The quality and quantity of social support: Stroke recovery as psycho-social transition

Sot. Sci. Med. Vol. 34. No. II. pp. 1249-1261. 1992 Printed in Great Britain. All rights reserved 0277-9536192 S5.00 + 0.00 Copynght G I’992 Pergamon Press Ltd THE QUALITY AND QUANTITY OF SOCIAL SUPPORT: STROKE RECOVERY, AS PSYCHO-SOCIAL TRANSITION THOMAS A. GLASS’.* and GEORGE L. MADDOX~ ‘Department of Epidemiology and Public Health, Yale University, School of Medicine, 60 College Street, P.O. Box 3333, New Haven, CT 06510-8034, U.S.A. and 2Duke University, Department of Sociology, Durham, NC 27706, U.S.A. Abstract-The impact of various types and amounts of social support is examined in the context of recovery from first stroke. We conceptualize the rehabilitation process as a psychosocial transition. In a longitudinal design, 44 patients were followed for 6 months following first stroke. Growth-curve analysis (repeated measures MANOVA) was utilized to examine the impact of three types of social support on changes in functional status during recovery. While all three types of support (emotional, instrumental and informational) were shown to be significantly related to recovery of functional capacity, substantial differences were found in the nature of those effects. The impact of social support does not appear during the first month of rehabilitation, indicating the importance of longitudinal designs and longer observation. Patients reporting high level of emotional support showed dramatic improvement despite having the lowest baseline functional status. Instrumental support is most closely related to positive outcomes when provided in moderate amounts. Unlike the other two types, the effect of informational support is mediated by disease severity. Key wjords-stroke, social support, rehabilitation, stressful events INTRODUCTION During the last two decades, the concept of social support has enjoyed the attention of social and behavioral scientists to a greater extent than any other psychosocial variable. It has been heralded as a magic bullet and attacked as a myth. There is little doubt that an association exists between social inter- actions and mortality [l-3] as well as morbidity [4-91. However, the nature of this association and the mechanisms through which it operates are still much debated. Recently, questions have been raised about the continued usefulness of the concept. Coyne and Bolger [IO] have argued, for example, that the litera- ture on social support rests on untenable assump- tions. Pearlin [ 1 l] has recently argued further that, to the extent the study of social support is separated from the institutional and network contexts in which it exists, the concept lacks ‘sociological substance’. Many investigators apparently agree the next step must be to explore the mechanisms through which social contacts exert an influence on health status. Only then will the concept of social support be rescued from becoming over generalized and non-specific. This task will require explicit testing of competing theories and the use of focused hypotheses. Toward this end, this paper has two purposes. First, using the work of Parkes [12], a conceptual model is developed for understanding the role of social support in recovery from stroke. In this model, ‘To whom requests for reprints should be addressed. we treat recovery from stroke as a ‘psycho-social transition’. We contrast this model with the dominant ‘stress-mitigation model’ showing that the two models lead to differing hypotheses about how and why social contacts are supportive. Secondly, using data from a prospective study of outcome in first stroke, we test one small part of this larger model by examining how the impact of social support varies by the type (quality) and amount (quantity) of support. In a prospective design, we explore the differential impact of social support on change in functional status (i.e. the capacity for self care and role perform- ance) in the aftermath of serious illness by differen- tiating effects among emotional, instrumental and informational types of support, while controlling for illness severity. Stroke as a psycho-social transition In a paper which sought to draw together theoreti- cal work in stress research, crisis studies, the soci- ology of disasters and life events, C. Murray Parkes defined psycho-social transitions as “. . . major changes in life-space which are lasting in their effects, which take place over a relatively short period of time and which affect large areas of the assumptive world” [ 12, p. 1031 (italics in the original). The concept of ‘the life-space’, which originates in the work of Lewin [ 131, refers to the totality of the sphere of interaction in which an individual operates. Parkes, following Lewin, stresses the directly physical aspects of individuals in context by referencing “ . . . other persons, material possessions, the familiar world of home and place of work, and the individual’s body 1249 and mind” [12, p. 1031. In a more general sense, Lewin spoke of the life-space as comprising the totality of actual and potential social facts that had or might in the future shape the actions of an individual. The second important element in Parkes’ definition is the concept of the ‘assumptive world’. He writes, The assumptive world is the only worid we know and it includes everything we know or think we know. It includes our interpretation of the past and our expectations of the future, our plans and our prejudices. Any or all of these may need to change as a result of changes in the life space (12, p. 1031. The individual’s assumptive world is that which is taken for granted, or assumed to exist on the basis of past experience [14]. The concept bears some resemblance to Mead’s concept of the generalized other in that it may be thought of as a set of attitudes that may be taken in different contexts. But unlike the generalized other which is public and shared, the assumptive world of an actor is his or her own and may never be fully known by another. A major ill- ness such as stroke challenges existing assumptions about the identity self-concept, and role capability. Examples of other events likely to initiate psycho- social transitions include job loss, the death of a spouse (or close family member) or an accident which leaves permanent disabilty. While much of Parkes’ work was concerned with events involving loss, such as bereavement, a psycho-social transition is not necessarily initiated only by a loss. To a person that believes all they have is bad luck, winning the lottery may be a profound challenge to the assumptive world. individual. Stroke takes place over a relatively short period and often strikes without warning. Stroke can alter the individual’s capacity for social role function- ing (i.e. as spouse, as lover, as parent, as employee, as storyteller, or as singer in the church choir). Survivors of stroke are often at risk for a loss of functional independence. They must adjust to new definitions of self, and new limitations in physical, psychological and social capacity including cognitive deficits, disturbances in affect, limitations in mobility and movement or any combination of the three [IS, 161. These impairments constitute the loss of an individual’s vital capacities, which in turn have potentially devastating implications for the patient’s perception of competence and efficacy. Stroke can render untenable a self-concept grounded in a sense of autonomy and personal efficacy. While most stroke survivors are eventually able to live in the community, many face a lengthy, difficult and demanding rehabilitative challenge [17]. The essence of that challenge is to reestablish functional indepen- dence while incorporating residual deficits into a new personal identity and assumptive world. The import- ance of identity in stroke recovery is well articulated in the work of Gold [18] and others within the phenomenological tradition who point to processes of impression management during rehabilitation. Because the effects of stroke are potentially stigmatiz- ing, the stroke survivor learns techniques of behavior and self-presentation which allow him to minimize or counter the effects of stigma, and at the same time, reestablish continuity betweeen his/her prestroke identity and postsroke reality. Seen in this light an event initiating a psycho-social transition calls upon an individual to make important and potentially lasting adjustments in their assump- tive world in order to accommodate to the conditions imposed by the transition. Psychologically, the challenge of a transition is one of acceptance and realization that the relationship between self and significant others is no longer the same. Parkes cites the phenomenon of the ghost limb among amputees of an example in which that acceptance and real- ization of a change in life-space can be delayed or incomplete. Sociologically, the challenge of a psychosocial transition involves a shift in social roles and identities. The transition of retirement involves the loss or work related roles and the changed indentities which accompany those roles. A psycho- social transition calls upon an individual to bring into alignment his or her assumptive world with the newly modified life space. The effectiveness or this necessary realignment constitutes effective coping and adaptation. The model presented of stroke recovery as a psychosocial transition has several implications for the empirical study of the recovery process and the role of psychosocial factors in response to traumatic disease events more generally. By conceptualizing stroke recovery as a psychosocial transition, we are able to better understand the psychological and social challenges posed by the disease and how social support is involved in the recovery process. This has the effect of focusing our attention on two factors which are often inadequately addressed in studies of recovery and rehabilitation: the quality and timing of psycho-social resources. The role of social support in psycho-social transitions In research on social support and illness, most studies have focused on the stressful nature of an acute event rather than the process of long term adaptation. In studies of illness and social support, the dominant theoretical approach has been the stress-mitigation model, or what Cohen calls “stress- centered models” [9, 19-211. This approach is summarized briefly here in order to suggest why we believe our psycho-social transition model sheds different and useful light on the recovery process. In the stress-mitigation approach, potentially positive rehahilitive nutcomes are threatened bv the stress The occurrence of a stroke (or cerebrovascular accident) may be described as a potentially useful and instructive event for studying psycho-social tran- sitions because of its profound and potentially lasting impact on the life space and assumptive world of the ._._ -- ._... _ .~.. ...~~~ I250 THOMAS A. GLASS and GEORGE L. MADWX Social support and stroke recovery 1251 associated with the events surrounding illness, result- ing in the disruption of information processing [21], unmet needs resulting in psychological distress [20], reduction in motivation [22] or non-compliance with medical regiments [23] or some other intermediate factor. Many studies have documented the role of psychosocial factors in coping with (or mitigating the deleterious effects of) the stress associated with preg- nancy [24], with myocardial infarction [25-291, with cancer [3@-341 and with stroke [35,36, 171. Out of this large literature has grown a central concern for whether the beneficial effects of social support during recovery are direct or buffering [37-391. While this model has proven to be useful, the psycho-social transition model differs from the stress-mitigation model in its shifts away from illness events as inherently stressful, to a focus on the effectiveness of coping reactions to loss (or gain) and adaptation to change. The psychosocial transition model also calls attention to the importance of a temporal dimension of coping processes which is not a logical require- ment in a stress-centered approach. Our preferred approach emphasizes the rather sudden occurrence of an event, followed by the alteration of life space and the adjustment of the assumptive world. Each phase is logically and temporally related to the previous phase. The analysis of ‘griefwork’ by Bowlby [40], in which the grieving process is divided into four stages (a. numbing, b. yearning and searching, c. disorgan- ization and despair, and d. reorganization) serves as an example. While the stress mitigation models tends to focus on the acute phase of the illness experience, the psychosocial transition model focuses on later and longer term outcomes. We agree with Mestrovic [41] that events are not inherently stressful, but become traumatic only as they relate to the meanings and interpretations attached to them by actors [42]. Similarly, the literature on stroke recovery indicates clearly that physical parameters (e.g. stroke severity) alone do not fully explain variations in perceptions of stress. For example, in an analysis of data from the Framingham study, Labi et al. [43] found that a sizable proportion of stroke survivors reported significant levels of disability despite complete physical restoration. The concept of a psycho-social transition implies a gateway between one identity and another. That aspect of identity which is of particular relevance to our research is demonstrated capacity for self care and social role competence. The use of a general marker of functional status as our primary outcome measure is not without problems. It provides no *The literature on stroke recovery describes a period of approximately 30 days after stroke onset in which spontaneous recovery occurs as a function of the brain’s natural healing mechanisms [ 151. Biological events alone appear to guide this phase of rehabilitation. The impact of psychosocial forces appears only after this acute phase of the illness, including hospitalization and inpatient rehabilitation. direct view of the purely psychological aspects of psychosocial transitions which was central to Parkes analyses of bereavement. However, such a measure is well suited to examining the process of regaining functional independence in the aftermath of an acute illness event. Survivors of stroke are at risk for functional disability and dependency on caregivers for self-care. Optimally, the successful rehabilitative effort allows the stroke survivor over time to regain functional independence, while at the same time making adjustments in the assumptive world which allow for an acceptance of the new range of potential. Our approach focuses not on the perceived adequacy of supportive interactions, or on the stress-buffering functions of support, but rather on the direct impact of various types of supportive actions in promoting greater functional independence. We assume that return of functional capacity is the result of two processes, one biological and one social. Some vari- ation in outcome is due to the severity of the stroke in biological terms.* Remaining variation results from features of the social environment which promote or retard the development of functional independence. Social support is that feature of the environment whose effect is of particular interest here. Stroke and social support: types and timing in transition A modest body of research has sought to examine the impact social support on recovery after stroke [35,36,44-47]. Friedland and McCall [35], for example, have shown that social support from friends, from the community, and from a close personal relationship each have a beneficial effect in stroke patients. In general, however, these studies have fallen short of providing answers to questions surrounding the mechanisms through which this as- sociation operates. Few of these studies have used widely accepted and state of the art measures of both physiological and psychosocial variables, and fewer still have utilized longitudinal methods. In a recent study of psychosocial adjustment following stroke, the authors call for longitudinal designs in studies of stroke recovery [48]. Only then can we begin to learn the answers to central questions. What makes social interaction supportive? Which dimensions of social support are most important, and when during the transition are different types of support most beneficial? A small number of studies have examined the effects of different types of support in stroke recovery. There is modest evidence that emotional support plays a particularly important role in recovery from stroke. Evans and Northwood [44, p. 631 found that support was most effective when it was judged by the patient to meet their emotional needs. A study by Robertson and Suinn [36] has shown a relationship between the rate of progress in stroke rehabilitation and the extent to which families showed a greater 1252 THOMAS A. GLA~.~ and GEORGE L. MADWX degree of one particular component of emotional support; empathy. Much of this work has been done in relative isolation from theories or conceptual models of social support in rehabilitation. Different theories offer different accounts of the mechanisms that underlie the relationship between various types of support and outcomes. Similarly, differing theories make competing predictions about which factors (or types of support) are likely to have the most significant effects. What role does emotional sup- port play in the management of the psychosocial transition? First, Parkes finds a connection between the loss of a body part or physical capacity and feelings or loss and grief. In summarizing the literature, he argues “Most of the physicians and psychiatrists who have written about psychological reaction to loss of a limb maintain that ‘grief’ is to be expected” [IZ, p. 1091. The loss of use of a limb, or sensory organ as the result of paralysis (hemiparesis), the loss of speech (aphasia) or the malfunction of a host of other physical systems may be analogous in stroke. As a result, depression and irritability are often the major stumbling blocks to successful recovery [ 17,491. Emotional support in the form of nurturence, affec- tion, reassurance and the opportunity to ventilate negative feelings, may be an important factor in permitting patients to resolve the grieving process effectively. Or, emotional support may operate by enhancing self-efficacy. Successful rehabilitation appears to require the construction of a new assump- tive world which includes a sense of mastery despite deficit. This may be the heart of the challenge of rehabilitation. Through encouragement and the reinforcement of positive self appraisals, emotional support may play a uniquely important role in facil- itating the transition to a sense of mastery despite deficit. For example, Pearlin et al. [50] found that emotional support reduced depressive symptoms by intervening to increase mastery and self-esteem. Our psychosocial transition model leads us to anticipate that emotional support will be more important than instrumental or informational support during recov- ery. The latter types may promote dependency result- ing in lower self-efficacy. Research in a nursing home setting showed that encouragement (emotional sup- port) was associated with better performance on a cognitive task while direct assistance (instrumental support) was not [5l]. Emotional support may be more important for patients attempting to come to terms with the loss of certain physical limitations, while at the same time learning new techniques of self-care and social role functioning. Moreover, we hypothesize that emotional support will be dtyeren- tially related to functional outcome in stroke over time, and that its beneficial impact rc*ill be greater than other types. Other authors have found a significant relationship between recovery and instrumental support [l7, 52). In the transition to a new assumptive world, the availability of instrumental support poses a dilemma. On the one hand, assistance with daily chores and compensatory role performance is a necessary part of the support process [44]. On the other hand, instrumental support can promote an assumptive world based not on mastery and independence, but on the assumption of helplessness and dependency. For example, McLeroy et al. [52] found a significant negative association between instrumental support and A.D.L. (activities of daily living, a global measure of functional status) at 6 months but did not investigate variations in quantity of support. There may in fact be a threshold effect in the operation of instrumental support. Too little support and patients are confronted with daunting tasks for which they are not well equipped, while with too much support, a self-concept based on mastery and competence is less likely to develop, resulting in a potential atrophy of capability. There are other possible negative consequences associated with a ‘surplus’ of instrumental support. Evidence suggests that the quality of supportive interactions begins to decay over time. For example, in a prospective study of stroke rehabilitation, Schulz and Tompkins [53] report a two-fold increase in negative types of social interaction over a 6 month period of time. Evidence reported by Labi et al. [43] suggests further that patients who live alone reintegrate more favorably than those living in a family context. Overprotective family members may inhibit the construction of an assumptive world based on autonomy, self-worth and mastery despite deficit. In our interviews, several patients voiced similar complaints about unwanted assistance, smothering attention from relatives, or patronizing attitudes among hospital personnel (see also Refs [46, 471). Therefore, we hypothesize that instrumental support will be associated with greater functional independence when provided in moderate amounts as compared to larger amounts or smaller amounts. The beneficial impact of social support is contingent upon the capacity of the individual to make use of the support offered [44]. The nature and timing of social resources, research on rehabilitation increasingly suggests, is most beneficial when targeted to the emerging needs and capabilities of the stroke patient. Informational support, more than the other two types, requires the patient to be able to process information in decision making operations. If such potential is intact, the provision of information reinforces an assumptive world in which the patient is capable of making autonomous decisions and acting independently. If the patient is cognitively impaired or unable to communicate as the result of language deficits, the provision of information may reinforce an assumptive world in which mastery and autonomy are impossible. Thus we hypothesize that the impact of informational support on functional independence will be contingent on (interaction effect) lecel of illness set’erity. Social support and stroke recovery I,53 In summary, while there is evidence for the claim that social support has a generally beneficial effect on recovery from stroke, little is known about how the timing or the amount of available support relate to changes in functional capacity. The effect of support may be delayed or may vary over time during the course of recovery. To date no research exists which has systematically compared the effects of differing types of support or the amount of support provided over time, while controlling for the severity of the stroke itself. It is in the hope of filling these gaps that the research described below was undertaken. While a comprehensive test of the psychosocial transition model is beyond the scope of this analysis, what follows is a first attempt to examine empirically one piece of a larger puzzle. It is our hope that this will demonstrate the utility of a transitional model. DATA ASD MEASURES This research was conducted as a component of a study on prognostic factors in stroke recovery at the Duke/Veterans Affairs Cerebrovascular Research Center. An inception cohort design was used to collect data on first-time stroke patients. In the inception cohort design, all incidence of a particular disease (in this case first stroke) within a particular geographic region over a particular period of obser- vation are eligible for inclusion in the sample at point of disease onset. The sample was assembled from admissions records at three local hospitals; (I) Durham Veterans Affairs Medical Center, (2) Duke University Medical Center and (3) Durham County General hospital. Any hospital admission for stroke in the county will occur at one of these three hospitals (refer to Table I in the results section for a breakdown of patients by hospital). Patients with a significant history of heart disease or who did not receive rapid medical attention after onset of symptoms (less than 24 hr) were excluded. The full sample is comprised of a panel of 64 patients who were admitted to one of the three hospitals with a first stroke. Of that group, 11 died or left the study area, 4 refused to participate in the social support component of the study and 3 of the remaining patients failed to return at least one questionnaire despite repeated attempts at follow-up. Complete data were collected on 46 stroke patients who survived at least six months. Rates of non- compliance with social support questionnaires were comparable to other studies (14% in Ref. [44]). While the resulting cohort is small, it is adequate for purposes of this suggestive inquiry into stroke rehabilitation. This sample contains sufficient vari- ation to allow the testing of our hypotheses. In addition, the high quality of measures used and the availability of repeated measures of both support and the outcome variable, strengthen the quality of the sample. Measurement of key variables Two key measures in research on stroke recovery are stroke severity and functional status. Access to standard and widely accepted measures of these two important parameters constitutes a major strength of this study. While generally less familiar to medical sociologists, the measures of stroke severity and functional status used in this study are con- sidered reliable and standard measures by medical investigators in stroke research. Assessment of sub- jects were made by physician’s assistants under the supervision of physician investigators from the larger study. Stroke severity was measured by the widely used Oxbury Level of Consciousness (LOC) Scale and was collected upon initial examination [54]. The (LOC) scale is assessed upon admission, and is coded from 0 to 5 (0 = active and alert, 5 = unconscious). Mean (LOC) scores for those patients who survived to 6 months was 0.39 reflecting generally milder strokes. This is to be expected given the exclusion of patients who died prior to the end of the study or who were too severely impaired to participate in the study. Changes in functional status, the principal out- come measure, were assessed using the Barthel Index (BI) of activities of daily living collected at 5 days, 30 days, 3 months and 6 months. The (BI) is a well documented and widely used measure of both mobil- ity and ADL skills [55-571. Evidence has suggested that repeated measures of this construct produce a highly efficient and reliable measure of global func- tioning [55]. Four repeated measures were combined to form a single trajectory, or growth curve of recovery. Received social support is assessed using a self- administered questionnaire given to either the patient or, when necessary, a proxy at I, 3 and 6 months after onset. Based on the Inventory of Socially Sup- portive Behavors (ISSB) designed by Manuel Barrera [58, 591, this instrument taps the patient’s perceptions of the frequency of available social support in the previous four weeks. This instrument provides separate scales for three distinct dimensions of social support: (I) Emotional, (2) Instrumental and (3) Informational [cf. 601. The choice of these three subscales is based in part on research by Evans and Northwood [44] who argue that in stroke, social interaction is supportive when (a) support is available to address the patient’s emotional needs (b) patients are assisted with general tasks of dai/.v hing, and (c) patients are given access to specific information about services, employment assistance and rehabilitation. The (ISSB), which consists of items which refer to specific helping behaviors, was attractive because it provides separate subscales which correspond to these three elements. Also, the behavioral specificity of the items (e.g. how often in the last 4 weeks has someone driven you shopping or to an appointment?) requires less interpretation, and is therefore more 1254 THOMAS A. GLAND and GEORGE L. MADLWX Table I. Means. oroDortions and standard deviations bv reswndent source Sample Demographic cariables Age in years Race (% non-white) Sex (% female) Marital status (% non-married) Full (n = 46) 68.9(11.0) 23.9 43.5 52.2 Patient (n = 27) 66.6 (12) 22.2 55.5 59.3 Respondem proxy (n = 19) 72.4 (92) 26.3 26.3 42. I Hospiralirotion uoriables Length of stay in days IX.2 (27) 10.8 (6) Hospital Duke 25 (58.7%) 17(63%) VA 8(17.4%) 5 (19%) DCGH I I (23.9%) 5(190/o) Physiological coriables Stroke severity 0.39 (0.68) 0. I I (0.32) (LOC) scale 5 pt. cont. scale (0 = alert/conscious, 5 = uncon.) Baseline ADL (BI) 60.2 (40. I) 77.4 (36) 6 Month ADL (BI) 80.9 (37.2) 91.1 (32) Received social supporr: (ISSB) Emolional support (sum) 35.5(7.1) 35.0 (7.3) Instrumental support (sum) 16.8 (3.7) 16.6 (4.0) Informational support (sum) 10.4(3.4) 10.3 (3.2) Standard deviations in parentheses. optimally suited to applications in which proxy respondents are used. The advantages of the (ISSB) include high inter- tester reliability and good sensitivity. A recent review by Heitzmann and Kaplan [61] reports that the (ISSB) has test-retest reliability of 0.88 and internal reliability of 0.93 on first admission. The (ISSB) has been adapted to fit the needs of this research. Irrelevant items have been omitted. This procedure follows the recommendation of several researchers who recommend modifying items from established scales to meet specific research needs [62, 631. STATISTICAL PROCEDURES In the following analyses, repeated measures MANOVA (also known as ‘growth-curve’ models) was used to examine the impact of social support resources on changes in Activities of Daily Living (ADL) over a 6 month period following first time stroke. The repeated measures design is robust with small samples, and has the advantage that it separates variability between subjects from experimental error. This has the effect of making maximum use of the information contained in the data (641. MANOVA was used to analyze determinants of the shape of trajectories of functional status in the aftermath of serious illness. Hypotheses about how these trajectories differ by levels of support and by levels of severity across support types were tested using ‘MANOVA using polynomial contrasts assumes Type H covariance structures. For each model reported here, a sphericity test was performed to determine whether this assumption was met. In all cases, the test confirms that the matrices are Type H. 28.9 (40) IO (53%) 3 (16%) 6 (32%) 0.79 (0.85) 34.4 (34) 64.7 (40) 36.3(7.1) 17.1 (3.6) 10.6 (3.7) univariate tests for within-subjects effects in a multivariate analysis of repeated measures [65, 64].* Patients were divided into two groups according to level of severity (mild vs moderate/severe) according to (LOC) scale scores. Patients were then divided at the 30th and 70th percentile into three levels of received support for each type (low, medium and high). This yields a between subjects design matrix of 2 x 3 or six cells. Also, polynomial contrast tests were used to test for trends in the data over time. First linear, then quadratic and higher order trend lines were fit to the four measures of the BI. This analysis produced two kinds of information. For each term in the model, the F-test is presented for the overall significance of the effect (only the more conservative Wilkes Test is shown). Also presented is the Roy- Bargman Step-down F, which tests whether univari- ate trend components (linear, quadratic and cubic) are non-zero [64, p. 223,661. If the over-all F-test for the term is not significant, then the trend analyses are irrelevant and uninterpretable. RESCLTS Table 1 presents descriptive statistics for the sample of patients. Means and sample distributions are presented for both the full sample and broken down by respondent source (patients vs proxy). Assessing the representativeness of a sample of stroke survivors is difficult in the absence of standard characterizations of representative populations avail- able in the literature. In general, comparison of this cohort with data from other studies suggests that our sample has suffered slightly less severe stroke (mean (LOC) score of 0.39 compared with 0.77 in Ref. [52]), and is more likely to be white (24% non-white). In other respects, these data appear to be quite similar. Social support and stroke recovery 1255 Comparisons among those patients who were primary respondent and proxy respondents are reassuring. The group with proxy respondents were slightly older (72 compared to 67) more likely to be men (73% vs 44%) and had longer hospital stays (1 I vs 29 days). In addition, as one would expect, the group with proxy respondents suffered more severe strokes on average and scored higher on the (LOC) scale and lower on both baseline and 6 month BI. However, none of the above variables relied on proxy responses; the (LOC) and the BI are all collected by the P.A. on the study. Only the social support variables were collected by proxy. Patient and proxy responses on the three dimensions of social support are almost identical. This is partly the result of using the (ISSB) which features items which are behaviorally specific and are thus less vulnerable to variations in interpretation. Inspection of corre- lations in Table 2 provides evidence for the view that while closely related, these three types of support are not identical. All correlations are moderately strong ranging from 0.499 to 0.63 1. Our measures provide a view of three types of overlapping, yet analytically separable support. Results of our analyses are presented in two steps. First, in order to examine the effect of social support by types, results of MANOVA models are presented. Secondly, the same results are shown graphically for the bivariate relationships in order to shed light on the importance of timing during recovery. Support specificity: growth curve analyses Emotional support. Table 3 presents Wilks’ Lambda tests for the analysis of the impact of emotional support. In model I, Term (1) indicates that stroke patients vary across time in their level of functional capacity as measured by the Barthel index Table 2. Simple corrcla~ions among types of social support Variable Emotional Instrumental Informational Emotional 1.00 0.600 0.63 I Instrumental 1.00 0.499 Informational 1.00 N = 46. (P = 0.000). Term (2) suggests the presence of an interaction among time since stroke and stroke severity (P = 0.000). Stroke patients vary according to level of severity of initial injury. Trend analysis extracts a significant cubic trend in this relationship. Term (3) is also significant (P = 0.006) indicating that mean trajectories of functional status vary signifi- cantly across levels of social support. However, unlike stroke severity, emotional support is best modeled with a straight line indicating a linear trend in this relationship. Differences in trends in the trans- formed dependent variable are due to differences in the shape of effect over time (see discussion). Term (4) is not significant indicating that no three-way inter- action exists between emotional support and severity. Instrumental support. Table 4 presents results of a similar analysis examining the unique impact of instrumental support. In this model, as in model 1, the effect of time and severity are both significant (P = 0.00 and P = 0.016 respectively). A significant quadratic trend is detected in the polynomial contrast tests (P = 0.024). In model 2, term (3) indicates that mean trajectories vary significantly across levels of instrumental support (P = 0.016). Again, only the linear component of the trend is found to be signifi- cant. Finally, in model two, the three-way interaction among instrumental support, severity and time is not significant. Neither the association between instru- mental nor emotional support are contingent upon stroke severity. Table 3. Multiple analysis of variance (growth curve analysis) on trajectories of recovery. Averaged tests of significance using sum-of-squares and cross-products matrices: Source of variation: Term: Emotional support (Model I) Exact F (Wilks’ Test) Significance of F (I) Time Polynomial Contrast F-Test (2) Severity x Time Polynomial Contrast F-Test (3) Emotional Support x Time Polynomial Contrast F-Test (4) Severity x Support x Time Polynomial Contrast F-Test 17.4 (0.42) F Linear 47.4 Quadratic 26. I Cubic 4.88 4.94 (0.42) F Linear 9.8 Quadratic 6.7 Cubic 5.5 3.29 (0.63) F Linear 7.5 Quadratic I.9 Cubic 2.2 -not significant- F Linear NA Quadratic NA 0.000’ Sig. o.ooo* o.ooo* 0.002’ o.ooo* Sig. 0.003* 0.013’ 0.024’ 0.006. Sig. 0.002’ 0.374 0. I25 Sig. *Significant at P Q 0.05 190’0 880’0 P80’0 ‘S!S +sso.o L09’0 08E’O fgg0 990’0 990’0 .ZEO’O l SOO’O ‘%!S l ZlO'O OS I’0 l 000‘0 l OOO’O ‘%!S l OOO'O 6’2 3!9”3 9’i >!,eJpen~ 9.5 JFs”!-j _d ,sal_j ,sEd,“0~ le!WOU6&, (ZL‘O) L I 'Z aru!l x wddns x .+.mwg (p) SO 5!9”3 0’1 3!leJpenb 5’9 ,&?a”!~ .lJ Isal-, WJ,“o3 le!“JOU~[Od (EL'O) 80’2 aro!l x uoddns ,euo!,sur~o_,u~ (E) 9’E s!q”3 6’P J!IeJpen~ 6’8 ,W”!l zl ,sal_d ,SW,"O~ [t?!UOUXlO,, (PL'O) 8 I't xm~ x Ll&mS (Z) SI’Z s!q”3 L'li s!,eJpPnb P’W ,eaug 2/ ISal- d ,SW"O~ lE!UIOUdlOd (WO) 8‘51 am!1 (I) .1loddns [EUO!lWLLlOJU! ql!M SUO!l3EJa)U! 2ql U! pUnOJ 3J3M SpuaJl ou ‘Ll!JaAas pue wI!lJO slsagaaql JOJ pawlap aJaM spua~~ ywpenb pue nau!~ W?XJ@!S a[!q~ .~1!JahasaSE!aS!p ~0 ~euO!,!puo~s! ‘sadiil OM1 ~aqloaql aygun ‘IJoddns ~t~o~~w~~o~u~~o Isedur! aq, ‘suogel -sadxa JnO ql!M )WlS!SUO~ 'aLLI!] pUE @J2AaS SS%Ifl! ‘)Joddns [~uo!~eu~~oju! ~UOLUE UO!I~EJ~)U! REM-aaJy) lUe3L@!!S iCI[WI$!Jl?W E JOJ 23Uap!ha ap!AoJd Elep 3Saql '(SSO'O= d)BOW 50.0 aql 01 JaSO[3 UaAa S! UO!]3EJ3lU! I&M %Jql aql (fL[‘Z)d JO an[eA Jaq%!q dpq%lS e qI!M XU!l pUe ~!:~ACJS ']Joddns ~euo!l -r?UIJOJU! %IOLUE UO!PEJ3lU! h'M-aaJql lUE3I@+ e ~0 hg!q!ssod aql sisal (p) UJaL '(990'0 = d) q%!q A[qelda3sEun s! 0 .+?nl3&? s! d!qsuogE[aJ paz!sa -qlodLq aqi ieqi dig!qeqoJd aqi ieqi apn[3uo3 01 pal all? aM (SO.0 = D) a(nJ UO!S!Zk?p 6J&?Jl!qJE UE %I!Sn ye~y!u8~s IOU ~[~WI!%JELII s! ‘IJoddns [\?UO~?LLLIO~I! Jo [aAal JOJ K’?Jja U!“” aql ‘(f) UX’L ‘(ZIO’O = (i) SalEJ %X!Jag!p ]E JaA03aJ A]!JaAaS JO SIaA2l lUaJag!p ql!M sluayed leql %U!lSa%nS ~IIE~~!U~!S os[e s! uoye -~alu! L~!J~A~S Lq auy aqL ~(000~0 = d) lue3y!u%!s s! au1!1 JO 13a#a U!I?LLI aql ‘KI!I?~V 3suarJadxa ssau[l! aql JO Sla3E'~ Jay]0 hr! JO ~JaAO3aJ ‘CCfOJlS lnOq'l3 SpUa!JJ pUE I(l!UI&?j ‘JJE')S Aq pap!AOJd UO!lWLIOJU! JO isedru! aq, sdel S!S~[BUB s!qL .IJoddns leuo!i -ElUJOJU! JO S)32#2 aql %J!U!UIEXZJ ‘[apOUI pJ!qI JnO JO sl[nsaJ sMoqs s a[qeJ_ .uoddns ]vuopnu~o/ul 'S!S 00’ I 058’0 *I 10’0 ‘S!S P90’0 .PZO’O r900’0 ‘%!S 801'0 .000'0 l OOO’O ‘B!S VN 5!9"3 VN ye,penb VN ,ea"!l Ll -lue3y!u8!s ,O"- PP.Z J!W3 Z’O 3!,e.Jpenb O‘S Jea”!~ d (L9’0) 08’2 P9‘E JFl”3 S’S J~,rJpen~ P’8 Jea”!+j d (LL'O) 88X Z8.Z 2!9"3 S'ZZ s!,eJpen~ SW nxl”!l _4 (EP'O) 9x-I Social support and stroke recovery 1257 Support timing: graphic anal_wis Having established that each of three types of social support has a significant effect on recovery of functional status among survivors of first stroke, we may wish to ask further questions regarding the shape and nature of these relationships. Figures l-3 present in graphic form the bivariate relationships between support and outcome. However, due to the presence of a three-way interaction among time, severity and informational support caution must be taken in interpreting Fig. 3. The impact of the remaining two types of support is more readily interpretable in the bivariate form. Emotional support. Figure 1 shows a fairly clear pattern in which patients reporting high and medium levels of emotional support progress from baseline on an upward trajectory of recovery across the four measurement periods. Early in recovery, differences in ADL among levels of support are quite small. This may account for why studies which look only at predictors of discharge functional status have missed the sizable impact of social sup- port. Trend analysis presented above suggests that the impact of social support is gradual and linear over time. Patients reporting a high level of emotional sup- port achieve the steepest slope and the greatest final level of functional independence as compared with the other two groups. This is unexpected given that patients with high support begin with the lowest average functional status. More severely impared patients appear to either garner more emotional support, or are more positively impacted by the support they are provided. The latter interpretation would be consistent with past research showing 90 - _^_-~---- -.-- 85 - ,/ __+-------__ J 80 - Level of emottonol support BoselIne 1 2 3 4 5 6 Time since stroke (months) Fig. I. Trajectories of recovery from stroke changes in functional status by emotional support. 105 _ Level of Instrumental support %I- 0 Low support 45 - 0 Medwm support A High support 40- 35 - 30 d- I I I I I BoselIne I 2 3 4 5 6 Time since stroke (months) Fig. 2. Trajectories of recovery from stroke changes in functional status by instrumental support. that supports are more effective for those in greatest need [50]. Patients with the lowest levels of emotional support exhibit a similar pattern of recovery for the first 1-2 months. Then, after reaching an apparent plateau, this group begins to decline gradually. The period between 30 and 60 days post-stroke may be of crucial importance in the timing of social support resources. lnstrumenlal support. Figure 2 presents results of the analysis of instrumental support. This figure 105 L loo - 95 - 90 - 65 - Level of lnformotlonal support 0 LOW support : 55- 0 Medium support 50 - d. High support 45 - 40- 35 - 30 I I I I I BaselIne 1 2 3 4 5 6 Time since stroke (months) Fig. 3. Trajectories of recovery from stroke changes in functional status by informational support. 1258 THOMAS A. GLASS and GEORGE L. MADDOX provides compelling evidence for the need to disentangle types of social support. Patients reporting low levels of instrumental support exhibit a pattern which is similar to the group low on emotional support (normal progress in the first month, leading to an early plateau and later, a decline in functional status). In the remaining two groups we see evidence of what may appear to be a paradoxical effect. Patients reporting moderate amounts of instrumental support show a curve which is both steeper and has a higher plateau than the high support group. Unlike emotional support, the benefits of moderate levels of instrumental support appear to outweigh those of higher levels (see Discussion). Informational support. Figure 3 presents results for informational support. The impact of informational support is less pronounced in this representation. Again, for ease of presentation, the impact of severity is not presented. Consistent with the other two support types, patients reporting low levels of infor- mational support do as well as others prior to one month post-stroke, at which point recovery slows, and begins to decline. The overall shape of average trajectories of recovery is quite similar for high and medium support groups. However, due to the significant three-way interaction we urge caution in interpreting this figure. DISCUSSION Overall, we found that trajectories of recovery differ dramatically across levels of social support in this panel of stroke survivbrs. No discernable variability in outcome across support groups was detected during the first month (typically the inpatient phase), supporting our belief in the import- ance of the timing of social support. The contribution of support to recovery may be eclipsed or diminished in cross-sectional designs or in research which uses discharge measures as end points. The impact of social support appears to be developmental. Patients with high or medium amounts of support, regardless of type, exhibit better trajectories of recovery over time. A significant upward linear trend was detected in all analyses. Those with inadequate support experi- ence spontaneous recovery, but stop improving and in some cases decline in functional capacity. Among the most significant findings of this analysis is that the nature of the relationship between support and recovery varies across support types. Designs which fail to differentiate among support types may miss significant variations in these effects. When differentiated into types and examined longitudinally, a clearer picture of the complex relationship between .social integration and recovery from illness begins to emerge. Emotional us instrumental support Consistent with our first hypothesis, emotional support was found to be a powerful predictor of the shape and extent of recovery of function. These findings strengthen other reports which suggest the importance of emotional support [44, 361. Emotional support may help patients adjust to the disfiguring consequences of paralysis. Labi et al. [43] speculate that, in stroke, body image may be a central under- lying mechanism in the support process. Emotional support may meet a specific need by reinforcing self-esteem and allowing stroke patients to feel accepted despite an altered body image. As hypothesized, the impact of instrumental sup- port is quite complex. These data suggest that the benefits of instrumental aid accrue from moderate quantities of support. As with emotional support, the impact of material aid, physical assistance and help around the house seems to be gradual and develop- mental. But, as expected, these analyses reveal that patients who report moderate levels of support do better than those who report the highest or the lowest levels of support. Mulley (671 appears to be correct in concluding that large amounts of instrumental support, particularly in less severe stroke may, at times, play a negative role. Patients who receive high levels of instrumental support may be less active than those patients who in receiving moderate amounts of support, must do more for themselves. Findings indicating a high prevalence of over-protectiveness and unrealistic expectations in spouses of stroke patients lends support to this interpretation. Modest amounts of instrumental support may result in greater stimulation, and more optimal rehabilitative outcomes. Too much instrumental support may engender dependency (i.e. learned helplessness) and lack of motivation in achieving rehabilitative goals. Theoretically, these findings suggest that in the transition to a reintegrated identity, instrumental support and emotional support may play differing roles. In addition, it is suggestive that emotional support is associated with more rapid and more extensive recovery. Informational support considered. Consistent with our third hypothesis, the role of informational support does appear to be mediated directly by ill- ness severity. While only marginally significant, the interaction among time, severity and informational support suggests that the efficacy of this dimension of support may be uniquely contingent upon the patient’s cognitive capabilities (e.g. memory, language, coping and appraisal processes). Patients who have suffered more severe strokes are less able to make use of informational support as a result of a greater level of impairment. This is the only evidence available in these data of a buffering effect (i.e. stress-mitigation model). The impact of the remain- ing two types of support are more direct and less sensitive to illness severity (stress) as measured here. This interpretation is intuitively appealing but must remain speculative. We cannot rule out the possibility that emotional and instrumental support may operate not through cognitive pathways, but rather through Social support and stroke recovery 1259 either physiological ones (e.g. endocrine processes) or more strictly psychosocial mechanisms. Source of support as possible mediating factor. The impact of emotional and instrumental support may differ because the two types of support come from different sources. Seeman and Berkman [68] have suggested, for example, that older adults living in the community rely more on children for instru- mental support and on close friends for emotional support. In addition, they report interestingly that the presence of a spouse is not significantly related to the availability of instrumental support, as one might predict. Rather, their data suggest that the most important predictor of the availability of instru- mental support was the presence of a confidant (the spouse in only 37% of married persons in their sample). Role theory might suggest that for older persons who rely on their children for instrumental support, the receipt of this support may conflict with role expectations relating to being a parent. This conflict, or role strain, may reduce the efficacy of instrumental support when provided by children. Several researchers have noted the importance of source of support in studies of illness recovery [28]. In an analysis of the sources of social support in stroke recovery, Friedland and McCall [35] have shown that support from the community (non-kin) is more closely associated with outcome than from other sources. In addition, they note a conspicuous absence of the impact of professional sources of support. Unfortunately, this is beyond the scope of our data. A valuable addition to these analyses would be to incorporate source of support (kin, non-kin, spouse, friends, religious and business associates) into our model. Issues of causal order. It is increasingly clear that the availability of social support can be considered neither stable, nor independent of transition events themselves [69]. One limitation of our study is the lack of premorbid measures of social support. These results, while robust, may be the result of con- founding influences. We are unable to determine the extent to which stroke itself damages or potentiates the mobilization of social support. The literature is equivocal. One study concluded that supports avail- able to both stroke patients and their primary sup- port providers decline substantially over time [53]. The occurrence of a serious illness such as stroke, however, need not lead to a decline in social sup- port. In a longitudinal study of the effect of major depression on the availability of support, Blazer [70] found that support increased during the course of illness. In our data, reported levels of support remain quite stable at least between stroke onset and the end of the observation interval. At least one other longitudinal study has found social support to remain stable over time [71]. Clinical significance. Stroke has long resisted clinician’s attempts to predict outcome and manage patients toward a return to premorbid level of func- tioning. Several studies (which have not generally included measures of social support) have found the characteristics of stroke itself (size and site of lesion) are poorly linked to functional outcome [72, 731. In addition, the evidence regarding the efficacy of rehabilitation therapy remains equivocal [15]. If characteristics of the illness itself are poor prognostic factors, what explains variations in rehabilitative outcomes? We argue that these data support the need for greater attention to psychosocial factors in the process of recovery and rehabilitation. In view of the downward trajectory of patients with low social support, clinicians may be advised to screen for social isolation as a risk factor for poor out- come in stroke rehabilitation. These data suggest that both clinicians and researchers need to pay greater attention to the particularities of support types, the amount or quantity of support received, and the timing of support provision during illness course. Moreover, case managers should be wary of underestimating the rehabilitative potential of patients who do poorly on biomedical prognostic indicators. In the data of this study, with the aid of adequate social support, even patients who had severe initial injury were able to achieve favorable outcomes. Conclusion Our results confirm earlier finding that in general, social support is an important predictor of outcome in recovery from stroke [35,36,44,46]. Further, these analysis have added to an understanding of the role of social support by examining the quantity, types and timing of support, Researchers who fail to take account of the dynamic quality of the rehabilitation process, as well as the specificity of support type may be lead to underestimate the impact of psychosocial factors. In addition, these data support the hypoth- esis that the impact of support varies by amount. Findings indicate that with emotional support ‘more is better’. In contrast, for instrumental support, ‘moderation in all things’ may be a better rule of thumb. Further, the efficacy of informational support depends on illness severity while it does not for the other two types. The psychosocial transition model is well suited for the exploration of specific mechanisms in the association between psychosocial resources and recovery from illness. Our model focuses on longer term outcomes, on the direct impact of social sup- port (rather than stress-buffering) and on the signifi- cance of the type, timing and amount of received support. A psychosocial transition model may have wide application in the study of other illnesses which induce important changes in the life-space of individuals. A complete and formal test of the transitional model is beyond the scope of these data and this analysis. We hope instead to offer a reason- able supplement to the dominant stress-mitigation approach. The shift in attention toward issues of self I260 THOMAS A. GLAXS and GEORGE L. MADDOX definition, social role identity, and related social- psychological mechanisms offers new hope for a more theoretically informed approach to the study of social support and health. Ackno,ciedgemenrs-This research was supported through the Veterans Administration Health Services Research and Development Field Program, Durham, NC. Substantial assistance and support were provided by Dr David B. Matchar and Dr John R. Feussner of the Durham VA- HSRD. The authors wish to thank Dr Fred Wolinsky for helpful comments on an earlier dFaft. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. Il. 12. 13. 14. 15. REFERENCES 25. House J. S.. Robbins C. and Metzner H. L. The association of social relationships and activities with mortality: prospective evidence from the Tecumseh community health study. Am. J. Epidemiol. 116, 123-140, 1982. Berkman L. F. and Syme S. L. Social networks, host resistance, and mortality: a nine year follow-up study of Alameda County residents. Am. J. Epidemiol. 2, 186-204, 1979. Blazer D. Social support and mortality in an elderly community population. Am. J. Epidemiol. 115,684694, 1982. 26. 27. 28. House J. S., Landis K. R. and Umberson D. Social relationships and health. Science 241, 540-545, 1988. Cobb S. Social support as a moderator of life-stress. PJyclrosom. Med. 38, 300-314, 1976. Hibbard J. H. Social ties and health status: an examin- ation of moderating factors. Hlth Educ. Q. 12, 23-34, 1985. 29. 30. Cohen S. and Wills T. A. Stress, social support and the buffering hypothesis. Psychol. Bull. 98, 310-357, 1985. Kessler R. C. and McLeod J. D. Social support and mental health in community samples. In Social Support and Healfh (Edited by Cohen and Syme). Academic Press, Orlando, 1985. DiMatteo R. M. and Hays R. Social support and serious illness. In Social Networks and Social Support (Edited by Gottleib B). Sage, Beverly Hills, CA, 1981. 31. 32. 33. Coyne J. C. and Boiger N. Doing without social support as an explanatory concept. J. Sot. clin. psych. 9, 148-158. 1990. Pearlin L. I. The sociological study of stress. J. Hlfh sot. Behac’. 30, 241-256, 1989. Parkes C. M. Psycho-social transitions: a field of study. Sot. Sci. Med. 5, 101-115, 1971. Lewin K. A Dynamic Theory of Personality. McGraw, New York, 1935. Parkes C. M. Attachment and the prevention of mental disorders. In The Place of Attachmenr in Human Behau- ior (Edited by Parkes C. M. and Stevenson-Hinde J.), pp. 295-309. Tavistock Publications, London, 1982. Dombovy M. L., Sandok B. A., Basford J. R. Rehabilitation for stroke: a review. Srroke 17, 363-369, 1986. 34. 35. 36. 37. 16. Gresham G. E. Stroke outcome research. Swoke 17, 358-360, 1986. 17. Brocklehurst J. C.. Morris P.. Andrews K.. Richards B. 38. and Laycock P. Social effects of stroke. Sic. Sci. Med. 15A, 35-39, 1981. 18. Gold S. J. Getting well: impression management as stroke rehabilitation. Quaf. Social. 6, 238-254, 1983. 19. Cohen S. Psychosocial models of the role of social support in the etiology of physical disease. Hlth Psychol. 7, 269-297, 1988. 20. Jacobson D. E. Types and timing of social support. J. Hlrh sot. Behac. 27, 250-164, 1986. 21 22. 23. 24. 39. 40. 41. Hamburg B. A. and Kill&a M. The surgeon general’s report on health promotion and disease prevention. Washington DC. In Healthy People: Background Papers. U.S. Government Printing Office, DHEW (HHS) Pub. # 79-5507lA. 1979. Hyman M. D. Social psychological determinants of patient’s performance in stroke rehabilitation. Arch. Phys. Med. Rehabil. 53, 217-226, 1972. Evans R. L., Bishop D. S., Matlock A. L., Stranahan S., Smith G. G., Halar E. M. Family interaction and treatment adherence after stroke. Arch. Phys. Med. Rehabil. 68, 513-517, 1987. Nuckolls K. G., Cassel J. and Kaplan B. H. Psycho- social assets, life crisis and the prognosis of pregnancy. Am. J. Epidemiol. 95, 431441. 1972. Finlayson A. Social networks as coping resources: lay help and consultation patterns used by women in husbands’ post-infarction career. Sot. Sci. Med. 10, 97-103, 1976. Ruberman W., Weinblatt E., Goldberg J. D. and Chaudhary B. S. Psychosocial influences on mortality after myocharidal infarction. N. Engl. J. Med. 311, 552-559, 1984. Garrity T. F. Vocational adjustment after first M.I.: comparative assessment of several variable suggested in the literature. Sot. Sci. Med. 7, 705-717. 1973. Croog S. H., Lipson A. and Levine S. Help patterns in severe illness: the roles of kin network, non-family resources, and institutions. J. Marriage Family 34, 32-41, 1972. Siegrist J., Dittmann K., Rittner K., Weber 1. The social context of active distress in patients with early myocardial infarction. Sot. Sci. Med. 16,443-453, 1982. Bloom J. R. Social support, acomodation to stress and adjustment to breast cancer. Sot. Sci. Med. 16, 1329-1338, 1982. Peters-Golden H. Breast cancer: varied perceptions of social support in the illness experience. Sot. Sci. Med. 16, 483-491, 1982. Funch D. P. and Mettlin C. The role of support in relation to recovery from breast surgery. Sot. Sci. Med. 16, 91-98, 1982. Lindsey A. M., Norbeck J. S., Carrieri V. L., and Perry E. Social support and health outcomes in post- mastectomy women: a review. Cancer Nursing 4, 377-384, 1981. Maher E. L. Anemic aspects of recovery from cancer. Sot. Sci. Med. 16, 907-912, 1982. Friedland J. and McCall M. A. Social support and psychosocial dysfunction after stroke: buffering effects in a community sample. Arch. Phys. Med. Rehabil. 68. 475480, 1987. Robertson E. K. and Suinn R. M. The determination of rate of progress of stroke patients through empathy measures of patient and family. J. PsIchosom. Res. 12, 189-191, 1968. Cohen S. and McKay G. Social support, stress and the buffering hypothesis: a theoretical analysis. In Hand- book of Psychology and Health, Vol. 4, Social Psycho- logical Aspects of Health (Edited by Baum A., Taylor S. E. and Singer J. E.). Lawrence Erlbaum Associates, Hillside, 1984. Lin N., Woelfel M. W. and Light S. C. The buffering effect of social support subsequent to an important life event. J. Hlth sot. Behav. 26, 247-263, 1985. Wilcox B. L. Social support, life stress, and psycho- logical adjustment: a test of the buffering hypothesis. Am. J. Commun. Psych. 9, 371-386, 1981. Bowlby J. Arrachmenl and Loss, Vol. I: Auachment. Pelican, Harmondsworth, 197 1. Mestrovic S. G. A sociological conceptualization of trauma. Sot. Sci. Med. 21, 835-848, 1985. Social support and stroke recovery 1261 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. Mestrovic S. G. and Glassner B. A Durkheimian hypothesis on stress. .Soc. Sci. Med. 17, 1315-1327, 1983. Labi M. L. C., Phillips T. F. and Gresham G. E. Psychosocial disability in physically restored long-term stroke survivors. Arch. Phys. Med. Rehabil. 61,561-565, 1980. Evans R. L. and Northwood L. K. Social support needs in adjustment to stroke. Arch. Phys. Med. Rehabil. 64, 61-64, 1983. Christie D. and Lawrence L. Patients and hospitals: a study of the attitudes of stroke patients. Sot. Sci. Med. 12, 49-51. 1978. Kinsella G. J. and DulTy F. D. Psychosocial readjust- ment in the spouses of aphasic patients. Stand. J. Rehabil. Med. Il, 129-132,-1979. _ Stenhens M. A. P.. Kinnev J. M.. Norris V. K. and R&hie S. W. Social networks as assets and liabilities in recovery from stroke by geriatric patients. Psychol. Aging 2, 125-129. 1987. Thompson S. C., Sobolew-Shubin A., Graham M. A. and Janigian A. S. Psychosocial adjustment following a stroke. Sot. Sci. Med. 28, 239-247, 1989. Parikh R. M., Robinson R. G., Lipsey J. R., Starkstein S. E., Federoff J. P. and Price T. R. The impact of poststroke depression on recovery in activities of daily living over a 2-year follow-up. Archs Neural. 47, 785-789, 1990. Pearlin L. I., Menaghan E. G., Lieberman M. A. and Mullan J. T. The stress process. J. Hllh Sot. Behav. 22, 337-356, 1981. Avorn J. and Langer E. J. Induced disability in nursing home patients: a controlled trial. J. Am. Ger. Sot. 30, 397400, 1982. McLeroy K. R., DeVellis R., DeVellis B., Kaplan B. and Toole J. Social support and physical recovery in a stroke population. J. Sot. Person. Relafionships 1, 395-413, 1984. Schulz R. and Tompkins C. A. Life events and changes in social relationships: examples, mechanisms and measurement. J. Sot. clin. Psych. 9, 69-77. 1990. Oxbury J. M.. Greenhall R. C. D. and Granger K. M. R. Predicting the outcome of stroke: acute stage after cerebral infarction. Br. Med. J. 3, 123-127, 1975. Granger C. V., Hamilton B. B., Grasham G. E. and Kramer A. A. The stroke rehabilitation outcome study: part II. Relative merits of the total Barthell Index score and a four-item subscore in predicting patient out- comes. Arch. Phys. Med. Rehabil. 70, 100-103, 1989. Granger C. V., Dewis L. S., Peters N. C.. Sherwood C. C. and Barrett J. E. Stroke rehabilitation: analysis of repeated Barthell Index measures. Arch. Phys. Med. Rehabil. 60, 14-17, 1979. Wood-Dauphinee S. L., Ivan Williams J. and Shapiro S. H. Examining outcome measures in a clinical study of stroke. Stroke 21, 731-739, 1990. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. Barrera M. A method for the assessment of social support networks in community survey research. Connecfions 3, 8-13, 1980. Barrera M., Sandier 1. N. and Ramsey T. B. Preliminary development of a scale of social support: studies on college students. Am. J. Communiry Psych. 9, 435-7, 1981. House J. S. Work Stress and Social Support. Addison- Wesley, Reading, MA, 1981. Heitzmann C. A. and Kaplan R. M. Assessment of methods for measuring social support. Hlth Psychol. 7, 75-109, 1988. Tardy S. Social support measurement. Am. J. Communiry Psychol. 13, 187-202, 1985. House J. S. and Kahn R. Measures and conceots of social support. In Social Support and He&h (Edited bv Cohen and Svme). Academic Press. New York, 1985. I I Bock D. R. Univariate and multivariate analysis of variance in time-structured data. In Longifudinal Research in the SIU~V of Behavior and Development (Edited by Nesselroade J. R. and Baltes P. B. Academic Press, New York, 1979. Rogosa D. Myths about longitudinal research. In Methodological Issues in Aging Research (Edited by Schaie, Campbell, Meredith and Rawlings). Springer, New York, 1988. Campbell R. T., Mutran E. and Parker R. N. Longitudinal design and longitudinal analysis: a com- parison of three approaches. Res. Aging 8, 48&504, 1986. Mulley G. P. Practical Management of Srroke. Medical Economics, Oradell, NJ, 1985. Seeman T. E. and Berkman L. F. Structural character- istics of social networks and their relationship with social support in the elderly: who provides support. Sot. Sci. Med. 26, 737-749, 1988. Atkinson T., Liem R., Liem J. H. The social costs of unemployment. J. Hlth Sot. Behav. 27, 317-331, 1986. Blazer D. G. Impact of late-life depression on the social network. Am. J. Psychiaf. 140, 162-166, 1983. Aneshensel C. S. and Frerichs R. R. Stress, support and depression: a longitudinal causal model. J. Community Psychol. 10, 363-376, 1982. Santus G. A., Ranzenigo A., Caregnato R. and Inzoli M. R. Social and family integration of hemiplegic elderly stroke patients I year after stroke. Stroke 21, 1019-1022, 1990. Feigenson J. S., McDowell F. H., Meese P., McCarthy M. L. and Greenberg S. D. Factors influencing out- come and length of stay in a stroke rehabilitation unit: part I. Analysis of 248 unscreened patients-medical and functional prognostic indicators. Slroke 8,651-662, 1977. SW Y,ll-o