| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLE |
From the Department of Psychology (A.J.C., R.D., S.L.E), The University of Iowa, and Department of Psychiatry (S.K.S.), The University of Iowa College of Medicine, Iowa City, Iowa.
Address reprint requests to: Alan J. Christensen, PhD, Department of Psychology, E11 Seashore Hall, The University of Iowa, Iowa City, IA 52242. E:mail: alan-christensen{at}uiowa.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONThe Present StudyMETHODSDiscussionNOTESREFERENCES |
|---|
METHODS: Social histories were abstracted from the medical records of a cohort of 133 deceased schizophrenic patients admitted for inpatient treatment between 1934 and 1944. Two independent raters assessed the quantity and quality of support available in each patient’s social environment.
RESULTS: Cox regression analysis revealed that higher quantity of social support was significantly related to survival time (p < .05) after controlling for marital status and quality of support. The Cox model indicated that a 1-point increase in the support quantity rating was associated with a proportional 25% decrease in the hazard rate.
CONCLUSIONS: The present findings suggest that social environment, specifically the quantity of social support available to the patient, may impact longevity in psychiatric populations.
Key Words: schizophrenia, • survival, • social support, • psychiatric populations.
Abbreviations: ICD-9 = International Classification of Disease;; SPSS = Statistical Package for the Social Sciences.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONThe Present StudyMETHODSDiscussionNOTESREFERENCES |
|---|
An emerging body of evidence suggests that the beneficial effect of social support on physical health and mortality extends to a variety of medical populations (4, 5). Various measures of the quantity or perceived quality of the social environment have been linked to lower mortality rates among patients with coronary heart disease (6), cancer (7), and end-stage renal disease (8). For example, Christensen et al. reported 5-year mortality rates among renal dialysis patients classified as low in family support to be nearly three times higher than mortality rates for their counterparts with high support.
The possible association between social support and mortality has led to an impressive amount of research in both community and medical populations. However, among psychiatric populations, this potential link has been largely ignored. Past research has documented that patients with major psychiatric disorders have significantly higher mortality rates compared with nonclinical populations (9, 10). Some of the most compelling evidence of earlier mortality in a psychiatric population has involved schizophrenic patients (11–13). For example, a study of 688 schizophrenic patients observed for 10 years reported an age- and sex-adjusted mortality rate more than 2.5 times higher than the general population (11).
Despite a well-documented increase in mortality for schizophrenic patients, little is known about potential influences on survival in this population. There is some evidence that demographic factors, such as age, gender, and education, may influence survival (14, 15). However, little data has addressed the possible influence of social support or social environment on survival among individuals with schizophrenia or any other major psychiatric disorder.
|
The Present Study |
|---|
|
TOPABSTRACTINTRODUCTIONThe Present StudyMETHODSDiscussionNOTESREFERENCES |
|---|
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONThe Present StudyMETHODSDiscussionNOTESREFERENCES |
|---|
|
Social Resources Assessment
Each patient’s social history as recorded at the time of admission was abstracted from the medical record. The social history record contained detailed descriptive information about the patient’s family and social environment. Each patient’s available social resources were evaluated by two independent raters using two 5-point interval rating scales. One scale assessed the quantity of social resources available to the patient. Ratings ranged from 1 ("No support providers or close relationships can be identified") to 5 ("Daily interactions with support providers, many close relationships"). Another scale assessed the quality or nature of social interactions reported by the patient. Ratings on this scale ranged from 1 ("Frequent negative interactions or conflict in social and family environment") to 5 ("No evidence of conflict, very positive social interactions"). The interrater reliability was acceptable for both the quantity (r = .57) and quality (r = .81) assessments. The two ratings were aggregated (averaged) to form a single rating of support quantity and quality. The aggregate rating of support quantity had a mean of 2.03 (SD = 0.80; range 1–5). The aggregate support quality rating had a mean of 2.47 (SD = 0.89; range 1–5). The two aggregate ratings were modestly intercorrelated (r = .21, p < .02). One advantage of aggregating ratings is a marked increase in the estimated reliability of the composite rating. For example, using the Spearman-Brown formula the interrater correlation of .57 obtained for the support quantity rating yields a reliability estimate of 0.73, whereas the .81 interrater correlation for support quality yields a reliability estimate of 0.90 (16).
Overview of the Survival Analysis
Multivariate survival analysis was performed using the Cox proportional hazards regression procedure program from SPSS-Windows, Release 7.0 (17). Unlike some other survival analytic procedures, the Cox regression procedure considers the effect of multiple predictors simultaneously (18). Because all patients were deceased at the time of follow-up, no cases were censored from the analysis.
Preliminary Survival Analysis
A preliminary Cox regression procedure was conducted to examine the association of patient age at admission, years of education, gender, marital status, and institutionalization status to survival time. Effect coding (ie, +1 or -1) was used for the categorical gender, marital status, and institutionalization status variables. Predictor variables meeting a liberal (p < .10) significance level were allowed to enter the regression model. Using this criteria, only marital status entered the model, 
2(1)= 3.59, p < .06, ß = .17. This marginally significant effect indicated that married patients exhibited somewhat longer survival times than did unmarried patients.1
Primary Survival Analyses
In the primary analysis, all predictor variables were entered into the Cox regression equation simultaneously. Elapsed survival time measured from the initial assessment made at hospital admittance until the date of death served as the dependent variable for the analysis. Results of this analysis are reported in Table 2.
|
2(3)= 9.13, p < .03. The only significant term in the model was quantity of social support, 2(1)= 5.22, p < .03. Neither quality of support or marital status exhibited a significant unique effect on survival time (p values > .10). 2 After exponentiating, the regression coefficient for quantity of support indicates that a 1-point increase on the support rating is associated with an average 25% hazard reduction across the follow-up period. To illustrate the influence of support quantity on survival, survival functions contrasting high and low support quantity were estimated from the Cox regression model. Support scores 1.0 SD above and below the mean were used to reflect representative high and low values. As shown in Figure 1, the estimated survival function for patients with high versus low support quantity reflects greater estimated cumulative survival across the large majority of the follow-up period for high support patients.
|
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONThe Present StudyMETHODSDiscussionNOTESREFERENCES |
|---|
In contrast to the findings for the frequency of social interaction, the rated quality of patients’ social interactions was not significantly associated with survival. This pattern seems consistent with community-based studies that have suggested that the degree of social integration has a more important influence on longevity than do qualitative aspects of the social environment (3). It is also possible that our relatively generalized and simplistic rating of the overall quality of social interactions failed to adequately capture the multifaceted nature of social support. Past research has suggested that different characteristics of social support may have quite different adaptational correlates (1, 3).
In addition to examining the effects of social support, the predictive influence of a number of demographic factors on survival was also examined. In a preliminary analysis of demographic variables, only marital status (ie, married vs. unmarried) approached significance (p < .10) as a predictor of survival. However, the unique effect of marital status was substantially lower (p > .40) when added to the primary regression controlling for quantity of social support. Marital status was significantly correlated with the rating of support quantity (r = -.25, p < .005). The lack of a significant unique association between marital status and survival may have, in part, been due to overlapping variance with the support quantity rating in the prediction of patient longevity. Patient age was also not a significant predictor of survival. Age and marital status were significantly intercorrelated in the present sample (r = .40, p < .0001), suggesting that older patients were more likely to be married than younger patients. This confound between marital status and age may have attenuated the unique predictive effects of these two characteristics.
Causes of death in this sample largely mirrored those of the general population. Cardiovascular and cerebrovascular disease (two of the three most common causes of death in the general population) together accounted for nearly half of the deaths in the sample. Only three patients in the sample (2.3%) had died of a confirmed suicide. Past research involving schizophrenic populations has reported suicide rates of approximately 10% (eg, Refs. 19, 20). One factor that may have contributed to the relatively low rate of confirmed suicide in the present sample involves a bias in the coding of cause of death. Specifically, there may have been a reluctance during this historical time period (1930s and 1940s) to list suicidal behavior as a cause of death in the preparation of death certificates. This possibility is consistent with past reports that suicide is typically underreported in studies relying on death certificates to determine cause of death (21). By utilizing collateral sources of information regarding cause of death, future research will be able to more clearly delineate the extent to which deficits in social resources are related to increased risk of death from suicide.
The mediational mechanisms underlying the effect of social resources on survival in this population are yet to be identified. Various models have been proposed to account for an association of social resources and health (5). For example, there is considerable evidence that social support influences cardiovascular, immunological, and neuroendocrine responses (1, 22). These psychophysiological processes may partly account for the influence of support on health. It is also possible that the effect of social resources on survival in the present sample was due to a reduction in negative health behaviors among patients with more frequent social contact. This interpretation is consistent with research involving other populations that have identified an association between social support and health practices (23, 24). Negative health practices (eg, smoking and drinking behavior) are known to be particularly common among psychiatric patients (25, 26). Unfortunately, the archival database used in the present research did not provide a reliable assessment of substance use or other health practices. Additional research is needed to address the role that health behaviors may play in explaining the survival-support association among schizophrenic patients.
Causal interpretations of the present data should be made cautiously, due to the correlational and archival nature of the study methodology. An alternative interpretation of the present findings involves the possibility that more severe schizophrenia leads to decreased social support and it is actually the severity of the underlying schizophrenia that is associated with decreased longevity. Because social support was assessed at a single point in time, we were also unable to establish how changes in social environment over the extended follow-up period may have influenced mortality. More compelling causal evidence may require research examining the effects of manipulated social support on patient outcomes. This approach has already yielded provocative data regarding psychosocial treatments and survival among nonpsychiatric medical patients (27).
The archival nature of the data also presents an important methodological limitation for the assessment of social environment. Although extended follow-up periods are difficult without the use of archival records, future research would benefit from the use of social support measures with more established construct validity and reliability than the ratings used here. More specific assessment data regarding other aspects of the patient’s social environment, both during hospitalization and after discharge, would also be useful in providing a clearer understanding of the specific types of social interaction that are important correlates of longevity. For example, negative aspects of the family environment (ie, expressed emotion) have been linked to relapse and reinstitutionalization among schizophrenic patients (28, 29). Additional research is necessary to determine whether expressed emotion or related aspects of the family environment impact patient survival.
The above qualifications notwithstanding, the present data suggest that the frequency of social interaction plays a role in influencing longevity among schizophrenic patients. Identifying potentially modifiable influences on long-term health is particularly important, given the increased mortality risk often observed among psychiatric patients. Our data represent an early step in understanding factors that may influence this risk.
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONThe Present StudyMETHODSDiscussionNOTESREFERENCES |
|---|
2(1)= 2.73, p = .10. This effect reflects a trend toward longer survival times among younger patients. 
2 Adding patient age to the primary Cox regression analysis does not shift the pattern of results. The effect of support frequency on survival remained significant after controlling for age, 2(1)= 5.03, p < .03.
Received for publication June 14, 1998.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONThe Present StudyMETHODSDiscussionNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  W. S. Shaw, T. L. Patterson, S. J. Semple, M. C. Halpain, W. L. Koch, M. J. Harris, I. Grant, and D. V. Jeste Use of Community Support Services by Middle-Aged and Older Patients With Psychotic Disorders Psychiatr Serv, April 1, 2000; 51(4): 506 - 512. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. T. Tsuang Observational Versus Experimental Studies: Would the Results Be Similar? Psychosom Med, March 1, 1999; 61(2): 146 - 147. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
