Intended for healthcare professionals

Education And Debate

Socioeconomic determinants of health: Community marginalisation and the diffusion of disease and disorder in the United States

BMJ 1997; 314 doi: (Published 03 May 1997) Cite this as: BMJ 1997;314:1341
  1. Rodrick Wallacea, research scientist,
  2. Deborah Wallaceb, president
  1. a New York Psychiatric Institute, New York, NY 10032, USA
  2. b Public Interest, Scientific Consulting Service Inc, New York, NY 10027, USA
  1. Correspondence to: Dr R Wallace PISCS Inc, 549 W 123 St, Suite 16F, New York, NY 10027, USA.


    This article describes the cascading diffusion of “inner city problems” of disease and disorder in the United States–from the huge marginalised inner city communities of the largest municipalities, first along national travel routes to smaller cities, and then from central cities into surrounding more affluent suburbs–following the pattern of the daily journey to work. Public policies and economic practices which increase marginalisation act to damage the “weak ties” of the community social networks which bind central city neighbourhoods into functioning units. Spreading disease and disorder can be interpreted as indices of the resulting social disintegration, which is driven by policy. This “failure of containment” in the United States should serve as a warning for cities in Europe against reducing the municipal and other services that they provide to “unpopular” subpopulations.


    Visitors from Europe who, by chance or design, encounter the marginalised poor communities of the large American cities are often stunned by the extent and intensity of physical and social deterioration. In New York City alone, some 600 000 people live in the devastated zones 1. This “bombed out” urban landscape is largely the creation of public policies of “planned shrinkage” and “benign neglect,” which led to reductions of services such as fire control and garbage collection in poor neighbourhoods.2 Following similar service cuts and related disinvestment, much rental housing was rapidly destroyed during the 1970s in many large American cities. Presently, nearly a quarter of American children are, according to official figures, growing up in poverty, many within these collapsing communities.3

    A principal underlying cultural assumption of the policies creating these conditions is that the effects of this devastation are being, and will be, confined largely to the targeted communities, and thus they will be separated from the suburban counties in which most affluent people now live and in which political power now lies. Only a quarter of Americans live in central cities; half now live in the surrounding suburbs.

    A second underlying assumption is that, even if there is suburban “leakage” from the decaying central city neighbourhoods of New York, Los Angeles, Chicago, and some other large cities, this will not greatly affect the nation as a whole.

    These cultural assumptions were epitomised by a study titled “The social impact of AIDS in the United States,” published in 1993 by the National Research Council. The study concluded, on the basis of a single cross sectional map of AIDS in postal zones of New York city, that

    Many geographical areas and strata of the [American] population are virtually untouched by the epidemic and probably never will be [touched]; certain confined areas and populations have been devastated and are likely to continue to be … HIV/AIDS will “disappear,” not because, like smallpox, it has been eliminated, but because those who continue to be affected by it are … beyond the sight and attention of the majority population.


    Geographic diffusion of disease and disorder

    Such a statement flies in the face of a century of studies of geographic diffusion on several scales of space, time, and population, well summarised by Abler et al.5 Almost needless to say, no geographers or spatial ecologists participated in the National Research Council's study. Geography, history, economics, anthropology, ecology, sociology, and epidemiology all study how rumours, fads, and technical and social innovations–as well as epidemics–spread in space and time and between social groups. Three mechanisms, acting at different scales, have been found to characterise such spread: hierarchical diffusion, spatial contagion, and network diffusion.

    Hierarchical diffusion describes a cascading hopscotch transmission from socially dominant larger cities to smaller ones along the national transportation network. Two places may be geographically distant but they will be “close” in their probability of interaction if many people travel between them frequently. On a slightly smaller scale, spatial contagion (or expansion diffusion) describes radial spread along local travel routes from a central city epicentre into adjacent communities; this is often described as “a wine stain on a tablecloth.” Network diffusion usually occurs on a still smaller scale; the term describes spread along personal, domestic, and community social nets which, when they have a geographic focus, can be characterised as sociogeographic structures.

    Our recent work has quantified the spread of AIDS between the standard metropolitan statistical areas of the 25 most populous American cities, containing a total of 113 million people.6 We found a hierarchical structure for the national AIDS epidemic: a top-down pattern of spread from the initially infected epicentres of New York City and San Francisco to other urban regions. Using data from the US Census Bureau on migration between metropolitan areas, we calculated the probability of contact between each region and all the others. The cumulative number of people with AIDS through 1995 within these metropolitan regions was closely predicted (r2=94%, the percentage of total variance in AIDS cases predicted through regression) by a model based on three logarithmically transformed variables: probability of contact with New York; probability of contact with San Francisco; and regional 1991 rate of violent crime per unit population. The spread of AIDS among the 25 largest metropolitan areas was thus determined by the intersection of local social disintegration (indexed by violent crime) and the probability of contact with the two most heavily infected epicentres.

    The national and regional scales are bridged by the two determinants of the epidemic's structure: the links between regions and the socioeconomic structure, function, and history of the individual regions. Thus, contrary to cultural assumption, large metropolitan regions with high prevalence of urban decay, such as New York, constitute great epicentres from which disease and disorder spread nationally.

    When we examined population rates of AIDS, violent crime, and tuberculosis for the 24 counties constituting the New York City standard metropolitan statistical area, we found that a single composite index which convolved the area density of the workday commuting pattern with the local county poverty rate predicted well over 90% of the variance for each of these variables.7 The intensity of the commuting pattern was determined from census data at county level on the daily journey to work; a step by step interaction was allowed to continue until a 24 element “equilibrium distribution” was reached.7 8

    Breakdown of the AIDS data by time period (before 1985, 1985-7, 1988-90) showed that as the incidence of AIDS in Manhattan, the commuting center, rose, the incidence in all other centres rose correspondingly and in proportion (on a log-log scale, indicating a power law7 8). Figure 2) shows county rates of AIDS cases as a function of distance from Fifth Avenue and 42nd Street for 1982 and 1984, illustrating the nature of the relation, and figure 3) maps the spread of AIDS from the travel centre into those suburbs. Analysis for tuberculosis (1985-7, 1988-92) gave a similar result: as the incidence of AIDS in the dominant travel centre, Manhattan, rose, so in exact proportion did incidence in the counties in the entire metropolitan region.7 8 The incidence of low birth weight, although not parallel to the three other markers, also was strongly predicted (r2=92%) by commuting pattern and poverty rate.


    Commuting patterns predict the spread of diseases outward from inner cities to suburbs

    Fig 1
    Fig 1

    Cases of AIDS per 100 000 population for counties in or near New York City as a function of their distance from Fifth Ave and 42nd Street in Manhattan, 1982 and 1984. Rates for Manhattan rose from 131/100 000 in 1982 to 348/100 000 in 1984. Reprinted courtesy of Professor Peter Gould, Pennsylvania State University

    Fig 2
    Fig 2

    Cases of AIDS per 100 000 population by county for the New York metropolitan region (a) to end 1984; (b) 1985-7, (c) 1988-90

    Using the same approach, we analysed eight large standard metropolitan statistical areas in the United States, together containing 54 million people, for four public health problems: patterns of incidence of AIDS and tuberculosis, low weight birth babies per 10 000 live births, and the incidence of violent crime.7 We characterised the public health problem of the central city as regionalised throughout the area if the statistical significance of the correlation data was not destroyed by removing the point of the commuting centre from the regression of the log incidence on the log of the commuting density per unit area. This omission of the travel centre constitutes a more rigorous condition, since such centres strongly dominate overall pattern. Regionalisation means that incidence in the central city determines the incidence in the surrounding counties, as modulated by the area density of the commuting pattern.

    Different patterns of regionalisation emerged from our analysis of the eight areas.7 Although all four public health problems showed regionalisation in the New York area, the other problems were regionalised only in some areas. At the other end of the spectrum from New York, the San Francisco area was regionalised only for tuberculosis.

    Some areas showed trends toward regionalisation which indicate a strong influence of the central city on the outlying counties–for example, violent crime in the Washington DC and St Louis areas, and low weight births in the Detroit area. Some of these metropolitan areas are also characterised by a central city that is small in comparison to the total suburban population. Essentially, the travel centre tail wags the regional dog: disease rates in the core city and the local pockets of poverty in the county determine disease rates in suburban counties via the economic linkages within the region as indexed by the commuting pattern. Indeed, the workplaces of the metropolitan area mix the diverse populations. The area's single socioeconomic system is the reality: that city and suburb are totally separated is a public health myth.

    Urban decay, social networks, and diffusion

    Lives of individuals and families are deeply affected by influences on a neighbourhood scale: the neighbourhood embodies the “weak ties” through which the larger society channels information, support, and social control to families and individuals. These are relations of occupation, common interest, and neighbourliness beyond the “strong” ties of kinship, ethnicity, or peer group which bind small groups tightly and exclusively together into isolated “equivalence classes.” Strong ties cannot easily serve larger community purposes, a paradox which Granovetter characterises as the “strength of weak ties.”9

    As a neighbourhood disintegrates under the assaults of public policies of planned shrinkage and benign neglect, those weak ties begin to fray. Families leave, people are afraid to congregate on the streets, and legitimate economic activity (and the fraternity of occupation which it embodies) declines. As weak ties erode, possibilities for individuals and families narrow, and family groups are thrown back on their own resources.

    Youth behaviours such as doing well in school, getting a regular job, avoiding substance abuse, and maintaining stable relationships become more difficult as the neighbourhood structures that value such attainments dissolve. Negative acts such as violent behaviour, multiple sexual conquests, and drug taking are messages that can be more easily “heard” in a dissolving community than positive acts. If such “bad” behaviours damage a community's weak ties further–for example, by making street life more dangerous–the result may be destabilising positive feedback between community disintegration and antisocial behaviour.10

    Neighbourhood processes affect families and individuals. Individuals and families who would otherwise have retained their housing become homeless due to the combination of the housing losses and the fragmentation of social networks. Individuals and families who would otherwise have remained independent and off the welfare roles have to receive public assistance as a result of a lack of low income housing and the disruption of community. Children who would otherwise have had one parent, if not two, become orphans from the epidemics of violence, substance abuse, and AIDS.

    We find that public health at every scale of population is largely driven by contagious phenomena affecting socioeconomic processes, disease patterns, and behavioural processes at the neighbourhood level. Poor neighbourhoods in large central cities, suffering greatly from urban decay triggered and sustained by policy, have a disproportionate influence on the health, safety, and wellbeing of a huge proportion of the American population, including rich people.

    With approaches from geography, demography, and ecology, the geographical patterns of disease can be modelled and predicted at a variety of scales of spatial distance, population, and socioeconomic distance. Disease and behavioural relationships between populations, whether purely spatially distant or merely socioeconomically distant, can be established.

    In his seminal paper the noted ecologist CS Holling 11 described how, at each scale of space, time, and population, certain processes are crucial for stability and resilience. These processes are nested and linked between scales. In particular, Holling called attention to the “mesoscale,” the familiar realm of population and community from a few metres to a few kilometres: at this scale, contagious processes funnel the impacts of events at the level of the individual and small group up to larger scales and also mediate events downward from the large scale to the micro, the individual.

    Our data and analyses show that human ecology also includes fundamental processes, each of which have characteristic scales of population, geography, and time, and which are also nested and linked. As in natural ecosystems, contagious processes occur at the mesoscales of neighbourhood and city–one to a few kilometres–magnifying the impacts from the small to large and mediating impacts from the large down to the micro, the individual and the family.

    In the United States, the keystone population which determines public health and public order at larger and smaller scales is the poor urban neighbourhood. If this structure cannot, for reasons of public policy and private interests, engage in the keystone community processes, all populations in the country suffer deterioration of health and safety.

    The belief that subpopulations in one country are separate and do not operate as a single ecosystem, affecting each other, has propelled the United States into a crisis of social and economic structure and of public health and public order which is so severe that even such crude measures as life expectancy show deterioration.12 It reflects a profound error: concentration is mistaken for containment. Fundamental processes at and across the mesoscale ensure that concentration causes diffusion. Public policies or economic practices which marginalise vulnerable communities within Europe may be expected to create a crisis similar to that now raging in the United States.


    Funding: The writing of this paper was supported under an Investigator Award in Health Policy Research from the Robert Wood Johnson Foundation.

    Conflict of interest: None.


    The term “failure of containment” as applied to our analysis of the spread of social disintegration from poor urban neighbourhoods was first coined by Gregory Pappas.


    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.
    6. 6.
    7. 7.
    8. 8.
    9. 9.
    10. 10.
    11. 11.
    12. 12.