Decomposing the Association Between Neighborhood Homicide Rates and Stress-Related Health Outcomes Among Black Men in Chicago

HIGHLIGHTS

• •Neighborhood homicide rate was associated with hair cortisol concentration.
• •Perceived stress mediated the homicide rate and hair cortisol relationship.
• •Neighborhood violence may affect physiologic stress response.

Abstract

Introduction

Individuals living in neighborhoods with high levels of violence are exposed to chronic stress, which can have deleterious effects on health. In addition, subjective appraisal of stress may influence health outcomes.

Methods

The authors conducted structural equation modeling. The authors used data collected between May and August 2022, which included 127 older Black men who resided in Chicago, to explore the extent to which perceptions of stress and violence exposure mediate the relationship between actual neighborhood homicide rate and health outcomes.

Results

Bivariate statistics showed that the upper quartiles of homicide rates were associated with higher hair cortisol concentration (p=0.01) but not with depression and post-traumatic stress disorder. Both perceived stress (p=0.01) and perceived violence exposure (p=0.001) were significantly associated with the homicide rate quartiles. Structural equation modeling results showed that with perceived stress as a mediator, there was a significant direct effect (p=0.03) and total effect (p=0.04) of homicide rate on hair cortisol. Although perceived stress was significantly associated with depression (p=0.01) and post-traumatic stress disorder (p=0.03), the direct and total effects of the homicide rate were not statistically significant. With perceived violence exposure as a mediator, there were no significant direct or indirect effects of the homicide rate on hair cortisol, depression, and post-traumatic stress disorder.

Conclusions

Even when older Black men do not perceive or report stress and violence exposure, neighborhood violence still affects their physiologic stress response. The impact of stress can be more accurately assessed by considering both actual and perceived stress measures as well as incorporating stress biomarkers.

Graphical abstract

INTRODUCTION

The homicide rate in the U.S. spiked in 2021 but has since begun to steadily decline.¹ In 2019, the national homicide rate was 12.7 per 100,000, decreasing to 12.0 in 2024, a 6% decline. However, this downward trend is not consistently experienced across all cities. For example, Chicago, IL, has seen a 15% increase in homicide rates between 2019 and 2024.^1,2 Chicago has long reported a higher homicide rate than other cities across the U.S., with a murder rate of 21.5 per 100,000 population in 2024.^3,4 Black Americans are disproportionately affected by this violence.5, 6, 7 The higher rates of homicide among Black Chicagoans reflect the city’s high level of racial residential segregation. Geographic clusters of homicides overlap with racial disparities, which, consequently, leave some neighborhoods in Chicago facing homicide rates nearly 68 times higher than in areas with the lowest rates.⁵

Individuals living in high-crime communities experience chronic stress from ambient crime and violence. Although the body’s stress response is equipped to regulate acute, short-term stress, chronic stress can dysregulate the hypothalamic–pituitary–adrenal axis.⁸ The hypothalamic–pituitary–adrenal axis releases glucocorticoids to regulate immediate, short-term physiologic responses to stress. Chronic stress often leads to persistent elevation of cortisol levels, which is associated with various chronic diseases, such as cardiovascular disease, obesity, and cancer.9, 10, 11, 12, 13, 14 Hair cortisol concentration (HCC) is a marker for retrospective stress response.^15,16 Studies documented a positive association between HCC and chronic stress, which has been shown to be related to perceived social stress, discrimination, and depression.^17,18 A study showed that HCC was significantly higher among Black men living in Chicago community areas with higher homicide rates than among those living in communities with low homicide rates.¹⁹ However, to be sure, there are studies documenting stress-induced declines in cortisol output.^20,21 These studies indicate that severe stress-associated mental health conditions, such as post-traumatic stress disorder (PTSD), may result in low cortisol levels.²² Although factors explaining these discrepancies are not well explained, these cases are often associated with early childhood stress experiences, and trauma in early life may contribute to blunted glucocorticoid metabolism.^23,24

The effects of violence exposure on mental health, such as anxiety, depression, and PTSD, have been documented. For example, in primary care patients, violence exposure was associated with a 46.2% lifetime prevalence of PTSD and 36.7% of major depressive disorder.²⁵ Residents residing in violent-crime hot spots are shown to have nearly twice as many reaching clinical thresholds for depression and PTSD diagnosis as those residing in low-crime areas.^26,27 Similarly, direct exposure to gun violence was significantly associated with increased risk of suicidal ideation and behaviors, with higher risk observed as exposure accumulates.²⁷ Perceptions of stress and violence, that is, the subjective appraisal of the level of crime and stress, influence physical and mental health outcomes. Studies have found that heightened perceptions of stress and exposure to violence are associated with negative mental health outcomes, such as depression and PTSD.^28,29 Other studies have examined the mediating role of family and community support between perceptions of stress and violence. Notably, it is the perception of support rather than the actual support received that appears to have a stronger influence on health outcomes.^30,31 Previous studies have explored whether or to what extent actual neighborhood or perceptions of crime and violence negatively affect physical and mental health outcomes, such as elevated cortisol, depression, and PTSD.32, 33, 34

Previous studies on the indirect effects of perceptions of crime and violence on health outcomes have primarily focused on youth or women.35, 36, 37, 38, 39, 40, 41 Relatively little research has examined stress and health outcomes among adult men, despite the potential long-term cumulative effects of chronic violence exposure among older men.⁴² Men respond to stress differently than women, which may differently affect not only physical and mental health but also perceptions of stress.⁴³ Perceptions of neighborhood crime/violence, stress, and social support mediate the effects of actual violence on health outcomes among various populations of youth and women.

However, limited evidence is available on whether the mediating effects of perceptions significantly influence health outcomes for older men. This study aims to bridge this gap in knowledge by elucidating the indirect effects of perceived stress and violence exposure on physical and mental health outcomes. The authors explore the direct effects of living in neighborhoods with high homicide rates on outcomes of hair cortisol, depression, and PTSD. The authors further examine the mediating effects of perceptions of stress and violence exposure between actual homicide rates and outcome measures. Protective factors, including community collective efficacy (CCE), family coping, and social support, are also evaluated as intermediate confounders.

METHODS

Study Sample

Black men who were aged ≥45 years, were former or current smokers, and resided in Chicago were recruited from 4 barbershops in predominantly Black communities.^19,44 Those who were taking steroids or other corticosteroid medications were excluded. Research staff administered surveys after informed consent. Participants completed an in-person survey at the barbershop and received haircuts from the barbers who collected hair samples for this study. Participant data included demographic characteristics, residential community areas, perceived stress, exposure to violence, and family and community support factors. Participants received $150 upon finishing the survey. Recruitment occurred between May 23, 2022, and August 1, 2022. A convenience sample of 161 older Black men was recruited. This analysis was part of a research project titled “Reducing lung cancer by decoding the link between neighborhood exposure and biological responses (RECODE),” funded by the National Institute on Minority Health and Health Disparities (R01MD014839). The protocol was approved by the University of Illinois Chicago IRB (IRB Number 2022-0301). The detailed recruitment strategy was described elsewhere.^19,44

Measures

Figure 1 presents the analytic framework of the structural equation model (SEM) to inform the mediation analysis.⁴⁵ The main exposure of interest was violence exposure, measured by neighborhood homicide rates. The authors evaluated the direct and indirect effects on 3 outcomes: HCC, depression, and PTSD. The mediators were perceived stress and perceived violence exposure. Intermediate confounders were 3 levels of support factors, including collective efficacy at the community level, family coping at the family level, and social support at the individual level. External confounders—age, smoking status, marital status, and employment status—were controlled for in all SEMs tested.

Figure 1.

Analytic framework for mediation analysis with intermediate confounding.

The exposure variable, violence exposure, was measured using neighborhood homicide rates per 100,000 population at the respondent’s residential community area level. Homicide cases were retrieved from the Cook County Medical Examiner’s Case Archive.⁴⁶ The population size was obtained from the U.S. Census Bureau.⁴⁷ Homicide rate per 100,000 as a continuous variable was used in SEM. The homicide rates were also grouped into quartiles for descriptive statistics.

In terms of outcome variables, for HCC, physiologic stress was measured using HCC, collected from participants' hair samples. Additional details of hair collection and laboratory procedures can be referenced in the parent study to this paper.¹⁹ Because HCC was not normally distributed, the log of HCC (lnHCC) was used for the analysis, ranging from 0.82 to 4.77.

For depression, the Patient Health Questionnaire 9 (PHQ-9) was used to measure depression. PHQ-9 is a 9-item survey to assess the severity of depression. Responses ranged from not at all (0) to nearly every day (3), and the score ranged from 0 to 27.⁴⁸ PHQ-9 is a widely used tool with consistently high validity, with a pooled Cronbach’s alpha of 0.86 from 60 studies.⁴⁹

For PTSD, the Adult National Stressful Events Survey PTSD Short Scale (NSESSS) was used to measure the presence of traumatic stress symptoms, which is a 9-item survey rated on a 5-point scale, from not at all (0) to extremely (4). The total score ranges from 0 to 36, with higher scores indicating a greater severity of trauma symptoms.⁵⁰ Items include questions about whether having flashbacks, feeling very upset when something reminded a stressful experience, trying to avoid thoughts or feelings, losing interest in activities that the participant used to enjoy, and being super alert or on guard. The validity of NSESSS has been shown to be very high, with a Cronbach's alpha of 0.91, supporting its use as a brief screening tool for PTSD.⁵¹

In terms of mediator variables, for perceived stress, Perceived Stress Scale (PSS) was used. The PSS is a well-validated instrument designed for use within community samples, which includes 10 items with a 5-point Likert scale. Scores can possibly range from 0 to 40.^52,53 Cronbach’s alpha for this study was 0.851. For perceived violence exposure, Screen for Adult Violence Exposure (SAVE) was used, which has been validated for its internal consistency, test–retest reliability, and validity for adolescents and adults.^54,55 The survey includes 30 questions with a 5-point Likert scale. Total scores can range from 0 to 120. The Cronbach’s alpha for internal consistency in this study was 0.956.

In terms of intermediate confounder variables, community-level support was measured using the CCE tool, a 10-item survey that measures participants’ perception of how well their community works together and deals with community issues.⁵⁶ The CCE includes items on informal social control, social cohesion, and trust. Item scoring uses a Likert scale from 1 to 5 points, with a Cronbach alpha over 0.80.⁵⁶ For family-level support, the Family Coping Index (FCI) is a 24-point questionnaire about the participants’ family unit’s responses to cope with problems or difficulties. Items were on a 5-point Likert scale, ranging from 24 to 120.⁵⁷ Cronbach’s alpha for FCI in this study was 0.948. The Medical Outcomes Study (MOS) Social Support Survey was used to measure individual-level support.⁵⁸ MOS is a well-validated measure instrument for individual-level support,⁵⁸ with an item correlation over 0.90.⁵⁹ The MOS score was calculated by taking the mean of 19 items, with scores ranging from 19 to 95.⁶⁰

Statistical Analysis

Of the total of 161 participants, 3 cases were excluded owing to missing residential community information. In addition, HCC scores above 119.55 pg/mg were deemed outliers on the basis of the median absolute deviation method,⁶¹ thus excluded from the analysis. A total of 127 participants were included in the final analytic data set. Owing to missing responses, the analytic samples varied by the outcomes: HCC (n=120), PHQ-9 (n=118), and NSESSS (n=119). The authors employed the SEM to estimate the direct and indirect associations between neighborhood homicide rates and stress-related outcomes, including HCC, depression, and PTSD. Path coefficients represent the effects of each variable on another within the SEM. In all SEMs, the authors controlled for external confounders, including age, smoking status, marriage status, and employment status.

Figure 1 shows the combination paths for direct, indirect, and total effects for this analysis. Pathway c represents the direct path from the homicide rate to HCC, whereas pathways a and b are the indirect effects. The indirect effect is from the homicide rate to perceived stress (a) and perceived stress to HCC (b). To assess whether the mediation effect is present, the authors combined individual path estimates to quantify the total, direct, and indirect effects as well as the proportion mediated for perceived stress and perceived violence exposure. Table 1 describes direct and indirect effect paths. The first indirect effect mediated through perceived stress or violence exposure was the sum of the indirect path through perception $\left(\mathrm{a}*\mathrm{b}\right)$and the indirect path through the intermediate confounder and perception $\left(\mathrm{f}*\mathrm{e}*\mathrm{b}\right)$. The second indirect effect captures the path from homicide to the outcome through the intermediate confounder only $\left(\mathrm{f}*\mathrm{d}\right)$. The total effect of the homicide rate on HCC was calculated by combining the direct effect and both indirect effects: $\mathrm{c}+\left(\mathrm{a}*\mathrm{b}\right)+\left(\mathrm{f}*\mathrm{e}*\mathrm{b}\right)+\left(\mathrm{f}*\mathrm{d}\right)$. Finally, the proportion mediated was calculated as the indirect effect through the mediator divided by the total effect: $\left(\left[\mathrm{a}*\mathrm{b}\right]+\left[\mathrm{f}*\mathrm{e}*\mathrm{b}\right]\right)/\left(\mathrm{c}+\left[\mathrm{a}*\mathrm{b}\right]+\left[\mathrm{f}*\mathrm{e}*\mathrm{b}\right]+\left[\mathrm{f}*\mathrm{d}\right]\right)$. These parameters present how much of the total effect of neighborhood homicide on HCC, depression, and PTSD is mediated through indirect pathways of perceived stress and violence exposure and the 3 levels of support factors.

Table 1.

Description of the Indirect and Direct Paths of the Homicide Rate on Outcomes

Parameter	Paths	Description
Indirect effect 1	$\left(\mathrm{a}*\mathrm{b}\right)+\left(\mathrm{f}*\mathrm{e}*\mathrm{b}\right)$	Total indirect effect through the mediators and the intermediate confounders
Indirect effect 2	$\mathrm{f}*\mathrm{d}$	Indirect effect through community collective efficacy directly to outcomes, bypassing mediators
Direct	c	Direct effect of homicide rate on HCC, depression, or PTSD
Total	$\mathrm{c}+\left(\mathrm{a}*\mathrm{b}\right)+\left(\mathrm{f}*\mathrm{e}*\mathrm{b}\right)+\left(\mathrm{f}*\mathrm{d}\right)$	Total effect of homicide rate on HCC, depression, or PTSD
Proportion of mediated effect	$\frac{\left(\right[\mathrm{a}*\mathrm{b}]+[\mathrm{f}*\mathrm{e}*\mathrm{b}\left]\right)}{(\mathrm{c}+[\mathrm{a}*\mathrm{b}]+[\mathrm{f}*\mathrm{e}*\mathrm{b}]+[\mathrm{f}*\mathrm{d}\left]\right)}$	Proportion of the mediation effect of the total effect

HCC, hair cortisol concentration; PTSD, post-traumatic stress disorder.

RStudio, Version 2023.6.0.421, was used for the analysis.⁶² Using the sem function in the lavaan package,⁶³ the authors bootstrapped the SEM by 1,000 repetitions to obtain estimates for different combinations of the SEM. Coefficients were estimated for mediators perceived stress (PSS) and perceived violence exposure (SAVE). CCE, FCI, and social support (MOS) were examined separately as the intermediate confounders. Findings are presented for CCE, and the results for FCI and MOS are provided in the Appendix (available online). Statistical significance for SEM estimation was set at a p<0.05.

RESULTS

A total of 127 participants were included in this analysis. Table 2 summarizes the characteristics of the sample. The mean age for the study participants was 56.5 years. The mean age did not differ by the homicide rate quartile. Participants who lived in areas with lower homicide rates were more likely to be married. Over 52% of those living in Quartile 1 homicide rates were married, whereas 27.3% of those in Quartile 2, 20% in Quartile 3, and 22.2% in Quartile 4 were married. Overall, 34.1% of participants were unemployed. The unemployment rates were substantially higher for those living in Quartile 3 (58.1%) and Quartile 4 (35.7%) homicide rates. The proportion of current smokers was highest for Quartile 3 (71.0%) but lowest for Quartile 4 (35.7%).

Table 2.

Demographics Sample Characteristics by Homicide Rate Quartile (N=127)

	Homicide rate quartile
Characteristics	Total (N=127)	Q1 (n=44)	Q2 (n=24)	Q3 (n=31)	Q4 (n=28)	p-value
Mean homicide ratea	71.9 (38.3)	33.4 (22.5)	58.4 (9.3)	88.2 (5.9)	126.0 (4.3)	-
Outcomes
Hair cortisol concentrationa	31.9 (28.4)	21.5 (20.2)	36.0 (32.4)	36.0 (26.6)	40.2 (34.1)	0.01
Depressiona	6.1 (5.9)	4.9 (5.8)	5.1 (4.8)	8.2 (6.2)	6.5 (6.1)	0.06
Post-traumatic stress disordera	9.7 (6.8)	8.8 (6.9)	8.5 (6.4)	11.4 (7.0)	10.1 (6.8)	0.30
Individual characteristics
Age in years	56.5 (8.7)	57.4 (10.4)	55.8 (10.7)	56.0 (4.4)	56.3 (7.4)	0.76
Marriedb	41 (33.3)	23 (52.3)	6 (27.3)	6 (20.0)	6 (22.2)	0.01
Unemployedb	43 (34.1)	10 (22.7)	5 (21.7)	18 (58.1)	10 (35.7)	0.01
Current smokerb	58 (45.7)	17 (38.6)	9 (37.5)	22 (71.0)	10 (35.7)	0.01
Support factors
Collective efficacya	29.0 (6.3)	31.5 (7.6)	27.9 (5.8)	27.8 (3.7)	27.4 (5.8)	0.04
Family copinga	84.6 (18.3)	89.0 (17.4)	81.7 (19.2)	85.9 (17.1)	79.1 (19.2)	0.30
Social supporta	68.7 (20.2)	75.7 (19.7)	63.2 (22.8)	67.0 (15.7)	64.0 (20.5)	0.02
Perceptions
Violence exposurea	55.1 (7.2)	52.8 (6.5)	54.0 (8.2)	59.2 (6.1)	55.2 (6.8)	0.001
Perceived stressa	17.2 (5.3)	15.2 (5.0)	16.9 (6.2)	19.6 (5.0)	18.0 (4.1)	0.01

^aKruskal−Wallis H test used for non-normally distributed continuous variables and ANOVA test used for normally distributed continuous variables with mean (SD).

^bChi-square test used to calculate p-value for categorical variables with n (%).

Q, quartile.

Regarding the outcomes, HCC was higher for the upper quartiles of homicide rates (p=0.01). However, there were no statistical differences for depression and PTSD by homicide rate quartile. Both perceived stress (p=0.01) and perceived violence exposure (p=0.001) were significantly higher for Quartile 3 and Quartile 4 homicide rates. In terms of support factors, CCE (p=0.04) and individual social support (p=0.02) were lower for the higher homicide quartile. Family coping, although higher for lower homicide quartiles, was not statistically significant.

Table 3 presents the estimated path coefficients with perceived stress (PSS) and perceived violence exposure (SAVE) as mediators and CCE as an intermediate confounder. Appendix Table 1 (available online) summarizes the findings for the FCI, and Appendix Table 2 (available online) summarizes the findings for the MOS social support survey (MOS) as intermediate confounders. Coefficients for pathways can be interpreted as a 1-unit increase in the predictor corresponding to a β-unit change in the outcome.

Table 3.

Estimated Coefficients From Regression Models Examining the Association Between Neighborhood Homicide Rate and Health Outcomes

	Outcome
	lnHCC (n=120)		Depression (n=118)		PTSD (n=119)
Pathway	Estimate (95% CI)	p-value	Estimate (95% CI)	p-value	Estimate (95% CI)	p-value
Mediator: perceived stress (PSS)
Homicide rate (c)	0.16 (0.01, 0.30)	0.03	0.02 (−1.04, 0.99)	0.97	−0.24 (−1.60, 0.92)	0.71
PSS (b)	0.00 (−0.03, 0.03)	0.99	0.43 (0.27, 0.61)	<0.001	0.39 (0.15, 0.65)	0.001
CCE (d)	0.00 (−0.03, 0.03)	1.00	−0.10 (−0.24, 0.04)	0.18	−0.26 (−0.46, −0.05)	0.01
Homicide rate → PSS (a)	1.02 (0.42, 1.65)	0.002	1.00 (0.30, 1.66)	0.003	1.02 (0.41, 1.63)	0.001
CCE → PSS (e)	−0.05 (−0.19, 0.11)	0.55	0.05 (−0.20 0.13)	0.53	−0.05 (−0.20, 0.11)	0.56
Homicide rate → CCE (f)	−1.31 (−2.34, −0.15)	0.02	−1.51 (−2.65, −0.36)	0.01	−1.29 (−2.41, −0.19)	0.03
Mediator: perceived violence (SAVE)
Homicide rate (c)	0.13 (0.004, 0.28)	0.07	0.18 (−0.83, 1.21)	0.72	−0.15 (−1.28, 0.98)	0.80
SAVE (b)	0.02 (−0.01, 0.05)	0.13	0.24 (0.06, 0.40)	0.01	0.38 (0.23, 0.54)	<0.001
CCE (d)	0.01 (−0.02, 0.03)	0.64	−0.07 (−0.23, 0.06)	0.31	−0.20 (−0.40, 0.00)	0.04
Homicide rate → SAVE (a)	0.88 (−0.12, 2.01)	0.11	0.10 (−0.04, 2.12)	0.07	0.88 (−0.20, 1.92)	0.10
CCE → SAVE (e)	−0.18 (−0.35, −0.003)	0.05	−0.16 (−0.33, 0.04)	0.10	−0.18 (−0.35, −0.002)	0.04
Homicide rate → CCE (f)	−1.27 (−2.42, −0.09)	0.02	−1.47 (−2.53, −0.35)	0.01	−1.27 (−2.43, −0.13)	0.03

Note: Exposure is neighborhood homicide rate, outcome is natural lnHCC, mediator is perceived stress (PSS) and perceived violence (SAVE), and intermediate confounder is CCE.

CCE, community collective efficacy; lnHCC, log-transformed hair cortisol concentration; PSS, Perceived Stress Scale; PTSD, post-traumatic stress disorder; SAVE, Screen for Adult Violence Exposure.

Figure 2 shows the SEM results explaining the associations between homicide rates and depression, with the 2 mediators, PSS (Figure 2A) and SAVE (Figure 2B). For example, pathway b in Figure 2A shows that a 1-unit increase in perceived stress score (e.g., PSS score from 26 to 27) was associated with a 0.43-unit increase in the depression score (β=0.43, p<0.01). Similarly, pathway d in Figure 2A shows that a 1-unit increase in the CCE score (e.g., CCE score from 10 to 11) is associated with a 0.10-unit decrease in PTSD score (β= −0.10, p=0.18). Appendix Figures 1 and 2 (available online) present the SEMs for the 2 other outcomes (hair cortisol and PTSD) with the 2 mediators (PSS and SAVE).

Figure 2.

Structural equation model results explaining the association between homicide rate and depression, with mediators (perceived stress and perceived violence exposure).

Note: Pathway (coefficient), *p<0.05 and **p<0.01.

CCE, community collective efficacy; lnHCC, log-transformed hair cortisol concentration; NSESSS, Adult National Stressful Events Survey PTSD Short Scale; PHQ-9, Patient Health Questionnaire 9; PSS, Perceived Stress Scale; SAVE, Screen for Adult Violence Exposure.

With PSS as the mediator (Appendix Figure 1, available online), there was a significant positive association between the homicide rate and lnHCC (β=0.16, p=0.03) but not with depression or PTSD, suggesting that a higher neighborhood homicide rate is associated with increased physical measurements of stress but not mental health measurements of stress. There was a positive association between neighborhood homicide rates and PSS in all the models, indicating that higher homicide rates are associated with increased perceived stress.

Although neither PSS nor CCE was significantly associated with lnHCC, PSS was significantly associated with depression (p<0.001) and PTSD (p<0.01), indicating that higher perceived stress was associated with more severe mental health outcomes. Although CCE had no significant association with depression (p=0.31), it was negatively associated with PTSD (p<0.05), suggesting that higher collective efficacy is associated with lower levels of PTSD.

With SAVE as a mediator (Figure 2B and Appendix Figure 2, available online), neighborhood homicide rate was not associated with lnHCC (β=0.13, p=0.07), depression (β=0.18, p=0.72), and PTSD (β= −0.15, p=0.80). Furthermore, there were no significant associations between neighborhood homicide and the mediator variable SAVE in any of the models. Unlike PSS as a mediator, CCE was negatively associated with SAVE for 2 of the 3 models: lnHCC (β= −0.18, p=0.05) and PTSD (β= −0.18, p=0.04), which means that higher CCE is associated with lower HCCs and PTSD but not depression. SAVE has a significant positive association with depression (β=0.24, p=0.01) and PTSD (β=0.38, p<0.001) but not with lnHCC (β=0.02, p=0.13).

Table 4 summarizes the direct, indirect, and total effects of neighborhood homicide rate on lnHCC, depression, and PTSD with CCE as the intermediate confounder. With PSS as a mediator, there was a statistically significant total effect between homicide rate and lnHCC (β=0.16, p=0.04), which is explained entirely by the direct effect (β=0.16, p=0.03) with no significant indirect effect through either perceived stress (indirect effect 1) or CCE (indirect effect 2). This result suggests that the association between neighborhood homicide rate and HCC operates primarily through exposure to violence rather than perceived stress or CCE.

Table 4.

Estimation of the Direct, Indirect, and Total Effect of Neighborhood Homicide Rate on Health Outcomes

	Outcome
	lnHCC (n=120)		Depression (n=118)		PTSD (n=119)
Pathway	Estimate (95% CI)	p-value	Estimate (95% CI)	p-value	Estimate (95% CI)	p-value
Mediator: perceived stress (PSS)
Indirect effect 1	0.00 (−0.03, 0.04)	0.99	0.46 (0.12, 0.88)	0.01	0.42 (0.11, 0.84)	0.03
Indirect effect 2	0.00 (−0.04, 0.04)	0.99	0.15 (−0.07, 0.51)	0.32	0.33 (0.01, 0.80)	0.12
Direct	0.16 (0.01, 0.30)	0.03	0.02 (−1.04, 0.99)	0.97	−0.24 (−1.60, 0.92)	0.71
Total	0.16 (0.02, 0.30)	0.04	0.63 (−0.37, 1.66)	0.22	0.51 (−0.85, 1.63)	0.42
Proportion mediated	0.002 (−0.47, 0.53)	0.99	0.74 (−4.55, 9.06)	0.94	0.83 (−7.66, 7.19)	0.96
Mediator: perceived violence (SAVE)
Indirect effect 1	0.02 (−0.01, 0.07)	0.25	0.29 (0.02, 0.64)	0.08	0.42 (0.01, 0.93)	0.07
Indirect effect 2	−0.01 (−0.05, 0.03)	0.67	0.11 (−0.08, 0.41)	0.40	0.26 (−0.02, 0.67)	0.16
Direct	0.13 (0.004, 0.28)	0.07	0.18 (−0.83, 1.21)	0.72	−0.15 (−1.28, 0.98)	0.80
Total	0.14 (0.01, 0.29)	0.05	0.58 (−0.40, 1.60)	0.23	0.54 (−0.76, 1.67)	0.39
Proportion mediated	0.15 (−0.07, 0.99)	0.97	0.50 (−3.10, 3.94)	0.99	0.79 (−8.60, 4.51)	0.98

Exposure is homicide rate, mediator is perceived violence exposure, intermediate confounder is community collective efficacy, indirect effect 1 is the effect mediated by perceived stress or perceived violence exposure, and indirect effect 2 is the effect mediated by community collective efficacy and not through perceived stress or perceived violence.lnHCC, log-transformed hair cortisol concentration; PSS, Perceived Stress Scale; PTSD, post-traumatic stress disorder; SAVE, Screen for Adult Violence Exposure.

For depression and PTSD as the outcomes, the direct and total effects were no longer significant. However, the indirect effect through perceived stress and CCE (indirect effect 1) was significant in both depression (β=0.46, p=0.01) and PTSD (β=0.42, p=0.03) models. With perceived violence exposure as a mediator, the total effect between neighborhood homicide rate and lnHCC was statistically significant (β=0.14, p=0.05), but, similar to the PSS model, it can be entirely explained by the direct effect (β=0.13, p=0.07). Unlike the results from the PSS mediator models, there were no statistically significant indirect effects through perception of violence on depression or PTSD. Appendix Table 3 (available online) summarizes the models with the FCI instead of CCE, and Appendix Table 4 (available online) summarizes the models with the MOS social support survey. The results were similar to the models with CCE.

DISCUSSION

In this analysis, the authors examined how residing in neighborhoods with high levels of violence may be associated with physical and psychological stress outcomes, particularly focusing on the extent to which perceptions of violence and stress may mediate the relationship between actual neighborhood violence and stress outcomes. The authors found that higher homicide rates have a direct effect on hair cortisol, a marker for chronic stress, but not on the psychological outcomes, both depression and PTSD. In addition, there was no indirect effect mediated by perceptions of violence exposure and stress, suggesting that the total effect of social stress exposure was carried entirely through the direct path to stress outcomes. This finding may suggest that even when no significant perceived stress or violence exposure is present, neighborhood violence may still affect physiologic stress response. Studies have documented the phenomenon of skin deep resilience, which describes individuals’ ability to maintain high levels of self-control and adjustment, whereas the apparent resilience, all the same, costs physical health.^64,65 The finding seems to corroborate the cost of sustaining the discrepancy between self-reported stress and actual stress exposure, which still has physical consequences.

Although there was no significant direct or total effect of homicide rate on mental health outcomes, the authors also observed that perceived stress was significantly associated with depression and PTSD. This mechanism has been documented in prior studies, where perceived neighborhood disorder and violence were associated with higher levels of depression and PTSD.^26,55,56 The authors found no significant direct effects of actual violence on depression and PTSD, which may be due, in part, to men generally underreporting symptoms of mental health challenges.⁶⁶ Social expectations of masculinity may interfere with the expression of emotions and with seeking help; consequently, men may be less likely to disclose mental health symptoms owing to shame and stigma. Symptoms of depression may also present differently in men, either as aggression, alcohol use, or other substance use, and are not always aligned with assessment questions, such as the PHQ-9, which contributes to the underdiagnosis of mild-to-moderate depression in men.⁶⁶

Current literature on trauma and mental health predominantly explores the experiences of women and adolescents. The findings with older men may indicate that men may perceive and/or report violence exposure and stress differently.35, 36, 37, 38, 39, 40, 41 For example, women may report stress and anxiety more openly.⁶⁷ Previous studies also documented that men and women adopt different strategies to deal with stress and difficult emotions^67,68 and perhaps experience and perceive stress differently.⁶⁹ Furthermore, older adults who have lived in areas with high levels of violence for a longer period of time might be desensitized to the persistent threat, which then leads to altered perceptions of stress.^70,71 Although further research comparing different demographic groups is required, the findings suggest that models for stress exposure and health outcomes for older men may take pathways different from those for women and younger men. Thus, to account for differences in expectations and norms around stress for different population groups, it would be beneficial to reconsider existing models and assessment tools to better capture men’s experiences and responses to social stress.

Limitations

This study has several limitations. First, the authors used cross-sectional data for this analysis; thus, the authors could not establish a temporal sequence of events in a dynamic mediation relationship. The authors hypothesized that actual neighborhood violence influences perceived violence exposure and stress and ultimately affects health outcomes. The data did not allow for the temporality of these measures, although the directions of these variables were less problematic, because there was no confusion around whether perceptions of stress and violence exposure would affect the actual homicide rate, the exposure variable. However, the associations between depression and PTSD as outcomes and perceived stress and violence exposure as mediators are less clear. Future prospective studies may tease out this relationship or perhaps bidirectional relationships between perceptions and mental health outcomes.

Second, given the known gender differences in the expression and self-reporting of mental health symptoms, perceptions of stress and violence exposure among men in this study may not have fully captured the mediating factors linking actual violence exposure to health outcomes. Alternative appraisal of stress, particularly behavioral responses such as maladaptive coping, substance use, or aggression, may capture relevant mediators for men. Relatedly, the authors used the homicide rate as the social stress exposure measure in this analysis, but other forms of social stress could be considered, for example, neighborhood persistent poverty, racial segregation, or physical disorder. Although there are strong correlations between these neighborhood characteristics, expanded analysis of a range of potential social stressors can help examine the relative importance of these highly correlated and yet different aspects of stressors.

Third, although the authors adjusted for a range of relevant sociodemographic and contextual factors, they were not able to account for certain health behaviors, such as physical activity and sleep, which may be potential confounders. These health behaviors have been shown to influence both mediators and mental health outcomes.72, 73, 74, 75, 76 Thus, the analysis may have led to residual confounding. Future research should incorporate potential health behaviors that may be associated with both neighborhood conditions and health outcomes.

Finally, this analysis was done with a relatively small sample of Black men primarily from 4 community areas in Chicago. Although this is a unique strength in addressing an underrepresented population, the findings may not be generalized to the wider population of men. In addition, these community areas have substantially higher rates of homicides than the overall rate for the city of Chicago.¹⁹ However, it should also be noted that the effect of high neighborhood violence cannot effectively be examined when the exposure to violence is rare for the overall population. For this reason, the authors focused on Black men residing in Chicago’s community areas that have historically been affected by high crime and violence, which are also highly segregated Black communities. Future research may explore a wider range of crime and violence to explore the full dose response.

CONCLUSIONS

Exposure to chronic stress has downstream physiologic consequences. High levels of neighborhood violence affect the health of residents, in part, through the body’s stress response and, subsequently, elevated cortisol, which is associated with a myriad of chronic illnesses. People living in neighborhoods with high crime and violence may exhibit high levels of hair cortisol as a marker for chronic stress. On the other hand, even when older Black men do not perceive or report stress and violence exposure, neighborhood violence still affects their physiologic stress response. Men and women may perceive and process stress differently. The impact of social stress could be more accurately assessed by considering gender-specific appraisal and incorporating stress biomarkers.

CRediT authorship contribution statement

Evgenia Karayeva Pertsch: Conceptualization, Data curation, Formal analysis, Methodology, Software, Writing – original draft, Writing – review & editing. Sage J. Kim: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Validation, Writing – original draft, Writing – review & editing.