A NEIGHBORHOOD-LEVEL POTENTIAL HEALTH IMPACT SCORING TOOL TO SUPPORT LOCAL-LEVEL HEALTH IMPACT ASSESSMENTS

Abstract

Health Impact Assessments (HIAs) have quickly become a widely utilized tool for integrating health and health-related evidence and data into decision making processes across a range of projects and polices. Integrating and utilizing the wide range of available data can be daunting. To support communities seeking to engage in health impact assessments, we developed the Neighborhood Potential Health Impact Score (NPHIS) methodology. We present the NPHIS method’s four step process, and how this process was applied to an HIA focusing on the rebuilding of public housing following a natural disaster. We discuss developing the boundary definition, selection and definition of indicators, calculation of the NPHIS, and interpretation and utilization of the scores. Findings were validated using feedback from a community stakeholder advisory board as well as through feedback collected from focus groups of community residents. NPHIS methodology has proven to be a useful resource in better understanding the complex sources of potential health impacts facing communities, and in being an evidence-based, data-driven resource for HIA decision-makers and their stakeholders in our specific application. Other groups seeking to integrate similar data into their decision-making processes could benefit from replicating the NPHIS in their efforts.

INTRODUCTION

Health impact assessment (HIA) is a structured process used to incorporate health considerations into decision making processes occurring outside of the traditional realm of the health sector.^1,2 Wernham (2011) states that HIAs provide a platform for integrating scientific data, public health expertise and principles, and stakeholder input to identify the potential health effects of a proposed policy, program, project, or plan and to craft health-based recommendations.³ Although widely utilized in many other countries, HIAs have only recently started to see widespread use in the United States (14 HIAs were documented in the United States between the years 1999 and 2006; at least 376 were documented between 2007 and early 2016).⁴ To date, HIAs have been used to support the decision making of projects and policies related to housing development, access to food, goods movement, paid sick leave, and a host of other topics.³

An HIA’s success hinges, at least in part, on robust sources of data, stakeholder input, and achieving balance between optimal and realistic scenarios. Further challenging the HIA process is that communities are often faced with multiple decisions that potentially impact health, and amongst each of these decisions are a number of possible intervention strategies that could prevent or mitigate negative health impacts and/or promote positive health impacts. It can be difficult to know where to begin.

HIA’s operate within the limits of best available evidence with an eye towards the likely impacts of a given policy, project, intervention, or other activity. Mindell and colleagues note that among some of the HIA process’ current limitations are the lack of quality in available evidence, including quantitative data for analyses. Issues around rigor of collection, quality and analytical approaches are also noted.⁵ While useful, quantification of health impacts (or even potential health impacts) is often not attempted, due to a number of factors, including lack of data, availability of tools, etc.⁶

While no standard tool is available for every HIA, there are a number of tools that may be adapted for use by HIA practitioners seeking to quantify potential health impacts. Lhachimi and colleagues provide a review of a number of tools, in which they identified 6 as being generic enough for application to diverse diseases, risk factors, and settings. However, these 6 were either too technically advanced, not accessible, or oversimplified.⁷ As noted by Veerman, et al., ideally, tools are needed to support health impact assessments by aiding in quantifying the effects of social determinants of health, simulating impacts, and summarizing measures of public health data and expert opinion.⁶

To support communities seeking to engage in health impact assessments, we developed the Neighborhood Potential Health Impact Score (NPHIS) methodology. This paper presents the NPHIS method for examining these complex issues facing HIA practitioners. This method may be a useful tool for empowering communities to assess several different sources of risk to health in their communities, understand the potential cumulative impacts of these diverse sources of risk, integrate data from existing data sources and expert opinion, and estimate potential improvements for identified candidate interventions or policy solutions through simulation. The end product can also serve as a source of baseline data, a tool for beginning the screening and scoping process, a tool for estimating various solutions’ potential effectiveness, and a source of data to help inform targets for the evaluation of an HIA’s impact.

MATERIALS AND METHODS

The NPHIS was developed as part of an HIA that examined where and how to rebuild public housing following a hurricane.⁸ For this project, multiple possible threats to the health of future residents needed to be considered when determining the best place to rebuild housing. Further, specific locations of where to rebuild were not yet determined. These two challenges led to the development of the NPHIS method. The process of developing and applying the NPHIS method was reviewed and approved by the University of Texas Medical Branch Institutional Review Board.

The NPHIS method relies heavily on geospatial and statistical analyses. For these projects, geospatial analyses were conducted using ArcGIS 10.2 (a geographic information system software package available from ESRI, Redlands, California, USA), and statistical analyses were first developed using Microsoft Excel, and were replicated in Stata 14.1 (a statistical analysis package available from StataCorp, College Station, Texas, USA). However, the NPHIS method could be implemented using a variety of free, open-source software packages, such as QGIS, LibreOffice Calc, and R. The NPHIS method, described in detail below, consists of 4 steps: 1) boundary definition, 2) indicator selection and definition, 3) calculation, and 4) interpretation and utilization (which is presented in the results section) (see Figure 1). Further discussion of the more technical aspects of this process are presented elsewhere.⁸

Figure 1:

Flowchart of the NPHIS process

Stakeholder Engagement

We convened a stakeholder advisory board comprised of leaders from multiple sectors across the community. This included municipal government representation (e.g., urban planning, police, etc.), local neighborhood leadership (e.g., neighborhood associations), commerce (e.g., port, chamber of commerce), human development (e.g., families and children advocates), housing (e.g., local public housing authority, housing advocates from different perspectives), and others. This group provided input and feedback on several aspects of this project, including boundary definition, identification of indicators and data sources, weighting of variables, interpretation of analyses, and dissemination efforts.

We also engaged a targeted group of stakeholders consisting of residents from lower-income areas of the city. Three groups of 9–12 residents (31 in total) participated in these interactive focus groups that included a mapping exercise. This exercise engaged participants to highlight areas in the city that were more salutogenic and areas that were more pathogenic to the health of their neighbors, families, and themselves. Results from this exercise were used to corroborate findings from our application of the NPHIS method to support identifying sites for the rebuilding of public housing.

NPHIS boundary definition

Coinciding with the scoping phase of health impact assessments,² two boundaries need to be defined at the onset of using the NPHIS. First, the geographic boundary defining the limits of the study area needs to be established. This boundary is termed the community boundary. Examples include political boundaries of a given city, county/parish, or state, though any boundary can be utilized. This boundary sets the universe from which each smaller neighborhood will be compared against other neighborhoods. For our project, the political boundary of [Removed for Review] served as the community boundary.

The second definition is the neighborhood boundary. Neighborhoods here are defined as smaller units within the community and serve as the units of analysis for the NPHIS. Example neighborhood boundaries include census tracts or block groups within a county or parish or counties/parishes (within states/provinces). As with the community boundary definition, any definition of boundary for neighborhood can be utilized so long that the geographic units are smaller than and contained within the community boundary. However, due to the statistical methods used in calculating the NPHIS, we recommend selecting a neighborhood boundary definition that will yield more, rather than fewer, neighborhoods within the community boundary. All neighborhoods within the community boundary should be included in the calculation of the NPHIS. Further, neighborhood boundaries should be contained wholly within the community boundary. This can be problematic in circumstances where neighborhood boundaries and community boundaries are not congruent, such as with some instances of Zip Code Tabulation Areas (ZCTAs) crossing political boundaries of cities, counties/parishes, etc. One possible solution is to define the community boundary as the area containing a set of targeted neighborhoods. In our application of the NPHIS, census blocks served as neighborhood boundaries.

NPHIS indicator selection and definition

Sources of potential health impacts (both pathogenic and salutogenic⁹) are then identified using a collaborative and community engaged process with the goal of identifying a wide a range of sources of social and environmental contributors to health in the community (in our case, with the support of a community advisory board, described earlier). There are several resources available that provide a range of example candidate sources of potential health impact and appropriate data sources for consideration (e.g., the San Francisco Indicator Project¹⁰). Once contributors to health are identified, the team identifies specific indicators and data sources for each source. Operationalization of these indicators can be informed by metrics described in the literature, through input from local community expertise, or by practicality based on data availability. Note that data required for the NPHIS calculation are geospatial in nature. Data sources identified need, at a minimum, a geographic identifier (such as census block, zip code, etc.) in order to be utilized. Common geographic resolutions include point or parcel-level data, census block or tract, or ZCTA.

Table 1 highlights the sources of potential health impacts used in our initial development of the NPHIS, as well as the indicators that were utilized, their associated data sources, and their geographic resolution. Note that not all sources of potential risk identified by our team were able to be paired with a specific indicator or data source, and therefore were not able to be included in our NPHIS analysis. When presenting findings, noting indicators that could not be included due to unavailable data highlights areas where future data collection efforts are needed.

Table 1:

Indicators used to calculate the NPHIS in [Removed for Review]

Source of Potential Health Impact	Indicator	Type	Data Source	Geographic Resolution	Weight & Direction of Impact in Calculation
Population density	Number of people / area of census block	Density	2010 Census	Census block	8 / +
Density of poverty	Percentage of all households that are single parent households	Density	2010 Census	Census block	10 / −
Racial / ethnic segregation	Number of racial or ethnic minorities / total population	Density	2010 Census	Census block	8 / −
Living within a flood prone area	Is the census block within the FEMA 100-year flood plain	Binary	U. S. Federal Emergency Management Agency	Census block	6 / −
Living near a public park	Distance to nearest public park	Proximity	Municipal Data	Parcel	7 / −
Living near a public recreation center	Distance to nearest public recreation center	Proximity	Municipal Data	Parcel	8 / −
Living near a public elementary school	Distance to nearest public elementary school	Proximity	Texas Education Agency	Parcel	7 / −
The number of licensed child care providers in the area	Number of licensed child care providers / total population	Density	Texas Department of Family Protective Services	Parcel	5 / +
Living near health care services	Distance to nearest hospital or federally qualified health center	Proximity	U.S. Health Resources and Services Administration	Parcel	8 / −
Living near healthy, affordable food sources vs. less healthy food sources	Density of supermarkets and grocery stores vs. fast food and convenience stores	Density	U. S. Dept. of Agriculture	Parcel	9 / +
Density of businesses permitted to sell alcohol for off-site consumption	Density of permitted facilities	Density	Texas Alcoholic Beverage Commission	Parcel	9 / −
Presence of pedestrian safety measures	Presence of reduced speed limits, speed humps	Binary	Municipal Data	Point	9 / +
Living near a truck route or major traffic route	Distance to nearest major highway or designated truck route	Proximity	Census	Line	8 / −
Living near an industrial zone	Distance to an area zoned for industrial use	Proximity	Municipal Data	Parcel	10 / +
Living in an area with environmental hazards	Distance to nearest facility in Toxics Release Inventory or Facility Registry Service	Proximity	U.S. Environmental Protection Agency	Parcel	8 / +
Living near a bus route	Distance to nearest bus route/stop	Proximity	Municipal Data	Line	6 / −

^{a –}This row indicates whether an indicator’s value was added or subtracted in the calculation of the score. For example, being in close proximity to a park was determined to be health promoting, and increased distance from a park would be less desirable. Thus, distance from a park was subtracted from the score. Conversely, living near an environmental hazard is less protective of health, so increased distance would be desirable. The distance from such sites was added to the score.

NPHIS calculation

The NPHIS uses three types of geospatial calculations. First, indicators can be operationalized in terms of distance from a given neighborhood. Examples include proximity to a major traffic route¹¹ or proximity to a park¹², calculated using the Near tool on an appropriately projected surface in ArcGIS 10.2. Second, some indicators are operationalized based on the density of the indicator. Examples include the density of alcohol vendors (for consumption off premises) within a neighborhood¹³ and the density of healthier versus less healthy food outlets¹⁴, and are calculated by creating buffers around the sources of risk at distances appropriate for each indicator (as determined through current findings in the literature, current applicable regulations, and stakeholder input), and then counting the number of features overlapping a given neighborhood. The count can be standardized using the number of people living in the given neighborhood (e.g., per 1,000 residents) if appropriate. Lastly, some indicators were simply binary. Examples include whether the neighborhood was located in a U.S. Health Resources and Services Administration-designated medically underserved area¹⁵ or within the U.S. Federal Emergency Management Agency (FEMA) 100-year flood plain¹⁶. Regardless of type, each indicator should be calculated for each neighborhood. To put each indicator into a common unit, each indicator should be standardized using z-scores.

Next, the strength of the potential impact to health needs to be accounted for. For example, while both proximity to a park and proximity to a major traffic route both have the potential to impact the health of residents in a given neighborhood, the severity of that impact to the health of the neighborhood’s residents may not be equivalent. To account for this difference in magnitude of potential impact, each indicator is multiplied by an impact weight. In our example, weights ranged from 0.1–1.0, but weights can be set as desired based on individual case context (see Table 1). Considered weighting schemes were a 3-point system (low, medium, high), a 5-point scale, and a 100-point scale. Determination of each indicator’s weight should account for the strength of the evidence for each impact in the literature, the quality of the indicator and associated data set relative to how each are typically operationalized in previous studies and consider input from community stakeholders. Input from our stakeholder advisory board was very influential in determining the final weights of each indicator, as this group provided information on the local context of each indicator’s role in determining health locally. For example, weighting for access to licensed childcare was lowered from our initial recommendation due to the concurrent implementation of a human capital plan for public housing residents which included provisions for ensuring access to childcare resources.

Now that each indicator is in a common unit, and the magnitude of potential health impacts is accounted for, we create a cumulative potential impact score by adding all weighted standardized indicators together. The direction of the cumulative score calculation should be consistent. In our example, an increasing score represented a relatively more salutogenic potential health impact. So, for this example, the weighted standardized value for distance to a grocery store was subtracted from the score, as the literature demonstrated that increasing distance from a grocery store was negative in terms of health impact, which was corroborated by input from local stakeholders. Conversely, distance to a major traffic route was added to the cumulative score, as increasing distance from a major traffic route was noted as being positive in terms of health impact in both the research literature and by community stakeholders. There was discussion on whether interactions amongst indicators should be multiplicative or additive. There is evidence that additive models are more appropriate when indicators are linked to health impacts through different processes, and this is typically the case when using disparate indicators of potential health impacts in the NPHIS.^17,18

RESULTS

The cumulative score reflects the relative potential health impacts of living in a given neighborhood, due to the indicators used to create the score, compared to all other neighborhoods within the defined community. As the value of the score itself is arbitrary, to aid in interpretation, calculating and utilizing the percentile rank for each neighborhood score is recommended. To aid in interpretation, percentiles could be calculated for only neighborhoods that are populated, neighborhoods in specific areas that are of interest, or across all neighborhoods, depending on how the score is being utilized. This score is not an indication of whether a given neighborhood is “healthy” or “unhealthy”; rather the score reflects the potential health impacts of the social and physical/built environment in the proximate surrounding area (e.g., neighborhood), relative to other neighborhoods in the study area (e.g., community). The interpretation is limited to the potential for impacts on health attributed by the indicators included in the calculation of the cumulative score. Thus, it is critical to have a wide variety of indicators to get a broad understanding of the cumulative sources of health impacts.

The score is impacted by the weighting scheme utilized for each indicator. Several scores could be calculated using various weighting schemes (such as those determined by the research team, weights determined by the residents of the community or sub-sets of residents, etc.). To perform a sensitivity analysis, the team can then create various sets of scores based on various weighting schemes could yield useful information about differences in perceived risk and threats, as well as differences in the understanding of the impacts of salutogenic (i.e., health promoting) and pathogenic (i.e., health detracting) factors in the neighborhood.

While the percentile ranking of each neighborhood can aid in locating areas with desired, or undesired, relative health impacts, this scoring process also provides an opportunity to identify priorities for policy development and public health intervention. First, for any given neighborhood, the individual indicators of the overall NPHIS score for that neighborhood can be examined to determine which specific indicator(s) are negatively (or positively) impacting the score. This examination can aid in determining which indicators should be addressed to improve the overall NPHIS cumulative impact score. This information can inform policy and intervention strategies that potentially have greater potential at improving the health impact of a specific neighborhood (or neighborhoods). For example, if proximity to a traffic route is a key driver of a given neighborhood’s score, intervention strategies aimed at mitigating the impact of this proximity, such as redirecting traffic, may be better investments for improving the overall neighborhood’s potential health impact.

In our example, the results of the scoring process largely matched the expectations of community stakeholders. Members of a community stakeholder advisory committee formed to advise this project reviewed the scores for various areas of the city (visualized using a color-shaded map that presented NPHIS scores by quintiles). The committee agreed with the findings. Further, three focus groups (each with approximately 9–12 participants) were held with lower-income residents in [Removed for Review] to further examine the results of the NPHIS process in our example. During these focus groups, participants were presented a map of the city and, using adhesive colored flags, asked to indicate areas that were health promoting (salutogenic, green flags) and health compromising (pathogenic, red flags). The areas marked by the participants strongly correlated with the scores generated by the NPHIS process, offering further face validity to this approach.

The results from this initial application of the NPHIS were translated into several dissemination products, including a technical report, a four-page executive summary, and a Microsoft PowerPoint slide deck.⁸ In addition to these resources, the provision of specific indicator and NPHIS score data for any given census block in the study area was made available to decision makers for use in their deliberations into where to consider rebuilding of public housing following a natural disaster. These data could be used not only to select specific areas or rule out others, but also to determine what potential built or social environmental interventions might be feasible to enhance the potential salutogenic impact in that area.

In summary, the NPHIS supported each of step in our HIA process. Further, the NPHIS process served as a catalyst for building a collaborative sense of shared ownership among stakeholders engaged in the process. The NPHIS process provided stakeholders with a more concrete understanding of what the issues of concern were behind the policy decisions at play and greater understanding of the rationale behind recommendations being presented. Because their input, debate, and discussion were accounted for throughout the process, our stakeholders were confident that the HIA’s final recommendations were not only evidence-based and data-driven, but also more aligned with community needs and values. Ultimately, the HIA’s recommendations were incorporated into official informative materials provided to the actual decision makers and the contractors tasked with replacing new public housing following this natural disaster.

Implementation of and Challenges in Applying the NPHIS

Once initiated, this project took a total of 18 months to complete. The timeline was extended for a number of reasons. First, the specific plan for rebuilding of public housing was in a state of flux, so determining the scope of our application of the NPHIS, including our boundary definitions, shifted over the first several months of this project. Ultimately, rather than examining specific sites to be considered, we opted to apply the NPHIS method to include all census blocks across the entire city. This ensured that the results from our NPHIS calculations would be applicable in the event of any future shifts to the rebuilding plan.

Second, we sought to engage a broad range of stakeholders onto our advisory committee. This included groups that may be at odds politically on certain issues. Over the course of 3 months, we met with different groups of stakeholders individually to build a common ground of interest around the health of all residents and neighborhoods. Once common ground was established, a series of collective meetings were held with the entire group of stakeholders. These meetings occurred as the project met significant milestones, but generally were held at each stage of the process. We held an initial orientation and on-boarding meeting, one to identify boundaries and solicit indicators and data sources approximately one month later, one to present candidate indicators and data sources and solicit weights another month later, once to present findings from the NPHIS calculation and receive feedback on its validity approximately 3 months later, and finally a series of meetings to vet dissemination products (e.g., reports, presentations, etc.) and identify dissemination strategies

DISCUSSION

In general, this methodology can support health impact assessment (HIA) efforts at multiple stages of the HIA process.² In screening and scoping, the NPHIS may assist with identifying and narrowing geographic scope and support prioritizing sources of positive or negative potential health impacts related to the issue or issues in question. In the assessment stage, the NPHIS process can provide a valuable source of baseline data, as well as comparison data to compare a geographic area’s potential health impact to other surrounding areas. In the recommendation and reporting stages, NPHIS scores can be used to identify potential drivers behind health impacts, aid in narrowing the search for evidence-based intervention solutions, and provide the potential for identifying potential intervention impacts by modelling changes to the baseline NPHIS score.⁸ A demonstration of this type of NPHIS application will be reported elsewhere (article forthcoming). Finally, in the monitoring and evaluation stage, a longitudinal approach to the NPHIS could provide data on changes in potential health impacts within areas of concern, as well as in other areas that may not have been intended targets of the intervention or improvement process.

However, there are some specific challenges in using the NPHIS tool that communities seeking to implement it within the scope of an HIA should be aware of. First, not all communities may have the needed technical expertise on hand. Partnering with collaborators experienced in working with public data sets, geographical information systems, and stakeholder engagement or partnership building will greatly enhance the ability of a community seeking to use this tool. Second, it is critical to have multiple perspectives within the stakeholders engaged in the NPHIS process. In some instances (e.g., a politically contentious issue), it may be difficult to engage stakeholders from reasonable yet competing perspectives. Investment in time to build common ground to engage these opposed sides is most critical. Any ignored reasonable perspectives would potentially bias the scoring process, add difficulty in interpreting the findings and disseminating results.

Limitations

The NPHIS method is not without limitations. There are instances when indicators are desired, but data sources may not be available. This was encountered in our application of the NPHIS (e.g., local crime statistics), and was recognized as a limitation of the process. Also, there may be indicators that are not uniformly agreed upon by all stakeholders, which creates a need for additional deliberation. Second, data collection can vary across different areas (e.g., state-to-state) and leave data sources not comparable for a given indicator. Thus, comparing broad areas (such as states), or a small geographic area that happens to cross a major political boundary (such as metropolitan area that spans state lines) may be problematic. Data may be limited, or assumptions may have to be made about the comparability of datasets in these cases. Generalizability of the score beyond the community boundary is also a limitation. Scores generated by the NPHIS process for each neighborhood are relative to the other neighborhoods and are dependent on the weighting scheme used in each process. Thus, the scores are not comparable to other scores unless they are calculated in an identical and pooled manner.

Practice Implications

This paper presents a methodology used in a community setting to promote community decision making on the potential health impact of a neighborhood due to a multitude of diverse sources of pathogenic and salutogenic health drivers in the surrounding environment. The method is particularly useful when conducting a formal Health Impact Assessment (HIA). By creating relative scores for neighborhoods within a community, decision-makers, advocates, and other stakeholders can better understand the relative potential health impacts facing residents in a given neighborhood or set of neighborhoods relative to all other neighborhoods, as well as gain a better understanding of what may be driving these potential health impacts. This process is limited by, in part, the indicators of potential health impact (and the data used to evaluate these indicators), and comparability and generalizability beyond the study area. Overall, however, the Neighborhood Potential Health Impact Score methodology may be a valuable resource in better understanding the complex sources of potential health impacts facing communities, and in being an evidence-based, data-driven resource for HIA decision-makers and their stakeholders.