This Article Abstract Figures Only Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Rutledge, T. Articles by Cauley, J. A. Search for Related Content PubMed PubMed Citation Articles by Rutledge, T. Articles by Cauley, J. A. Related Collections Social Support Other Epidemiology

Social Networks and Marital Status Predict Mortality in Older Women: Prospective Evidence From the Study of Osteoporotic Fractures (SOF)

From the Departments of Psychology (T.R., K.M.) and Epidemiology (J.A.C.), University of Pittsburgh, Pittsburgh, Pennsylvania; and the Department of Epidemiology and Biostatistics (L-Y.L.) and School of Medicine (K.L.S.), University of California, San Francisco, California.

Address reprint requests to: Thomas Rutledge, PhD, Psychology Service (116B), VA San Diego Healthcare System Medical Center, 3350 La Jolla Village Drive, San Diego, CA 92161. Email: dr.tom{at}medscape.com

Received for publication April 1, 2002; revision received August 5, 2002.

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
OBJECTIVE: To assess the relationship between social network size and prospective mortality risk among a large sample of older, Caucasian women.

METHOD: The study included 7524 Caucasian community-dwelling women, age 65 or older (mean age = 74.1), who participated from four U.S. communities. Study participants completed a protocol that included anthropomorphic and health assessments at baseline and the Lubben Social Network Scale at year 2. We followed participants for an average of 6 years after they had completed the year-2 assessment. We used hospital records and a copy of the participant’s official death certificate to document mortality and cause of death in accordance to ICD-9 revision codes.

RESULTS: A total of 1451 deaths (19.3% of sample) were observed over follow-up, 215 (3.4%) due to cardiovascular causes. Higher social network scores were a robust predictor of lower multivariate-adjusted mortality (RR = 0.92, 95% CI = 0.86–0.98), controlling for age, comorbid disease, body mass, smoking, depression, and education. However, social network benefits were attenuated after controlling for marital status. Married participants showed lower total (RR = 0.83, 95% CI = 0.74–0.94) and CVD (RR = 0.59, 95% CI = 0.43–0.81) covariate-adjusted death rates compared with unmarried participants.

CONCLUSIONS: Social network scores and marriage were each associated with reduced prospective mortality risk among older women. The relationships shown here suggest that much of the protection afforded by larger social networks in older women results from marriage rather than other forms of social relationships. Mechanisms at the physiological or behavioral level explaining social relationship benefits remain important areas for future research.

Key Words: marital status, • social networks, • mortality, • osteoporotic fractures.

Abbreviations: BMI = body mass index;; CVD = cardiovascular disease;; LSNS = Lubben Social Network Scale;; SOF = Study of Osteoporotic Fractures.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
Decades of research provide evidence in favor of an association between social relationships and mortality (1–3). A series of independently conducted prospective studies showed reduced rates of mortality among participants with larger or more diverse social networks (4–10), and the apparent magnitude of this protective effect rivals that of standard mortality risk factors such as hypertension or diabetes. Although the potential mechanisms for the association remain poorly understood, recent experimental findings are also suggestive, showing reduced blood pressure responses and alterations in neuroendocrine reactivity under stressful circumstances among those in the presence of supportive friends or confederates and an increased resistance to cold infections (11–15).

The strength of the evidence for social networks as an epidemiological risk factor is tempered by three additional findings. First of all, social networks seem to predict mortality but not disease incidence, indicating that the influence of social relationships may be limited to those with existing disease (6, 16). Second, many studies have demonstrated gender differences, such that social network effects were present only - or in stronger form - among men (4–5, 8). Finally, because marital status is an important component of most social network measures, it is necessary to demonstrate that social networks provide mortality benefits independent of marriage (4).

We assessed social networks and marital status as predictors of prospective all-cause and cardiovascular (CVD) mortality risk in a large cohort of older (age 65+) women in testing the following hypotheses:

(a) Higher social network scores and marriage would be associated with improved survival in the SOF cohort independent of standard mortality risk factors. We projected that the relationship would replicate across all-cause and CVD mortality categories.

(b) The survival benefits of a larger social network were predicted to be robust to marital status. We assessed this hypothesis by including marital status as an additional control variable and by testing the interaction between marital status and social network scores on 6-year mortality.

(c) Higher social network scores were anticipated to predict improved survival for participants with and without a history of significant disease (diabetes, stroke, hypertension, or previous myocardial infarction).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
Participants
The study recruited 9704 Caucasian women at least 65 years of age from populations in Baltimore (Maryland), Minneapolis (Minnesota), Portland (Oregon), and the Monongahela Valley (Pennsylvania) between September 1986 and October 1988 for the Study of Osteoporotic Fractures (SOF) (19). Participants consisted of community-dwelling, ambulatory women with no history of bilateral hip replacement. The SOF investigation focused on Caucasian women due to the high rates of falls and fall-related deaths in this racial group (20). During a second visit (1988–1990), 7524 of these women (80% of survivors) completed the Lubben Social Network Scale (17). The remaining 20% of the SOF sample did not complete the Lubben scale and are not included in the analyses described in this paper. All participants provided written informed consent, and institutional review boards at each participating site gave approval to the SOF study.

Measurements
SOF participants completed a battery of self-report questionnaires assessing factors such as disease history, health behaviors, fall history, and psychological functioning and physical evaluations for characteristics such as height, weight, bone mineral density, and blood pressure. More detailed descriptions of the methods and measures used in the SOF investigation can be found in previous publications (19).

Social Networks
During the second visit (1988–1990), participants completed the Lubben Social Network Scale (LSNS) (17), a validated 10-item self-report inventory assessing family relationships (3 items regarding size and frequency of contact); relationships with friends (3 items, similar to family questions); and interdependent relationships (4 items) such as the presence of a confidante. A validation study found that the LSNS possessed acceptable internal consistency levels (21) and showed associations between LSNS scores and self-reported behavioral health outcomes including extended hospital stays, mental health scores (Life Satisfaction Index), and a checklist of health practices (17). In order to create a total social network score, we then summed the responses to the individual questions (possible scores on the LSNS range from 0–50). Higher scores indicate a larger social network and/or more frequent social contact. The LSNS showed modest internal consistency with this cohort ( = 0.55); however, with the exception of item 6 (r = 0.26; How often do you see or hear from friend with most contact?), all items correlated 0.40 or greater with the total scale score, indicating reasonable shared item content by psychometric standards.

Other Self-report Measures
Participants also provided responses to questions regarding age; marital status (coded for married, widowed, separated, divorced, or never married); education; medical history; perceived health status; and smoking that were administered as part of the baseline test protocol. Due to the older age of the sample, the criterion for hypertension based on systolic values was set at 160 mm Hg, diastolic blood pressures higher than 90 mm Hg, or reported use of a thiazide diuretic. Supine blood pressure measures were collected using a standard clinic protocol at the participant’s right brachial artery.

During the second visit, anthropomorphic measures (height, weight, waist circumference) were collected using a balance beam scale (weight) and stadiometer (height). We calculated a body mass index score for each participant as a function of weight (in kilograms) divided by height (in squared meters). Depression symptoms were assessed using the 15-item Geriatric Depression Scale (range 0–15) (18). Based on scale development studies describing this instrument (22), we defined the presence of depression by the presence of six or more symptoms (approximately 6.3% of the sample met this criterion).

Mortality
Mortality was determined for an average of 6 years after the initial social network measure. During participation, morbidity and mortality assessments were completed at 4-month intervals by having participants (or, in the event of death, participant’s family or contacts) return postcards. If a participant died, we obtained a copy of the official death certificate and hospital discharge summaries, if available. From this information, cause of death was assigned by a SOF physician investigator who was blind to the participant’s social network score or other measures. Cause of death was assigned in accordance with International Classification of Diseases, Ninth Revision codes (cardiovascular = 394–402, 410–414, 424–444, and 798); deaths from all codes were also collapsed into a single dichotomous mortality score to assess relationships with total mortality. Total mortality and mortality from cardiovascular causes (the largest single cause of death in the SOF sample) were used as outcome measures for the current study.

Statistical Analyses
We assessed social network effects as a continuous, dichotomized (high-low standing based on a median split of the total score) and quartiled variable in our analyses. All relationships between social network scores and subsequent mortality were tested using Cox regression methods in which risk factors [smoking status (yes-no); BMI (continuous scores); age; depression (<6 or 6); education (dichotomized as 12 or >12 years of education); history of stroke, diabetes, and hypertension] were force entered at Step 1, followed by the social network score at Step 2. Our initial set of covariates also include history of myocardial infarction; however, data for this variable was missing for approximately 2500 participants. Preliminary analyses showed that controlling for heart attack history had no effect on the Cox regression models, and we subsequently removed the heart attack variable from the list of covariates for the final model calculations in order to assess social network effects in the complete sample (ie, the nearly one-third of SOF participants without heart attack data were otherwise excluded from the models). All analyses were completed using SPSS 10.0 software (SPSS Inc., Chicago, IL).

To determine the predictive value of the social network scores independent of marital status, we completed a Cox regression analysis incorporating all baseline covariates with the addition of marital status (the marriage covariate was a married vs. not married dichotomized variable) as Step 1 followed by social network scores at Step 2. Where indicated, we calculated social network x marriage scores using the dichotomized form of each variable, thereby producing a total of four groups.

We examined marital effects in which the health covariate terms (including, in this case, social network scores) described above were entered at Step 1 followed by a dichotomized (0 = married, 1 = unmarried) marital status variable. In preliminary testing, we also assessed mortality relationships within broader categories of unmarried participants (0 = married, 1 = widowed, 2 = separated, never married, or divorced) that maintained reasonably close sample sizes for the groups. These calculations indicated no differences among the unmarried participant groups and we therefore reverted back to a simple dichotomized marriage variable (0 = married, 1 = unmarried) for final models. Resulting hazard ratios [risk ratios (RRs)] were calculated using the married group as the reference category.

In order to address the predictive value of social network scores relative to pre-existing disease status, Cox equations were performed in which baseline risk factors (including disease measures of diabetes, stroke, myocardial infarction, and hypertension) were entered at Step 1, followed by social network scores at Step 2, and finally a social network x existing diseases interaction term at Step 3. In producing the interaction term, we first summed the four disease variables for each participant (4 dichotomous 0–1 variables) to compute the number of disease conditions for each person in a single variable that ranged from 0 (for participants with no history of diabetes, stroke, hypertension, or myocardial infarction) to 4 (for participants with all four conditions) and then multiplied this variable with the participant’s social network score.

The dichotomous social network variable was used to create the interaction term in order to simplify the interpretation of the interaction results. A significant interaction suggested that the relationship between social network scores and mortality risk differed for participants with and without a history of the four disease conditions.

Power Analyses
Preliminary power analyses, based on a two-tailed level set at 0.05 and a sample size of 7524, indicated that our ability to detect effect sizes exceeding r = 0.10 (approximately equal to a risk ratio of 2.0 in this sample) (23–24) was greater than 0.95.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
Overview
Table 1 provides a description of the SOF cohort on a number of demographic, disease history, and psychosocial variables. Univariate relationships with total and CVD mortality are also shown. A total of 1451 participants were confirmed dead by the end of follow-up, 215 due to cardiovascular causes. As illustrated in Table 1, nearly all of the baseline risk variables were reliable predictors of subsequent mortality.

View larger version (44K): [in this window] [in a new window]   Fig. 1. Unadjusted () and age-adjusted () total mortality rates per 100 participants (ie, total deaths/total number of participants) across social network quartile groupings.

View larger version (43K): [in this window] [in a new window]   Fig. 2. Unadjusted () and age-adjusted () CVD mortality rates per 100 participants across social network quartile groupings.

Social Networks Versus Marriage
Question 10 from the LSNS queries the living situation of the respondent (eg, alone, live with spouse, live with others, etc.). Although even married participants reported living alone in some instances due to separation or illness of the spouse (eg, husband in a nursing home), we anticipated that this question would be highly correlated with the participant’s marital status. As a result, we computed LSNS scores for the complete scale and with the first nine items only in assessing mortality relationships for possible differences, in addition to directly controlling for marriage in the analyses reported in this section.

Social network scores and marital status were interrelated [r = 0.36 ( coefficient test for dichotomous variables), p < .001], indicating moderate overlap in the measures. Social network scores were reliably associated with total mortality after adjusting for age and marital status (RR = 0.82, 95% CI = 0.73–0.93), but the relationship was not significant after adjusting for marital status and all covariates (RR = 0.89, 95% CI = 0.76–1.0). Similarly, the relationship between social networks and CVD mortality disappeared after adjusting for age and marital status (RR = 0.84, 95% CI = 0.62–1.4, p > .10) and marriage and all covariates (RR = 0.86, 95% CI = 0.68–1.4). Marital status itself was associated with reduced total mortality and CVD mortality risk after controlling for all covariates -including social network scores - [RR value = 0.83 (95% CI = 0.74–0.94), 0.59 (95% CI = 0.43–0.81), respectively, for total and CVD mortality].

Among married participants, social network scores were not associated with total or CVD death after covariate adjustment [RR value = 0.92 (95% CI = 0.75–1.1), 1.1 (95% CI = 0.73–1.8), respectively, for total and CVD mortality]. However, higher LSNS scores did predict lower mortality rates among unmarried participants [RR value = 0.83 (95% CI = 0.72–0.96), 0.75 (95% CI = 0.54–1.0), respectively, for total and CVD mortality], suggesting that a larger social network benefited unmarried participants.

We completed a final pair of Cox regression models to assess the presence of an interaction between social network scores and marital status. Table 3 summarizes these findings, showing that, for both total and CVD mortality categories, married participants reporting high social network scores had the lowest mortality rate over the follow-up interval. In contrast, the mortality rates for the group reporting low social network status and being unmarried were highest for both mortality categories. The middle groups (low social network status and married; high social network status and unmarried) showed mortality rates between these groups. Only the low-unmarried and high-married group mortality rates differed at the .05 significance level.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
This study addressed three hypotheses regarding the relationships between social networks, marriage, and 6-year mortality risk. Initially, the current study represents the largest and oldest female cohort in which social network effects have been tested, and the number of deaths observed permitted inferential testing with high levels of statistical power. Our results indicated that a larger social network was associated with a reduced risk of subsequent mortality, a finding that held across all-cause and CVD mortality categories. The relationship with all-cause mortality remained significant even after controlling for an established set of covariate predictors that included markers of socioeconomic status (years of education) as well as biomedical variables. Our set of covariates also included depression scores from the Geriatric Depression Scale, which were shown to be powerful predictors of mortality and fall risk in the SOF sample in previous studies (25–26). Equally important, the mortality risk associated with social network scores was generally linear. Compared with the highest scoring quartile, mortality risk showed a dose-response increase among the 75th, 50th, and 25th percentile groups. That said, most - though not all - of the benefits of social networks in this sample seemed attributable to marriage.

Because measures of social relationships - including social network scales - often include marital status, some investigators have suggested that the benefits of social networks might be explained, at least in part, on the basis of marriage. Kaplan (4), however, provided results in favor of an independent effect for social networks. In contrast, our results indicated that marital status explained much of the relationship of mortality with social networks. The LSNS includes a single item that correlates highly with marital status. Removing this single item from the scale scores, however, did not attenuate the overall effects of the social network scores.

When marital status was included as an additional covariate term in Cox regression models, social network scores were no longer a significant predictor of covariate-adjusted all-cause or CVD mortality. These findings suggest that the protective effects of social networks among women were largely explained by marital status in this cohort. Results provided in Table 3, in which we coded participants for high-low social network status and married-unmarried status, did suggest some benefits from social networks independent of marriage. The latter findings indicated that women who reported both being married and high social network status had the lowest mortality rates over follow-up, whereas those who reported only high social network status or being married showed comparatively higher mortality rates for both total and CVD mortality categories. Finally, the group reporting being unmarried and with low social network status showed the highest mortality rates over follow-up. Therefore, both marriage and larger social networks may provide a protective effect on their own, whereas the combination of the two seems to be most beneficial.

Prior studies have also shown that social network measures are reliable predictors of mortality but not of disease incidence (6), indicating that social network benefits might apply most strongly to those surviving with pre-existing disease. We collected information regarding disease status (including diabetes, hypertension, history of stroke and myocardial infarction) at baseline testing and included these variables as an interaction term with social network scores to assess the equality of social network benefits among those with and without a history of these four diseases. Our results gave no indication of group differences. Thus, although we did not report associations with prospective disease risk as a potential mediator of mortality risk, the increased longitudinal mortality risk shown here as a function of lower social network scores held true for participants irrespective of disease history.

A question of central importance to the findings reported here and in previous social network studies concerns the mechanisms through which the benefits of social relationship networks and marriage are achieved. Experimental and animal studies offer mechanistic findings that may be valuable to understanding social network effects (13–15), but the artificiality of many of these results limits their applicability to the real world dynamics of human social relationships. The results from the current study, as well as from more than a dozen previous prospective investigations, indicate that social network effects are not simply a proxy for pre-existing physical health, socioeconomic status, or psychological well-being. However, the suggestion of a causal link between social networks and health is equally unproven. The recurrent theme of social networks benefiting health outcomes across samples differing widely in terms of age, gender, and ethnic origin clearly warrants continued investigation, although the focus of future research should move away from global descriptions of relationship size in favor of more precise measures of relationship quality and support that are the most likely candidates for explaining the health benefits of larger social networks.

Limitations
Despite the large sample size, breadth of measurement, and length and detail of follow-up of the SOF study, the value of the findings described herein is affected by several methodological limitations. Initially, our sample comprised older Caucasian women and cannot necessarily be generalized to other ethnic groups or to male populations. The social network scores in this sample also represent size and frequency of interpersonal contact and should not be interpreted to infer potential effects of relationship quality. Many of the variables assessed in SOF were likewise based on self-report descriptions, which permit the usual criticisms given to subjective measures, with the addition of possible influences from age-related cognitive declines in an older sample (although it should be emphasized that the Lubben Social Network Scale was specifically designed for and normed on older individuals). Finally, our attempts to associate social network and marriage effects with disease history (ie, hypertension, stroke, diabetes, etc.) were limited to a modest extent by incomplete measurement of these variables; we did not measure disease severity nor did we track progression or incidence of these disease variables over follow-up.

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
We reported prospective associations between social network scores, marital status, and mortality risk over a 6-year follow-up period among older (post-65) women. Both higher social network scores and marriage at study baseline were potent predictors of lower total and CVD mortality across follow-up. These benefits were largely independent of demographic variables, pre-existing disease, and other psychosocial measures, but the measures showed considerable overlap with each other. In this sample, marital status - and not social network scores - was the most consistent predictor of subsequent mortality, and marriage explained much - but not all - of the mortality relationships with social network scores. These findings extend previous work by demonstrating clinically relevant associations within an exclusively female cohort and further reinforce ongoing research to identify biological and behavior mediators of social and interpersonal relationship benefits.

	ACKNOWLEDGMENTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
This study was supported by US Public Health Service Grants AG05394, AG05407, AR35582, AR35583, AR35584, and NS36016.

	REFERENCES

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 

1. Cassel J. The contribution of the social environment to host resistance. Am J Epidemiol 1976; 104: 107–23.[Free Full Text]
2. House JS, Landis KR, Umberson D. Social relationships and health. Science 1988; 241: 540–5.[Abstract/Free Full Text]
3. Berkman LF, Syme SL. Social networks, host resistance, and mortality: a 9-year follow-up study of Alameda County residents. Am J Epidemiol 1979; 109: 186–204.[Abstract/Free Full Text]
4. Kaplan GA, Salonen JT, Cohen RD, Brand RJ, Syme SL, Puska P. Social connections and mortality from all causes and from cardiovascular disease: prospective evidence from Eastern Finland. Am J Epidemiol 1988; 128: 370–80.[Abstract/Free Full Text]
5. House JS, Robbins C, Metzner HL. The association of social relationships and activities with mortality: prospective evidence from the Tecumseh Community Health Study. Am J Epidemiol 1982; 116: 123–40.[Abstract/Free Full Text]
6. Vogt TM, Mullooly JP, Ernst D, Pope CR, Hollis JF. Social networks as predictors of ischemic heart disease, cancer, stroke, and hypertension: incidence, survival, and mortality. J Clin Epidemiol 1992; 45: 659–66.[CrossRef][Medline]
7. Schoenbach VJ, Kaplan BH, Fredman BH, Kleinbaum DG. Social ties and mortality in Evans County. Am J Epidemiol 1986; 123: 577–91.[Abstract/Free Full Text]
8. Welin L, Tibblin G, Svardsudd K, Tibblin B, Ander-Peciva S, Larsson B, Wilhelmsen L. Prospective study of social influences on mortality. The study of men born in 1913 and 1923. Lancet 1985; 1: 915–8.[CrossRef][Medline]
9. Orth-Gomer K, Johnson J. Social network interaction and mortality. A 6-year follow-up study of a random sample of the Swedish population. J Chronic Dis 1987; 40: 949–58.[CrossRef][Medline]
10. Blazer DG. Social support and mortality in an elderly community population. Am J Epidemiol 1982; 115: 684–94.[Abstract/Free Full Text]
11. Cohen S, Doyle WJ, Skoner DP, Rabin BS, Gwaltney JM. Social ties and susceptibility to the common cold. JAMA 1997; 277: 1940–4.[Abstract]
12. Kissel S. Stress-reducing properties of social stimuli. J Pers Soc Psychol 1965; 2: 378–84.[CrossRef]
13. Levine S. The influence of social factors on the response to stress. Psychother Psychosom 1993; 60: 33–8.[Medline]
14. Thomas PD, Goodwin JM, Goodwin JS. Effect of social support on stress-related changes in cholesterol level, uric acid levels, and immune function in an elderly sample. Am J Psychiatry 1985; 142: 735–7.[Abstract/Free Full Text]
15. Unden AL, Orth-Gomer K, Elofsson S. Cardiovascular effects of social support in the workplace: 24-hour ECG monitoring of men and women. Psychosom Med 1991; 53: 50–60.[Abstract/Free Full Text]
16. Linden W, Chambers L, Maurice J, Lenz J. Sex differences in social support, self-deception, hostility, and ambulatory cardiovascular activity. Health Psychol 1993; 12: 376–80.[CrossRef][Medline]
17. Reed D, McGee D, Yano K. Social networks and coronary heart disease among Japanese men in Hawaii. Am J Epidemiol 1983; 117: 384–96.[Abstract/Free Full Text]
18. Lubben JE. Assessing social networks among elderly populations. Fam Community Health 1988; 11: 42–52.
19. Yesafage JA. Geriatric Depression Scale. Psychopharmacol Bull 1988; 24: 709–11.[Medline]
20. Cummings SR, Nevitt MC, Browner WS. Risk factors for hip fracture in white women: Study of Osteoporotic Fractures Research Group. N Eng J Med 1995; 332: 767–73.[Abstract/Free Full Text]
21. Anastasi A, Urbina S. Psychological testing. 7th ed. Upper Saddle River, NJ: Prentice Hall; 1997.
22. Shiekh JI, Yesavage JA. Geriatric Depression Scale (GDS): recent evidence and development of a shorter version. Clin Gerontol 1986; 5: 165–73.
23. Cohen J. Statistical power for the behavioral sciences. 2nd ed. Hillsdale (NJ): Erlbaum; 1988.
24. Rosenthal R, Rosnow RL, Rubin DB. Contrasts and effect sizes in behavioral research: a correlational approach. Cambridge, UK: Cambridge University Press; 2000.
25. Whooley MA, Browner WS. Association between depressive symptoms and mortality in older women. Arch Internal Med 1998; 158: 2129–5.[Abstract/Free Full Text]
26. Whooley MA, Kip KE, Cauley JA, Ensrud KE, Nevitt MC, Browner WS. Depression, falls, and risk of fracture in older women. Arch Internal Med 1999; 159: 484–90.[Abstract/Free Full Text]

This article has been cited by other articles:

				L. Garand, M. A. Dew, B. Urda, J. H. Lingler, S. T. DeKosky, and C. F. Reynolds Marital Quality in the Context of Mild Cognitive Impairment West J Nurs Res, December 1, 2007; 29(8): 976 - 992. [Abstract] [PDF]

				E. D. Eaker, L. M. Sullivan, M. Kelly-Hayes, R. B. D'Agostino Sr, and E. J. Benjamin Marital Status, Marital Strain, and Risk of Coronary Heart Disease or Total Mortality: The Framingham Offspring Study Psychosom Med, July 1, 2007; 69(6): 509 - 513. [Abstract] [Full Text] [PDF]

				H. Litwin and S. Shiovitz-Ezra Network Type and Mortality Risk in Later Life Gerontologist, December 1, 2006; 46(6): 735 - 743. [Abstract] [Full Text] [PDF]

				H. Litwin Social Networks and Self-Rated Health: A Cross-Cultural Examination Among Older Israelis. J Aging Health, June 1, 2006; 18(3): 335 - 358. [Abstract] [PDF]

				T. Rutledge, S. E. Reis, M. Olson, J. Owens, S. F. Kelsey, C. J. Pepine, S. Mankad, W. J. Rogers, C. N. Bairey Merz, G. Sopko, et al. Social Networks Are Associated With Lower Mortality Rates Among Women With Suspected Coronary Disease: The National Heart, Lung, and Blood Institute-Sponsored Women's Ischemia Syndrome Evaluation Study Psychosom Med, November 1, 2004; 66(6): 882 - 888. [Abstract] [Full Text] [PDF]

				Other articles noted: 25 Jul 03 to 7 Nov 03 Evid. Based Nurs., January 1, 2004; 7(1): e1 - 1. [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Rutledge, T. Articles by Cauley, J. A. Search for Related Content PubMed PubMed Citation Articles by Rutledge, T. Articles by Cauley, J. A. Related Collections Social Support Other Epidemiology