Abstract
Social stability is an understudied construct in public health that offers a useful framework for understanding social disadvantage across multiple domains. This study investigated prevalence and patterns of cooccurrence among a hypothesized set of social stability characteristics (housing, residential transition, employment, income, incarceration, and partner relationship), evaluated the possibility of underlying subgroups of social stability, and investigated the association between social stability and health outcomes. Data were from comprehensive interviews with primarily African-American low income urban women and their female social network members (n = 635) in Baltimore. Analysis included exploratory statistics, latent class analysis, and latent class regression accounting for clustered data using Stata and Mplus software. Social stability characteristics cooccurred in predictable directions, but with heterogeneity. Respondents had an average of three stability characteristics (S.D.: 1.4). Latent class analysis identified two classes of social stability: low (25%) and high (75%), with the higher class less likely to experience each of the included indicators. In controlled models, higher social stability was significantly correlated with social network characteristics and neighborhood integration. Higher social stability was independently associated with reduced risk of chronic illness (AOR: 0.54, 95% C.I.: 0.31, 0.94), mental illness history (AOR: 0.24, 95% CI: 0.15, 0.39), and current depressive symptoms (AOR: 0.35, 95% C.I.: 0.22, 0.57). The current set of social stability characteristics appears to represent a single construct with identifiable underlying subgroups and associated health disparities. Findings suggest a need for comprehensive policies and programs that address structural determinants of cooccurring social disadvantage and help to mitigate the likely spiral effect of instability experiences.
Keywords: Social stability, Social disadvantage, Health disparities, Chronic disease, Depressive symptoms, Urban health
Introduction
The literature on social conditions and health disparities points to a set of linked social circumstances that are common indicators of disadvantage and vulnerability, including poverty, homelessness, unemployment, incarceration, and social isolation. Social stability is an understudied construct in public health that offers a useful framework for understanding social disadvantage across multiple domains. Measured on an individual level, social stability refers to the range of life structure and reliable routine that is protective against further situational hazards and helps maintain connections with social resources and societal expectations. The construct is commonly assessed as the product of steady social circumstances within a defined set of domains, e.g., housing, employment, social ties, sufficient income, and lack of imprisonment. Most of the existing social stability measures assume an independent or additive relationship among indicators, which may not fully reflect the natural configuration of these factors within impoverished communities. There is considerable variation within the existing literature and prior investigations leave a number of questions unaddressed.
Social stability theoretical background
Importance of stable life circumstances has been discussed in the developmental and clinical psychology literature. From a psychological perspective, lack of stability due to divorce, unemployment, or household disorder has been viewed as a stressful life event that can impinge on child development and psychological well-being.1,2 The social drift hypothesis suggested that mental illness may lead to unemployment, social role disruption, and downward social mobility.
The construct of social stability in public health was first developed to explain the range of social functioning observed among male alcohol treatment enrollees.3 Early conceptualization drew on theories of social integration, social control, and social roles to suggest that social stability is an indication of life structure and appropriate participation in society. The domains of social stability were selected as those that provide meaning to life and indicate one’s ability to function within and be engaged with one’s environment.3 The capacity to fulfill societally determined social roles further instills certain demands and expectations, which simultaneously limit mobility and offer status and support.4
The social role premise of social stability is reflected in the most frequently referenced measures of social stability, which calculate one’s cumulative social stability score based on a predefined set of criteria. These most commonly include assessment of one’s employment, marital, and housing status, but also include criminal justice and mobility indicators. For example, Straus and Bacon measured social stability using a cumulative index of living in one’s own home, residential immobility for at least 2 years, living with wife, and having a steady job for at least 3 years.
An extension of the social role conceptualization emphasizes the aspects of constancy and resilience in social stability, both of which are common themes in crossdisciplinary stability research.5 These aspects are most evident in qualitative research, where observations demonstrate the cyclical and exacerbating associations among stability domains and associated feelings of constant change and uncertainty that challenge attempts to improve one’s social situation. For example, a study among drug users showed that some face daily life difficulties such as homelessness, lack of regular employment, poverty, and fragile relationships that contribute to a “chronic sense of uncertainty” and feeling “out of control” that contrasts stability.6 Similarly, among urban welfare recipients, a litany of daily struggles made it nearly impossible to maintain long-term daily routines given varying income, housing difficulties, employment wait lists, and health care issues. Respondents lived in circumstances of “constant uncertainty and instability” and could take little for granted on any regular basis.7
Taken together with the original understanding of social stability as a reflection of social integration and social role fulfillment, social stability can be understood as a state of life structure and constancy that functions in a protective way against further hazards and helps to maintain one’s connection with societal expectations.
Social stability assessment
Operationalization and construct measurement of social stability has varied extensively, making it difficult to compare across studies.8 Construct definitions vary in the type, specificity, and number of included indicators as well as the time periods of measurement. Looking across measures, social stability domains most commonly include housing, employment status, partnership, and legal status. Some definitions also include residential stability, income level, and occasionally other measures of social support.
Two studies have sought to discern which variables best fit the construct of social stability. A small cluster analysis study classified the following as social stability: residential status, months at last residence, occupational status, number of recent job changes, marital status, and number of recent close friends.9 Using correspondence analysis, one study assessed a set of hypothesized social stability indicators, resulting in a final continuous measure composed of five variables: unstable housing, unemployment, living on welfare, history of incarceration, and lack of stable relationship.10
Most measures have assumed that social stability is a continuous variable that ranges according to the number of stability domains in one’s life and thus assessed the construct cumulatively, reporting either mean scores or severity groupings according to the sample distribution. Though it is likely that the accumulation of stability would portend a more positive life situation, it is difficult to maintain that the measure is inherently a continuous one, such that the degree of instability increases proportionally with each additional indicator. This approach devalues the qualitative contribution of any particular stability indicator, makes it impossible to understand the intricacies of how the stability domains cooccur, and negates the possibility that there are underlying categorical patterns of social stability.
There is a need to improve the conceptualization and measurement of social stability, in order to enhance consistency, gain a better sense of its manifestation in different populations, make comparisons that can inform programs and policies, and understand the relationship between stability and health outcomes.
Interrelated domains
A variety of studies have demonstrated the dynamic and often cyclical relationships among domains of social stability. Employment and economic resources contribute to homelessness,11 whereas prison history and marital status predict employment among homeless individuals.12 Unstable housing creates difficulties for maintaining positive social and economic ties11,13 and homeless individuals face substantial barriers to employment and economic stability.14,15
Incarceration is a dominant characteristic within urban impoverished neighborhoods, especially communities of color.16 Lack of employment and economic uncertainty contributes to participation in informal economies and illicit resource generation, which increases incarceration risk.17–19 Incarceration increases the likelihood of relationship dissolution or divorce,20 is disruptive to permanent housing and job opportunities, and is common among individuals experiencing homelessness.21 Community reentry for those returning from jail is marked by challenges in reestablishing social ties, safe housing, and employment.22–24
Moving often reflects life course changes such as employment.25,26 Frequent mobility, common among populations experiencing homelessness,27–29 may be an indication of uncertain economic and social resources. Unemployment and underemployment is a primary contributor to economic instability, yet marital relationships help to buffer the negative impact of job loss and underemployment.30 There is some indication that individuals of lower socioeconomic status face economic barriers to marriage31 and receive less social support due, in part, to increased chronic social stressors.32
Despite these indications that social stability domains are not isolated from one another and, in fact, tend to exacerbate one another, few have evaluated the ways in which the experiences cooccur and assessed the health impact of cumulative or multidimensional instability. In this study, latent class analysis was used to evaluate the possibility of underlying clusters among the hypothesized social stability indicators. This approach takes the intricate relationships among variables into account, yet removes the question of causal or confounding relationships among the indicators. Instead, it focuses on the broader question of whether the domains of housing, residential transition, employment, income level, incarceration, and partner relationship compose a single multidimensional construct with identifiable clusters.
Social stability and health
Reviews of public health concerns indicate that the set of common social stability domains are also frequent challenges to health issues as wide ranging as prevention of HIV,33,34 tuberculosis,35 and sexually transmitted diseases.36 While many of these social characteristics have been well investigated as independent determinants of health status, few have sought to understand the implications of the interrelationships among this set of characteristics. Thus, important additive or synergistic effects that may be addressed through social and structural interventions may be obscured.
This study draws on the history of social stability research to evaluate relationships among a set of commonly cited social stability indicators among a community-based sample of low income urban women. Additionally, this research evaluated the premise that underlying subgroups of social stability could be identified from observed indicator patterns and that these subgroups would be associated with health-related outcomes.
Methods
Data collection
Data were from female baseline participants in the CHAT project, a 2005–2007 randomized HIV behavioral intervention among women at sexual risk for HIV and their social network members in Baltimore, MD. Index participant recruitment utilized outreach and advertisement. Eligible index participants were 18–55-year-old Baltimore City residents who reported past six-month heterosexual activity, no drug injection, and any one of the following: multiple partners, partner with sexual risk, STI diagnosis in the past 6 months, or noninjection cocaine or heroin in the past 6 months. Network members were recruited by the index and were eligible if they were Baltimore City residents; age 18 or above; and were either sex partners, injection drug users, or someone the index was willing to talk to about HIV/STIs. Eight hundred eighteen individuals were enrolled (490 index, 329 network participants). The 635 female study participants with data for all stability indicators were included in this analysis.
Trained interviewers administered a comprehensive 120-minute behavioral and social network survey using computer-assisted personal interview (CAPI) and audio computer-assisted self-interviewing (ACASI) software at a community research site. Participants received $35 for completion of the baseline visit. Respondents completed consent procedures and provided consent prior to participation. All study protocols were approved by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board prior to implementation.
Measures
Dichotomous social stability indicators were identified based on review of existing social stability measures and operationalized to reflect stability in housing (no experiences of homelessness), criminal justice experiences (no jail, prison, or correctional facility), residential transition (no residential moves), income level (monthly income higher than median sample income of $500), and employment (no unemployment) during the prior six-month period. The resulting measure closely resembled one prior measure,10 with a few exceptions. Income level was measured instead of welfare status, in order to account for the variety of income sources among the population. Residential transition was also included in the measure as a life experience linked to stress, disruption, and uncertainty that may exacerbate other domains.37–39 Each indicator was limited to a consistent six-month period to reduce recall bias and create a measure of recent social stability. A cumulative variable was also calculated by summing all positive responses to the stability indicator variables (range 0–6).
Hypothesized social stability correlates included demographic variables (age above sample mean of 41, completed high school education, and race/ethnicity), indicators of social integration (live with child under 18, have a child under 18 not in household, parent in social network, and weekly religious service attendance), indicators of neighborhood integration (completely satisfied with block as a place to live, feeling part of your block vs. feeling your block is just a place to live, and past six-month participation in neighborhood organizations).
Hypothesized chronic health correlates were assessed with the following question for each type of illness: “Have you ever been told by a doctor or other health professional that you had _________?” Indicators of chronic illness were diabetes (or problems with your sugar), heart disease, asthma, pneumonia, hepatitis or liver disease, and hypertension, as well as a composite dichotomous indicator of any chronic illness diagnosis. Hypothesized mental health indicators were anxiety, bipolar disorder, depression, and schizophrenia. These were assessed by first asking, “Have you ever been told by a doctor or other health professional that you have a mental illness—for example, depression, schizophrenia, or serious trouble with your nerves?” Respondents who answered favorably were given a list of possible diagnoses and asked to select all that apply. Additionally, the 10-item Center for Epidemiologic Studies Depression Scale (CES-D) was used to assess depressive symptoms during the past week.40
Data analysis
Descriptive statistics were used to identify the prevalence of each stability characteristic and sociodemographic and health characteristics of the study population. Bivariate associations among variables were statistically assessed with Pearson’s Chi-square test and logistic regression using Generalized Estimating Equations to account for possible intercluster correlation effects within networks.41
Latent class analysis using Mplus version 5.0 was used to identify subgroups of individuals based on patterns of cooccurrence among stability variables. The employment indicator was not included in this stage of analysis due to low prevalence and covariance with housing and incarceration. Models were clustered by index and all missing data was imputed using the missing data function of Mplus. A series of models was fit with increasing numbers of classes until the model that best fit the data was identified. Substantive evaluation and model fit statistics, including the Bayesian Information Criteria (BIC),42 Akaike’s Information Criteria (AIC) and sample size adjusted BIC, were used to determine the most appropriate number of classes. Latent class regression was used to assess the association between latent class membership and each of the covariates using bivariate models and demographic adjusted models. In latent class regression, the latent class and regression component of the model are estimated together in order to reduce measurement error.43 Analysis included bivariate and adjusted models.
Results
Demographic characteristics
The sample was primarily African-American (96.1%) with a mean age of 41 years. About half (47.9%) had completed high school. The majority (66.9%) had used cocaine, crack, or heroin in the past 6 months, although few (8.0%) were injection drug users, and less than half (43.5%) reported daily drug use.
Stability characteristics
Table 1 describes stability characteristics of the sample and bivariate associations among them. Most had not been homeless (70.4%) or incarcerated (85.2%). Only about 10% reported no unemployment. About half earned more than $500 from all sources in an average month and only 13% earned more than $1000. The majority (61.6%) had not moved in the past 6 months and most (61.1%) had a main partner. On average, respondents reported approximately three stability characteristics (mean: 3.37, S.D.: 1.38). Very few reported having none (2.4%) or all (3.2%) characteristics; about half reported between three and four characteristics, although combinations varied.
Table 1.
Social stability characteristics among female CHAT participants, total sample and by characteristic (n = 635)
| Characteristican (%) | Total sample (n = 635) | No homelessness (n = 447) | No incarceration (n = 541) | No unemployment (n = 63) | Income > $500 (n = 307) | No residential move (n = 391) | Main partner (n = 388) |
|---|---|---|---|---|---|---|---|
| Homelessness | |||||||
| Yes | 188 (29.6) | 0 | 144 (26.6) | 4 (6.4) | 67 (21.8) | 56 (14.3) | 47 (24.7) |
| No | 447 (70.4) | 447 (70.4) | 397 (73.4)* | 59 (93.7)* | 240 (78.2)* | 335 (85.7)* | 292 (75.3)* |
| Incarceration | |||||||
| Yes | 94 (14.8) | 50 (11.2) | 0 | 6 (9.5) | 40 (13.0) | 46 (11.8) | 48 (12.4) |
| No | 541 (85.2) | 397 (88.8)* | 541 (85.2) | 57 (90.5) | 267 (87.0) | 345 (88.2)* | 340 (87.6)* |
| Unemployment | |||||||
| Yes | 572 (90.1) | 388 (86.8) | 484 (89.5) | 0 | 255 (90.1) | 343 (87.7) | 349 (90.5) |
| No | 63 (9.9) | 59 (13.2)* | 57 (10.5) | 63 (9.9) | 52 (16.9)* | 48 (12.3)* | 39 (10.05) |
| Monthly income | |||||||
| ≤$500 | 328 (51.7) | 207 (46.3) | 274 (50.7) | 11 (17.5) | 0 | 184 (47.1) | 189 (48.7) |
| >$500 | 307 (48.4) | 240 (53.7)* | 267 (49.4) | 52 (82.5)* | 307 (48.4) | 207 (52.9)* | 199 (51.3) |
| Residential move | |||||||
| Yes | 244 (38.4) | 112 (25.1) | 196 (36.2) | 15 (23.8) | 100 (32.6) | 0 | 135 (34.8) |
| No | 391 (61.6) | 335 (74.9)* | 345 (63.8)* | 48 (76.2)* | 207 (67.4)* | 391 (61.6) | 253 (65.2) |
| Partnership status | |||||||
| Single | 247 (38.9) | 155 (34.7) | 201 (37.2)* | 24 (38.1) | 108 (35.2) | 138 (35.3) | 0 |
| Main partner | 388 (61.1) | 292 (65.3)* | 340 (62.9) | 39 (61.9) | 199 (64.8) | 253 (64.7)* | 388 (61.1) |
| Total number of stability characteristics | |||||||
| 0 | 15 (2.4) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) |
| 1 | 52 (8.2) | 6 (1.3) | 32 (5.9) | 0 (0.0) | 2 (0.7) | 5 (1.3) | 7 (1.8) |
| 2 | 106 (16.7) | 30 (6.7) | 83 (15.3) | 2 (3.2) | 27 (8.8) | 22 (5.6) | 48 (12.4) |
| 3 | 132 (20.8) | 97 (21.7) | 111 (20.5) | 2 (3.2) | 48 (15.6) | 74 (18.9) | 64 (6.5) |
| 4 | 193 (30.4) | 177 (39.6) | 179 (33.1) | 7 (11.1) | 98 (31.9) | 162 (41.4) | 149 (38.4) |
| 5 | 117 (18.4) | 117 (26.2) | 116 (21.4) | 32 (50.8) | 112 (36.5) | 108 (27.6) | 100 (25.8) |
| 6 | 20 (3.2) | 20 (4.5)* | 20 (3.7)* | 20 (31.8)* | 20 (6.5)* | 20 (5.1)* | 20 (5.2)* |
aAll measures based on past 6 months duration
*p < 0.05
Many stability characteristics were significantly associated with one another in bivariate analysis and homelessness was significantly associated with each of the other stability characteristics. Each characteristic was associated with accumulated stability in other domains, providing an indication of validity for the included measures. Average number of characteristics associated with stability in other domains ranged from 3.5 (S.D.: 0.05) among those who were not incarcerated to 4.05 (S.D.: 0.05) among those who were employed. In contrast, average number of characteristics associated with instability in each domain ranged from 1.9 (S.D.: 0.08) among those reporting homelessness to 3.2 (S.D.: 0.05) among those who were unemployed.
Patterns among stability indicators
Table 2 shows the most common patterns of cooccurrence overall and for each characteristic. A total of 43 different patterns were observed, indicating considerable variation within groups that is not well captured by the mean number of characteristics. The most common pattern overall (14.6%) reflected primarily economic instability; individuals with this pattern lacked employment and were in the lower income group, but still reported stability in each of the other domains. The second most common pattern (13.4%) was one of stability in all domains except employment, suggesting income sources other than employment. Cluster columns show the three most common stability patterns stratified by presence or nonpresence of stability for each domain. For example, those with housing stability were most likely to also report stability in criminal justice, residential transition, and main partnerships. Those without housing stability (those who reported any homelessness) were most likely to report having a main partner and no stability in any other domain.
Table 2.
Patterns among social stability characteristics (SSC), by characteristic, among female CHAT participants, Baltimore (n = 635)
| Mean number of SSC (SD) | Most common SSC clusters H/CJ/E/I/RT/MP | Number of SSC | n | Mean number of SSC (SD) | Most common SSC clusters H/CJ/E/I/RT/MP | Number of SSC | n | |
|---|---|---|---|---|---|---|---|---|
| Stability domain | Stability | Instability | ||||||
| Housing | Housed (n = 447) | Any homelessness (n = 188) | ||||||
| 3.83 (0.04) | 110011 | 4 | 93 | 1.93 (0.08) | 010001 | 2 | 37 | |
| 110111 | 5 | 85 | 010000 | 1 | 32 | |||
| 110010 | 3 | 41 | 010101 | 3 | 16 | |||
| Criminal justice | No incarceration (n = 541) | Any incarceration (n = 94) | ||||||
| 3.49 (0.05) | 110011 | 4 | 93 | 1.96 (0.13) | 000000 | 0 | 15 | |
| 110111 | 5 | 85 | 100111 | 4 | 11 | |||
| 110010 | 3 | 41 | 100011 | 3 | 9 | |||
| Employment | Employed (n = 63) | Any unemployment (n = 572) | ||||||
| 4.05 (0.12) | 111111 | 6 | 20 | 3.18 (0.05) | 110011 | 4 | 93 | |
| 111110 | 5 | 17 | 110111 | 5 | 85 | |||
| 111101 | 5 | 9 | 110010 | 3 | 41 | |||
| Income | Monthly income > $500 (n = 307) | Monthly income ≤ $500 (n = 328) | ||||||
| 3.97 (0.05) | 110111 | 5 | 85 | 2.60 (0.07) | 110011 | 4 | 93 | |
| 110110 | 4 | 39 | 110010 | 3 | 41 | |||
| 110101 | 4 | 28 | 010001 | 2 | 37 | |||
| Residential transition | No residential move (n = 391) | Any residential move (n = 244) | ||||||
| 3.92 (0.05) | 110011 | 4 | 93 | 2.23 (0.07) | 010001 | 2 | 37 | |
| 110111 | 5 | 85 | 010000 | 1 | 32 | |||
| 110010 | 3 | 41 | 110101 | 4 | 28 | |||
| Partner | Main partnership (n = 388) | No main partnership (n = 247) | ||||||
| 3.79 (0.05) | 110011 | 4 | 93 | 2.44 (0.08) | 110010 | 3 | 41 | |
| 110111 | 5 | 85 | 110110 | 4 | 39 | |||
| 010001 | 2 | 37 | 010000 | 1 | 32 | |||
Cluster columns show the three most common stability patterns stratified by presence or nonpresence of stability for each domain. For each stability domain (H/CJ/E/I/RT/MP), 1 refers to “yes” and 0 refers to “no”, e.g., 111111 refers to stability in all domains and 000000 refers to instability in all domains.
All measures based on past six-month duration. All differences in means are significant at p < 0.05
Latent class analysis
Table 3 shows the results of the latent class analysis of stability characteristics. Three models were fit, each with an increasing number of classes, until the model that best fit the observed data was identified. The independence model with only one class did not fit well with the observed data, indicating that the items are not independent, supporting the hypothesis that social stability indicators are interrelated in latent subgroups. Models with two and three classes had satisfactory goodness-of-fit-based likelihood ratio tests. However, the two-class model had a lower AIC, BIC and sample size-adjusted BIC, indicating better fit. Additionally, the small third class of the three-class model resulted in multiple estimated item prevalences of 0 and 1.0, which may reflect nonconvergence in model estimation despite using multiple start values.
Table 3.
Latent class analysis of social stability characteristics among female CHAT participants, Baltimore (n = 635)
| 1-Class model | 2-Class modela | 3-Class model | ||||
|---|---|---|---|---|---|---|
| Class 1 | “High” class 1 | “Low” class 2 | “High” class 1 | “Mixed” class 2 | “Low” class 3 | |
| Characteristic | ||||||
| No homelessness | 0.70 | 0.92 | 0.18 | 1.00 | 0.55 | 0.22 |
| No incarceration | 0.85 | 0.90 | 0.73 | 0.92 | 0.81 | 0.75 |
| Income > $500 | 0.48 | 0.56 | 0.30 | 0.51 | 1.00 | 0.14 |
| No residential move | 0.62 | 0.79 | 0.19 | 0.85 | 0.50 | 0.23 |
| Main partner | 0.61 | 0.67 | 0.46 | 0.68 | 0.68 | 0.44 |
| Class size | 1.00 | 0.74 | 0.26 | 0.65 | 0.09 | 0.26 |
| Pearson Chi-square | 234 (p < 0.00) | 22 (p < 0.36) | 12 (p < 0.61) | |||
| Log-likelihood | −1939 | −1856 | −1851 | |||
| AIC | 3888 | 3733 | 3736 | |||
| BIC | 3911 | 3783 | 3812 | |||
| Sample size-adjusted BIC | 3895 | 3747 | 3758 | |||
| Entropy | – | 0.67 | 0.66 | |||
aBest model given fit statistics and class size
The two-class model identified one class with substantially higher probabilities of responding positively to all stability indicators. This class composed almost 75% of the sample and was referred to as the “high stability” class. In contrast, members of the “low stability” class composed approximately 25% and were more likely to report instability in each of the five indicators. Members of the “high stability” class had high probability of being housed (0.92), not being incarcerated (0.90), not moving (0.79), and having a steady partner (0.67) and higher income (0.56). Members of the “low stability” class were very unlikely to have remained housed (0.18), to have stayed in the same residence (0.19), or to be in the higher income bracket (0.30). They were also relatively likely to have not been incarcerated (0.73) and to have a main partner (0.46), although both probabilities were lower than the higher stability group.
A variety of sociodemographic characteristics were associated with latent class membership, as shown in Table 4. Bivariate and adjusted models indicated that membership in the high stability class was significantly correlated with having children in the household, not having any children who lived outside the household, having a parent in one’s social network, feeling a part of one’s block, and participating in neighborhood associations.
Table 4.
Association between sociodemographic characteristics and high social stability class membership among female participants in the CHAT study, Baltimore (n = 635)
| Characteristic | Total (n = 635) n (%) | UOR (95% CI) | p-Value | AORa (95% CI) | p-Value |
|---|---|---|---|---|---|
| Sociodemographic characteristics | |||||
| Age: 42 years or older | 339 (53.4) | 1.32 (.83, 2.09) | 0.245 | 1.33 (0.83, 2.12) | 0.233 |
| Education >12 years | 303 (47.9) | 1.43 (0.91, 2.25) | 0.701 | 1.41 (0.87, 2.29) | 0.163 |
| Race/ethnicity: African-American | 610 (96.1) | 2.04 (0.78, 5.30) | 0.144 | 1.94 (0.75, 5.06) | 0.174 |
| Drug use in past 6 months | 425 (66.9) | 0.58 (0.36, 0.92) | 0.021 | 0.48 (0.28, 0.83) | 0.008 |
| Social integration | |||||
| Live with child under 18 years old | 171 (26.9) | 4.66 (2.18, 10.0) | 0.000 | 5.94 (2.79, 12.6) | 0.000 |
| Child under 18 not living with you | 170 (26.8) | 0.38 (0.23, 0.61) | 0.000 | 0.46 (0.27, 0.78) | 0.004 |
| Parent in social network | 255 (40.2) | 1.86 (1.06, 3.27) | 0.030 | 2.12 (1.30, 3.45) | 0.012 |
| Weekly religious service attendance | 145 (22.8) | 0.81 (0.50, 1.34) | 0.426 | 0.75 (0.61, 1.63) | 0.249 |
| Neighborhood integration | |||||
| Completely satisfied with block | 135 (21.3) | 1.53 (0.82, 2.84) | 0.178 | 1.50 (0.81, 2.80) | 0.200 |
| Feel you are a part of your block | 185 (29.2) | 4.65 (2.66, 8.12) | 0.000 | 5.24 (2.63, 10.45) | 0.000 |
| Participate in neighborhood organizations | 93 (14.7) | 3.63 (1.33, 9.89) | 0.012 | 3.21 (1.33, 7.74) | 0.009 |
aAdjusted for age, education, race/ethnicity, and drug use in past 6 months
Table 5 shows health-related correlates of social stability membership. Unadjusted and adjusted models show that members of the high stability class were significantly less likely to have ever been diagnosed with a chronic disease and with heart disease. After adjustment for demographic characteristics, the high stability group was half as likely to have been diagnosed with any chronic illness and with heart disease compared to the low stability class. Higher stability class membership was significantly associated with lower likelihood of pneumonia in bivariate analysis, but the relationship was slightly attenuated in the presence of demographic variables. Mental health indicators were strongly correlated with stability class membership. In bivariate and adjusted analyses, membership in the high stability class was associated with diagnosis of mental illness, anxiety, bipolar disorder, and depression. The high stability group was approximately three times less likely to have received a mental health diagnosis and two to three times less likely to have been diagnosed with anxiety, bipolar disorder, depression, or high depressive symptoms.
Table 5.
Association between health characteristics and high social stability class membership among female participants in the CHAT study, Baltimore (n = 635)
| Characteristic | Total (n = 635) n (%) | UOR (95% CI) | p-Value | AORa (95% CI) | p-Value |
|---|---|---|---|---|---|
| Chronic illness | |||||
| Ever diagnosed with any chronic illness | 445 (70.1) | 0.52 (0.29, 0.94) | 0.031 | 0.54 (0.31, 0.94) | 0.030 |
| Diabetes | 62 (9.8) | 0.85 (0.41, 1.76) | 0.658 | 0.79 (0.37, 1.66) | 0.531 |
| Heart disease | 58 (9.1) | 0.52 (0.27, 0.97) | 0.039 | 0.43 (0.22, 0.86) | 0.016 |
| Asthma | 221 (34.8) | 0.67 (0.42, 1.06) | 0.670 | 0.74 (0.47, 1.18) | 0.206 |
| Pneumonia | 158 (24.9) | 0.62 (0.38, 0.99) | 0.048 | 0.64 (0.39, 1.05) | 0.080 |
| Hepatitis or liver disease | 147 (23.2) | 0.78 (0.45, 1.35) | 0.373 | 0.80 (0.59, 1.09) | 0.473 |
| Hypertension | 234 (36.9) | 0.77 (0.45, 1.35) | 0.252 | 0.67 (0.42, 1.05) | 0.081 |
| Mental health | |||||
| Ever diagnosed with mental illness | 322 (50.7) | 0.25 (0.16, 0.49) | 0.000 | 0.24 (0.15, 0.39) | 0.000 |
| Anxiety | 43 (6.8) | 0.38 (0.18, 0.80) | 0.011 | 0.26 (0.11, 0.60) | 0.002 |
| Bipolar disorder | 118 (18.6) | 0.32 (0.19, 0.54) | 0.000 | 0.32 (0.18, 0.58) | 0.000 |
| Depression | 198 (31.2) | 0.41 (0.25, 0.67) | 0.000 | 0.39 (0.24, 0.64) | 0.000 |
| PTSD | 10 (1.6) | 0.53 (0.13, 2.16) | 0.376 | 0.37 (0.05, 2.75) | 0.334 |
| Schizophrenia | 35 (5.5) | 0.46 (0.20, 1.04) | 0.063 | 0.48 (0.20, 1.13) | 0.094 |
| Depressive symptoms (CESD score >23) | 311 (49.0) | 0.34 (0.23, 0.51) | 0.000 | 0.35 (0.22, 0.57) | 0.000 |
aAdjusted for age, education, race/ethnicity, and drug use in past 6 months
Discussion
This study evaluated the validity of social stability as a multidimensional construct with implications for health. The selected characteristics appeared to represent the construct of social stability well. Relationships among indicators and associations with covariates occurred in expected ways, demonstrating internal construct validity. Latent class analysis results further supported the premise of a single social stability construct with identifiable underlying subgroups. Regression results support the hypothesis that social stability is associated with health-related outcomes and particularly show a strong relationship between social stability and mental health.
This study demonstrated that social stability characteristics tend to cooccur, but with considerable diversity among low income women living in an urban environment. Stability in any one dimension was associated with increasing cumulative social stability, indicating an additive effect. However, heterogeneity underlying the cumulative scores suggests that cumulative assessment of social stability may obscure important population diversity. Use of latent class analysis enabled the assessment to capture heterogeneity among indicators while also distinguishing categorical subgroups of stability.
There are remaining questions about the measure that should be addressed in future research. Rates of stability indicators in this sample were consistent with similar Baltimore samples.44 However, the proportion of homelessness and incarceration among those who were employed was sufficiently low as to preclude inclusion of the employment variable in latent class analysis. This finding is not surprising given research showing the protective effect of employment on incarceration23,45 and barriers to employment for homeless populations.14 It is unlikely that the latent class model results distinguishing between high and low stabilities would be substantially changed if employment were included, but a larger sample with more diversity may have enabled further exploration of a multiple class model.
The income variable was determined according to the sample median response. Although the results support use of the $500 per month cutoff as an indicator of economic instability, this should not be interpreted as in indicator of economic sufficiency; this relative scale may be better represented by another income cut-point in other populations. Nonetheless, these findings indicate that even among a low income group, higher income is positively associated with stability in other domains.
The distinction between classes in likelihood of responding positively was most substantial for the homelessness and residential transition characteristics. This supports a need for priority attention to housing concerns. The challenge of achieving life stability while experiencing homelessness15 and housing transitions39 has been well demonstrated; housing is also a common thread in prior investigations of social stability. Interventions based on the premise that housing should be assured as a first step to addressing other health and social challenges have been positively evaluated46 and offer promise as overall social stability interventions.
Although incarceration was more common in the low stability group compared to the higher stability group, the prevalence and distinction between classes was lower than that of other instability domains. Prior investigations have raised doubts about whether criminal justice measures should be included in social stability measures.9,47 The extent of cooccurrence with other indicators and distinction between high and low stability groups in this study supports inclusion of criminal justice in social stability measurement. The low rate of incarceration in this sample may reflect lower rates of incarceration among women.48 However, future research may consider a wider measurement time frame in order to fully capture the likelihood of recent incarceration.
Main partnership varied between the classes, which is consistent with research on steady partnerships and social support49,50 and early conceptualizations of social stability based on performance of appropriate societal roles.3,51 However, these findings suggest partnership status is a less substantial contributor to overall social stability in this population. In the context of poverty and uncertain circumstances, partnership may be viewed by women as an ideal but may in fact be less prioritized, bring further stress, or limit support from other sources.52 Further research should explore the role of partnerships in the context of overall experiences of instability.
These findings should be viewed in light of a few limitations. Data were primarily self-reported, although health risk behavior questions were self-administered. Responses may have been subject to recall bias, although the time period was restricted to the previous 6 months partly in order to enhance recollection. There is some risk of sampling bias due to recruitment and data collection time periods and locations, despite the variety of recruitment sources and network recruitment. This sample was composed primarily of African-American low income women living in an urban setting. Inquiry into the extent to which these findings are generalizable would contribute to further understanding of the social stability construct.
These findings show that social network characteristics and neighborhood integration are associated with social stability. Stability on an individual level likely reflects that of one’s social environment, which may further exacerbate personal vulnerability and challenge efforts to stabilize situations. Individuals experiencing instability may also benefit less from available support and resources. Further research should explore additional dimensions of social network influence. Women with children outside the household were more common in the lower stability group. Stress associated with instability may compromise personal coping and parenting. There may also be additional cofactors such as social resources and addiction that help to explain this relationship. Additionally, more stable respondents were more engaged with their neighborhood, which suggests a relationship between individual stability and social capital that should be further explored. Future research might also examine ways to conceptualize stability at multiple levels, including social network and neighborhood, in order to better understand multidimensional influences on individual stability and associated health problems.
The social stability characteristics evaluated here are commonly cited challenges for a variety of health problems. This research showed that the combined measure of social stability was significantly associated with chronic illness, particularly heart disease and pneumonia, and substantially associated with a variety of mental health diagnoses. Future investigation should explore the causal direction of these associations and evaluate potential pathways through which social stability may influence health. Although in some cases instability may actually reflect shifts towards more favorable circumstances (e.g., leaving an abusive partner or undesirable living situation), instability across multiple domains is likely to be accompanied by accumulated stress, which is exhausting, creates demands on one’s social and functional life, and contributes to further stressors.53 Increased chronic social stressors often challenge the ability to maintain support networks and mobilize them as necessary.32 Stressors are more likely to result in social network disruption among disadvantaged populations because the stressful events tend to be frequent, clustered together, and more difficult to control.54 The uncertainty and lack of control associated with instability may also make it difficult to anticipate needs and upcoming challenges. Future research should evaluate the extent to which personal, network, and neighborhood instabilities are related and explore examples of successful mobilization of social resources. Additionally, future investigation should evaluate the extent to which social stability is a barrier to medical care and preventive health behaviors.
On a practical level, these findings provide some insight into patterns of instability. This information may prove useful for program planners or policy makers who aim to understand and address the scope of challenges among low income women. In most cases, instability in any one domain was compounded by at least one other. This suggests a need for comprehensive policies and programs that address cooccurring social needs and help to mitigate the likely spiral effect of instability experiences. Social stability items may be useful as evaluation indicators.
Processes of societal stratification and structural determinants contribute to disparities across populations such that social determinants of health have disproportionate effects on health among those whose social position already sets them in a position of disadvantage.55,56 Processes of stress proliferation may further compound experiences of disadvantage, as stressful events are likely to precipitate further chronic stressors, whether immediately or over the life course.56 Although enhancing individuals’ ability to cope with uncertain circumstances through service provision and improved support systems is necessary to ameliorate some of the psychosocial consequences of social instability, attention must also turn to the structural origins to have a lasting impact on disparities.
Efforts to enhance social stability may take a variety of forms ranging from targeted efforts to address single dimensions to multifaceted approaches with multiple social outcomes. Addressing the structural origins of social stability will require commitment to a host of societal fundamentals, including improved education, job availability, incarceration, and reintegration policies, low income housing, drug treatment, neighborhood revitalization, transportation, and physical and mental health care services. Attention should also be directed to those structures that may specifically contribute to exacerbated relationships among social stability domains, such as employment and housing restrictions for ex-felons, availability of emergency housing assistance, and low threshold housing in locations that facilitate access to employment and income generation opportunities. Priority should be given to determining and implementing the programs and policies necessary to address the accumulation of social disadvantage that contributes to inequalities in health.
Acknowledgments
The National Institute of Mental Health provided financial support (grants R01 MH066810 and F31 MH073430). Additionally, the authors wish to thank the study participants and acknowledge the contributions of the Lighthouse data collection team as well as the valuable insight from David R. Holtgrave, David C. Celentano, and George W. Rebok.
References
- 1.Pearlin LI, Lieberman MA, Menaghan EG, Mullan JT. The stress process. J Heal Soc Behav. 1981;22(4):337–356. doi: 10.2307/2136676. [DOI] [PubMed] [Google Scholar]
- 2.Compas BE. Stress and life events during childhood and adolescence. Clin Psychol Rev. 1987;7(3):275–302. doi: 10.1016/0272-7358(87)90037-7. [DOI] [Google Scholar]
- 3.Straus R, Bacon SD. Alcoholism and social stability; a study of occupational integration in 2,023 male clinic patients. Q J Stud Alcohol. 1951;12(2):231–260. [PubMed] [Google Scholar]
- 4.Gerard DL, Saenger G. Out-patient treatment of alcoholism: a study of outcome and its determinants. Toronto, Canada: University of Toronto Press; 1966. [Google Scholar]
- 5.Hansson SO, Helgesson G. What is stability? Synthese. 2003;136(2):219–235. doi: 10.1023/A:1024733424521. [DOI] [Google Scholar]
- 6.Ware NC, Wyatt MA, Tugenberg T. Adherence, stereotyping and unequal HIV treatment for active users of illegal drugs. Soc Sci Med. 2005;61(3):565–576. doi: 10.1016/j.socscimed.2004.12.015. [DOI] [PubMed] [Google Scholar]
- 7.Angel RJ, Lein L, Henrici J. Poor families in America's health care crisis. New York, NY: Cambridge University Press; 2006. [Google Scholar]
- 8.German D. Social Stability and HIV Risk: Exploring the Role of Multidimensional Social Vulnerability [doctoral dissertation]. Baltimore, MD: Johns Hopkins University; 2009. Available from Dissertations and Theses Database (UMI No. AAT 3357168).
- 9.Gibbs LE. A classification of alcoholics relevant to type-specific treatment. Int J Addict. 1980;15(4):461–488. doi: 10.3109/10826088009040031. [DOI] [PubMed] [Google Scholar]
- 10.Bouhnik AD, Chesney M, Carrieri P et al. Nonadherence among HIV-infected injecting drug users: the impact of social instability. J Acquir Immune Defic Syndr. 2002; 31(Suppl 3): S149–53. [DOI] [PubMed]
- 11.Zlotnick C, Tam T, Robertson MJ. Disaffiliation, substance use, and exiting homelessness. Subst Use Misuse. 2003;38(3–6):577–599. doi: 10.1081/JA-120017386. [DOI] [PubMed] [Google Scholar]
- 12.Zuvekas SH, Hill SC. Income and employment among homeless people: the role of mental health, health and substance abuse. J Ment Health Policy Econ. 2000;3(3):153–163. doi: 10.1002/mhp.94. [DOI] [PubMed] [Google Scholar]
- 13.Coumans M, Spreen M. Drug use and the role of homelessness in the process of marginalization. Subst Use Misuse. 2003;38(3–6):311–338. doi: 10.1081/JA-120017376. [DOI] [PubMed] [Google Scholar]
- 14.Zuvekas SH, Hill SC. Income and employment among homeless people: the role of mental health, health and substance abuse. J Ment Health Policy Econ. 2000;3(3):153–163. doi: 10.1002/mhp.94. [DOI] [PubMed] [Google Scholar]
- 15.Morrell JP. Voice from the street: truths about homelessness from sisters of the road. Portland, Oregon: Grey Sunshine; 2007. [Google Scholar]
- 16.Clear TR. The effects of high imprisonment rates on communities. Crime Justice. 2008;37(97):97–132. [Google Scholar]
- 17.Fagan J. Youth Gangs, Drugs, and Socioeconomic Isolation. In: Flannery DJ, Huff CR, eds. Youth Violence: Prevention, Intervention, and Social Policy. Washington, DC: American Psychiatric Press Inc; 1999. pp. 145-170.
- 18.Miller M, Neaigus A. An economy of risk: resource acquisition strategies of inner city women who use drugs. Int J Drug Policy. 2002;13(5):409–418. doi: 10.1016/S0955-3959(02)00119-6. [DOI] [Google Scholar]
- 19.Bourgois P. In search of respect: selling crack in El Barrio. New York, NY: Cambridge University Press; 1995. [Google Scholar]
- 20.Lopoo LM, Western B. Incarceration and the formation and stability of marital unions. J Marriage Fam. 2005;67(3):721–734. doi: 10.1111/j.1741-3737.2005.00165.x. [DOI] [Google Scholar]
- 21.Center for Poverty Solutions. Barriers to stability: homelessness and incarcerations's revolving door in Baltimore City. 2002. http://www.soros.org/initiatives/baltimore/articles_publications/publications/barriersstability. Accessed 13 Feb 2010.
- 22.Blankenship KM, Smoyer AB, Bray SJ, Mattocks K. Black–White disparities in HIV/AIDS: the role of drug policy and the corrections system. J Health Care Poor Underserved. 2005;16(4):140–156. doi: 10.1353/hpu.2005.0110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Freudenberg N, Daniels J, Crum M, Perkins T, Richie BE. Coming home from jail: the social and health consequences of community reentry for women, male adolescents, and their families and communities. Am J Public Health. 2005;95(10):1725–1736. doi: 10.2105/AJPH.2004.056325. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Olphen J, Freudenberg N, Fortin P, Galea S. Community reentry: perceptions of people with substance use problems returning home from New York City jails. J Urban Health. 2006;83(3):372–381. doi: 10.1007/s11524-006-9047-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Clark WAV. Intervening in the residential mobility process: neighborhood outcomes for low-income populations. Proc Natl Acad Sci U S A. 2005;102(43):15307–15312. doi: 10.1073/pnas.0507308102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Shumaker SA, Stokols D. Residential mobility as a social issue and research topic (USA) J Soc Issues. 1982;38(3):1–19. doi: 10.1111/j.1540-4560.1982.tb01767.x. [DOI] [Google Scholar]
- 27.Tomas A, Dittmar H. The experience of homeless women: an exploration of housing histories and the meaning of home. Hous Stud. 1995;10(4):493–515. doi: 10.1080/02673039508720834. [DOI] [Google Scholar]
- 28.Sosin M, Piliavin I, Westerfelt H. Toward a longitudinal analysis of homelessness. J Soc Issues. 1990;46:157–174. doi: 10.1111/j.1540-4560.1990.tb01804.x. [DOI] [Google Scholar]
- 29.Weitzman B, Knickman J, Shinn M. Pathways to homelessness among New York City families. J Soc Issues. 1990;46:125–140. doi: 10.1111/j.1540-4560.1990.tb01802.x. [DOI] [Google Scholar]
- 30.Dooley D, Prause J, Ham-Rowbottom KA. Underemployment and depression: longitudinal relationships. J Heal Soc Behav. 2000;41(4):421–436. doi: 10.2307/2676295. [DOI] [PubMed] [Google Scholar]
- 31.Edin K, Reed JM. Why don't they just get married? Barriers to marriage among the disadvantaged. Futur Child. 2005;15(2):117–137. doi: 10.1353/foc.2005.0017. [DOI] [PubMed] [Google Scholar]
- 32.Mickelson KD, Kubzansky LD. Social distribution of social support: the mediating role of life events. Am J Community Psychol. 2003;32(3–4):265–281. doi: 10.1023/B:AJCP.0000004747.99099.7e. [DOI] [PubMed] [Google Scholar]
- 33.Des Jarlais DC, Friedman SR. Fifteen years of research on preventing HIV infection among injecting drug users: what we have learned, what we have not learned, what we have done, what we have not done. Public Health Rep. 1998;113(1):182–188. [PMC free article] [PubMed] [Google Scholar]
- 34.Peretti-Watel P, Spire B, Schiltz MA, Bouhnik AD, Heard I, Lert F, et al. Vulnerability, unsafe sex and non-adherence to HAART: evidence from a large sample of French HIV/AIDS outpatients. Soc Sci Med. 2006;62(10):2420–2433. doi: 10.1016/j.socscimed.2005.10.020. [DOI] [PubMed] [Google Scholar]
- 35.Sumartojo E. When tuberculosis treatment fails. A social behavioral account of patient adherence. Am Rev Respir Dis. 1993;147(5):1311–1320. doi: 10.1164/ajrccm/147.5.1311. [DOI] [PubMed] [Google Scholar]
- 36.Farley TA. Sexually transmitted diseases in the Southeastern United States: location, race, and social context. Sex Transm Dis. 2006;33(7 Suppl):S58–S64. doi: 10.1097/01.olq.0000175378.20009.5a. [DOI] [PubMed] [Google Scholar]
- 37.Magdol L. Is moving gendered? The effects of residential mobility on the psychological well-being of men and women. Sex Roles. 2002;47(11):553–560. doi: 10.1023/A:1022025905755. [DOI] [Google Scholar]
- 38.German D, Davey-Rothwell MA, Latkin CA. Residential transience and HIV risk behavior among injection drug users. AIDS Behav. 2007;11(6 Suppl):21–30. doi: 10.1007/s10461-007-9238-3. [DOI] [PubMed] [Google Scholar]
- 39.Winstanley A, Thorns DC, Perkins HV. Moving house, creating home: exploring residential mobility. Hous Stud. 2002;17(6):813–832. doi: 10.1080/02673030216000. [DOI] [Google Scholar]
- 40.Radloff LS. The CES-D scale: A self-report depression scale for research in the general population. Appl Psychol Meas. 1977;1:385–401. doi: 10.1177/014662167700100306. [DOI] [Google Scholar]
- 41.Zeger SL, Liang KY, Albert PS. Models for longitudinal data: a generalized estimating equation approach. Biometrics. 1988;44(4):1049–1060. doi: 10.2307/2531734. [DOI] [PubMed] [Google Scholar]
- 42.Wiedenfeld SA, O'Leary A, Bandura A, Brown S, Levine S, Raska K. Impact of perceived self-efficacy in coping with stressors on components of the immune system. J Pers Soc Psychol. 1990;59(5):1082–1094. doi: 10.1037/0022-3514.59.5.1082. [DOI] [PubMed] [Google Scholar]
- 43.Bandeen-Roche K, Miglioretti DL, Zeger SL, Rathouz PJ. Latent variable regression for multiple discrete outcomes. JASA. 1997;92(440):1375–1386. [Google Scholar]
- 44.Latkin CA, Knowlton AR, Sherman S. Routes of drug administration, differential affiliation, and lifestyle stability among cocaine and opiate users: implications to HIV prevention. J Subst Abuse. 2001;13(1–2):89–102. doi: 10.1016/S0899-3289(01)00070-0. [DOI] [PubMed] [Google Scholar]
- 45.Zanis DA, Metzger DS, McLellan AT. Factors associated with employment among methadone patients. J Subst Abuse Treat. 1994;11(5):443–447. doi: 10.1016/0740-5472(94)90097-3. [DOI] [PubMed] [Google Scholar]
- 46.Tsemberis S, Gulcur L, Nakae M. Housing first, consumer choice, and harm reduction for homeless individuals with a dual diagnosis. Am J Public Health. 2004;94(4):651–656. doi: 10.2105/AJPH.94.4.651. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Skinner WF, Massey JL, Krohn MD, Lauer RM. Social influences and constraints on the initiation and cessation of adolescent tobacco use. J Behav Med. 1985;8(4):353–376. doi: 10.1007/BF00848369. [DOI] [PubMed] [Google Scholar]
- 48.Pew Center on the States. One in 100: behind Bars in America 2008. Washington, D.C.: Pew Charitable Trust; 2008.
- 49.House JS, Landis KR, Umberson D. Social relationships and health. Science. 1988;241(4865):540–545. doi: 10.1126/science.3399889. [DOI] [PubMed] [Google Scholar]
- 50.Turner RJ, Marino F. Social support and social structure: a descriptive epidemiology. J Heal Soc Behav. 1994;35(3):193–212. doi: 10.2307/2137276. [DOI] [PubMed] [Google Scholar]
- 51.Glaser B, Strauss A. The discovery of grounded theory. Chicago, IL:: Aldine; 1967. [Google Scholar]
- 52.Tsui E, Leonard L, Laird C, Ellen JE. Poverty and sexual concurrency: a case study of STI risk. J Health Care Poor Underserved. 2008;19(3):758–777. doi: 10.1353/hpu.0.0055. [DOI] [PubMed] [Google Scholar]
- 53.Pearlin LI, Aneshensel CS, Leblanc AJ. The forms and mechanisms of stress proliferation: the case of AIDS caregivers. J Heal Soc Behav. 1997;38(3):223–236. doi: 10.2307/2955368. [DOI] [PubMed] [Google Scholar]
- 54.Klovdahl AS, Potterat JJ, Woodhouse DE, Muth JB, Muth SQ, Darrow WW. Social networks and infectious disease: the Colorado Springs Study. Soc Sci Med. 1994;38(1):79–88. doi: 10.1016/0277-9536(94)90302-6. [DOI] [PubMed] [Google Scholar]
- 55.Braveman P, Gruskin S. Poverty, equity, human rights and health. Bull World Health Organ. 2003;81(7):539–545. [PMC free article] [PubMed] [Google Scholar]
- 56.Baker EA, Metzler MM, Galea S. Addressing social determinants of health inequities: learning from doing. Am J Public Health. 2005;95(4):553–555. doi: 10.2105/AJPH.2005.061812. [DOI] [PMC free article] [PubMed] [Google Scholar]