URI Working Paper #47

The author hereby grants permission for photocopying, microfilming, or computer electronic scanning of "Health and Community Based Urban Residential Restoration: An investigation into the utility of the traditional epidemiological approach" for the purpose of individual scholarly consultation or reference. This permission is not to be interpreted as affecting publication of this work or otherwise placing it in the public domain, and the author reserves all rights of ownership guaranteed under common law protection of unpublished manuscripts.

_________[SIGNATURE]_________

Alexis Dinno

______[Dated May 16, 2000)_______

Date

The author explored the question "How do residential urban blight and community-driven restoration effect changes in health?" by participating in and observing the processes of Urban Resource Initiative’s community-based abandoned lot restoration projects. An ontological framing of health served as a departure point for the selection and testing of health and environmental indicators. These tools were then employed in a simple cross-sectional pilot study intended to suggest possible relationships between pairs of indicators. Results from this study indicate that URI’s restoration activities may be related to several dimensions of emotional well being including stress and depression, perceived security and factors pertaining to the social cohesion of the neighborhood. The author found that the fundamental assumptions of causal relation in epidemiological methods are likely to be inappropriate to characterize and accurately quantify the relations between the various environmental and health factors assessed. The relevance of interdependent dynamic systems to modeling health/environment relationships underscores the need for statistical tools capable of characterizing and testing feedback, longitudinal data collection that can tease apart system dynamics and event sequencing and—given quantitative uncertainty—alternative approaches to data collection, interpretation and communication. The study closes with an examination of the ethical ramifications of the preceding paradigmatic reevaluation of health/environment relationships. The author concludes that evaluation of the health effect of urban community environmental programs must be informed by a health promotion perspective and that the interdependencies in such systems are likely to make accurate quantitative articulation unlikely using currently available tools. Included are assessment tools and a community outreach pamphlet produced in the course of the project.

—Merriam Webster’s Collegiate Dictionary

Special thanks to David Langevin, Leigh Shemitz, David Fisher and Roger Taylor for their insight, feedback and patience.

Introduction: the Urban Resources Initiative and the question of health
Health’s Ontology
Creating tools to explore health in the abandoned lot
Epidemiology’s Epistemological Paradigm
An epidemiological approach: the cross-sectional study
Rejecting the epidemiological paradigm: dynamic systems
Implications for modeling dynamic systems
New observational approaches of health and environment
Implications for environmental health science
Answering URI’s mandate
Conclusion
References
Appendix

List of Tables

TABLE 1—Illustration of pathogenic and salutogenic factors
TABLE 2—Division of directly and indirectly influenced environmental indicators
TABLE 3—Scoring for Positive & Negative Emotionality and Emotional Tendency summary variables
TABLE 4—Chi-Square test for association and prevalence odds ratio (POR)
TABLE 5—Non-Parametric tests for association from the SAS System’s PROC NPAR1WAY
TABLE 6—Correlation data from The SAS system’s PROC CORR
TABLE 7—A paradigmatic map of epidemiology and environmental health science

List of Figures

FIGURE 1—The epidemiological paradigm’s general model for hypotheses.
FIGURE 2—Sources of error (lack of precision) and bias (lack of accuracy) in the epidemiological model
FIGURE 3—Exposure definition boundaries for each assessment site.
FIGURE 4—A general case where the prevailing epidemiological paradigm breaks down.
FIGURE 5—Simplistic allegory for a complex, interdependent system.
FIGURE 6—Theoretical model of health and community blight restoration

Introduction: Urban Resources Initiative and the question of health

This paper examines some of the relationships between health and environment within the context of outdoor community-based urban restoration and the prospects for creating knowledge and action around these relationships. It does this within an epidemiological context and within environmental health science in particular. The goals of this project are threefold. First, it demonstrates the limitations of two paradigms inherent in epidemiological science to understanding health and environment within Urban Resource Initiative’s scope of action. Second, it describes possible alternatives to standard epidemiological tools for creating knowledge about processes in health and environment. Finally, it considers several ethical and epistemological implications of a broader human ecology.

This project springs from work done during a summer internship at the Urban Resources Initiative (URI) in 1999. The internship bore a mandate to answer the question "How do residential urban blight and community-driven restoration effect changes in health?" To undertake this I engaged New Haven communities as an participant observer of URI’s abandoned lot projects. In addition, during this time I conducted research into other studies and methods used to assess the way physical environments—both built and grown—interact with social, physical and mental dimensions of health and well being. The assessment tools used during this independent project were developed during the URI internship.

URI is a non-profit organization partnered with the Yale School of Forestry & Environmental Studies. Its decade-old mission is devoted to environmental education, land stewardship and community outreach. In pursuing these goals, URI engages with a variety of community groups, local, state and federal level agencies and the enthusiasm, perspective and skills of Forestry School faculty and students. The organization tends to act directly on the environment through street-scaping and abandoned lot projects. Street-scaping involves residents collaborating with URI to articulate a vision of their neighborhood environment and bring such vision into fruition. Community effort to grapple with an abandoned parcel of land on the block marks the abandoned lot projects. Such projects eliminate eyesore and structural problems, promote aesthetic enjoyment, formalize and alter parcel use. At the same time they also promote social interaction around social norms, enhance trust, facilitate expression through multiple levels of community and government and create opportunities for personal development among children and adults.

Throughout URI’s history many of the people involved in its projects—both internally and externally—have repeatedly expressed the idea that among the many contributions these programs provide are "environmental health benefits." However, the broader literature says little about precisely or generally how such programs affect health. This pattern of observation prompted the 1999 summer internship exploring ways to characterize and bolster the health effects of URI’s programs.

Health’s ontology

The start of this research begat an ontological question: how ought health to be thought of existing? Initial exploration of the question made clear that the familiar exposures and health effects with which environmental health scientists are so comfortable were not quite germane to the question at hand. Typical approaches to environmental health use the term brownfield to talk about abandoned parcels of land that pose a health threat through the combination of a toxic legacy of the land’s past use and an ambiguity as to present day jurisdiction and responsibility for the health effects of those toxins. But, questions of health effects of URI’s summer projects are not really about "traditional" environmental health concerns such as asthma or lead poisoning. The brownfield is the wrong metaphor in this case.

Wanting to proceed cautiously, I decided to start at a fundamental philosophical level and consider the definition of health and health effect instead of engaging in some haphazard measurement of indicators. The current World Health Organization (WHO) definition served as a ready departure point for this purpose. The WHO defines health as "a state of complete physical, mental and social well-being and not merely the absence of disease and infirmity." Moreover, many WHO professionals feel that health defined and considered as a process, rather than a state, is more appropriate. The inappropriateness of the reductive approach—generalizable by an emphasis upon a specific malady or category of malady—will unfold throughout the remainder of this report.

Accepting a broad definition of health at the start of the exploration permits movement beyond the constraints imposed by the traditional biomedical framing of health as the simple absence of injury, disease and death. The traditional biomedical viewpoint particularly limits the kinds of questions one can ask about urban blight, and the kinds of tools for assessment and evaluation that one may employ. The kinds of questions prompting the development of epidemiological tools over the past century have been preeminently those associated with addressing pathogenesis. By contrast few questions have driving health sciences to explore questions addressing salutogenesis.

Salutogenesis is a principle proposed by medical sociologist Aaron Antonovsky as a health-driving process. It may be understood to be a process that acts opposite to pathogenesis. Table 1 makes a small illustration between some pathogenic and salutogenic factors of my own choosing. The pathogenic factors on the left are those that diminish health (a disease—asthma, for example) or which reduce the diminishment of health (a therapy to treat asthma). Salutongenic factors on the right are those that promote health (a health gain resource—access to green-space) or which inhibit the promotion of health (the social and physical deterioration, or blighting of that space).

Both pathogenesis and salutogenesis are paradigmatic and not theoretical: they reflect a way in which we organize our perceptions of the world. Particularly, they reflect the way in which we believe health to exist and the nature of health’s existence. Neither paradigm models the phenomena that engage our perceptions. For example, one does not use an assessment technique to measure an actual physical property called pathogenesis. However one’s perspective that there is a pathogenic ontology of health may underlie the measurement and modeling of some physical phenomenon such as carcinogenicity or fracture.

Note that the salutogenic factors I have listed tend towards a generality. That is, they probably have many effects, and many of their effects are likely to be caused by other factors as well. Therefore they may be very difficult to isolate from a wealth of other factors and levels of analysis. By contrast the pathogenic factors I listed (at least some of them) lend themselves well to an analysis that isolates cause and effect to a relationship between a small number of variables. For example vaccines are protective of specific diseases, and tend to have narrowly bounded side-effects that can be measured relatively easily. On the other hand, community trust is likely to play out along many dimensions of health and well being ranging from social, economic and political expression to availability of help in time of need.

One may also be concerned about precisely which side of the table some factor appears on. For example why is "Sex Education" listed as a factor that is appropriate to the pathogenic paradigm (reduction of disease), when "Advertising promoting tobacco and alcohol" is listed as appropriate to the salutogenic paradigm (inhibition of health promotion)? The answer is that it may be useful to consider salutogenic and pathogenic paradigms as ends of a spectrum. Probably very few processes can be considered from only one or the other ontological lens. For example, among athletes and sports-oriented care providers, conventional wisdom holds that physical injury can actually make one healthier. The empirically improved understanding of the limits and needs of ones body provided through an active healing process make the injury (obviously well suited to a pathogenic frame of reference) the source of some kind of health promotion.

Traditionally epidemiology and public health concerns itself almost exclusively with the left side of the table: the left side of this illustration being factors that detract directly from well being like disease and injury and are characteristically biomedical. The pluses and minuses represent contribution to or erosion of such pathogenic forces (i.e. ill-making behaviors such as alcoholism, violence, medicine, rehabilitation, disease prevention, family planning, etc.). This report asserts that such facets of health lend themselves to a reductive analysis and epidemiological modeling of health and environment relationships.

The public health and medical establishments traditionally ignore the right side of Table 1, and the environmental health sub-field of public health is no exception to this tradition. Salutogenic factors that improve the quality of life include community cohesiveness and trust, literacy, security and other aspects of basic human rights, and the active shaping of the physical environment that is the substrate upon which lives are lived and in turn shaped. The pluses and minuses may include such factors as wealth disparity, access to green-space, architectural determinants of social function (i.e. is there a physical space that promotes or prohibits particular kinds of social interaction?), etc. These facets of health tend towards systemic, process-oriented and integrative articulation of health.

The flexibility afforded by our initial framing of health to include salutogenesis permits us to begin to explore several questions related to our initial question. The following questions are of particular concern:

• Do urban community redevelopment projects fit into the context of environmental health?
• How does such redevelopment or its lack affect health within a community?
• What is the scope for public health intervention in blighted urban environments?
• When is it ethically appropriate to intervene for salutogenic aspects of health when the pathogenic risks are unknown?

The value judgements inherent in choosing questions such as these engenders a disposition towards certain kinds of analytic methods when attempting to answer them. This report will return to these questions in the concluding discussion. The next section examines the basic epidemiological model employed within the pathogenic paradigm of health, explores the salient consequences of such a model and its methods and illustrates several important limitations of this model.

An assessment of the relationships between health and environment within the context of restoration and blighting processes necessitates several considerations for choosing state variables (indicators). An assessment that incorporates observations internal and external to the community should provide insights that are relevant to individuals from both spheres. Therefore, the creation of a tool to evaluate different dimensions of community health entailed research within the community of study and through the wider body of published (English) instruments. Environmental processes such as lot clean-up, community gardening, etc. were likewise explored for meaningfulness internal and external to the community. However, the environmental assessment tool is designed to be employed by individuals who are not from the community. The following survey served as an informal guide when making observations within New Haven communities and among non-resident participants in the URI projects (URI staff, funders, city employees, etc.). These observations informed further research and creation of the environmental and health indicator tools.

The variables that could possibly characterize blight and restoration processes are many and varied. They could be values such as the ratio between abandoned lot area and occupied lot area, the ratio of front edge length of abandoned lots and occupied lots, the proportion or absolute number of condemned buildings, various characterizations of detritus accumulation, degree of plant overgrowth, etc. The complexity of the relationships between such variables seems intuitively obvious, though the exact relationship may be unclear. The answer to a question of health’s ecology like "how do neighborhood social ties and degree of plant overgrowth influence one another?" is expectedly to be dynamic and complex.

The selection of specific variables to assess health and environmental processes may seem somewhat arbitrary. And, one might be moved to the criticism that the variables are truly arbitrary and can not justifiably represent any genuine relationship between health and environmental processes. This concern is answered in several ways. First, the selection of a loose or "fuzzy" set of indicators to define a process is well known in both biomedical and environmental disciplines. Two examples illustrate this. Asthma is diagnosed by using a variety of symptomatic indicators (some of which are non-specific to asthma). Clinical depression is likewise assessed by a subjective evaluation of a number of non-specific indicators. The common methodology to both these assessments is a body of ecological understandings: a collective interpretation of repeatedly observed patterns by experts in contributing medical sub-fields.

Second, the variables of interest reflect the values of the individuals involved: both individuals with experience in urban health and environment and everyday residents. These values have been carefully synthesized and coupled with intentional observations to produce both sets of indicators. In short the variables reflect the concerns of the community, rather than reflecting a preconceived (and uninformed) expectation of health/environment interactions.

In order to characterize blight and restoration I assessed those qualities in the spheres of environmental and health processes that I believed to be related. Because the mandate question deals with interrelated and varied health and environmental processes I attempted to limit the scope of assessment to those dynamics that occur within approximately the same period as the processes related to URI’s abandoned lot projects. Thus I chose health indicators for our survey that are explicitly framed within an appropriate time frame. Health indicators are divided among social, mental and physical aspects. The specific indicators that I choose reflect conditions that may be influenced by the physical environment. Likewise, the environmental indicators are chosen for temporal appropriateness to the summer-long time span of URI’s abandoned lot projects.

Both environmental measures were further limited in scope by delineating those variables that are more or less directly affected by the URI interventions and those which are more or less indirectly influenced by such. Table 2 illustrates this division. Thus each of the "Directly influenced" indicators are literally factors that are addressed in the course of URI’s work. The people working on a site will explicitly undertake the pruning of overgrowth, the design and construction of formal paths, etc. The indicators on the "Indirectly influenced" side of Table 2 are not directly subject to URI intervention, even though they may be influenced. For example the presence of absence of tag graffiti may relate to the perceived investment and "ownership" or lack thereof of a space by a community, but URI does not attempt to either engage in or prevent "tagging." Evaluation of indicators of environmental processes in subject communities was proscribed by a present/not present inventory of the "directly influenced" variables from Table 2.

Health indicators were constrained in a similar manner to the environmental indicators: by scope of URI action and temporal scale. For example, detectable shifts in levels of environment-related stress are expected to fit within this time frame while shifts in neighborhood mortality levels are not. Since this study attempts to address a broad question of health, it employs physical, mental and social indicators to describe the community’s condition of well-being. The questionnaire developed includes questions to address these three areas. It attempts to assess exercise, acoustic disturbance, safety and violence, a number of variables related to stress and depression developed from several studies, neighborhood cohesion and social perception of environment, including assessment of the degree to which a person is willing to move. While these indicators may provide an insight into patterns within a community, benchmarking runs the risk of creating numbers that are meaningless outside of the context that they were generated in. For example the World Bank notes:

However, the data might be used effectively to show simply whether or not there was a change along the measured dimensions, and in which direction the change occurred. Implicit in this study design is the intention to observe what patterns, if any, emerge from such measurements.

The conventional tools for assessing factors that would appear in the left side of the illustration in Table 1 are primarily quantitative and statistical in nature. Generally, epidemiological studies attempt to show a statistically significant association between a causally independent exposure and a dependent outcome (see Figure 1). A body of epidemiological studies addressing equivalent exposures and outcomes provides a weight of evidence about the relationships. This evidence is then interpreted as supporting or not supporting the modeled relationship between exposure and outcome. This general exposure Þ outcome model and its use in risk analysis and communication underlie epidemiology. I refer to it as the epidemiological paradigm.

Models constructed within the epidemiological paradigm have several related prerequisites to generate knowledge effectively. The first requisite is specificity of exposure and outcome variables. The second is the standardization of state variables. The third prerequisite is a function of resource allocation and pragmatism: the exposure and outcome must be "easy" to assess. The second and third requisites are relevant to the "state of the science" as a field and will not be discussed further with respect to this study.

Models in the epidemiological paradigm depend upon extremely precise and specific definitions of exposure and outcome in order to produce meaningful results. Traditional environmental health sciences require the supposed presence of either a narrowly defined independent exposure variable (such as hazardous concentrations of organic lead, UV radiation, benzo-a-pyrene, etc.), or a narrowly defined dependent outcome variable (i.e. serum lead level, melanoma, testicular cancer, etc.) to initiate a study. Problems such as describing what happens to people exposed to a dynamic mixture of potentially hazardous materials are not tractable to these methods. Similarly, generalized outcomes such as comfort effects, non-specific symptoms and broad social and behavioral impacts may not fit well with the predominant epidemiological approach. Small decreases in specificity can create large increases in the statistical power required to demonstrate association. The concerns reflected in our questions about broad indicators of health largely remove us from the domain of this approach, which require at least one of the following conditions:

• Very precisely defined and measured exposure and outcome variables to minimize error, and increase statistical precision.
• Extremely strong associations between exposure and outcome to overcome the bias towards the null-hypothesis derived from imprecision of definition and measurement. (see Figure 2)
• Large sample size, providing increased statistical power.

With the assumption that the epidemiological approach to modeling relations between health and environment is acceptable in the case of disadvantaged New Haven neighborhoods, the cross-sectional study design informed experimental observation over five weeks between September and October of 1999 to provide insight into associations between environmental and health indicators. This design has several advantages in this case:

• Unknown relationships
• No assumption about the direction for the causal arrow.
• Inexpensive to collect data for the wide array of variables in the surveys.
• Takes little time.

This study design has several general limitations. It cannot provide evidence to support the causal assumptions of the hypothesis, and is easily biased—especially through exposure or outcome-related entry or exit from the study population. Moreover, the use of a cross-sectional study to statistically explore unknown associations of a wide variety of variables should employ a split-population design. Such a design would randomly assign half the sample population to a general analysis of all variables. The results of this analysis would inform the a priori hypothesis about a narrowly constrained association to be tested in the remaining sample population. However, given the small resource pool and low expected sample size for this pilot study, the results are intended to suggest relationships to focus on in a more intensive study, rather than to model risk.

The study population is drawn from New Haven residents within neighborhoods serving as the location of a URI abandoned lot project. Eligible participants in the study were 18 years or older and spoke English. The source population were restricted to individuals living adjacent to abandoned lot projects or to the nearest abandoned lot on the same street no closer than two parcels distance from the URI project site (see Figure 3). This method was chosen with the assumption that the strongest associations between environmental and health variables were likely to be exhibited by proximity to the abandoned lot containing the surveyed variables.

---

This population was canvassed a week prior with a flier informing them of the impending solicitation for survey participants. The sample population was comprised of a single individual from each household meeting the above requirements who was available for interview when the interviewers approached her.

Environmental surveys were conducted prior to health surveys, but on the same day as the health surveys were conducted (for neighborhoods in which health surveys spanned several dates, the environmental survey was made on the first day of interviews). The five-week limit on surveys was chosen because seasonal changes in outdoor climate alter use and appearance of neighborhood landscape. The abandoned lot adjacency approach described above narrows the definition of exposure considerably, however it makes no attempt to characterize the relative difference between environmental variable scores for the abandoned lots assessed, and environmental characteristics of other parcels (abandoned on not) on the same block.

Three interviewers gave health surveys. Respondents were permitted to choose whether to participate orally or in writing. Both choices employed exactly the same survey tool. Respondents’ questions about acceptable answers to a specific question entailed repeating all available options. Respondents’ questions about the meaning or content of a specific question were answered by repeating the question and inviting the respondent to answer with her own interpretation of the question. For example, several respondents were concerned with the generality of question 19 "How satisfied with your neighborhood are you?" and were invited respond as they felt most appropriate. Response to question 26—the open-ended question—often took form as an engaged dialogue with the interviewer. Responses were answered with either yes or no, or with an ordinal scale (i.e. some used a Likert-like scale (0 = None to 4 = A lot, etc.).

Several sources of bias have been considered. Differential bias may result in one way from systemic socioeconomic differences between individuals exposed to a URI abandoned lot project versus those exposed to an abandoned lot without such a project if the capacity of a community to organize itself and partake of URI’s program is related to socioeconomic status. The selection of non-URI abandoned lots within close physical proximity is an attempt to control for such bias by attempting to select exposed and non-exposed respondents from as similar a socioeconomic population as possible. This assumes that variations in socioeconomic status are correlated with variations in geographic location of a household address. Selective (differential) bias due to socioeconomic factors such as income, class or education is assumed to be significantly eliminated by restricting the study population to similar neighborhoods. The effect of ethnicity may also be reduced in this case, although differences in ethnic composition between project sites varies (for example, Fair Haven is a predominantly Latin community, while Newhallville is predominantly Black, and Arch Street is more heterogeneous). A larger and more rigorous study might attempt to explore the effect of ethnicity as a covariant in analysis. A non-differential bias in the associations between health and environmental indicators may come from the imperfect correlation between social participation in URI’s projects and household distance from the site of intervention, particularly if health indicators are influenced by changes in the social environment facilitated by the URI project.

Of a potential total of 55 surveys in this study population (potential because English fluency in the uncontacted portion is unknown) 23 participated (41.8%), 2 refused to participate (3.6%) and 30 were unreachable for survey (55.6%). Households were approached weekday mornings, evenings and on one Saturday. This low response reflects the low person-hours available to collect data. It is possible that there is some systematic bias in responses depending on one’s employment status. The direction of such a bias is difficult to assess: unemployed persons may be more susceptible to stress and depression, though this is unclear. Likewise, the health promoting or inhibiting effects of the environment may vary with respect to employment status. While reaching someone at home on a weekday between 9:00 AM and 5:00 PM is not a definitive indication of employment, improved survey response rates are highly desirable to bolster external validity of any future study.

The environmental assessment and health assessment tools developed last summer are located in the Appendix. The environmental assessment contains 7 simple indicators evaluated as either present or absent. The health survey that is much more extensive encompasses questions about neighborhood social cohesion, self-perceived stress, self-perceived emotional response, several questions relating to the presence of violence and perceptions of safety, exercise activity, landscape value, and ideas about past and future trend. The very last item of the health survey is an open question which, while not especially suited for quantitative analysis, provided many insights from respondents.

I have also employed several summary variables in an attempt to create understanding about categories of health and environment by grouping similar responses within health and environmental categories. In all cases an individual who refuses to respond to a question contributing to a summary variable is not scored for that summary. Some summary variables reflect a positive response, and in these cases "Can’t Say" responses can contribute to the denominator (i.e. they do not contribute to the score for that summary variable). The interpretation of summary variables reflects positive responses contributing to the numerator.

Positive Emotionality (S1), Negative Emotionality (S2), and Total Emotionality (S3) are the first three summary scores. They assess positive, negative and total emotional response associated with living in one’s neighborhood in the month prior to the survey. Total emotionality scores for any affirmative answer to a sub-question of 21 (which is the set of self-perceived emotional response). Positive and negative emotionalities are each scored as shown in this table for an affirmative response to the sub-questions of 21. The terms positive emotionality and negative emotionality reflect a generalization of "up" and "down" emotions respectively (see Table 3).

A fourth summary variable for emotionality is Emotional Tendency (S4). This summary score combines affirmative answers to "positive" emotions and negative answers to "negative" emotions. The range is from 0 (i.e. neither affirmation of any positive emotions, nor negation of any negative emotions) to 12 (i.e. affirmation of all positive emotions and negation of all negative emotions). This variable is scored as shown on Table 3.

An environmental summary variable termed Environmental Control Summary (E8) is employed to try to grasp something of the breadth and depth of an individual abandoned lot project. As a summary this probably reflects more of the breadth of environmental intervention and it may not be a good proxy for community intent in shaping the landscape because community members often desire limited kinds of change in their street’s landscape. For example, the group with which URI works might want to remove rubbish and establish a system of waste removal, but might not want to create a space in which people will loiter or hang out.

Assuming an epidemiological model describes these relationships one would use a Prevalence Odds Ratio (POR) to describe the likelihood that exposure and outcome occur together in a cross-sectional study. The strongest of the associations in Table 4 (tests for association and POR for dichotomous exposure and outcome variables) suggest that recent URI efforts (abandoned lots projects generally and garden maintenance in particular) and a lack of neighborhood induced irritation or temperamentality occur together. They also suggest that URI efforts (formal memorial objects, formal front-edge boundaries and garden maintenance specifically) and feeling like one is part of a solid community occur together. Table 4 illustrates the 15 associations significant at the 80% confidence level (weak association). Weak significance was explored because the purpose of this study is to direct future research, rather than to establish definitive evidence. Another interpretation of the analysis in Table 3 is: garden maintenance (E4), formal memorial objects (E6) and formalized front edge boundaries (E7) are associated (both positively and negatively) with several dimensions of emotional response in residents since of the 15 associations, 14 of them occur with these exposure variables and 12 of the outcomes are emotional response indicators. The enormous upper boundaries on the 95% confidence intervals of several of the associations probably reflects the role of chance in a study with such a small sample size.

Table 5 illustrates all weakly significant tests for difference between dichotomous exposure variables (E1-E7) and ordinal outcome variables. Non-parametric tests found a strongly significant difference between presence or absence of overgrowth (E3) and both sense of belonging in the neighborhood (H17) and reported acknowledgement of neighborhood residents when meeting in passing (H16). A strongly significant difference was also found between the presence or absence of formal memorial objects (E6) and positive emotionality (S1). An assumption of ordinality for summary variables for emotional response was assumed to be appropriate because it is not clear that values reflect a quantitative relationship of effect. For example, while a response of 4 for positive emotionality does reflect twice the number of positive affirmations of the "positive" emotions in Table 3, the value of the emotions is probably not cardinally quantifiable. Of the 21 findings of weak significance, 6 were with Formal Memorial Objects (E6), 5 with waste removal (E2) and 4 with garden maintenance (E4).

Table 6 illustrates no significant correlation found between the environmental control summary (E8) and the number of days an individual exercised in the past month. H7 was analyzed both for correlation as a continuous variable and as a categorical variable in Table 4.

Statistical analyses of the data generated by this small pilot cross-sectional study are suggestive of several things. All of these points are bounded by the time of the study which was late summer or early fall. First, there is some indication that some environmental variables indicating degree of community control over the local environment may more important in driving outcome variables. Second, emotional summary variables—as tools that allow analysis of responses by category of question—may provide a useful way to analyze relationships between URI’s abandoned lot projects and mental health outcomes. The variety of associations found with different self-reported emotional variables (H21a-H21l) supports this. Third, the neighborhood social ties (NST) variables (H11-H17) also exhibited some significance for further, more specific study. A method to summarize NST responses might prove useful in facilitating further study. Fourth, this analysis did not illuminate any significant trend in relationships between exposure variables and indicators of violence and safety (H8a-H8b, H22a-H22e and H24). Fifth, little statistical evidence was found for indicators of both adult and child exercise and landscape use (H5-H7, H9 and H18) relative to environmental indicators. Better methods of generally indicating physical use of space outside formal exercise may be warranted.

Answering the original question—How do residential urban blight and community-driven restoration effect changes in health?— when employing epidemiological tools is undertaken as a series of hypothesis tests about statistical measures of association (or tests of difference in means for non-parametric tests) between an exposure and an outcome. The results are interpreted as providing evidence (or failing to provide evidence) for the causal relationship diagramed in Figure 1. In a cross-sectional study, the evidence is presumed to be unable to describe the direction of the causal arrow. The outcome of interest might actually cause the exposure of interest (the classic example of this being exposure to clean outdoor air in the Southwest and respiratory distress outcomes such as asthma). In this study the epidemiological tools have suggested associations between several possible exposures (components of URI’s abandoned lots programs, especially waste removal, formal memorial objects and garden maintenance) and a wide array of possible health outcomes (particularly among the emotional response, and to a lesser extent among indicators of security and social cohesion). The precise manner in which these relationships play out, specifically in a causal sense, requires more intensive and focused study. The inadequacy of the epidemiological paradigm to inform such models is the topic of the next section.

The question "How do residential urban blight and community-driven restoration effect changes in health?" does not lend itself to modeling dependent outcome variables driven by independent exposure variables. This is important because it reveals a limitation not with a specific theoretical model within the epidemiological paradigm, but a failure of the paradigm itself as outlined in Figure 4. The key difference in this paradigm is the bi-directionality of the causal arrow. The epidemiological paradigm is limited because it cannot make hypotheses about factors that are causally fed back upon one another.

The epidemiological model fails to capture three characteristics of this general case. The first of these characteristics is interdependency. The interdependent nature of the state variables violates the basic assumption of a dependent outcome variable driven by independent exposure variables used in regression modeling. The statistical methods for assessing associations are inappropriate for such relationships (some Bayesian statistical methods incorporate terms for pair-wise interdependence of variables and might prove more appropriate, but this is beyond the scope of my analysis).

Third, the concept of allostasis—long-term and intermediate-term adaptation-induced stress of a system moving regain stability—may bear upon an interdependent system. Essentially, the contribution of various risk factors to any outcome may change relative to the history of stress in the system. This is another kind of extreme sensitivity in chaotic systems the details of which are well beyond the scope of this discussion.

Neither increasing statistical power nor increasing precision of variable definition and assessment within any epidemiological model can address these three characteristics because they arise from a fundamentally different paradigm. A simplistic example will serve to illustrate what I mean by a system characterized by feedback between interdependent variables.

Consider a system for which three variables serve as indicators (Figure 5). One variable will simply be Litter, which can serve allegorically as an environmental assessment. A second variable might capture something of the litter producing or litter abating behavior of individuals within the population of study. This allegory for a behavioral indicator will be labeled Hedonic Litter Capacity. And the third variable will allegorically capture something of the mental condition related to the Litter in the environment. It will be Environmental Gruntlement. Now clearly Hedonistic Litter Capacity drives Litter. And Clearly Litter drives Environmental Gruntlement. And, there is no surprise that an individual’s feelings about the environment will drive her attitude towards acting in it. That is feedback. The causal pathway does not proceed through some environmental factor or factors and confounders to arrive at some final outcome destination. In fact complex processes can occur in both directions. Perhaps I get really upset when I observe someone trashing my favorite space. Causal feedback can exist between two variables.

If one accepts a dynamic system model of causality (something that probably looks a lot like Figure 6) as a more appropriate description of the complex of health and environment in which URI is intervening, then the tools required for quantifying the model have two primary qualities. First, the statistical tools employed must be appropriate to the causal assumptions of interdependence between many factors in the system (as those common to epidemiology are not). If risk models are to transcend the limitations of the epidemiological paradigm, they must also be able to integrate different kinds of causal linkages (i.e. one or bi-directional causality, indirect causality, multi-factorial causality in these different modes, etc.) Second, such tools must be able to characterize different kinds of feedback. This requirement implies longitudinal data that can capture details about changes in indicators over time and sequence. Such longitudinal data could be generated by panel-studies that are essentially a hybrid between prospective cohort and cross-sectional designs. One novel approach afforded by this kind of study is that the unit of study could be the union of study population and the environmental locus. This way instead of tracking a population as specific individuals, health data is tabulated and careful accounting of emigration and immigration data are carefully tracked to characterize the relationship dynamics that occur there. In this way the co-evolution of health and environment may be explored.

FIGURE 6—Approximation of the theoretical model of health and community blight restoration

---

---

 

New observational approaches for health and environment

The current inadequacy of environmental health science to address questions such as the URI’s mandate begs the question of how to inform perspectives on how the human interrelates with its environment (what might be called human ecology). The requirement by some audience—whether a funding agency, governmental body or professional group—of an epidemiologic articulation of the relationship between health and environment is a bias against research guided by different modes of data collection. This section advises alternative directions for the scientific study of health and environment. An argument is made for employment of scientific observation outside the scope of that associated with the quantitative modeling currently in vogue with the proponents of evidence-based policy (though none of the following inherently exclude attempts at modeling). An observational project outlined around such alternative modes of data collection concludes the section.

Acceptance of salutogenesis as an appropriate dimension of health conception implies new categorization and description of health and environmental processes. Just as the collective body of observations of malady among clinicians leads to common expectations about the etiology of diseases, a collective body of observations about health promotion should produce shared expectations about the natural history of salutogenic factors. The likelihood of complexity in salutogenic factors may make this kind of categorization challenging. Articulating some salutogenic processes may be fundamentally different than the articulation of certain pathogenic processes such as gene expression, quantitative description of some toxicological pathways or many cancer histologies. But the operational description of complex processes already occurs in certain kinds of immune disorders and allergic sensitivities (again, asthma is an excellent example). Moreover, psychology and psychiatry both reflect operational definitions of many disorders that do not mesh well with more classically biomedical diseases (anxiety disorders, for example). Looking at health through a salutogenic lens will produce different images of health than a pathogenic lens.

The question of what "observational data" means is also important to understanding health’s ecology. Currently the bias is unequivocally on the side of cardinally or ordinally quantified measurement. "Show me numbers," the audience implores "Then I will listen and fund your research. Numbers will justify approval of this grant for intervention." But numbers may not be available. The lack of appropriate statistical tools and the difficulty of cost-effectively producing statistical power may both make quantitative justification problematic. Alternatives include uncounted qualitative questions such as the closing question in the health indicator survey in this study. Visual data include comparative photo-journal studies of different environments, community-produced representations of neighborhood environments (such as those employed in rapid rural appraisals) or visual expressions of social and environmental processes through art (a practice with an historic counterpart in disease research). Employing metaphor, anecdote and detailed case study can provide deep insights through storytelling data. The old apothegm "A picture is worth a thousand words" ought to be turned on its head. A thousand well-chosen words can bring understanding—supplanting another individual power-point ready graph in an endlessly expanding population of such images—that will do much to counter the understanding that is reduced inversely to the proportion of that population.

A study seeking to elucidate the natural history of urban blight could make excellent use of such methods. A population of newly abandoned residential lands could be documented photographically. Regular intervals (weekly or monthly) of well-defined perspectives of the properties over the course of five years or a decade could be coupled with interviews with local residents about attitudes and use of the neighborhood landscape. Potential and opportunity for restoration within the studied neighborhoods may be realized by the skilled weaving of needs and resources into stories shared between community members, enabling organizations and external resources such as government or private agencies.

This study’s approach to exploring the health effects of URI’s projects has produced four primary insights into environmental health science and some limitations in its applicability understanding health and environment:

The current emphasis of EHS as a discipline may be generalized as concern about environmental media as vehicles for pathogens of various sorts. In many applications this approach is highly suited for non-interdependent models of causality. However, health and environment also interact in other ways better understood in terms of dynamic systems.

The health/environment relationships modelable within the epidemiologic paradigm identify independent leverage points that may be used to control an outcome. If one pulls on the string of some independent exposure, whether a hazard or therapy, the marionette (at least in a population sense) will respond to some degree as directed. However, in dynamic systems pulling on a single string may produce no effect, the desired effect or actually exacerbate the problem. And the manner effect produced is likely to change, even radically depending on the state of other factors in the system.

My preliminary conclusion is that barring more useful knowledge of the system, effective intervention becomes an action that simultaneously exerts pressure on all variables of critical import to conform to desire. URI’s mode of fostering social cohesion and functional capacity, personal development and environmental improvement simultaneously serves well as an example. This implication may be considered form two perspectives. First, intervention in such systems should explicitly target the processes of interest. Second, when intervening in such a system, all interdependent processes ought to be evaluated for impact, whether or not a particular process is of primary concern as an outcome.

Finally, because dynamic systems function as iterations over time, the nature of interventions within them might also tend towards implementation as processes rather than events. While many interventions generated within the epidemiologic paradigm occur over a period of time (i.e. a steady course of medication, etc.), interventions within dynamic systems may reflect more complex and nuanced actions. These may reflect the importance of timing signals within the system or sensitivity to feedback conditions.

Multi-factorial models of risk in health and environment are fundamentally inappropriate to characterizing dynamic systems of health’s ecology.

Risk cannot be characterized accurately with predictive utility that extends for any non-myopic distance if the risk model does not employ a probability model appropriate to dynamic systems. As discussed above the issue of regression modeling versus dynamic systems modeling arises from fundamentally different constructions of causality. Risk models generated under the assumptions of the former and applied to systems better described by the later are invalid regardless of statistical power of association.

Examples of this shift in ethical perspective are URI’s abandoned lot projects. The analysis of health risk and benefit for such interventions suffers. How does one compare the potential for harm through traditional EHS exposure and outcome scenarios (dust and asthma, soil-lead in childhood poisoning, etc.) with the potential for social, mental and physical weal arising from the restoration efforts that could lead to these exposures? While one may feel comfortable arguing that "Good work is understood to be good" the example of US city police departments discouraging arborculture within neighborhoods is at odds with research detailing child developmental and general social benefits of neighborhood green spaces. From another perspective, if salutogenic factors and processes tend to be those that are interdependent, then current scientific tools may not be able to provide evidence that can inform decisions to fund salutogenic research rather than research into pathogenic factors and processes.

Equipoise is an ethical standard most common to the realm of biomedical clinical trials. It is derived from the prescription for physicians to "do no harm." In essence equipoise is the idea that given human subjects in an experiment testing an intervention (typically a therapy such as a drug, or a surgical technique) of unknown efficacy, the comparison to this intervention must be the known best alternative therapy. Only in the absence of an alternative therapy about which there is some measured evidence for its efficacy a placebo comparison is appropriate. If some kind of knowledge about the health promoting efficacy of an intervention (for example, URI’s programs) a moral obligation exists to engage in that intervention baring some better alternative (including a "do nothing" alternative). This is a moral argument for advocacy, and one rise in the high ground on top of which environmental health promoters may stand. The weight of this argument rests in part on whether an acceptance of the salutogenic paradigm translates the prescription to "do no harm" to "maximize health."

The exact association between variables that are to varying degrees interdependent is obscured by the incapacity to effectively model dynamic systems (see Figure 4). A large sample size is unlikely, particularly as observations likely to be ecological in nature (URI’s programs are bounded by the context of New Haven resources, limitations and culture) and reflect accumulated knowledge of largely unique events. Attempts to analytically explore this question are also likely to encounter a kind of irreproducibility in which experimental conditions are so large and complex, that they cannot be replicated. Geologists and ecologists often encounter such situations. Therefore a predictive model describing the health effects of particular URI activities is unknown, and may not in fact be knowable.

Notwithstanding this daunting lack of a risk or benefit model, several conclusions are available from this study. Returning to four questions posed earlier:

Environmental health science is currently constrained by inappropriately narrow definitions of health and methods for describing health relationships. URI’s attempt to grapple with the question of health in its environmental program is appropriate. It is also forward thinking work that does not have a large body of research methodology to draw upon. Continued efforts to understand URI’s role as a health actor within the environment are warranted, but present many new challenges.

Despite the lack of a dose-response articulation of URI’s effect on health several health effects are apparent. First, URI’s projects directly affect both social and mental aspects of health. The restoration of a neighborhood contributes to positive feelings and to the reduction of stressful and negative feeling about the daily environment. At the same time, the program fosters social interaction by providing opportunities for new social interactions, bolstering of the physical space in which many social interactions take place and through mobilizing resources for the improvement of social skills. The abandoned lot projects also create an environment of shared exercise and learning that enhances personal development for community members of all ages. Stress and anger-reducing effects are likely to provide health benefits in physical terms as a growing body of evidence shows that immune system function is strongly tied by several biochemical pathways. Second, these health-giving influences are interactive and may combine to produce an effect greater than the sum of its parts. Such system effects are probably in the roots of the neighborhood wisdom that "daisies beat drug deals." Put succinctly such redevelopment affects health because it is broadly salutogenic, and because it is an ongoing process that is constantly future-seeking. While the indoor stresses of daily life in a socially or economically marginalized community may in many ways overwhelm the magnitude of the outdoor landscape’s contributions to distress, rehabilitation and blight are part of the very process of marginalization and therefore are an excellent leverage point for intervening.

Urban residential blight is a complex set of processes that—at least in part—reflect a divestiture of both control of an outdoor environment and articulation of shared use values by its residents. Such processes include a de-formalization of social use of the landscape (i.e. property maintenance within the population of properties diminishes and is replaced by individual or limited social landscape use with no maintenance such as dumping or prostitution), an incapacity to control "organic" processes such as plant overgrowth or accreted damage to physical structures, and a disintegration of the social interconnections required for the sharing of vision, feelings and opportunity. Just as the rigid conceptual constraints of mind-body duality within western medicine are giving way to an integrated and holistic sense of the individual, so is recognition of the enmeshment between social and physical environments. The scope of intervening for health within a context of urban blight should be considered as such. While the physical environment can certainly be altered with little or no input or participation between the residents and environment-altering agency, such approaches are unlikely to foster a personal sense of investment, future vision or social skills within and beyond the community. At the same time, landscape interventions such as URI’s are only one of many demonstrations of realized community values. Because the effect of URI’s projects are so bounded by time and locale, they can provide a "proof of concept" for neighborhoods struggling for common meaning and hope.

To paraphrase Vishnu in the Bhagvad Gita, "Uncertainty cannot abrogate the fact that one must decide whether to act." When discussing action within the environment, epidemiologists have been guided by the Precautionary Principle. Particularly, they give consideration to the idea of default when some action such as setting pesticide use levels, or emission standards could cause harm. However, as discussed above salutogenic and pathogenic risks may be quantitatively unknown and unknowable. Such uncertainty leaves two factors to decide: intuition and informed consent.

Three ethical guidelines should direct URI within a health sense in such cases. First, keep participants apprised of the best understanding of salutogenic and pathogenic risks involved in a URI project. For example, distributing an introductory pamphlet describing URI’s understanding of their role in community health, and providing access to more detailed information upon request could serve to keep participants informed.

Second, articulate the costs and difficulties of enhancing knowledge about such risks in the face of different interventions. A parent of young children living adjacent to a URI site might seek to allay concerns about possible exposure to toxins agitated by site cleanup. The range of possible action progresses through casual dismissal of such concerns to research into land-use records to internship-bracketed monitoring of URI interns to expensive sampling and lab testing. The resources available to generate better knowledge of risks must be weighed against the resources available to actually make improvements in the community.

Third, URI should advance its understanding of human ecology. This guideline has several implications. URI’s staff training should incorporate best knowledge about health and its projects. URI should also observe and there are a wealth of ways it can do so. URI health observations can be added to the formal rubric of project reports. URI summer interns are in an excellent position to collect data using tools like the ones employed in this study. Contributions from such observation can provide a considerable longitudinal database of cross-sections as the years pass. At an institutional level, URI can also seek channels of communication with other health or urban landscape-oriented agencies and students and faculty to further inform its own understanding of health and restoration.

Ultimately my efforts to explore this small facet of human ecology offer few answers and many questions. But the questions are different kinds of questions, and trying to answer them may lead to new kinds of understanding. Many environmental health scientists I know share with me a concern about the limits of knowledge and application inherent in environmental health science as it is traditionally taught and practiced. My hope is that the ideas I have presented here—first the salutogenic paradigm of health’s ontology, and second that there is a dynamic system paradigm of epistemology (see Table 7)—help justify an exploration of different questions about environment and health and provide affirmation of broad perspectives of health ecology.

(baseline indicators of environment)

Cast off mattresses, discarded garbage bags not set out for removal, furniture, appliances, automotive detritus, etc. Waste associated seasonal or non-periodic dumping.

Filled and tied garbage bags, stacked disposables, material curbed and awaiting pickup, obviously engaged in waste gathering like clippings, permanently installed garbage cans, etc.

Overgrowth ___ ___

Occlusion of paths, drives, doors, porches etc. by shrubs and trees, colonization of open spaces (as opposed to at edges and fences) by pioneer species like Ailanthas or bamboo, grass and weeds gone to seed in places where they are normally kept trimmed, etc.

Designed garden features like perennials planted this season, maintained trees or hedges, lawn, etc. Assessors ought to be sensitive to climatic factors such as drought or winter that may detract from even the best maintained gardens.

Lined or paved walks, gateways, driveways or parking spaces, etc.

Mural graffiti, sculpture, monuments, benches, plaques, birdbaths, etc.

Cyclone fences, LCI corral fencing, berms, etc.

¡ ¡