variables in epidemiologic studies, thus incorporating multiple levels of determination in the study of health outcomes. These types of analyses, which have been called contextual or multi-level analyses, challenge epidemiologists to develop theoretical models of disease causation that extend across levels and explain how group-level and individual-level variables interact in shaping health and disease. They also raise a series of methodological issues, including the need to select the appropriate contextual unit and contextual variables, to correctly specify the individual-level model, and, in some cases, to account for residual correlation between individuals within contexts. Despite its complexities, multilevel analysis holds potential for reemphasizing the role of macro-level variables in shaping health and disease in populations.

Context, composition and heterogeneity: using multilevel models in health research

This paper considers the use of multilevel models in health research. Attention focuses on the structure and potential of such models and particular consideration is given to their use in elucidating the importance of contextual effects in relation to individual level social and demographic factors in understanding health outcomes, health-related behaviour and health service performance. Four graphical typologies are used to outline the questions that multilevel models can address and the paper illustrates their potential by drawing on published examples in a number of different research areas.

People, places and coronary heart disease risk factors: a multilevel analysis of the Scottish Heart Health Study archive

The Scottish Heart Health Study (SHHS), which recruited 5123 men and 5236 women between 1984 and 1986, was set up in part to investigate geographical variation in coronary heart disease in Scotland. Multilevel models are particularly appropriate for such hierarchical data, in which the individuals in the study can be represented by the lower level and the districts in which they live by the higher level. Multilevel models are presented for four coronary heart disease risk factors-diastolic blood pressure, cholesterol, alcohol consumption (defined both as units of alcohol consumed per week and as being a non-drinker) and smoking, for men and women separately. Significant district level variance was found for three out of the four variables studied, after controlling for socioeconomic and other variables considered at the level of the individual. These were for diastolic blood pressure, cholesterol and alcohol. Although the large majority of the variance was present at the individual level, the existence of significant variance at the district level is evidence that places may have a role in the distribution of coronary heart disease risk. Health policy aimed at reducing coronary heart disease should therefore consider the characteristics of places as well as individuals.

Multilevel models: applications to health data

This paper presents an introductory account of multilevel models, highlighting the potential benefits that may be gained by the use of these methods. It draws on recent applications in health services research that have appeared in the literature. Methodological advances in these statistical techniques have taken place in the field of education, where empirical studies have mainly been concerned with comparing pupil achievement across different schools by exploring the relationship between individual and institutional factors. Although recent widespread availability of suitable software packages has enabled other disciplines to adopt these methods, to date they have received little attention in the health services research literature (the investigation of effects of geographical areas on health being a possible exception) despite their obvious application in many areas of current interest. Key areas that could benefit greatly from these techniques include the exploration of variations in clinical practice, comparisons of institutional performance and resource allocation.