Component Director: Paul J. Gruenewald, Ph.D.
Three truisms about drinking and problems:
- Drinking must take place some place.
- Any problem related to drinking must take place some place.
- These may be the same or different places.
It is an oddity of prevention epidemiology that these simple facts are typically overlooked in assessments of neighborhood risks for alcohol related problems. The characteristics of people living in neighborhoods, like income or education, and characteristics of neighborhoods themselves, like housing and retail business, can affect rates of problems within those neighborhoods or in neighborhoods quite some distance away. These direct and indirect impacts arise because characteristics of one neighborhood may affect problems in another. Mathematical models of community systems and statistical analyses of neighborhood data are deeply challenged by this fact. In this component we further develop the mathematical and statistical tools necessary to elucidate such effects across neighborhoods of cities in California. We use the demand for alcohol and incidents of alcohol-related problems as our ‘test bed' for this work.
Here is an outline of the kind of problem we address: The dark areas in the left-hand figure show the neighborhoods of Chico, California, where the demand for alcohol was greatest in 2009. The darker areas in the right-hand figure show the neighborhoods where alcohol-related car crashes were most frequent in the city. Locations of alcohol outlets are indicated by the colored dots. As the figures show, demand is greatest in wealthy suburban downtown areas, outlets appear to arise near areas of high demand, and crashes seem to be most frequent around outlets, but also seem frequent in outlying areas.
So our critical research question is: How do we measure these spatial relationships?
Research Goals and Activities
In order to answer this critical question we are taking a series of shorter research steps:
- First we need “highly resolved” spatial data over a long time so we can distinguish direct “neighborhood” effects from other indirect spatial effects.
For example, we have identified and mapped the geographic location of every car crash in California over 10 years. We also have available the location of every alcohol outlet, detailed demographic, economic, housing, transportation and business data for Census block group areas over the same period of time.
- Then we need a conceptual model of how we might expect some alcohol-related problem to be related to drinking in different places, say at home or at a bar or restaurant.
Geographic relationships between places at which people drink and locations of car crashes related to drinking are very difficult to specify. Primary sources of drinkers and drinking locations may be either private residences, bars or restaurants. Drinkers may drive to or from these locations either before or after drinking. Drinking drivers are more likely involved to be in a crash due to their drinking, of course, but also due to roadway conditions along the travel paths they follow.
- Then we must to represent these relationships in a quantitative model that relates characteristics of places to rates of problems.
The challenge here is to describe the spatial topology of community areas, recognizing that they are connected by transportation networks along which drinking drivers drive their cars. These transportation networks can be fully characterized using matrix techniques and spatial relationships between “sources” of drivers (say, at home) to “sinks” at which they drink (say a bar) statistically assessed using “Bayesian Poisson conditional autoregressive space-time models.”
- Finally, we must attack the problem of assessing “causal” relationships in these data from a statistical point of view.
We do this by measuring how much the changing characteristics of a place, like the number of bars in a neighborhood, affect outcomes in the same or different neighborhoods over time. This “variance components” analysis gives us our first insights into the degree to which the effects we observe are extended in time.
So, what have we found out so far?
Numbers of alcohol outlets grow in response to alcohol demand:
- Alcohol outlets appear most often in areas of communities near to but not actually within areas of high alcohol demand.
- Most often the places where outlets are located are generally the low income section of the city. Since higher income areas exclude alcohol outlets to keep property values higher, more outlets appear in nearby lower income minority areas. There are two distinct geographic economic processes occurring:
- Demand effects: Outlets will locate as close as possible to meet greater alcohol demand. These will be near to higher income dense population neighborhoods.
- Land rent effects: Outlets will be excluded from neighborhoods where land and housing values are high. In part due to outlet owners desire to minimize overhead costs of running an outlet; lower rents generally mean greater profits. Outlets are strongly excluded from high income residential neighborhoods to maintain housing values that are negatively impacted by the presence of outlets.
- These geographic economic processes are a natural social source of health disparities in exposures to risks associated with alcohol outlets.
There are two critical population sources of drinking and driving crashes:
- The first source, shown in the figure on the left, are bars and pubs located in downtown areas of communities. Drinking in these places is related to a 15-fold increase in risks for drinking and drunken driving.
- The second source, shown in the figure on the right, are residential areas of communities. Here the risks for drinking and drunken driving related to any one drinking event are quite small, but the number of potential drinking drivers very large.
- There is a third source not shown: Drinking in restaurants is also a considerable risk, laying between the other two.