Validation of the DSSI-10 to Measure Social Support in a General Population

Abstract

Objective: The purpose of this study was to validate the 10-item DSSI as a brief measure of social support for use in diverse adult populations. Methods: EFA was performed on 2010 Arizona Health Survey (AHS) data (n = 8215). Confirmatory Factor Analysis (CFA) then confirmed the factors structure by gender, ethnicity, and age, as well as for the total population. DSSI-10 and subscales were compared with variables related to social support. Results: CFI confirms this structure exhibits a good model fit. Low self-reported health status and low self-reported quality of life were related to lower DSSI scores. Living alone was significantly negatively related to the DSSI-10. Conclusions: Researchers may confidently use DSSI-10 to measure social support for diverse adult populations. This instrument can be used in epidemiological studies to increase understanding of mental and physical health in relationship to social supports in the general population.

Keywords

health care field of practice mental health quantitative methodological article adults population factor analysis psychometric study reliability study

People are social beings living in interconnected communities, a context with an effect on both physical and mental health (Smith & Christakis, 2008). Coleman (1990) suggested that social relationships are a form of capital and these social relationships influence health in a variety of ways including access to health-related information and emotional well-being (Wellman & Berkowitz, 1988). Much evidence exists that social support is an important factor in physical health, mental health, and general well-being (Cohen, Gottlieb, & Underwood, 2000; House, Landis & Umberson, 1988; Koenig, George, & Titus, 2004; Pachana, Smith, Watson, McLaughlin, & Dobson, 2008). Moreover, Cohen and Willis (1985) argued that social support provides a buffering effect against stress and depression as well as supportive influences during times of need. Together this suggests that social support is an important mechanism that may enhance positive coping during stressful times and moderate the effects of stress on both physical and mental health (Smith & Christakis, 2008). But to better understand these effects, there remains a need for an acceptable, brief, and valid measure of social support in the general adult population (age 18 and older); thus this study aims to evaluate the responsiveness of the 10-item Duke Social Support Index (DSSI) and its subscales in a diverse adult population.

McDowell and Newell (1996) defined social support as “the availability of people whom the individual trusts, on whom he can rely, and who make him feel cared for and valued as a person” (p. 125) or in more simple terms, social interaction and social satisfaction. Smith and Christakis (2008) suggest that social networks can be conceptualized as the number of social contacts a person has and the person’s perception of how well those contacts provide social support. Thus, social support scales should include measurement of the number and quality of social relationships as a basic assessment of social support (McDowell & Newell, 1996).

The authors were involved in two simultaneous efforts in Arizona to measure this important health-related factor of social support. This concept and related programming emphasizes a strengths based, resiliency approach and provides another positive measurable characteristic that can be addressed as part of health and wellness strategies. As consultants to the Arizona Health Survey (AHS), the authors suggested the DSSI because it addressed the construct, was brief enough to add to an already lengthy survey, replaced several miscellaneous nonvalidated questions in the 2008 survey, and provided the opportunity to validate the measure with an adult population.

Other measures were considered but determined to measure social support more in the context of mental health as opposed to the context of health prevention. The Centers for Disease Control and Prevention Behavioral Risk Factor Surveillance System asked only general context questions (health status/stress) or social and emotional support from any source or general life satisfaction. The Rosenberg Self-Esteem scale, the Connor-Davidson Resilience scale, and the WHO (Five) Well-Being Index, do take the resiliency perspective but froman individual viewpoint rather than in the wider social context.

The 35-item DSSI was originally designed to assess subjective social support in elderly populations. Landerman, George, Campbell, and Blazer, (1989) used principle component analysis (PCA) with iteration and varimax rotation to confirm five factors or dimensions of social support: social network size (5 items), social interaction (4 items), social satisfaction (10 items), instrumental support (13 items), and 3 additional items. The three additional items were (1) Are you satisfied with how often you see your friends and relatives? (2) Is there at least one person with whom you have a close, lasting relationship? and (3) Are you presently married or currently living with someone as though married? requiring a yes/no response (Landerman, George, Campbell, & Blazer, 1989; Table 1). Only items with factor loadings ≥ .40 were included in the five subscales. The five factors had eigenvalues >1 and explained 82% of the variance among the items (Landerman et al., 1989).

Table 1.

Summary of the Questions Included in Each Version of the DSSI

	DSSI items
Q# in original scale	35	23	11	10	Questions in scales
					Social Network Size
1	X				Number of parents and grandparents who live within 1 hr’s travel
2	X				Number of brothers and sisters who live within 1 hr’s travel
3	X				Number of children who live within 1 hr’s travel
8	X				Amount of time spent talking with other people at work or school
34	X				Household size
					Social Interaction
4	X	X	X	7	Number of family members within 1 hr that you can depend on or feel close to
5	X	X	X	8	Number of times past week spent with someone not living with you
6	X	X	X	9	Number of times in past week talked with friends/relatives on the telephone
7	X	X	X	10	Number of times in the past week attended meetings of clubs, religious groups, or other groups that you belong to (other than work)
					Social Satisfaction
10	X				How often do you feel lonely?
11	X	X	X	2	Do family and friends understand you?
13	X	X	X	3	Do you feel useful to family and friends?
14	X	X	X	5	Do you know what’s happening with family and friends?
15	X	X	X	4	Do you feel listened to by family and friends?
16	X	X	X	1	Do you feel you have a definite role in the family and among friends?
17	X				Can you count on family and friends in times of trouble?
18	X	X	X	6	Can you talk about your deepest problems?
19	X	X	X		How satisfied are you with your relationships with family and friends?
33	X				Do you need additional help?
					Instrumental Support
					Does family or friends ever help in any of the following ways? (response yes/no)
20	X	X			When you are sick?
21	X	X			Shop or run errands?
22	X	X			Give you gifts?
23	X	X			Help you out with money?
24	X	X			Fix things around your home?
25	X	X			Keep house for you or do household chores?
26	X	X			Give you advice on business or financial matters?
27	X	X			Provide companionship to you?
28	X	X			Listen to your problems?
29	X	X			Give you advice on dealing with life’s problems?
30	X	X			Provide transportation for you?
31	X	X			Prepare or provide meals for you?
32	X				Help take care of small children?
					Other items
9	X				Are you satisfied with how often you see your friends and relatives?
12	X				Is there at least one person with whom you have a close, lasting relationship?
35	X				Are you presently married or currently living with someone as though married?

Note. Duke Social Support Index taken from Koenig et al. (1993) and Landerman, George, Campbell, and Blazer (1989).

The validation of the abbreviated 23-item version of the DSSI was conducted using cross-sectional data from Wave II (1983–1984) of the National Institute of Mental Health Epidemiologic Catchment Area survey in a sample of 2,954 community-dwelling adults including a sample of 1,294 adults over age 60 (Koenig et al., 1993). The 5-item social network subscale was removed as it was weakly related to psychological symptoms and the size of the social network was deemed inconsequential. The abbreviated 23-item scale included a 12-item instrumental support subscale, 7-item social satisfaction subscale, and a 4-item social interaction subscale (Table 1). The item, Do you need help taking care of small children? was removed from the instrumental support scale due to a very low response rate, especially for elderly individuals. The instrumental support subscale asked about the presence or absence (yes/no) of instrumental assistance using the stem of, Do family or friends ever help in any of the following ways: (1) Help out when you are sick? or (2) Listen to your problems? This subscale was dropped as instrumental support may not be as relevant for community-dwelling adults as it is for those people who suffer from a chronic physical illness. Three questions from the original 35-item scale were eliminated from the social satisfaction subscale (1) How often do you feel lonely? (2) Can you count on family and friends in times of trouble? and (3) Do you need some additional help? All four social interaction items from the original 35-item scale were retained because they were related to both major depression and physical health (George, Blazer, Hughes, & Fowler, 1989; House, Landis, & Umberson, 1988). The inclusion of this subscale was further supported by evidence that having more social contacts moderate physical and mental illness in elderly people (Arling, 1987; Revicki & Mitchell, 1990).

Factor analysis using orthogonal and oblique rotations were conducted on the 23-item scale to determine whether the original items would reproduce in this new sample (Koenig et al., 1993). The 7 items in the social satisfaction subscale had factor coefficients of > .45. Analysis of the social satisfaction subscale in various populations revealed Cronbach’s α for the entire sample on the 10 items (.77) and 7 items (.75). Further analysis in subgroups revealed adequate Cronbach’s α in 10-item and 7-item subscales (respectively) for healthy young (.78 and .76), sick young (.83 and .81), and healthy elderly (.76 and .75). The 10-item and 7-item social satisfaction subscales were similarly analyzed in a sample of 409 chronically ill elderly adults and produced Cronbach’s α of .71 for both versions (Koenig et al., 1993). To further reduce the scale, the 3 items that did not factor with any of the subscales were eliminated; these items are noted in Table 1.

The 12-item instrumental support subscale registered little variation in responses and was only weakly associated with depression or health outcomes (Koenig et al., 1993). Therefore, this 12-item instrumental support subscale was dropped when creating the 11-item abbreviated index (see Table 1). Koenig et al. (1993) recommend conceptualizing the 11-item version as a single scale composed of two dimensions of social support; subjective support (7 items) and social interaction (4 items). It is not advisable to use each subscale as a separate measurement of social support.

The abbreviated 11-item DSSI has been primarily used to assess social support for elderly populations (Koenig & George, 1998; Koenig et al., 2004; Powers, Goodger, & Byles, 2004). In a large longitudinal survey of older women, Pachana, Smith, Watson, McLaughlin, and Dobson (2008) eliminated one question from the scale, How satisfied are you with relationships with family and friends? and reported that this DSSI 10-item scale remained responsive to changes in the lives of older women. Cronbach’s α were .60 for the social interaction subscale and .80 for the social satisfaction subscale. Pachana et al. (2008) concluded that even though the subscales of the DSSI may not address the complexity of social support, they are useful in community-based epidemiological studies.

Goodger, Byles, Higganbotham, and Mishra (1999) used PCA to assess the 11-item DSSI in an elderly Australian population (n = 298) of adults 70 years and older (Cronbach’s α = .77). Goodger et al. (1999) found that the 11-item DSSI was supported for use in health promotion strategies and was useful as a brief measure of support. Powers et al. (2004) conducted exploratory factor analysis (EFA) and PCA on 10 items of the DSSI in elderly (70 years and older) Australian women (n = 12,223). The Cronbach’s α were .58 for the social interaction subscale and .80 for the social satisfaction subscale.

In addition, several studies found significant relationships between social support and living alone, general health status, and perception of quality of life (Cohen et al., 2000; House et al., 1988; Koenig et al., 2004). Social support was negatively correlated with living alone, poorer health status, and lower quality of life (Goodger et al., 1999; Powers et al., 2004).

Recently, the 11-item abbreviated DSSI was used with younger populations. Ebeling-Witte, Frank, and Lester (2007) used the abbreviated DSSI to measure social support in 88 undergraduate students. Poleshuck, Giles, and Tu (2006) utilized the 11-item DSSI to measure social support in a low-income, urban, African American female population. Internal reliability was not reported for either of these studies, however, so whether or not the DSSI-11 scale was an adequate measure of social support in these studies remains unknown.

Social support is important for people of all ages and ethnicities as a factor related to both physical and mental health (Koenig et al., 2004; Smith & Christakis, 2008). Because there remains a need for an acceptable, brief, and valid measure of social support in diverse adult populations, this study aimed to evaluate the responsiveness of a 10-item DSSI and its subscales in a large, diverse, adult population.

Method

Institutional Review Board approval was obtained for this secondary data analysis. The 2010 AHS data were collected through telephone interviews of 8,215 community-dwelling adult heads of household living in Arizona (18 years and older). A professional research service (Westat, Inc., Rockville, Maryland.) administered the telephone survey using Random Digit Dialing, a procedure that excludes businesses while including unlisted residential telephone numbers. Data were weighted to represent the population demographics of Arizona and to allow for generalization across a statewide context.

The weighted sample was comprised of 49.7% males and 50.3% females. Ethnicity was self-identified as Caucasian (67.3%), Hispanic (22%), African American (4%), Native American (4%), and Asian Pacific Islander (2%).

Measures

The main targets for analysis were the 10-item DSSI (see Table 1) as well as the two subscales of (a) social satisfaction and (b) social interaction. There was one question dropped from the DSSI-11 as it was not measuring a particular construct related to social satisfaction. The question How satisfied are you with your relationships with family and friends? was omitted at the discretion of the survey administrator. While Koenig et al. (1993) conceptualized the scale as one factor, two subscales can be calculated through factorial analysis. The social satisfaction subscale has 6 items while the social interaction subscale is comprised of 4 items.

The social interaction subscale asks questions regarding the number of social interactions the person had within the past week. For example, How many times during the past week did you spend time with someone who does not live with you? The social satisfaction subscale asks about the subjective quality of those relationships. For example, When you are talking with your family or friends, do you feel you are being listened to? Thus, the total DSSI-10 is a combination of a number of social interactions along with the quality of those relationships, with social satisfaction reverse scored before summation. Therefore, higher scores indicate a higher level of social support.

The total score for the DSSI-10 ranges from 10 to 30 (M = 24.39, SD = 3.28), with higher scores indicating a stronger perception of social support. The social satisfaction subscale ranges from 6 to 18 (M = 16.05, SD = 2.36) and the social interaction scale ranges from 4 to 12 (M = 8.34, SD = 1.62).

Many studies found significant relationships between social support and living alone, health status, and quality of life (Cohen et al., 2000; Goodger et al., 1999; House et al., 1988; Koenig et al., 2004; Powers et al., 2004). Therefore, measures used in the bivariate correlation and linear regression analysis to confirm construct validity included living alone = 0 or living with others = 1, self-reported general health status (1 = poor, 2 = fair, 3 = good, 4 = very good, and 5 = excellent), and quality of life (1 = poor, 2 = fair, 3 = good, 4 = very good, and 5 = excellent).

Data Analysis

The DSSI-11 was derived using EFA by Koenig et al. (2004); in this analysis all of the variables were assumed to be continuous. EFA is a data-reduction method where the covariance structure is unspecified. EFA, which is based on eigenvalues and vectors of the covariance structure, was used to determine factor pairings. Once EFA was performed, confirmatory factor analysis (CFA) then confirmed the factors structure. The 10-item DSSI was analyzed using EFA assuming all items are of continuous measurement.

This 10-item DSSI factor structure was previously confirmed on the 2010 AHS using principal component analysis (PCA). The results for PCA and CFA should be comparable, since the factoring results for both are based on the eigenvector values of the covariance matrix. To check these results, the sample was randomly separated and the EFA was performed on one part of the data and the CFA on the other part. Both analyses confirmed the factor structuring for 10-item DSSI for the 2010 AHS data. For further verification of the factor structure with this data, CFAs were performed using the 10-item DSSI for the population by gender, ethnicity, and age group (39 and under, 70 and older) as well as for the total population. The results of fit for the subgroups are displayed in Table 2 and the best fit model for the total population is displayed in Table 3. MPLUS 5 was used to perform EFA and CFA using the standard defaults corresponding with each method. For example, to handle missing data the full information maximum likelihood method was used to handle missing data within MPLUS. This was an acceptable choice since the data were missing at random. Also, an orthogonal rotation was used for the EFA.

Table 2.

Goodness-of-Fit for the Confirmatory Factor Analysis Models

Statistical test	Male (n = 2,838; two factors)	Female (n = 5,212; two factors)	Under 39 (n = 1,328; two factors)	Over 70 (n = 2,313; two factors)	Hispanic (n = 1,483; two factors)
CFI	.951	.988	.977	.943	.939
TLI	.935	.984	.970	.925	.919
RMSEA	.037	.018	.033	.031	.037

Note. CFI, Comparative Fit Index, acceptable: >.90; TLI, Tucker–Lewis Index, acceptable: >.90; RMSEA, root mean-square error of approximation, acceptable <.05.

Table 3.

Confirmatory Factor Analysis Loadings for the Total Population

	CFA
	Social Satisfaction	Social Interaction
DSSI-1	.514	0
DSSI-2	.560	0
DSSI-3	.495	0
DSSI-4	.540	0
DSSI-5	.509	0
DSSI-6	.421	0
DSSI-7	0	.562
DSSI-8	0	.447
DSSI-9	0	.458
DSSI-10	0	.395

Note. CFA, confirmatory factor analysis; DSSI, Duke Social Support Index.

Results

Of the 8,215 participants surveyed, 8,003 (97%) completed the entire 10-item DSSI scale. As anticipated, factor analyses revealed two factors: social satisfaction (6 items) and social interaction (4 items). The first factor (social satisfaction) relates to the degree of satisfaction with social support and the second factor (social interaction) relates to the number of people included in social interactions as confirmed by the EFA and CFA (Table 2). Acceptable fit was observed, with the female subgroup factor structure presenting the best goodness of fit, most likely due to the large sample (Table 2). This factor structure was used for the total population leading to the results displayed in Table 3 where the comparative fit index was .984, Tucker–Lewis index was .979, and the root mean square error of approximation was .021. All of these values were in the range such that this structure exhibits a good model fit (Hu & Bentler, 1995). However, the assumption that all measures were continuous is not valid, since factors 7–10 were categorical. To appropriately validate the scale in this instance as well as the previous examples, items in the social satisfaction subscale must be recoded into categorical variables so that the model measures match in the analysis, which is further discussed under the limitations section.

Bivariate correlation was conducted using the variables included in the multiple linear regression model (Table 4). The social interaction subscale and the social satisfaction subscale were significantly positively correlated (.32, p < .01). General health status and quality of life were significantly negatively related to the DSSI-10 and subscales (p < .01). This demonstrates that low self-reported health status and low self-reported quality of life were related to lower levels of social support. Living alone was significantly negatively related to the overall DSSI scale and the social satisfaction subscale (p < .01), but not significantly related to social interaction. However, the direction of correlation was consistent with the findings by Cohen et al. (2000).

Table 4.

Bivariate Correlations for Variables Included in the Regression Model

	1	2	3	4	5	6
1. Total DSSI-10	–
2. Social satisfaction	.88**	–
3. Social interaction	.73**	.32**	–
4. Living alone	–.28	–.05**	.01	–
5. Health status	–.23**	–.19**	–.18**	.06**	–
6. Quality of life	–.31**	–.29**	–.21**	.03	.44**	–

Note. DSSI, Duke Social Support Index.

**p < .01.

To assess construct validity, scores for the overall DSSI scale and each subscale were compared with variables known to be related to social support: (1) living alone; (2) general health status; and (3) quality of life. The multiple linear regression model consisted of these three independent factors (Table 5). Positive correlations were hypothesized with reported health status and quality of life scores, as higher reported health status would be correlated with higher social support. Living alone was hypothesized to be related to lower DSSI scores.

Table 5.

Multiple Linear Regression to Assess Validity and Reliability of the 9-Item DSSI in a General Population

	Total scale		Social satisfaction		Social interaction
	β	SE	β	SE	β	SE
Construct Validity
Living alone	–.02	.11	–.05***	.08	.03*	.06
Health status	–.09***	.03	–.08***	.02	–.07***	.02
Quality of life	–.32***	.03	–.30***	.03	–.21***	.02
Variance explained (R ², %)	13.5		12		6
Reliability
(Cronbach’s α)	.74		.77		.45

Note. *p < .05. ***p < .001.

Construct validity of the 10-item DSSI was supported theoretically by consistent associations between the DSSI and living alone, general health status, and quality of life as shown in Table 5. Living alone was associated with lower overall social support; higher self-reported general health status was positively associated with higher levels of social support; and higher quality of life was positively associated with higher levels of social support.

The reliability of the DSSI was assessed using Cronbach’s α to measure internal consistency of the total scale (.74) and the social satisfaction subscale (.77). Cronbach’s α provides a measure of internal consistency, the extent that items in a scale are measuring the same concept (Tavakol & Dennick, 2011). While the reliability of the social interaction subscale (.45) was low, when used in conjunction with the social satisfaction subscale, the total measure of social support provides a more comprehensive measurement of social support. Thus, the social interaction subscale should only be used as a part of the total DSSI scale.

Discussion

The large, weighted sample from the 2010 AHS allowed for generalizability and created an excellent venue to explore the validation of the 10-item DSSI in a general adult population. Social support emerged as an important measurement to include in epidemiological studies as related to various aspects of physical and mental health (Cohen et al., 2000; House et al., 1988; Koenig et al., 2004; Pachana et al., 2008).

The robustness of the results across subgroups provides confidence in the reliability of the results. The subscales are not to be independently used as reliable measurements for each dimension of social support; however, combined, they provide a valid measurement of two important constructs related to social support. The DSSI 10-item scale is a short, acceptable scale for use in the general population to measure social support.

Limitations

The data used in this analysis were cross sectional, meaning they were subject to issues inherent for secondary data. There may be aspects related to social support not available in these data. The entire AHS 2010 questionnaire was long and delivered via the telephone, which may have contributed to interviewee fatigue and missing data. Moreover, telephone surveys may involve subject bias as certain people are more likely to engage in phone surveys and the increasing use of cell phones as the primary telephone excludes some populations. In addition, the social interaction subscale was less robust than would be optimal. For this reason, it should only be used in conjunction with the social satisfaction scale to measure social support. Also, previous studies’ assumptions on the data that all variables are continuous is incorrect as applied to the factor analysis. It may be better to convert scale variables into categorical/dichotomous variables, and then EFA and CFA could provide more comprehensive insight into the capabilities of the scale.

Implications

Some well-known validated scales (e.g., Kessler 6—K6, World Health Organization Disability Assessment Schedule) used in the National Survey on Drug Use and Health (NSDUH), focus on mental health, distress, and depression issues (Hedden et al., 2012). The Center for Mental Health Services’ Uniform Reporting System, the Treatment Episode Data Set, and the Drug Abuse Warning Network also focus on mental health data. The Substance Abuse and Mental Health Services Administration implementation of National Outcome Measures has over time included development and use of measures of social connectedness to measure the outcome of increased social supports/social connectedness for mental health, substance abuse treatment, and substance abuse prevention. The current prevention measure, Family Communication Around Drug Use, is taken from the NSDUH survey and examines the parent/child relationship more than social connectedness among adults (Substance Abuse and Mental Health Services Administration, Office of Applied Studies, 2006).

The necessity of having a short, reliable measure of social support is important for both practice and research. Large epidemiological surveys would benefit from having a brief measurement of social support for diverse populations. The original DSSI scale was created to assess social support in older populations, especially chronically ill elderly. Now researchers may confidently extend the use of this abbreviated DSSI to measure social support for younger adults and ethnically diverse adult populations. This instrument can be used in cross-sectional and longitudinal epidemiological studies to increase the understanding of both mental and physical health in relationship to social supports in the general population.

Practitioners can be comfortable with the 10-item DSSI as a quick tool to accurately assess two important aspects of social support, social interaction, and social satisfaction. Knowing the baseline information for individuals could lead to conversations about the possibility of increasing social support as a means to improve quality of life and health status.

The DSSI-10 is appropriate for a variety of situations and use with a general adult population. In Arizona, its continued use in the statewide adult survey allows for comparison when also used by the Arizona Department of Health Services, Division of Behavioral Health Services prevention programs as a core outcome measure for block grant reporting on state-level goals. With growing recognition of the importance of social connectedness as part of integrated health and preventative care, validating and using the DSSI-10 scale become an increasingly important translation of research findings to the work of social work/services and health care providers.

Footnotes

Acknowledgement

The authors wish to thank St. Luke's Health Initiative for generously sharing the AHS 2010 data in the analysis for this manuscript.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

References

Arling

(1987). Strain, social support, and distress in old age. Journal of Gerontology, 42, 107–113.

Cohen

Gottlieb

B. H.

Underwood

L. G.

(2000). Social relationships and health. In Cohen

Underwood

L. G.

Gottlieb

B. H.

(Eds.), Social support measurement and intervention (pp. 3–25). New York, NY: Oxford University Press.

Cohen

Wills

T. A.

(1985). Stress, social support, and the buffering hypothesis. Psychological Bulletin, 98, 310–357.

Coleman

J. S.

(1990). Foundations of social theory. Cambridge, MA: Harvard University Press.

Ebeling-Witte

Frank

M. L.

Lester

(2007). Shyness, internet use, and personality. Cyberpsychology Behavior, 10, 713–716.

George

L. K.

Blazer

D. G.

Hughes

D. C.

Fowler

(1989). Social support and the outcome of major depression. British Journal of Psychiatry, 154, 478–485.

Goodger

Byles

Higganbotham

Mishra

(1999). Assessment of a short scale to measure social support among older people. Australian and New Zealand Journal of Public Health, 23, 260–265.

Hedden

Gfroerer

Barker

Smith

Pemberton

M. R.

Saavedra

L. M.

Novak

S. P.

(2012). Comparison of NSDUH mental health data and methods with other data sources. Center for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration (SAMHSA). Retrieved from http://www.samhsa.gov/data/NSDUH/2k12Findings/CBHSQDataReviewC2MentalHealth2012.htm

House

J. S.

Landis

K. R.

Umberson

(1988). Social relationships and health. Science, 241, 540–545.

10.

L.-T.

Bentler

(1995). Evaluating model fit. In Hoyle

R. H.

(Ed.), Structural equation modeling. Concepts, issues, and applications (pp. 76–99). London, England: Sage.

11.

Koenig

H. G.

George

L. K.

(1998). Depression and physical disability outcomes in depressed medically ill hospitalized older adults. American Journal of Geriatric Psychiatry, 6, 230–247.

12.

Koenig

H. G.

George

L. K.

Titus

(2004). Religion, spirituality, and health in medically ill hospitalized older patients. Journal of the American Geriatrics Society, 52, 554–562.

13.

Koenig

H. G.

Westlund

M. S.

George

L. K.

Hughes

D. C.

Blazer

D. G.

Hybels

(1993). Abbreviating the Duke Social support index for use in chronically ill elderly individuals. Psychosomatics, 34, 61–69.

14.

Landerman

George

L. K.

Campbell

R. T.

Blazer

D. G.

(1989) Alternative models of the stress buffering hypothesis. American Journal of Community Psychology, 17, 625–642.

15.

McDowell

Newell

(1996). Measuring health: A guide to rating scales and questionnaires. New York, NY: Oxford University Press.

16.

Pachana

N. A.

Smith

Watson

McLaughlin

Dobson

(2008). Responsiveness of the Duke Social support sub-scales in older women. Age and Ageing, 37, 666–672.

17.

Poleshuck

E. L.

Giles

D. E.

(2006). Pain and depressive symptoms among financially disadvantaged women’s health patients. Journal of Women’s Health, 15, 182–193.

18.

Powers

J. R.

Goodger

Byles

J. E.

(2004). Assessment of the abbreviated Duke Social Support Index in a cohort of older Australian women. Australasian Journal on Ageing, 23, 71–76.

19.

Revicki

D. A.

Mitchell

J. P.

(1990). Strain, social support, and mental health in rural elderly individuals. Journal of Gerontology, 45, S267–S274.

20.

Smith

K. P.

Christakis

N. A.

(2008). Social networks and health. Annual Reviews of Sociology. 34, 405–429.

21.

Substance Abuse and Mental Health Services Administration, Office of Applied Studies. (2006, April). National outcome measures (NOMs) for co-occurring disorders. Retrieved from http://oas.samhsa.gov/dual.cfm

22.

Tavakol

Dennick

(2011). Making sense of Cronbach’s alpha. International Journal of Medical Education, 2, 53–55.

23.

Wellman

Berkowitz

S. D.

(1988). Social structures: A network approach. Cambridge, England: Cambridge University Press.