A systematic review of socioeconomic status measurement in 13 years of U.S. injury research

Abstract

Objective:

The purpose of this review was to assess the impact of socioeconomic status (SES) on injury and to evaluate how U.S. injury researchers have measured SES over the past 13 years in observational research studies.

Design & methods:

This systematic review included 119 US injury studies indexed in PubMed between January 1, 2002 and August 31, 2015 that used one or more individual and/or area-level measures of SES as independent variables. Study findings were compared to the results of a previous review published in 2002.

Results:

Findings indicate SES remains an important predictor of injury. SES was inversely related to injury in 78 (66%) of the studies; inverse relationships were more consistently found in studies of fatal injury (77.4%) than in studies of nonfatal injury (58%). Approximately two-thirds of the studies (n = 73, 61%) measured SES along a gradient and 59% used more than one measure of SES (n = 70). Studies that used a gradient measure of SES and/or more than one measure of SES identified significant relationships more often. These findings were essentially equivalent to those of a similar 2002 review (Cubbin & Smith, 2002).

Conclusions:

There remains a need to improve measurement of SES in injury research. Public health training programs should include best practices for measurement of SES, which include: measuring SES along a gradient, selecting SES indicators based on the injury mechanism, using the smallest geographic region possible for area-level measures, using multiple indicators when possible, and using both individual and area-level measures as both contribute independently to injury risk. Area-level indicators of SES are not accurate estimates of individual-level SES.

Practical applications:

Injury researchers should measure SES along a gradient and incorporate individual and area-level SES measures that are appropriate to the injury outcome under study.

1. Introduction

The association between socioeconomic status (SES) and health outcomes has been observed for decades, and continues to be a major area of investigation in many areas of public health (Glymour, Avendano, & Kawachi, 2014). One of the four overarching goals of Healthy People 2020 is to achieve health equity, eliminate disparities, and improve the health of all groups; within this goal, disparities are defined as a type of health difference that is closely linked with social, economic, and/or environmental disadvantage (U.S. Department of Health and Human Services Office of Disease Prevention and Health Promotion, 2015). Despite research and prevention efforts, injuries continue to disproportionately affect individuals and neighborhoods with lower SES (Centers for Disease Control and Prevention [CDC], 2013; Cubbin, LeClere, & Smith, 2000).

1.1. Definition of injury

The World Health Organization defines injury as “the physical damage that results when a human body is suddenly or briefly subjected to intolerable levels of energy” (Holder et al., 2001). In the United States in 2015, injuries were the leading killer of people ages 1 to 44, accounting for 214,008 deaths and 31.8 million emergency room visits (CDC-National Center for Injury Prevention and Control, 2017). Over 3 million of these individuals received further hospital or rehabilitation care; and injuries were responsible for 671 billion dollars of total lifetime medical and work loss costs (Florence, Haegerich, Simon, Zhou, & Luo, 2015).

1.2. Definition of SES

While there is no single definition of SES, and there is no one standard for measurement (Oakes & Rossi, 2003), for the purposes of this review this general definition is used: “SES is an indicator of an individual’s social and economic standing in society and often is determined by a combination of ratings on occupational status, income level, and education (Cralley, 2007).” SES is considered an important predictor of health and wellbeing across public health disciplines (Glymour et al., 2014). Individuals with high SES are likely to have more advanced education, work in prestigious positions, and earn higher salaries than individuals with low SES. These individuals have greater access to resources that can contribute to their success and to the perpetuation of similar benefits for their families (Cralley, 2007, p. 928).

SES shows a consistent inverse relationship with many types of injury (Bell, Arrington, & Adams, 2015; Denney & He, 2014). The pathways between SES and injury, as well as other health outcomes, may be causal, can indicate reverse causation (e.g., greater exposure to injury lessens one’s ability to earn money or advance one’s education), and can be confounded by third variables (Glymour et al., 2014). For example, individuals with lower educational attainment and family wealth are more likely to work in jobs with higher injury risk, such as construction, production, and mining than individuals from higher SES backgrounds (Krieger, 2010); these jobs also pay less, and during times when workers are injured, they may earn nothing at all. Children from families with lower SES are less likely to be restrained in child safety seats than children from families with higher indicators of SES (Macy & Freed, 2012). Lower safety seat utilization among families with fewer financial resources is attributable to a multitude of socioeconomic factors, including: income, which can affect the family’s ability to purchase a seat (Winston, Chen, Smith, & Elliott, 2006); education, which can impact both using the seat correctly and access to the recommendations for child passenger safety (Bilston, Finch, Hatfield, & Brown, 2008; Wegner & Girasek, 2003; Winston et al., 2006); and sociocultural norms about use of child safety seats in the community (Johnston et al., 2008; Macy & Freed, 2012), among other factors. In other words, there is not always a clear causal link between particular measures of SES and injury, rather there is a “web of causation” between SES and injury.

1.3. Methodological considerations for the measurement of SES

A number of individual- and household-level variables have been used to assess SES in injury research (Cubbin et al., 2000), including income, educational attainment, occupational or employment status, health insurance status, and wealth. Studies also employ area-level measures of SES such as poverty, concentrated poverty, income inequality, unemployment rates, and educational attainment. Each of these factors varies in terms of their ease of measurement, usefulness, meaning and relationship to health status overall and injury risk specifically.

Use of inadequate measures of SES in health and injury research is widespread (Braveman et al., 2005; Cubbin et al., 2000; Shavers, 2007). This occurs for many reasons, but commonly this is due to a lack of available data or insufficient inclusion of appropriate measures (or conceptualization of the measures) for the causal pathway under investigation (Braveman et al., 2005; Shavers, 2007). Additionally, while SES is often operationalized in a binary way that compares individuals considered to be poor to those who are not, it is important to note that SES generally has a graded association to health outcomes (Adler et al., 1994).

Studies of racial/ethnic disparities in injury require accounting for the role of SES. Severe socioeconomic disparities between racial/ethnic groups persist, causing potential confounding by socioeconomic variables in studies comparing injury and health outcomes between racial groups (Kaufman, Cooper, & McGee, 1997). As SES is often inadequately measured and/or poorly conceptualized in studies of health, racial/ethnic differences in health often persist after ‘controlling’ for socioeconomic variables (Cubbin & Smith, 2002). As stated by Oakes and Rossi (2003, p. 770), in the absence of appropriate measurement of SES, “racial/ethnic disparities may continue to be construed as signs of genetic differences or behavioral choices rather than powerful clues about how forms of racial discrimination and structural constraints, past and present, harm health.”

1.4. Purpose statement

Cubbin and Smith (2002) reviewed epidemiological, population-based studies investigating the link between SES and injury from 1960 to 2001. This review was intended to provide a critical examination of the methods and measures of SES in studies of injury; the results elucidated important methodological concerns. The purpose of the current study is to update this review to assess whether the U.S. injury research field has improved along the lines recommended by Cubbin and Smith in terms of accounting for the effects of SES. The topic remains important to the field, as there is evidence that SES disparities persist and may be widening across many health outcomes and across racial lines (Braveman, Cubbin, Egerter, Williams, & Pamuk, 2010; Laflamme, Hasselberg, & Burrows, 2010; Singh & Kogan, 2007).

2. Methods

2.1. Locating articles for review

On September 15, 2015 a search of the PubMed database (National Library of Medicine, 2012) was conducted using the following key injury Medical Subject Heading (MeSH) terms: “wounds and injuries,” “suicide,” “homicide,” “violence,” and “poisoning” and these SES terms: “socioeconomic factors,” “wealth,” “deprivation,” “crowding,” “housing,” “occupation,” and “rent.” Within PubMed the MeSH term “wounds and injuries” encompasses all types of bodily injury, as well as burns, submersions and drownings, asphyxia (suffocation) and a number of additional terms. Similarly, the MeSH term “socioeconomic factors” includes: employment and employment characteristics (e.g., career mobility, unemployment), family characteristics, education, income, poverty, and social class and conditions, among other relevant terms. The search terms “US, USA or United States” were added to eliminate many of the studies from other countries. This search returned 4172 studies. Study abstracts, and the full studies when necessary, were reviewed to determine whether they met the inclusion criteria.

2.2. Inclusion criteria

Original, peer-reviewed studies from the United States were retained for analysis if they were: indexed in PubMed between January 1, 2002 and August 31, 2015, focused on predictors/correlates of an injury outcome, and included one or more individual or area-level measures of SES as independent, mediating, moderating, or control variables. Articles were eliminated if they were intervention studies, case reports, comments, controlled trials, reviews or meta-analyses. Studies with an outcome variable such as violence, crime, abuse/maltreatment, suicide attempt, or collision (as opposed to actual injuries) were eliminated because injuries did not necessarily occur during these events. Studies of alcohol or drug overdose with no physical injury were not included. Studies where there could be no or very little variation in SES, including studies of institutionalized populations (e.g., prisoners) or occupational studies of a single occupation, and studies which only conducted descriptive analyses of the SES variables, were not included. Studies from the military were retained if rank and/or another measure of SES was utilized. Similarly, in studies of occupational injury, the studies were retained if there was a hierarchical assessment of the occupations (as opposed to a listing of occupations) or another measure of SES. Studies of poisoning were included as the prevention of poisoning often falls within the purview of injury prevention professionals, and are included in the scope of the CDC’s National Center for Injury Prevention and Control (CDC - NCIPC, 2015). As the primary focus was individual and area measures of SES, studies that only examined large macro-level SES, such as gross domestic product or state/national unemployment rates, were eliminated. Studies of injuries secondary to a medical condition, such as cancer or osteoporosis, were eliminated.

2.3. Categorization, definitions, and analysis strategy

Studies were grouped into studies of fatal injury, studies of non-fatal injury, or studies that included both fatal and non-fatal injury. The results of the review were then categorized according to level of analysis (individual or ecological). Individual-level studies examined the injury and SES variables among individuals, including those that use area-based SES measures as a proxy for individual-level SES. Ecological studies examined the injury and SES variables at the area-level (e.g., county, zip code, Census tract). There were two multilevel studies that did not employ SES variables at both levels; these were reviewed as ecological studies (Frye et al., 2008; Hemenway, Barber, & Miller, 2010), as the SES variables of interest were used at the ecological level.

Studies were then assessed according to multiple criteria, and the individual, and area-level measures of SES were coded and tallied. Individual-level SES measures included: income, poverty, education, employment status, health insurance status, housing status (e.g., own a home, rent, homeless), job class, composite or ratio measures (e.g., income to needs ratio), and net worth or wealth. Area measures included income, poverty, education, employment, insurance status, property value/rent, housing status, and composite measures or indices of deprivation or inequality, measured at the block group, tract, ZIP code, city or county level. Study samples were categorized by age group (e.g., child, adult, all ages). The data source and sample size were identified. Geographic coverage of the data set was coded (e.g., national, state, county, hospital). Injury intent was classified as intentional, unintentional or all intents, which includes studies that did not specify an intent. National Health Interview Survey data do not specify an intent, but the injuries reported are most often unintentional injuries, so these studies were classified as studies of unintentional injury.

A study was considered to use income as an SES measure if the study included any measure of income, either by self-report or area-level census measure. If a study used level of income to assess whether an individual, family, or area was in poverty by determining if the income was above or below a poverty designation (e.g., the Federal poverty level), this was recorded as a measure of poverty, not income. Studies were also considered to measure poverty if they assessed whether participants qualified for Medicaid or other government assistance programs. Job titles alone were not considered measures of SES; however, classifications of jobs that could establish a hierarchy in terms of status or earnings were counted as measures of SES. Insurance status refers only to health insurance status, which is considered a measure of SES; however, meeting criteria to qualify for a public insurance program as opposed to actually having that type of insurance (e.g., meeting qualification criteria for Medicaid) was considered a measure of poverty. Index scores refer to ratios or composite measures, such as an individual-level income to needs ratio or an area-level deprivation score incorporating several measures into one.

Each study was reviewed to determine how SES was measured, and measurement strategies were classified in the following ways: whether SES was measured on a gradient; whether multiple measures of SES were used and if so; whether they included indicators of both individual- and area-level SES, and whether area-level measures of SES were used to proxy individual-level SES. Determination of whether the study measured SES along a gradient was determined by use of three or more categories in at least one of the SES measures, or by use of a continuous variable.

Results of each study were reviewed to assess the relationship between the SES measure(s) used as independent variables and the injury dependent variable. Within individual studies, relationships were coded as positive between the SES indicator and injury if as the SES indicator increased, so did the risk of injury. The study was considered to have an inverse association when an increase in the indicator of SES was associated with a decrease in the risk of injury, or vice versa. The association was defined as non-significant if the relationship was not statistically significant. The associations were found to be “mixed” if the findings were only significant (or in the same direction) for some groups in the study and not others (e.g., for males but not females). These relationships for individual indicators within studies are described in Tables 2, 3, and 4.

Table 2.

Studies of fatal injury.

Author	Year	Data Source	Geography	Sample	Age	Design	Intent	Mechanism	Results	Gradient
Individual-level studies
Jain et al.	2001	National Center for Health Statistics, 1989–91	National	3218	Infants	Cross-sectional	All	Injury death	Parental Education (—)	Yes
Putnam-Hornstein	2011	California CPS records and vital records for children born 1999–2006	State	1917	Children	Cohort, prospective	All	Injury death	Public health insurance (+) vs. private health insurance Maternal Education (—)	No
Steege et al.	2014	Bureau ofLabor Statistics Census ofFatal Occupational Injuries, 2005–2009	National	26,996	Adolescents and adults	Cross-sectional	All	Occupational related fatalities	Education (—) Wages (—)	No
Schnitzer & Ewigman	2005	Missouri Child Fatality Review, 1992–1999	State	447	Children	Case-control	Intentional	Homicide, Inflicted injury fatality	Education (—) Poverty (—)	No
Stiffman et al.	2002	Missouri Child Fatality Review, 1992–1994	State	471	Children	Case-control	Intentional	Homicide	Education (—) Poverty (n.s.)a	No
Chang et al.	2005	Pregnancy Mortality Surveillance System, 1991–1999	National	7342	Adults	Cross-sectional	Intentional	Homicide during or within 1 year of pregnancy	Education (—)	Yes
Duberstein et al.	2004	Medical examiner’s office of 2 New York counties, 1996–2001	Multi-county	172	Older Adults	Case-control	Intentional	Suicide	Income (—) Financial trouble (+) Employment (—) Education (n.s.)	No
Harris et al.	2012	National Trauma Data Bank, 2005–2008	National	137,658	Adults	Cross-sectional	Intentional	Homicide	Health insurance coverage (—)	No
Kposowa	2001	US National Longitudinal Mortality Study 1979–89	National	471,922	Adults	Cohort, prospective	Intentional	Suicide	Employment (—) Among men: education (—) and income (—)	Yes
Phillips et al.	2010	National Center for Health Statistics, 1979–2005	National	US pop.	Adults	Cross-sectional time series	Intentional	Suicide	Education (—)	Yes
Purselle et al.	2009	Medical Examiner's Office, Fulton County, Georgia, 1988–2003	County	1377	Adults	Cross-sectional	Intentional	Suicide	Mixed by age & racial group. Incomeb (—) among whites; Incomeb (+) among adolescents: Incomeb (—) among elderly	Yes
Wexler et al.	2008	Tribal Health Organization, Suicide Reporting Forms, 1990–2001	Region	254	Adults	Cross-sectional	Intentional	Suicide, suicide attempt	Employment (—) Education (—)	No
Wiebe	2003	National Mortality Followback Survey, 1993; Controls from National Health Interview Survey, 1994	National	9804-homicide, 15,494- suicide	Adults	Case-control	Intentional	Homicide and suicide fatalities	For homicide: Income (—) Education (—) For suicide: Income (—)	Yes
Willis et al.	2003	National Mortality Followback Survey, 1993	National	1338	Adults	Cross-sectional	Intentional	Suicide (African Americans)	Education (—)	No
Lo et al.	2015	National Mortality Followback Survey, 1993	National	1409	Adolescents, adults and older adults	Cross-sectional	Intentional	Homicide (younger age at death)	Homeownership (—) Employment (+) among White victims, Employment (—) among Hispanic and Black victims	No
Carlberg et al.	2012	Linked birth and death certificates, 2000–2002	National	11.57 M	Infants	Cohort, retrospective	Unintentional	Death due to suffocation, asphyxiation or strangulation in bed (ASSB)	Education (—)	Yes
Johnson, Wilson, et al.	2009	Pennsylvania Dept. ofHealth Death Certificates, 1993	State	118	All ages	Cross-sectional	Unintentional	Heat-related death	Educationb (n.s.) Povertyb (n.s.)	No
Saluja et al.	2006	Death certificates, medical examiner reports, newspaper clippings from CPSC, 1995–1998	National	678	Children and Young Adults	Cross-sectional	Unintentional	Drowning death	Incomeb (—) when race not included in regression model; when race included incomeb (n.s.)	Yes
Ecological studies
Votruba & Kling	2009	National Death Index, fatalities among (Chicago) Gautreaux Assisted Housing Residents placed through 1994	Neighborhood	Tracts, not specified	Children, Adolescents, Young Adults	Cohort, prospective	All	Injury fatality	Tract: Education (—) Poverty (n.s.) % households receiving government assistance (n.s.)	No
Cerdá et al.	2010	Office of the ChiefMedical Examiner, NYC, 1990–1999	City	74 police precincts	Adults	Cross-sectional time series	Intentional	Homicides, gun-related	Tract: Education (n.s.)	No
Clouston, et al.	2014	National Center for Health Statistics, 1968–2009	National	3110	Adults	Cross-sectional time series	Intentional	Suicide	County: 5 indicators of SES as an aggregate measure (—)	Yes
Frye et al.	2008	Office of the Medical Examiner, NYC, 1990–1999	City	1042	Adults	Cross-sectional	Intentional	Homicide (IPV)	Community districts: Income (—) Education (n.s.) Job Status (n.s.) Unemployment (n.s.)	Yes
Gibbons et al.	2005	National Vital Statistics, CDC, 1996–1998	National	All US counties	Adults	Cross-sectional	Intentional	Suicide	County: Median income (—)	Yes
Johnson, Gruenewald, et al.	2009	California Death Public Use Files, 1995–2000	State	581 ZIP codes	Adults	Cross-sectional time series	Intentional	Suicide	ZIP Code: Income (—) in rural, more White areas, (n.s.) for other groups	No
McCall et al.	2008	FBI’s Uniform Crime Reports and Supplemental Homicide Reports (1970–2000)	National	83 cities	Adults	Cross-sectional time series	Intentional	Homicide	City: Deprivation (+), Employment (n.s.) Income (—) Poverty (+)	Yes
Miller et al.	2005	Office of the Chief Medical Examiner, NYC, 1996	City	827	Adults	Case-control	Intentional	Suicide	Community districts: Income inequality (+) for ages <35 Mean income (n.s.)	Yes
Robinson et al.	2009	Los Angeles County Dept. ofHealth Svcs. Mortality Database, 1994–2002	County	255 ZIP codes	Adults	Cross-sectional	Intentional	Homicide	ZIP code: Employment (—) High school dropout rate (+), Income (n.s.)	No
Spriggs Madkour et al.	2010	North Carolina Violent Death Reporting System 2004–2006	State	100 counties	Adults	Cross-sectional	Intentional	Homicide (IPV)	County: Disadvantage score (+)	No
Wu	2009	California Vital Statistics and Homicide Data, 1990–99	State	32,103	Adults	Cross-sectional	Intentional	Homicide (IPV)	ZIP code: Social deprivation (+) Median income (—) among non-Asian groups	Yes
Branas et al.	2004	Firearm injury reporting system in three counties, containing Allentown PA, Youngstown OH, and Cedar Rapids, IH, 1994–1998	Multi-county	468 tracts/3 counties	All ages	Cross-sectional	Intentional	Homicide and suicide due to firearms	County: Education (mixed by county), Income (n.s.) Unemployment (mixed by county)	Yes
Shai & Lupinacci	2003	Office of the Fire Marshal, Philadelphia Fire Department 1989–2000	City	324 tracts	Children	Cross-sectional	Unintentional	Fire related death	Tract: Education (n.s.) Employment (n.s.) Income (—)	Yes

^an.s. refers to a non-significant relationship.

^bIndicates an area-level SES measure used to proxy individual-level SES.

Table 3.

Studies of non-fatal injury.

Author	Year	Data source	Geography	Sample	Age	Design	Intent	Mechanism	Results	Gradient
Individual-level studies
Crandall et al.	2010	Fragile Families and Child Wellbeing Study, three year follow up of 1998–2000 cohort	National	3153	Young children	Cohort, prospective	All	Injury requiring medical attention in the first and third year oflife	Education (—) Poverty (n.s.)a Employment (n.s.)	No
Crandall et al.	2006	Fragile Families and Child Wellbeing Study, 1998–2000 cohort	National	3808	Young children	Cohort, prospective	All	Injury requiring medical attention in the 1st year of life	Employment (n.s.)	No
Grunwaldt et al.	2011	Emergency Department records, Children’s Hosp. ofPittsburgh, 2000–2005	Hospital	772	Adults	Cross-sectional	All	Facial fracture	Income (—) Poverty (+)for violence-related fractures:	Yes
Johnston & Martin-Herz	2010	Hospital trauma registry, 2003–2004	Hospital	258	Children	Cohort, prospective	All	Re-injury to index child or sibling	Education (—) Poverty (+)	Yes
Lee et al.	2008	National Survey of Children’s Health 2003–2004	National	15,990	Children	Case-control	All	Injury requiring medical attention in previous year	Poverty (n.s.)	Yes
Leininger et al.	2009	National Evaluation of Welfare-to-Work Strategies (NEWWS) Child Outcomes Study, 1992–1999	Multi-city	1280	Children	Cohort, prospective	All	Injury	Education (—) for 1of3 models; Employment (n.s.) Income: poverty ratio (n.s.) Welfare as child (n.s.) Health insurance (n.s.)	Yes
Marcin et al.	2003	University of California, Davis, Children’s Medical Center (UCDCMC) Trauma Registry 1988–1998	Hospital	3855	Children	Cross-sectional	All	Injury hospitalization	Income (—)b, Poverty (+)b	Yes
Mericli et al.	2011	Emergency Department records, Children’s Hosp. ofPittsburgh, 2000–2005	Hospital	793	Children	Case-control	All	Craniofacial fracture	Incomeb (—) Povertyb (+),	Yes
Schneiderman et al.	2012	National Survey of Child and Adolescent Well-being, 1999–2001	National	3440	Children	Cross-sectional	All	Injuries requiring medical attention in 15 months following CPS investigation	Education (n.s.) Health insurance coverage (n.s.) Income (n.s.)	Yes
Schwebel & Brezausek	2007	National Institute ofChild Health and Human Development (NICHD) Study of Early Child Care	National	7999	Young children	Cohort, prospective	All	Injury requiring medical treatment	Income: needs ratio (n.s.)	No
Simon et al.	2008	National Health Interview Survey, 1997–2003	National	2.24 M	Children	Cross-sectional	All	Injury (including poisoning) requiring medical advice or treatment in prior 3 months	Area Povertyb (n.s.) Income: Federal poverty level ratio (n.s.) Poverty (n.s.)	Yes
Sridharan and Crandall	2011	Fragile Families and Child Wellbeing study, enrollment 1998–2000, followed through 2005	National	2397	Children	Cohort, prospective	All	Injury requiring medical attention in the fifth year of life	Education (n.s.) Income (—) Type ofhealth insurance (n.s.)	Yes
Allareddy et al.	2014	Nationwide Emergency Department Sample (NEDS), 2008–2010	National	336,124	Children and adolescents	Cross-sectional	All	Facial Fractures	Incomeb (+) for falls & assaults Incomeb (—) for Firearm, motor vehicle	Yes
Houston et al.	2013	National Survey of Home Health Aides Survey, 2007	National	3377	Adults	Cross-sectional	All	Home health aides work-related injury	Hourly pay rate (+)	No
Kaufman et al.	2007	Hospital medical records and patient interviews from four burn centers, 1994–2005	Multi-hospital	2062	Adults	Case-control	All	Burn (assault vs. non assault)	Education (n.s.) Employment (—) Housing status (—)	Yes
Keough et al.	2001	Hospital medical records and patient interview, 1997–1998	Hospital	100	Adults	Cross-sectional	All	Injury recidivism	Education (—) Health insurance coverage (—)	Yes
Leff et al.	2003	Survey ofColorado residents, 1999–2000	State	2713	Adults	Cross-sectional	All	Injury requiring medical attention in the previous year	Education (n.s.) Poverty (n.s.)	Yes
Xiang et al.	2007	National Epidemiologic Survey on Alcohol and Related Conditions, 2001–2002	National	43,093	Adults	Cross-sectional	All	Injury requiring hospitalization	Among immigrants, Poverty (+)	Yes
Frencher et al.	2010	NY State Hospital Discharges, SPARCS, 2000 to 2002	City	1.53 M	All ages	Cross-sectional	All	Injury requiring hospital treatment	Housing status (—)	No
Sinclair, Smith & Xiang	2006	2000–2003 National Health Interview Survey	National	384,467	All ages	Cross-sectional	All	Injury requiring medical attention in past three months	Education (n.s.) Poverty (n.s.) Health insurance coverage (—) among foreign born persons	Yes
Xiang et al.	2012	Nationwide Emergency Department Sample (NEDS), 2007	National	57,909	All ages	Cross-sectional	All	Drug-related poisoning requiring ED admission	Incomeb (—)	Yes
Cheng et al.	2003	Injured adolescents in 2 emergency departments, 1997–2000	Multi-hospital	478	Adolescents	Case-control	Intentional	Assault injury	Education (n.s.)	No
Faulkner et al.	2001	Virginia Dept. of CriminalJustice, police records 1995–1996	Multi-city	571	Adults	Cohort, prospective	Intentional	Injury sustained during robbery	Building value (—) Rent (—)	No
Lipsky et al.	2004	Patient interview, 2001	Hospital	384	Adults	Cross-sectional	Intentional	Violence-related injury	Education (n.s.) Employment (n.s.)	No
Rich & Sullivan	2001	Survey	Clinic	177	Adults	Cross-sectional	Intentional	Assault Injury	Education (—)	Yes
Bishai et al.	2008	HealthySteps, 1996–1997	National	3449	Young children	Cohort, prospective	Unintentional	Injury requiring medical attention	Education (—) Income (n.s.)	Yes
Garzon et al.	2008	Parental self-report in Emergency Department, 2004	Hospital	124	Children	Case-control	Unintentional	Child Injury	Income (n.s.) Education (n.s.)	No
Jones et al.	2009	NHANES Phase III (Survey 1999–2004 compared to 1988–1991 and 1991–1994)	National	4817	Children	Cross-sectional	Unintentional	Elevated blood lead level	Poverty (+)	No
Kim et al.	2002	Jefferson Co. Lead Poisoning Prevention Program Registry, 1991–2000	County	34,798	Children	Cross-sectional	Unintentional	Elevated blood lead level	HousingValueb (—)	No
Malcoe et al.	2002	Prospective field data collection	Community	245	Children	Cross-sectional	Unintentional	Elevated blood lead level	Education (—) Poverty (+)	No
Ni et al.	2002	National Health Interview Survey 1997–1998	National	38,458	Children	Cross-sectional	Unintentional	Recreational injury requiring medical attention	Education (+) Poverty (—)	Yes
Nriagu et al.	2011	Detroit-area WIC clinic medical records and surveys, 2006–2007	City	429	Children	Cross-sectional	Unintentional	Elevated blood lead level	Paternal education (—) Maternal education (n.s.) Home ownership (—) Poverty (n.s.)	No
Polivka et al.	2002	National Health Interview Survey, 1997–1999	National	196	Children	Cross-sectional	Unintentional	Poisoning	Income (n.s.)	No
Schwebel et al.	2012	Healthy Passages participants, 2004–2006	Multi-city	3218	Children	Cross-sectional	Unintentional	Injuries requiring medical attention	SES composite (—) Health insurance coverage (n.s.)	No
Schwebel, Brezausek, & Belsky	2006	National Institute ofChild Health and Human Development (NICHD) Study of Early Child Care	National	1225	Children	Cohort, prospective	Unintentional	Injury requiring medical attention	Income to need ratio (n.s.)	Yes
Sinclair & Xiang	2008	National Health Interview Survey, 1997–2005	National	242,796	Children	Cross-sectional	Unintentional	Injury requiring medical attention	Education (n.s.), Poverty (n.s.)	Yes
Vaughan et al.	2004	Parent interview	Clinic	150	Children	Cross-sectional	Unintentional	Injury	Education (+)	No
Xiang et al.	2005	National Health Interview Survey, 2000–2002	National	57,909	Children	Cross-sectional	Unintentional	Injury requiring medical attention	Education (+) Health insurance coverage (+)	Yes
Rauscher and Myers	2008	Survey ofone High School	School	1430	Adolescents	Cross-sectional	Unintentional	Occupational eye injury	Education (—) Employment (+)	Yes
Schwebel & Brezausek	2009	California Health Interview Survey, 2003	State	12,536	Children and Adolescents	Cross-sectional	Unintentional	Injuries requiring medical treatment	Health insurance coverage (—)	No
Boyer et al.	2009	Workforce rosters and worker’s compensation claims for employees of one hospital in Northeastern Massachusetts.	Hospital	1483	Adults	Cross-sectional	Unintentional	Occupational injury resulting in worker’s compensation claim	5 category SES classification scheme (—)	Yes
Brown et al.	2006	Behavior Risk Factor Surveillance System, 2003	National	207,776	Adults	Cross-sectional	Unintentional	Sunburn	Education (+) Employment (+) Income (+)	Yes
Crandall et al.	2014	Study ofWomen’s Health Across the Nation	Multi-city	2167	Midlife adults	Cohort, prospective	Unintentional	Incident fracture	Difficulty paying for basics (n.s.) Income (n.s.) Poverty-to-income ratio (n.s.) Education (—) among non-Caucasian women	Yes
Dembe, Erickson & Delbos	2004	National Longitudinal Survey ofYouth 1998	National	7271	Adults	Cohort, prospective	Unintentional	Occupational injury	Education (—) Homeownership (—) HourlyPay (—) Income (—)	Yes
Forrest & Cali	2009	National Health Interview Survey, 2002	National	28,913	Adults	Cross-sectional	Unintentional	Occupational eye injury	Education (—) Income (n.s.)	Yes
Holman et al.	2014	National Health Interview Survey, 2010	National	24,970	Adults	Cross-sectional	Unintentional	Sunburn	College education (+) Private health insurance (+) (vs. other types of insurance)	Yes
Hsiao et al.	2010	Defense Medical Surveillance System, 1998–2007	Military	9299	Military	Cross-sectional	Unintentional	Patellar dislocation	Rank (—)	No
Hsiao et al.	2012	Defense Medical Surveillance System, 1999–2008	Military	15,514	Military	Cross-sectional	Unintentional	Clavicle fracture	Rank (—)	No
Luo et al.	2012	Behavioral Risk Factor Surveillance System (2005–2007)	National	43,510	Adults	Cross-sectional	Unintentional	Occupational eye injury	Education (—) Income (—); both among men only	Yes
Price et al.	2012	National Health Interview Survey, 2006–2010	National	183,676	Adults	Case-control	Unintentional	Injury requiring medical treatment	Education (—) for occupational injuries Health insurance coverage (+) for non-occupational injuries	No
Rastogi et al.	2007	Hospital medical records, 1999–2005	Hospital	6880	Adults	Cross-sectional	Unintentional	Elevated blood lead level	Education (n.s.) Poverty (n.s.)	No
Roy et al.	2012	Survey	Hospital	593	Military	Cross-sectional	Unintentional	Musculoskeletal injury	Rank (+)	Yes
Sprince et al.	2003	Agricultural Health Study, 1998	Multi-state	552	Adults	Case-control	Unintentional	Fall-related injury	Education (n.s.)	No
Sprince et al.	2008	National Health Interview Survey, 1997–2005	National	195,282	Adults	Cross-sectional	Unintentional	Occupational injury requiring medical treatment	Education (n.s.)	No
Sprince et al.	2007	Agricultural Health Study, Iowa	State	514	Adults	Case-control	Unintentional	Back injury	Education (+)	No
Strong and Zimmerman	2005	National Longitudinal Survey ofYouth (10 survey waves from 1988 to 2000, participants originally surveyed in 1979)	National	9294	Adults	Cohort, prospective	Unintentional	Occupational injury	Education (n.s.) Job status (—) Rate/income (n.s.)	Yes
Taylor et al.	2011	Center for Medicare and Medicaid Services Chronic Condition Warehouse, Claims data, 2000–2005	National	1.69 M	Older Adults	Cohort, retrospective	Unintentional	Fracture	Incomeb (—)	Yes
Wilson et al.	2006	Asset and Health Dynamics Survey, 1993 and 1995	National	5630	Older Adults	Cohort, prospective	Unintentional	Hip fracture	Education (—) Lack of supplemental insurance coverage (+) Income (n.s.)	Yes
Wirtz et al.	2012	National Health Interview Survey, 2004–2010	National	96,915	Adults	Cross-sectional	Unintentional	Occupational injury	Education (—)	Yes
Wolinsky et al.	2009	Survey on Asset and Health Dynamics	National	5511	Older	Cohort,	Unintentional	Hip fracture following	Education (—)	Yes
		Among the Oldest Old, linked to Medicare claims, 1993–2005			Adults	prospective		hospitalization	Income (—)
Zhang et al.	2009	National Health Interview Survey, 1997–2005	National	401,667	Adults	Cross-sectional	Unintentional	Occupational injury	Education (—) Health insurance coverage (n.s.) Poverty (n.s.)	Yes
Ecological studies
Freisthler et al.	2008	0–17 year olds in California Hospital Discharge Data, COSHPD, 2000	State	1646 ZIP codes	Children	Cross-sectional	All	Injury due to accident, assault, or child abuse by ZIP code	ZIP code: Education (—) for accidental injuries, (+) for child abuse Poverty (+) for assault injuries	Yes
Rewers et al.	2005	Colorado Trauma Registry, 1998–2001	State	1105 tracts	Children	Cross-sectional	All	Femur fracture	Tract: Income (—) among youngest children Single family housing (—) among 4–12 year olds	No
Zarzaur et al.	2010	Hospital Trauma Registry, Shelby County, TN, 1996–2005	Hospital	214 tracts	Adults	Cross-sectional	All	Crude injury admission rates, mechanism, severity	Tract: Poverty (+)	Yes
Edelman et al.	2010	Burn victims in five Utah state databases (EMS, ED, Hospital Discharge, Burn Database, HD) 1997–2001	State	29 counties	All Ages	Cross-sectional	All	Burn requiring treatment	County area: Education (—) Employment (—) Income (—) Poverty (+)	Yes
Krieger et al.	2003	Massachusetts Dept. ofPublic Health, 1995–1997	State	6500 block groups/1586 tracts/544 ZIP codes	All Ages	Cross-sectional	All	Non-fatal weapons related injury	Block group, tract, and ZIP code: Multiple SES indicators used, Increased risk ofinjury for those with the lowest resources compared to those with the most.	Yes
Reed et al.	2003	Hospital Trauma Registries, two Sedgewick County, Kansas Trauma Centers, 1990/1–1999	Multi-hospital	1 county	All Ages	Cross-sectional time series	All	Penetrating trauma requiring treatment	County: Employment (—)	Yes
Pabayo et al.	2014	Boston Youth Survey (BYS), 2008; Boston Neighborhood Study (BNS)	City	1878	Adolescents	Cross-sectional	Intentional	Assault related injury	Tract: Income inequality (+)	No
Boyle & Hassett-Walker	2008	Six hopsitals in Newark, New Jersey	Multi-hospital	262 block groups	All Ages	Cross-sectional	Intentional	Assault related injury	Block group: Poverty +	Yes
Chakravarthy et al.	2012	California Statewide Integrated Traffic Records System, Orange County, 2000–2004	County	577 tracts	All Ages	Cross-sectional	Unintentional	Pedestrian injury	Tract: Income (—) Education (—)	Yes
Haley & Talbot	2004	New York State Health department Laboratory Reporting System, 1994–1997	State	952 ZIP codes	Young children	Cross-sectional	Unintentional	Elevated blood lead level	ZIP code: Education (—)	Yes
Kaplowitz et al.	2010	Michigan Dept. ofCommunity Health database ofblood lead level tests, 1998–2005	State	533,647	Children	Cross-sectional time series	Unintentional	Elevated blood lead level	ZIP code vs. block group: Education (—) Medicaid insurance status (+) Poverty (+)	Yes
Shenassa et al.	2004	Hospital discharge data, Illinois, 1990–2000	State	11,735	Children	Cross-sectional	Unintentional	Injury occurring at home	ZIP code: Concentrated poverty (+)	Yes
Vivier et al.	2011	Rhode Island Department ofHealth, Lead Elimination Surveillance System, 1993–2005	State	808 block groups	Children	Cross-sectional	Unintentional	Elevated blood lead level	Block group: Poverty (+)	Yes
Chakravarthy et al.	2010	California Statewide Integrated Traffic Records System, Orange County, 2000–2004	County	577 tracts	All Ages	Cross-sectional	Unintentional	Pedestrian injury	Tract: Education (—) Income (—)	No
Istre et al.	2001	EMS, Fire Department, ME and Hospital Data, Dallas, 1990–1997	City	255 tracts	All Ages	Cross-sectional	Unintentional	Injury related to house fires	Tract: Income (—)	Yes
Shai	2006	Office of the Fire Marshal, Philadelphia Fire Department (1993–2001)	City	324 tracts	All Ages	Cross-sectional	Unintentional	Fire injury rates by census tract	Tract: Education (n.s.) Employment (n.s.) Income (—)	Yes

^an.s. refers to a non-significant relationship.

^bIndicates an area-level SES measure used to proxy individual-level SES.

Table 4.

Studies of fatal and non-fatal injury.

Author	Year	Data source	Geography	Sample	Age	Design	Intent	Mechanism	Results	Gradient
Individual-level studies
Martin et al.	2012	Vanderbilt pediatric trauma registry and Davidson Co. Medical Examiner records, 1998–2010	Hospital/County	201	Children	Cross-sectional	All	Firearm injury	Incomeb (n.s.)a	Yes
Schecter et al.	2012	Trauma center medical records, 2000–2009	Multi-hospital	598	Children	Cross-sectional time series	All	Penetrating injury	Incomeb (—)	No
Lee et al.	2013	Kid’s Inpatient database, 2006–2009	National	19,233	Children and adolescents	Cross-sectional	All	Firearm injury	Income (—)	Yes
Kesler et al.	2008	Pennsylvania Department of Public Welfare database, 1996–2002	State	327	Young children	Cross-sectional	Intentional	Abusive head trauma (AHT)	Education (—)	Yes
Istre et al.	2002	EMS, Fire Department, ME and Hospital Data, Dallas, 1991–1998	City	76	Children	Cross-sectional	Unintentional	Injury related to house fires	Incomeb (—)	Yes
Senier et al.	2002	Total Army Injury and Health Outcomes Database, 1980–1998	Military	28,352	Adults	Cross-sectional	Unintentional	Fall-related injury	Education (—) Rank (—)	No
Ecological studies
Pomerantz et al.	2001	Hospital Trauma Registry, Hamilton County, Ohio, and County Coroner death records	Hospital/County	209 tracts	Children	Cross-sectional	All	Injury requiring hospitalization or resulting in death	Tract: Education (—) Employment (—) Median household income (—) Poverty (+)	Yes
Williams et al.	2003	Medical records at two children’s hospitals, 1995	City	988 block groups	Children	Cross-sectional	All	Burn	Blockgroup: Deprivation (+) (spatial analysis)	No
Cinat et al.	2004	Trauma registries of Los Angeles County and Orange County, 1993–2002	Multi-county	2 counties	All Ages	Cohort, retrospective	All	County trauma rates	County: Number unemployed (+) Unemployment rate (+)	No
Anderson et al.	2014	Ohio Trauma Acute Care Registry, 2003–2011	State	984	Young children	Cross-sectional	Intentional	Contusion, intracranial hemorrhage, skull fracture, femur fracture, burn	County: Income (—) Poverty (+)	Yes
Donner	2007	John Hart’s Severe Plot v. 2.0, 1950–2000	National	Tracts, not specified	All ages	Cross-sectional	Unintentional	Tornado - related injury and death	Tract: Deprivation index (n.s.)	Yes

^an.s. refers to a non-significant relationship.

^bIndicates an area-level SES measure used to proxy individual-level SES.

For analysis of the studies as a group, the study results were coded as demonstrating an inverse, positive, or mixed relationship between SES and injury (for instance, when a study found injury rates were elevated in areas in poverty, the study was coded as demonstrating an inverse relationship between SES and injury). A chi-square test was used to determine whether studies that used more than one measure of SES were significantly more likely to find a measureable difference than those that used only one measure.

3. Results: SES measurement in studies reviewed

3.1. Overview of the studies reviewed

A total of 119 studies met the inclusion criteria (Fig. 1). Study characteristics are presented in Table 1. Thirty-one studies (26.1%) focused on fatal injury outcomes, 77 on non-fatal injury (64.7%), and 11 on both fatal and non-fatal injuries. There were 85 individual-level studies (71.4%) that measured SES and injury at the individual level (or using an area-based SES measure to proxy individual SES), and 34 ecological studies (28.6%) that measured both injury and SES at the area-level. The majority of the fatal injury studies were focused on intentional injuries (n = 23, 74.2%). The non-fatal injury studies, and the studies of both fatal and non-fatal injury, were more focused on unintentional injuries (55.3%) and injuries from all intents (35.2%). In terms of life stage under study, 43.7% of studies focused on children and/or adolescents, 42.0% on adults, and 14.3% on people of all ages.

Fig. 1.

Inclusion of articles in review.

Table 1.

Characteristics of reviewed studies by injury outcome.

		All studies (n = 119)		Non-fatal (n = 77)		Fatal (n = 31)		Both fatal and non-fatal (n = 11)
		n	%	n	%	n	%	n	%
Study type	Ecological	34	28.6%	16	20.8%	13	41.9%	5	14.7%
	Individual-Level	85	71.4%	61	79.2%	18	58.1%	6	7.1%
Intent	All Intents	37	31.1%	27	35.1%	4	12.9%	6	16.2%
	Intentional	31	26.1%	6	7.8%	23	74.2%	2	6.5%
	Unintentional	51	42.9%	44	57.1%	4	12.9%	3	5.9%
Age group	Infants	2	1.7%	0	0.0%	2	6.5%	0	0.0%
	Children	44	37.0%	32	41.6%	5	16.1%	7	15.9%
	Adolescents	6	5.0%	5	6.5%	0	0.0%	1	16.7%
	Adults	50	42.0%	30	39.0%	19	61.3%	1	2.0%
	Multiple Groups/All Ages	17	14.3%	10	13.0%	5	16.1%	2	11.8%
Study design	Cross-sectional	86	72.3%	54	70.1%	22	71.0%	10	11.6%
	Cohort	20	16.8%	15	19.5%	4	12.9%	1	5.0%
	Case-control	13	10.9%	8	10.4%	5	16.1%	0	0.0%
Measures of SES	Single	49	41.2%	32	41.6%	11	35.5%	6	12.2%
	Multiple	70	58.8%	45	58.4%	20	64.5%	5	7.1%
SES measure	Dichotomous	46	38.7%	27	35.1%	15	48.4%	4	8.7%
	Gradient (3 + categories)	73	61.3%	50	64.9%	16	51.6%	7	9.6%
Relationship between SES variable(s) and injury outcome	Positive (higher SES = increased injury)	8	6.7%	8	10.4%	0	0.0%	0	0.0%
	Inverse (higher SES = lower injury)	78	65.5%	45	58.4%	24	77.4%	9	11.5%
	Mixed (inverse relationship for some groups)	14	11.8%	9	11.7%	5	16.1%	0	0.0%
	Non-significant	19	16.0%	15	19.5%	2	6.5%	2	10.5%

The majority of the studies in the current review were cross-sectional in design (n = 86, 72.3%). There were 20 cohort studies (16.8%) and 13 case–control studies (10.9%).

3.2. Sample selection and generalizability

Forty-eight of the studies (40.3%) in the review utilized national samples. Thirty-three (27.7%) drew from single community sampling frames, representing a population of only a single hospital/clinic, community or city. Many of these studies used non-random sampling strategies and had smaller sample sizes, limiting generalizability. Variations in the demographics and cultures of single communities may mean that the effects of SES variables affect injury risk differently. In addition, smaller sample sizes contribute to greater sampling error and should be interpreted with care.

3.3. The measurement of SES

Cubbin and Smith (2002) made four primary recommendations for future injury research, which are used as a framework to analyze studies in the current review. First, studies should measure SES along a gradient. Second, injury research should use multiple measures of SES when possible, and account for the multidimensional nature of SES in both choice of variables and interpretation of the results. Third, Cubbin and Smith argue for the inclusion of area-level measures of SES, as they have been shown to have an independent contribution above and beyond those individual-level variables. Lastly, researchers should define and justify the variables they have chosen, and acknowledge their limitations, so that readers can more easily interpret their findings.

3.3.1. SES gradients

In the current review and in Cubbin and Smith (2002), studies were assessed to determine if the analysis of SES took into account variation along a gradient; if the SES variable was operationalized to include three or more categories or as a continuous measure, the study was coded as accounting for a gradient approach to measurement. Approximately 61% (n = 73) of the studies in our review used gradient measures of SES — nearly the exact same as the 60% found in the Cubbin and Smith review. This finding would suggest that U.S. studies of injury have not improved in the area of accounting for the effects of SES along a gradient since 2001. Studies that measured SES on a gradient found significant relationships between SES indicators and injury outcomes more often than studies that used dichotomous SES variables, χ²(1, n = 119) = 3.5, p = 0.06, 65.0% vs 35.0% respectively.

3.3.2. Multiple SES measures/level of measurement

One of the primary recommendations from the Cubbin and Smith (2002) review included accounting for SES through multiple measures when feasible, as SES is multidimensional in nature. Seventy of the 119 studies did include multiple measures of SES. Measurement at individual and area-levels contributes important information about both people and the places they exist within. However, in the current review, 49 of the 119 studies (41%) reviewed used only one measure of SES (this count does not include composite or ratio measures). Studies that used multiple measures of SES found significant relationships to injury 62.0% of the time, compared to 38.0% of the time for single SES-indicator studies; this difference was not statistically significant in a chi-square test of independence (p = 0.106).

The majority of the 49 studies using only one indicator of SES were individual-level studies (n = 38), 10 of which used area-level measures of SES to proxy individual SES. As Cubbin and Smith (2002) discussed, using area-level assessments in this way can create misclassification, as individual differences cannot be measured accurately with area-based proxies. Two studies compared various area-levels of measurement (Kaplowitz, Perlstadt, & Post, 2010; Krieger, Waterman, Chen, Soobader, & Subramanian, 2003). Both found that substantial measurement error may exist when larger areas (e.g., ZIP codes) are employed to approximate neighborhoods as they may consist of more heterogeneous groups of individuals with greater variety in SES; smaller areas (e.g., census block groups, census tracts) are more sensitive and account for more variance in the injury outcome as the individuals within tend to be more homogenous in terms of SES.

3.3.3. Selection of SES measures

Studies in this review continue to use problematic measures of SES, despite ongoing discussions of their inadequacy (Braveman et al., 2005; Cubbin et al., 2011) or their ill fit to the specific injury issue under study. Substantial misclassification exists with some variables used as measures of SES in our review (e.g., insurance status, enrollment in public assistance), as not all people who are experiencing low SES can be identified with such designations. As in Cubbin and Smith (2002), many of the studies in our review lacked justifications for their selection of SES measure; often, the measure appears to have been selected simply because it was readily available in an existing data set. In individual-level studies, for instance, two studies used only insurance status to approximate SES (Harris, Fisher, & Thomas, 2012; Schwebel & Brezausek, 2009). A total of 15 studies in the review employed insurance status as an independent variable. In another example, in 18 studies of inflicted intentional injuries (homicide, assault, and child maltreatment, not including suicide), five used only individual-level measures of SES, not accounting for the influence of neighborhood conditions.

4. Results: SES–injury relationships

The specific findings of each study are shown in Tables 2, 3, and 4. Overall, 84% of studies (n = 100) demonstrated a significant relationship between SES and the injury outcome variable. Of these, 78 studies found the relationship between at least one indicator of SES and injury to be inverse in nature, meaning as SES increased, injury decreased and vice-versa. An additional 14 of those studies demonstrated mixed results, finding a significant inverse SES–injury relationship for at least some portion of the sample. Eight studies found a positive relationship between SES and injury; of these, three of the injury outcomes were recreational in nature (Brown, Quain, Troxel, & Gelfand, 2006; Holman, Berkowitz, Guy, Hartman, & Perna, 2014; Ni, Barnes, & Hardy, 2002).

4.1. Studies of fatal injury

Over two-thirds (77%, n = 24) of the studies of fatal injury demonstrated a significant, inverse relationship between at least one indicator of SES and the fatal injury outcome (Table 2). An additional five studies (16%) found the relationship was significant for some sub-groups within the sample.

4.1.1. Intentional fatal injury studies

Most of the studies of fatal injury were of intentional injuries (n = 23); 7 of 11 ecological studies found significant inverse relationships between SES and fatal intentional injury, as did 11 of 12 individual-level studies.

4.1.1.1. Children & Adolescents.

Three statewide studies explored intentional injury deaths in children at the individual-level. Two studies examined data from the Missouri Child Fatality Review. One found negative relationships between both parental education and poverty and intentional injury death (Schnitzer & Ewigman, 2005), while one found a significant negative relationship only to parental education (Stiffman, Schnitzer, Adam, Kruse, & Ewigman, 2002).

4.1.1.2. Adults/all ages.

All individual-level studies of homicide found increased risk of homicide for individuals with lower SES indicators using national datasets. Lo, Howell, and Cheng (2015) found homeownership to be negatively associated with early death by homicide; among Hispanic and Black individuals, employment and education were also negatively associated with younger age at homicide. Among Whites in the dataset, employment was positively associated with younger age at homicide. In a study using Pregnancy Mortality Surveillance System data, educational attainment was inversely associated with femicide during or within one year of pregnancy (Chang, Berg, Saltzman, & Herndon, 2005). Homicides in the National Trauma Data Bank were negatively associated with having health insurance coverage (Harris et al., 2012). Homicide victims in the National Mortality Followback Survey had lower income than controls from the National Health Interview Survey (Wiebe, 2003).

Seven ecological studies focused on homicide. A 30-year study of FBI data from 83 cities demonstrated associations between homicide and city-level deprivation, poverty, and lower incomes; there was no significant relationship to employment rates (McCall, Parker, & MacDonald, 2008). In a Los Angeles study, rates of homicide were associated with unemployment and high school dropout rates, but not with median household income (Robinson et al., 2009). A New York City study found no association between gun-related homicides and education by Census tract (Cerdá et al., 2010).

Three of the 7 ecological homicide studies focused on fatalities from intimate partner violence. A study of 100 counties in North Carolina found intimate-partner homicide rates were significantly associated with county-level deprivation (Spriggs Madkour, Martin, Halpern, & Schoenbach, 2010). Wu (2009) found this as well for zip code-level deprivation in California; this study also found an inverse association with income among non-Asian women. Frye et al. (2008) found intimate partner homicides in New York City to be associated with lower neighborhood-level income, but not neighborhood-level education, job status, or unemployment.

Six individual-level studies examined relationships between suicide and SES, four using national datasets. All four national studies found at least one negative relationship to SES indicators. Kposowa (2001) found employment to be negatively related to suicide in the first three years of a longitudinal data from the U.S. National Longitudinal Mortality Study; among men in the study education and income were also negatively associated with suicide. In a longitudinal study of data from the National Center for Health Statistics, education was negatively associated with suicide (Phillips, Robin, Nugent, & Idler, 2010). Wiebe (2003) found lower household income was associated with increased risk of suicide in a national case–control study. Among African Americans, suicide was associated with lower educational attainment in the Willis, Coombs, Drentea, and Cockerham (2003) study of the National Mortality Followback Survey. In the three studies of data sets from smaller geographic regions, suicide was associated with lower income, financial trouble, and unemployment (Duberstein, Conwell, Conner, Eberly, & Caine, 2004; Wexler, Hill, Bertone-Johnson, & Fenaughty, 2008) in two studies, but in the third, income demonstrated mixed results among age and racial groups (Purselle, Heninger, Hanzlick, & Garlow, 2009).

Ecological studies of fatal suicide also found significant relationships to SES, but in some studies results were mixed. Using national, county-level data, Clouston, Rubin, Colen, and Link (2014) found an aggregate measure of SES (combining education, white collar employment, proportion above federal poverty level and access to a telephone) was negatively related to suicide rates after the introduction of selective serotonin reuptake inhibitors, which are protective against suicide but are not equally accessible across socioeconomic classes. In another national county-level study, suicide was negatively associated with county median income (Gibbons, Hur, Bhaumik, & Mann, 2005). In a California based study, Johnson, Gruenewald, and Remer (2009), found income at the zip-code level to be negatively associated with suicide in rural, more White areas but not in other communities. One study examined both suicide and homicide by firearm in three counties; unemployment rates demonstrated mixed associations by county, while income was not significantly related to homicide or suicide (Branas, Richmond, & Schwab, 2004). In New York City, income inequality in community districts was related to increased suicide rates (Miller et al., 2005).

4.1.2. Studies of unintentional fatal injury and fatal injury from all intents

4.1.2.1. Children & adolescents.

There were eight studies of fatal unintentional injuries and injuries from all intents among children. Using linked birth and death records, Carlberg, Shapiro-Mendoza, and Goodman (2012) found parental education was negatively associated with infant death due to suffocation, asphyxiation or strangulation in bed (ASSB). Using data from the National Center for Health Statistics, Jain, Khoshnood, Lee, and Concato (2001) found lower levels of parent’s education to be significantly related to higher risk of injury death from all intents in infants. Using national data, Saluja et al. (2006) found income to be negatively associated with child drowning deaths, but only when controlling for race. A study of whether children with a report of child maltreatment faced heightened risk for injury death from all intents controlled for both insurance status (private vs. public) and maternal education, as both were significant predictors of child injury fatality in the overall sample (Putnam-Hornstein, 2011). Istre, McCoy, Carlin, and McClain (2002) found child deaths in house fires in Dallas to be associated with median census tract income as a proxy for individual-level SES. Similarly, in an ecological study, Shai and Lupinacci (2003) found fire-related deaths in Philadelphia to be related to census tract-level income but not education or employment.

4.1.2.2. Adults/all ages.

Among adults, one national study found increased risk of injury for workers with lower educational attainment and those with lower wages (Steege, Baron, Marsh, Menéndez, & Myers, 2014). A state-level study in Pennsylvania showed SES variables such as poverty and education were not better predictors of heat-related injury death than measures of land surface temperature, but that indicators of poverty were associated with increased risk of death as temperatures increased (Johnson, Wilson, et al., 2009). One ecological study found injury fatalities from all intents among young black males in Chicago, who had been relocated to higher income neighborhoods as children and followed prospectively, were associated with lower census tract-level educational attainment; in this study, rates of poverty and percentage of houses receiving government assistance were not significant (Votruba & Kling, 2009).

4.1.3. Studies of non-fatal injury

Over twice as many studies of non-fatal injury met the inclusion criteria for this review compared to fatal injury (77 vs. 31; Table 3). Within the non-fatal studies, over half (n = 44, 57%) are studies of unintentional injury. The non-fatal studies reported a significant, inverse relationship with SES 58% of the time (n = 45), with an additional 12% finding significant relationships for some of the sample (n = 9). All eight of the studies in the review that showed positive relationships to SES are non-fatal studies, primarily of unintentional injury (7 studies of unintentional injury, 1 study of all intents).

4.1.4. Intentional non-fatal injury studies

4.1.4.1. Children & adolescents.

Only one study examined non-fatal intentional injuries to children at the individual level. In this study, maternal education was not a significant predictor of adolescent assault injuries in data from two hospitals (Cheng et al., 2003). An ecological study demonstrated income inequality at the census tract level was related to higher risk of assault-related injuries to adolescents in Boston (Pabayo, Molnar, & Kawachi, 2014).

4.1.4.2. Adults/all ages.

The three individual level studies of intentional injuries among adults, all conducted with data from a hospital or clinic, found mixed results. Faulkner, Landsittel, and Hendricks (2001) found injuries sustained during robberies were related to lower neighborhood building values and rents. Lipsky, Caetano, Field, and Bazargan (2004) found no significant associations between violence-related injury and education or employment status; however, this study’s sample had overall low educational attainment and high unemployment. Another study of assault injuries in a clinic demonstrated a negative relationship to educational attainment in a sample of young adult males (Rich & Sullivan, 2001). In an ecological study combining data from multiple hospitals, census block-group level poverty was associated with assault-related injury among adolescents (Boyle & Hassett-Walker, 2008).

4.1.5. Unintentional and all intent non-fatal injury studies

4.1.5.1. Children & adolescents.

Twenty-eight of the studies reviewed analyzed individual-level relationships between non-fatal child injury and SES due to unintentional injury or injuries from all intents. Nine of the 28 studies, eight of them with national sampling frames, found relationships between child injury and SES to be non-significant. All but one (Polivka, Elliott, & Wolowich, 2002) of the national studies finding non-significant relationships examined SES predictors of injury requiring medical attention in the past year, including: education and poverty among children with disabilities (Lee, Harrington, Chang, & Connors, 2007; Sinclair & Xiang, 2008), two studies on the influence of child care that also examined income-to-needs ratio to injury (Schwebel & Brezausek, 2007; Schwebel, Brezausek, & Belsky, 2006), parental employment (Crandall, Chiu, & Sheehan, 2006) and education, income and health insurance among high-risk families (Schneiderman, Leslie, Hurlburt, Zhang, & Horwitz, 2012), and measures of poverty (Simon, Emsermann, DiGuiseppi, Davidson, & Hambidge, 2008). Five of the remaining six national studies found at least one significant inverse relationship between the SES independent variable and the injury outcome (Allareddy et al., 2014; Bishai et al., 2008; Crandall, Sridharan, & Schermer, 2010; Jones et al., 2009; Sridharan & Crandall, 2011; Xiang, Zhao, Xiang, & Smith, 2012), the exception being a study of recreational injuries that had a positive relationship to SES (Ni et al., 2002).

The remaining 19 individual-level studies using non-national data varied widely in terms of sample size, with samples ranging from 124 to 34,798. Studies found mixed results in terms of associations between general unintentional injury with SES indicators: one study found no relationship to income or parental education (Garzon, Huang, & Todd, 2008); one study found an inverse relationship to a composite SES score but not to health insurance status among a sample of 5th grade children (Schwebel et al., 2012); a third study found education to be positively related to child injury among a small sample of only lower-income families (Vaughan, Anderson, Agran, & Winn, 2004). A study of eye injuries among adolescents found employment as an adolescent was positively related to injury, and parental education was negatively related (Rauscher & Myers, 2008). Three studies of elevated lead levels all found some association between at least one indicator of lower SES and increased risk of lead exposure (Kim, Staley, Curtis, & Buchanan, 2002; Malcoe, Lynch, Keger, & Skaggs, 2002; Nriagu, Senthamarai-Kannan, Jamil, Fakhori, & Korponic, 2011). Istre et al. (2002) found children in census tracts with lower median incomes were at increased risk of fire-related injury and death. Marcin, Schembri, He, and Romano (2003) found the median income and proportion of households in poverty in patient’s counties of residence indicated increased risk of pediatric injury hospitalization for injuries of all intents.

There were seven ecological studies investigating relationships between area-SES and pediatric injuries, all from individual states. Two looked at injuries from all intents and all found inverse associations between most area measures of SES and rates of injury (Freisthler, Gruenewald, Ring, & LaScala, 2008; Rewers et al., 2005). In the Freisthler et al. (2008) study of California Hospital Discharge Data, there was an inverse relationship between percentage of high school graduates by zip code and both abuse and unintentional injuries. Higher poverty was also related to assault and unintentional injuries in this study. Among the three ecological studies of elevated blood lead levels, one found associations with census block group poverty in Rhode Island (Vivier et al., 2011), one with education, poverty and Medicaid insurance in Michigan (Kaplowitz et al., 2010), and the last with zip code level education in New York state (Haley & Talbot, 2004). Chakravarthy, Anderson, Ludlow, Lotfipour, and Vaca (2012) found census tract income and education to be inversely related to child pedestrian injury rates in Southern California. Rates of unintentional injuries occurring at home were related to zip code concentrated poverty in a study of Illinois hospital discharge data (Shenassa, Stubbendick, & Brown, 2004). Femur fractures from all intents were related to census tract income and single family housing in families with young children in a Colorado study (Rewers et al., 2005).

4.1.5.2. Adults/all ages.

There were 27 individual-level studies of unintentional injuries to adults or persons of all ages. Nine of these studies examined non-fatal unintentional injuries to adults focused on occupational injuries; 8 of them used national data sets. Seven studies found a significant, inverse relationship between risk of occupational injury and at least one indicator of SES (Dembe, Erickson, & Delbos, 2004; Forrest & Cali, 2009; Luo et al., 2012; Price et al., 2012; Strong & Zimmerman, 2005; Wirtz, Lombardi, Willetts, Folkard, & Christiani, 2012; Zhang et al., 2009). A national study of home health aides found hourly pay rate was associated with higher risk of occupational injury (Houston, Young, & Fitzgerald, 2013). Sprince, Pospisil, Peek-Asa, Whitten, and Zwerling (2008) found no association between occupational injury and educational attainment in their multivariate model, although education was independently associated with occupational injury before adjusting for race and other factors. In a study comparing occupational to non-occupational injuries, Price et al. (2012) found education was negatively related to occupational injury, but not to non-occupational injuries. In this study, non-occupational injuries were significantly positively related to heaving health insurance coverage, perhaps due to ability to access care when not under workman’s comp coverage.

Five studies used national datasets to investigate varied injury outcomes; two of these found positive relationships between sunburn and SES indicators (Brown et al., 2006; Holman et al., 2014). Three studies investigated fractures in older adults. One found increased risk of hip fracture in older adults with lower education and lower income (Wolinsky et al., 2009), one found an inverse association between fractures and income (Taylor et al., 2011), and one found lower education and no supplemental health insurance were associated with increased risk of fracture, but income and net worth did not demonstrate significant relationships (Wilson, Chase, Chrischilles, & Wallace, 2006).

There were three studies of non-fatal, unintentional injuries in the military included in the review. In these studies, rank was inversely related to patellar (kneecap) dislocation (Hsiao, Owens, Burks, Sturdivant, & Cameron, 2010) and clavicle fracture (Hsiao et al., 2012). Rank was positively related to musculoskeletal injuries among soldiers deployed to Afghanistan (Roy, Knapik, Ritland, Murphy, & Sharp, 2012).

Individual level studies of adults with unintentional injuries from datasets with smaller geographic coverage found similar results. Boyer et al. (2009) found occupational injury to be negatively related to a measure of SES based on job category in a study of employees at one Massachusetts hospital. Studies of injuries to Iowa farmers found fall-related injuries were not significantly related to education (Sprince et al., 2003), but low back injury was positively related to educational attainment beyond high school (Sprince et al., 2007). Similar to findings in studies of children, Rastogi, Nandlike, and Fenster (2007) found elevated blood lead levels to be inversely related to income among pregnant women.

The seven ecological studies of injuries among adults and persons of all ages demonstrated consistent inverse relationships to SES. In a study of adults, census tract poverty was positively related to injury admission rates and severity in Tennessee (Zarzaur, Croce, Fabian, Fischer, & Magnotti, 2010). In studies of persons of all ages, area level indicators of SES were associated with non-fatal weapons-related injuries (Krieger et al., 2003), penetrating trauma (Reed, Smith, Helmer, Lancaster, & Carman, 2003), pedestrian injury (Chakravarthy, Anderson, Ludlow, Lotfipour, & Vaca, 2010) and fire- & burn-related injuries (Edelman, Cook, & Saffle, 2010; Istre, McCoy, Osborn, Barnard, & Bolton, 2001; Shai, 2006).

4.2. Studies of both fatal and non-fatal injury

Eleven studies examined SES and injury relationships in studies of injuries that were both fatal and non-fatal (Table 4). Seven of these involved pediatric samples. About half of these studies (n = 5) were ecological.

4.2.1. Studies of fatal & non-fatal intentional injury

Only two studies in this review addressed intentional injury with both fatal and non-fatal injury outcomes; both were focused on children. One found parental education to be significantly inversely related to abusive head trauma in young children in Pennsylvania (Kesler et al., 2008). The other, an ecological study, found intentional injuries to children were found to be associated with lower county mean income and higher percentages of children living in poverty in Ohio (Anderson, Pomerantz, & Gittelman, 2014).

4.2.2. Studies of fatal and non-fatal unintentional injuries and injuries from all intents

4.2.2.1. Children & adolescents.

Three of these studies focused on children in individual-level analyses. In two smaller studies using hospital data and zip code median income to estimate individual income, one found income to be inversely related to penetrating injury in children (Schecter, Betts, Schecter, & Victorino, 2012), while a study of firearm injury found no significant association to income (Martin et al., 2012). Lee, Moriarty, Tashjian, and Patterson (2013) found lower income to be related to higher risk of firearm injury death in children using data from the Kids’s Inpatient database.

There were two ecological studies of injuries to children, both demonstrating inverse relationships between SES and injury. A study of 209 census tracts in Ohio showed pediatric injury resulting in hospitalization or death was associated with lower tract-level education, income, employment, and with higher poverty (Pomerantz, Dowd, & Buncher, 2001).

Burn injuries and fatalities were associated with census block group-level deprivation in a spatial analysis of 988 block groups in St. Louis (Williams, Schootman, Quayle, Struthers, & Jaffe, 2003).

4.2.2.2. Adults/all ages.

An individual-level study of the military population demonstrated higher numbers of fall-related injuries and among individuals with lower educational attainment and lower military ranks in descriptive analyses (Senier, Bell, Yore, & Amoroso, 2002). An ecological study of injuries from all intents to people of all ages in two Southern California counties (Cinat, Wilson, Lush, & Atkins, 2004) demonstrated positive relationships between county traumatic injury rates and the county unemployment rate, as well as the number of unemployed individuals in the county. One study investigated injury deaths due to tornado events in a national dataset and found no association to census tract-level deprivation (Donner, 2007).

5. Discussion

The results of this systematic review of injury studies provides an opportunity to compare the methods and results of studies published from 2002 to 2015 to those reviewed by Cubbin and Smith (2002), which were published between 1960 and 2001. While the reviews have some differences (e.g., Cubbin & Smith included all English-language studies; the current review contains only studies from the United States), comparing them allows for consideration of how patterns and methods have changed or remained the same.

5.1. Socioeconomic disparities in injury

The results of this review demonstrate that disparities in injury by socioeconomic status persist, with 78 of 119 studies (66%) finding inverse relationships between at least one SES indicator and the injury outcome under study. Like Cubbin and Smith (2002), this review found that the strength and direction of the SES–injury relationship varies by fatal vs. non-fatal injury, by age of the sample, intentional versus unintentional injury, and by SES indicator.

5.1.1. Fatal injury

In fatal injury research, just over two-thirds of the 31 studies reviewed (77%) demonstrated that lower SES was a significant predictor of the dependent variable, with an additional 16% demonstrating significance for some, but not all, of the sample. Only two studies did not find a significant relationship; one used an area-level measure of SES to proxy individual-level SES in a study of heat-related deaths using a state-level sampling frame (Johnson, Gruenewald, et al., 2009; Johnson, Wilson, et al., 2009); the other found no relationship between tract-level education rates and gun-related deaths by police precinct in New York City (Cerdá et al., 2010). Cubbin and Smith (2002) also found consistent inverse relationships between fatal injury and SES. In their review, associations to suicide demonstrated more mixed results than our review, which found fairly consistent inverse associations for education (Kposowa, 2001; Phillips et al., 2010; Willis et al., 2003) and income (Duberstein et al., 2004; Kposowa, 2001; Wexler et al., 2008; Wiebe, 2003). Note that many of these studies addressed specific populations (e.g., African Americans, males, Native Americans) and that these studies only addressed completed suicides; there may be other important predictors of suicide attempts and suicidality that would not be reflected in either review.

5.1.2. Non-fatal injury

As in the Cubbin and Smith (2002) review, the findings of the current review indicated that the SES injury-relationship is less consistent across the 77 non-fatal injury studies than in fatal injury studies, demonstrating a significant, inverse relationship 58% of the time (n = 45) and mixed results for some sample sub-groups 12% of the time (n = 9).

Eight studies found a significant, positive relationship to SES, indicating that those with a higher status were at increased risk of injury. Three of these studies examined an injury associated with recreation, which would be expected to create higher exposure for individuals with more socioeconomic resources (Brown et al., 2006; Holman et al., 2014; Ni et al., 2002); the other five explored more minor injuries or work-related injuries. In one study, higher pay among home health workers, the only SES measure in the study, was related to higher injury (Houston et al., 2013); this could be problematic measure as it may reflect overtime pay, night-time pay or other conditions where the risk of injury could be higher. The studies all involved non-life threatening injuries (e.g., back injury, musculoskeletal injury, injury in the past year or requiring a medical visit).

Cubbin and Smith (2002) hypothesized that less consistent relationships among studies of non-life threating, non-fatal injuries and SES may be due to the fact that these injuries are often assessed by utilization of medical care. Many studies used the methodology from the National Health Interview Survey, which records self-report of injury requiring medical attention in a preceding period of time (which varied by study). In addition to concerns that accompany any self-report data, chiefly recall bias in this case is the medical care part of this definition. People of lower SES are less likely to have economic resources that allow for acute care. Therefore, some unknown portion of the samples in studies using this methodology are people who likely sustained injuries of similar severity but did not pursue medical care, resulting in selection bias. In other words, this conceptualization of the injury variable may miss occurrences of the dependent variable among low-SES participants, creating a false negative (Type II error) by demonstrating a null finding when there is actually a significant relationship.

5.2. Implications for future research

The findings of this review have implications for the conduct of injury research that involves assessment of SES variables. In summary, there are three primary takeaways for future injury research: (a) measures of SES should have as many categories as is feasible so that changes in injury risk across the SES gradient can be assessed; (b) there is value in utilizing multiple levels of SES assessment, so that relationships between both person and place characteristics and injury risk can be accounted for; and (c) measures of SES should be selected based on their relationship to the injury mechanism and their performance in previous research.

5.2.1. Assessing the gradient

Many studies in our review continued to dichotomize SES variables, which loses important information about the SES gradient. As Adler et al. (1994, p. 16) point out, “the relationship of [SES] to health is not simply a threshold effect in which morbidity and mortality increase only at severe levels of deprivation, but is a graded relationship occurring at all levels within the spectrum of social position.” In other words, it is not only the poorest group that is more likely to experience negative health outcomes; there are varying degrees of effects on health at all levels of SES, and improvements in health status at each step-up in level of SES. Studies in this review that measured SES on a gradient found significant, inverse relationships to SES nearly twice as often than studies that did not (65% vs 35%). Whenever possible, injury studies should allow variation in the measures of SES and look at changes in injury risk by level of SES, and should include samples large enough to allow for these assessments.

5.2.2. Measuring at multiple ecological levels

Ecological models are integral to examining the socioeconomic factors that influence injury outcomes. Individual, interpersonal (household), and area-level (neighborhood, community) socioeconomic factors have all been shown to be significantly related to injury and injury-related death (Cubbin et al., 2000; Cubbin & Smith, 2002) and the results of this review echo those findings. Measures of socioeconomic factors are most meaningful when they are measured at multiple ecological levels (Krieger, Williams, & Moss, 1997), according to the pathways of etiology for the injury outcome under consideration (Braveman et al., 2005). There were no studies in this review that could be considered multiple level studies, which would involve using both individual and area measures of SES in assessing risk for individual injury; in comparison, Cubbin and Smith identified only two. This is an area where injury research has significant room to grow.

5.2.3. Selection of measures

SES measurement is not a one-size-fits-all task (Braveman et al., 2005); various mechanisms of injury are linked more closely to certain SES predictors. The results of this review highlight some of these essential relationships, such as parental education, fatal child injury, area-level deprivation and violence-related injury and death. Other measures perform with less clarity, or only predict injury outcomes in certain groups. For instance, in suicide studies, lower income and education appear to be important predictor variables for older white males, but not among other groups. In Krieger et al. (2003), important differences in the performance of SES measures (in terms of detecting differences along the SES gradient) were found in intentional and unintentional weapons-related injury. In the intentional injury data, measures of deprivation (e.g., poverty, median household income) performed better than measures of wealth or income inequality. In unintentional injury studies, measures of occupational class and educational level were able to detect more detail in the SES gradient (which is less steep for unintentional injuries than intentional). In our review studies that included multiple measures of SES detected a significant relationship 62% of the time, compared to 38% in studies that used only one measure.

Injury researchers should select and conceptualize measures of SES carefully, with an eye to previous research, causal pathways, and context. Additionally, the results of this review do suggest that area-level measures of SES are important variables to include when assessing injury risk, and may predict injury even when individual-level measures do not.

5.3. Implications for policy

Much of the research presented in this review is conducted using secondary data. This data originates in survey datasets, hospital records, and medical examiners reports; each such location has an opportunity to collect meaningful data on SES that can contribute to eliminating health disparities. Policies from the agencies that regulate health care providers (and the government bodies that regulate those agencies) could integrate collection of additional SES variables to facilitate heath disparities research into existing systems. Hospital record keeping programs are upgraded fairly often and present a unique and rich opportunity for gathering data. Financial data are often gathered by hospitals and stored in separate databases used for billing purposes, or for assessing financial aid to patients. If regulatory agencies required reporting of financial data along with the medical information that is already reported to national clearinghouses, such as the National Trauma Databank, researchers would be able to identify additional valuable information that could contribute significantly to reductions in public health disparities. Many survey datasets could also be improved in order to enhance disparities research.

5.4. Limitations

The results of this review should be considered in light of its limitations. Study selection could be impacted by publication and selection biases. We made every effort to locate studies meeting our inclusion criteria indexed in PubMed. While PubMed is extensive, it may not catalog every injury study and we cannot eliminate the possibility of selection bias. Studies were selected with particular criteria for defining injury; only studies with actual injuries as the dependent variable were included. Certainly other types of injury studies, such as those that examine events that cause injury (e.g., intimate partner violence, motor vehicle collisions) also examine SES, however, those studies were excluded from this review.

6. Conclusions

In summary, this review was focused on exploring the relationship of SES to injury through a systematic review of recent injury studies, and a discussion of the methodological considerations, especially around conceptualizing and operationalizing SES measures in injury research. Injury remains a leading cause of death and disability in the United States, and socioeconomic disparities in injury persist.

By understanding the contribution of SES to injury risk, and by committing to measuring it effectively and holistically, injury researchers can contribute to lessening this gap. Although major concerns about the measurement of SES in injury research have been discussed previously, there is still a need to improve collection of these variables. In particular, injury research should identify and collect SES variables that are related to the particular injury outcome at hand, as demonstrated by previous research, theory and conceptual pathways, and local context (if applicable). In addition, SES variables will ideally be measured and analyzed in ways that allow for assessment of gradient effects; creating false dichotomies of the “poor” and “not poor” prevents injury researchers from understanding how risk increases by changing levels of SES, and/or where the threshold for increased risk may exist. Finally, both individual and area-level measures of SES affect a person’s risk for many injury outcomes, and whenever feasible, the SES of both the person and the place in which they live should be assessed.

Biographies

Paula Yuma-Guerrero, PhD is an Assistant Professor at the School of Social Work at Colorado State University. Prior to obtaining her PhD at The University of Texas at Austin in 2014, Dr. Yuma coordinated pediatric injury prevention programs at Level I Trauma Centers and served on local and state injury prevention committees. Dr. Yuma-Guerrero’s current research focuses on how community-level factors affect individual health and wellbeing.

Rebecca Orsi, PhD a Research Scientist and Statistician with the Social Work Research Center at the School of Social Work, Colorado State University. Her educational background includes a Master of Science in Statistics and a PhD in Research Methodology, both from Colorado State University. Dr. Orsi’s research interests include child welfare outcomes for special populations, such as adopted children or children with special health needs.

Ping-tzu Lee, PhD is a recent doctoral graduate of the School of Social Work at Colorado State University. Her research interests include resilience in children, critical issues for culturally diverse populations, equine-assisted therapy and mental health.

Catherine Cubbin, PhD is a Professor and Associate Dean for Research in the School of Social Work at the University of Texas at Austin, a Faculty Research Associate at the Population Research Center at the University of Texas at Austin, and a Faculty Associate at the Center for Women’s and Gender Studies, also at UT Austin. Dr. Cubbin received her PhD in Health and Social Policy from the Johns Hopkins University Bloomberg School of Public Health in 1998. Dr. Cubbin’s research focuses on using epidemiological methods to better understand social inequalities in health for the purpose of informing policy.