Incorporating community-level risk factors into traumatic stress research: Adopting a public health lens

https://doi.org/10.1016/j.janxdis.2022.102529Get rights and content

Highlights

  • Infusing community-level risk factors into trauma research broadens intervention targets.

  • Census data can be used to compute community-level risk factors.

  • Community-level risk factors predict elevated PTSD symptoms across the acute post-injury period.

Abstract

Infusing community-level risk factors into traumatic stress research can broaden intervention targets. The Neighborhood Deprivation Index (NDI) and the Index of Concentration at the Extremes (ICE) are two common community-level risk factors derived from U.S. census data. We provide R scripts facilitating the computation of these risk factors and demonstrate their relationship with PTSD symptomatology in 74 injury survivors assessed at 2-weeks, 6-weeks, and 3-months post-injury. The NDI and the ICE were computed using the Census Data Application Programming Interface, then matched to participants’ census tracts using their residential addresses. Results indicated that after controlling for person-level characteristics, both risk factors were associated with PTSD symptom severity at follow up time points (Cohen’s f2 =0.011,.14). This study provides an easy method for computing the NDI and ICE, demonstrates the increased mental health risk that they convey in the aftermath of injury, and highlights their value in intervention efforts.

Introduction

The field of public health aims to protect and improve the health of people and the communities in which they live (CDC Foundation, n.d.). At the center of public health is the acknowledgement of the importance of the broader ecosystem in which people live and the role that it plays in shaping the health of the individual. This orientation stems from contemporary epidemiological theories of human disease including ecological systems theory (McMichael, 1999), eco-epidemiology (Susser & Susser, 1996), and eco-social theory (Krieger, 1994). Common to these theories is the recognition that person-level health outcomes are a function of multiple, interacting levels existing outside of the individual, necessitating the consideration of contextual factors that shape disease trajectories. Given the diagnostic criterion requiring traumatic event exposure, PTSD represents a mental health disorder that is inextricably intertwined with environmental factors that reside outside of the individual (DiGangi et al., 2013). Indeed, the considerable variation present in the cross-national prevalence of PTSD in the World Mental Health Surveys underscores this point: lifetime PTSD rates range from 0.3% to 8.8% across the world (Koenen et al., 2017). This geographical variability provides an indication that community-level factors that co-vary with place may be instrumental in shaping the traumatic stress responses of the people who live in the regions where those factors are operating.

To date, the field of traumatic stress has primarily focused on person-level risk factors associated with PTSD (e.g., gender, coping efforts) and, consequently, person-level interventions targeting people experiencing PTSD (e.g., cognitive processing therapy, prolonged exposure). While these interventions are effective in reducing PTSD symptoms (Watkins, Sprang, & Rothbaum, 2018), they do not address the upstream community-level conditions that may set the stage for PTSD risk (e.g., neighborhood violence) or that make recovery difficult following trauma (e.g., treatment access difficulties associated with living in an impoverished area). One strategy to facilitate broader adoption of the public health perspective is to routinely include measures of community-level risk factors in trauma research protocols.

Three common methods for the measurement of community-level risk factors include: (1) administering traditional survey instruments collecting self-report data from participants, (2) conducting community audits with trained observers who perform structured evaluations of community features, and (3) using census records to create indices of community-level characteristics.

Self-report instruments assessing community-level characteristics often ask respondents to make a subjective rating of the perceived physical disorder or social cohesion that exists within their community (Coulton et al., 1996, Sampson, 1997). This approach has demonstrated that perceived community disorder increases risk for elevated PTSD symptoms among both White and Black community samples (Gapen et al., 2011, Monson et al., 2016) and that low levels of social cohesion increase risk for both prolonged distress and diagnostic PTSD levels following trauma exposure (Gapen et al., 2011, Johns et al., 2012). While these findings offer intriguing insights into the role played by community-level factors in shaping trauma responses, incorporating these additional measures into research studies may impose participant response burden. Furthermore, researchers may be reluctant to include these measures if community-level factors are not a primary aim of the research project.

By shifting data collection efforts onto researchers, community audits involve using trained observers to conduct structured evaluations of community characteristics – typically features of the built environment such as the presence of graffiti, trash, or vacant residential housing (Jones, Pebley, & Sastry, 2011). Technological advances have made “virtual” audits possible, whereby trained observers utilize digital maps to “walk through” communities at the street-level with the aid of a computer (Mooney et al., 2017). Although community audits offer a potentially viable approach to evaluating community characteristics, the validity and reliability of these methods is still being established (Marco, Gracia, Martín-Fernández, & López-Quílez, 2017). Furthermore, both in-person and virtual audits require a substantial amount of time, training, and personnel to implement.

The use of census records is a third approach to assessing community-level risk factors that overcomes many of the limitations associated with surveys and community audits. Although the United States performs a full census of its entire population every 10 years, it conducts the American Community Survey (ACS; United States Census Bureau, 2021) every year. The ACS is a smaller program of surveys performed with a representative sample of persons living in all 50 states, the District of Columbia, and Puerto Rico. Its use for assessing community-level risk has many advantages: it is free, available online, and directly accessible within many software packages through the use of the Census Data Application Programming Interface (API; Breakstone & Anderson, 2019).

Community-level risk factors derived from U.S. census data often involve aggregating social, economic, and demographic characteristics of the individuals within a designated area. Typically, this aggregation is performed at the tract level. Census tracts are small, relatively permanent subdivisions nested within counties that contain between 1200–8000 inhabitants (United States Census Bureau, 2018). They are the smallest geographical unit for which U.S. census data are reliably available. Following this tract level aggregation, the resulting community-level risk factors can then be matched to the tracts associated with participants’ residential addresses.

One approach to data aggregation is to combine person-level indicators of socioeconomic status among all persons living within a given community to produce a single Neighborhood Deprivation Index (NDI) reflecting community-level socioeconomic conditions. Typically, these person-level indicators reflect the education, occupation, housing, poverty, and employment statuses of all residents living within a geographic area (Messer et al., 2006). Relative to other measures assessing community-level socioeconomic conditions (e.g., the Area Deprivation Index; Kind et al., 2014), the NDI offers a highly flexible approach that can be tailored to specific features of a researcher’s study, including the use of census data that were collected during the exact year or range of years when a study took place.

Research employing the NDI has demonstrated that elevated tract level deprivation increases risk for preterm births, cardiometabolic markers, and premature mortality over and above the predictive effects of person-level features (Doubeni et al., 2012, Laraia et al., 2012). These findings clearly demonstrate the important role of community-level risk factors in shaping the health of individual persons.

A second approach to data aggregation is to assess the extent to which residents in a particular community are concentrated into groups defined by different levels of privilege and disadvantage. The Index of Concentration at the Extremes (ICE) uses this approach and is defined according to the distribution of people representing different income levels or racial identities within a designated area. These characteristics are often examined either singly or jointly, resulting in several different formulations of the ICE (i.e., ICE-I for income, ICE-R for race, ICE-IR for income/race). Thus, while the NDI provides a metric capturing community-level socioeconomic conditions, the ICE provides an indication of a community’s social spatial polarization.

Research using the ICE has shown that tracts with higher concentrations of lower income inhabitants confer greater risk for adverse childhood experiences among youth involved in the juvenile justice system (Baglivio, Wolff, Epps, & Nelson, 2017). Additionally, it has been shown that tracts consisting of higher concentrations of lower income or Black inhabitants increase risk for both fatal and non-fatal assault, with the highest risk observed in tracts consisting of a higher concentration of lower income, Black inhabitants (Krieger et al., 2017). As with the NDI, these findings persist after controlling for person-level features, demonstrating that factors defining the broader community in which a person lives are key determinants of health outcomes.

Matching the NDI and ICE community-level risk factors to participants only requires the collection of residential addresses, which is information routinely recorded by researchers. However, no study to our knowledge has considered how these risk factors may be related to post-traumatic symptomatology. One existing barrier that may deter researchers from using these metrics is the absence of a set of tools that make census data and the computation of the NDI and ICE easily accessible.

The purpose of this paper is twofold: 1) to provide a toolset consisting of R scripts (R Core Team, 2020) and accompanying instructions that will allow researchers to derive and apply the NDI and ICE to their own research using U.S. census data; and 2) to demonstrate the relationship that these community-level risk factors have with PTSD symptomatology over time among a sample of injury survivors. Examination of community-level risk factors is especially important in this population given that nearly a quarter of injury survivors experience a new psychiatric disorder in the first year post-injury (Bryant et al., 2010). Thus, we examined how the NDI and ICE predicted PTSD symptom severity at 2-weeks, 6-weeks, and 3-months post-injury. We hypothesized that greater community-level risk would be associated with greater PTSD symptom severity.

Section snippets

Participants

Participants consisted of 74 injury survivors who had been admitted to a Midwestern Level-1 trauma center due to injury and who were recruited during a routine medical follow up at an outpatient trauma clinic within 30 days of injury (median = 16.50, range = 3–29). Participants were predominantly male (58.1%; 41.9% female) and White (74.3%; Black 25.7%) with an average age of 35.78 years (SD = 12.25). The sample was evenly split between individuals who had received a high school diploma or less

Computing the NDI and ICE variables

The NDI and ICE variables were computed using R 4.0.0 (R Core Team, 2020) and the tidycensus (Walker, 2020), tidyverse (Wickham et al., 2019) and psych (Revelle, 2019) packages. For researchers interested in using these indices in their own research, annotated R scripts are available in the supplemental materials and at the Open Science Framework (https://osf.io/m2sav/). In addition, for researchers who may be unfamiliar with R, annotated scripts with accompanying instructions are available on

Descriptive statistics

At baseline, the mean score on the rescaled PDS was 32.52 (SD = 19.29). At 6-weeks post-injury, the mean score on the PCL-C was 39.14 (SD = 14.92), and at 3-months post-injury it was 32.85 (SD = 13.45); 29.0% of injury survivors at 6-weeks and 13.1% of injury survivors at 3-months post-injury met criteria for probable PTSD.

Regression Analyses6

The results of the regression models are presented in Table 2. In both Models 1 and 2, the NDI, ICE-R, and the ICE-IR were all statistically significantly associated with

Discussion

This study provides an easy method for incorporating a public health perspective into traumatic stress research by providing a set of R scripts and accompanying instructions that aid in the retrieval and calculation of two common community-level risk factors – the NDI and the ICE. The computation of these community-level risk factors only requires the collection of participants’ residential addresses, which are routinely collected in research protocols. Thus, the R scripts provided offer

Conclusions

Despite these limitations, this paper presents a novel and accessible method for deriving and applying community-level risk factors in traumatic stress research that can be broadly applied with the sole requirement of collecting participants’ residential addresses. This study provides detailed R scripts and accompanying instructions to allow researchers to incorporate these risk factors into their own research. Furthermore, this study provides evidence of the utility (above and beyond

References (56)

  • E.B. Blanchard et al.

    Prediction of remission of acute posttraumatic stress disorder in motor vehicle accident victims

    Journal of Traumatic Stress

    (1997)
  • Breakstone C. D., Anderson T. S. (2019). Census Data API User Guide (Version 1.6; p. 20)....
  • C. Brokamp et al.

    Decentralized and reproducible geocoding and characterization of community and environmental exposures for multisite studies

    Journal of the American Medical Informatics Association

    (2018)
  • R.A. Bryant et al.

    The psychiatric sequelae of traumatic injury

    American Journal of Psychiatry

    (2010)
  • CDC Foundation. (n.d.). What is Public Health? | CDC Foundation. What Is Public Health? Retrieved March 9, 2020, from...
  • J. Cohen

    A power primer

    Psychological Bulletin

    (1992)
  • R. Core Team. (2020). A Language and Environment for Statistical Computing. R Foundation for Statistical Computing....
  • C.J. Coulton et al.

    Measuring neighborhood context for young children in an urban area

    American Journal of Community Psychology

    (1996)
  • G. Cumming

    Inference by eye: Pictures of confidence intervals and thinking about levels of confidence

    Teaching Statistics

    (2007)
  • G. Cumming

    The new statistics: Why and how

    Psychological Science

    (2014)
  • L. DiGrande et al.

    Posttraumatic stress symptoms, PTSD, and risk factors among lower Manhattan residents 2-3 years after the September 11, 2001 terrorist attacks

    Journal of Traumatic Stress

    (2008)
  • D.N. Ditlevsen et al.

    The combined effect of gender and age on post traumatic stress disorder: Do men and women show differences in the lifespan distribution of the disorder?

    Annals of General Psychiatry

    (2010)
  • C.A. Doubeni et al.

    Health status, neighborhood socioeconomic context, and premature mortality in the United States: The national institutes of health–AARP diet and health study

    American Journal of Public Health

    (2012)
  • E.B. Foa et al.

    The validation of a self-report measure of posttraumatic stress disorder: The posttraumatic diagnostic scale

    Psychological Assessment

    (1997)
  • M. Gapen et al.

    Perceived neighborhood disorder, community cohesion, and PTSD symptoms among low-income African Americans in an urban health setting

    American Journal of Orthopsychiatry

    (2011)
  • Giannoulis C. (n.d.). Rescaling Sets of Variables to Be on the Same Scale. The Analysis Factor. Retrieved January 13,...
  • J.L. Horn

    A rationale and test for the number of factors in factor analysis

    Psychometrika

    (1965)
  • L. Irish et al.

    Trauma history characteristics and subsequent PTSD symptoms in motor vehicle accident victims

    Journal of Traumatic Stress

    (2008)
  • Cited by (0)

    1

    id https://orcid.org/0000-0002-8651-5827

    2

    id https://orcid.org/0000-0002-1431-9405

    3

    id https://orcid.org/0000-0002-5988-1645

    4

    id https://orcid.org/0000-0002-0538-8986

    5

    id https://orcid.org/0000-0002-9021-7064

    View full text