Methodological issues in social support and social network research

Methodological issues in social support and social network research

Sot. Sci. Med. Vol. 26, No. 8. pp. 863-873, 1988 Printed in Great 0277-953618853.00 + 0.00 Copyright 0 1988 Pcrgamon Rem pk Britain.All rightsreserv...

1MB Sizes 6 Downloads 147 Views

Sot. Sci. Med. Vol. 26, No. 8. pp. 863-873, 1988 Printed in Great

0277-953618853.00 + 0.00 Copyright 0 1988 Pcrgamon Rem pk

Britain.All rightsreserved

METHODOLOGICAL ISSUES IN SOCIAL SUPPORT SOCIAL NETWORK RESEARCH

AND

PATRICKo%ILLY Cardiovascular Institute, Boston University School of Medicine, 80 E. Concord St., Boston, MA 02118, U.S.A. Abstract-With the plethora of articles describing a relationship between social support and/or social network and health status, it was considered useful to take stock of the current status of research in this area, focusing on two critical methodological issues: clarity of definition, and validity and reliability of the measurement instruments. Of the 33 instruments reviewed only modest agreement was found in conceptual definition, and frequently the concepts were not defined or ill-defined. Of particular concern is the definitional confusion between social support and social network. Variables used to operationalire these concepts confirm this lack of specificity and ambiguity in definition. As for validity and reliability, many of the investigators reported no data on these issues; others provided information that only modestly supported the validity or reliability of their instrument. The conclusion of this assessment suggests the need to clarify the essential elements of social support and social networks in order to better distinguish between the behavioral (support) and structural (network) variables that may be affecting health status. A question is also raised as to the likelihood of a single questionnaire being designed that would accurately measure the perceptions of support or supportive behaviors in the variety of supportive research that will continue to be studied. Finally, more rigorous standards need to be used by investigators in establishing the validity and reliability of the instruments in order to improve their predictive utility. Key wordr-social

support, social network, conceptual definition, operationalization,

INTRODUCTION In the relatively

few years following Cassel’s [l] seminal paper on host susceptibility to illness and the social environment, theoretical statements [2-6], major reviews [7-l I], and numerous research studies have been published on the often ill-defined concepts of social support and social networks. Using a variety

of direct and indirect measures, investigators have noted an association between social support and social networks and overall mortality [ 12, 131,chronic diseases [14-l 91, depression and psychological distress [20-271, psychiatric disorders [28-321, bereavement [33,34], and other health and medical problems [3543]. While most of these studies have provided evidence of either the direct and/or indirect (‘buffering’) effects of social support and social networks on health status, the data have often been uncritically accepted. Under close scrutiny, a number of methodological weaknesses in this research becomes apparent. Some of these methodological issues were addressed in a recent article by Orth-Gomer and Unden [44], which provides a detailed examination of 16 instruments used to measure social support and social networks. The following review extends this type of analysis by examining first the use of conceptual and operational definitions in the development of 33 instruments and then, the validity and reliability of these instruments. Letters were sent to over 60 individuals who had reported either the development or use of instruments to measure these concepts or reported results of investigations of social support or social networks which implied the use of such instruments. The authors were asked to provide copies of the instru863

validity, reliability

ments used and to indicate how they had addressed the issue of validity and reliability. The following analysis, based on their responses as well as published reports, focuses primarily on 24 measures, summarized in Table 1, that purported to measure social support. In addition, we will briefly comment on the nine measures, summarized in Table 2, that purported to measure social networks. SOCIAL SUPPORT Conceptual definitions of social support

Conceptual definitions were provided for 14 of the 24 instruments listed in Table 1. Six of the definitions were derived from the writings of Caplan [3], Cobb [4], Kahn [5] and Weiss [6]. In looking for a concep tual link in the writings of these authors (as well as others), one can identify three common elements. Support is seen as (a) an interactive process in which (b) particular actions or behaviors (c) can have a positive effect on an individual’s social, psychological, or physical well-being. The differences in definition develop over what actions, interactions, and effects should be considered as related to social support. Caplan [3], Kahn (51 and Weiss [6] include cognitive, affective, and instrumental behaviors. Cobb’s [4] definition is more circumscribed. He excludes instrumental assistance, and also sees support as functioning predominantly in crisis situations. Other definitions of social support used in the studies listed on Table I include such network dimensions as size, source and frequency of contact, and accessibility to members [45-49]. Finally, 10 of the 24 researchers (41%) had either no or very unclear definitions of social support.

el al.

ef al.

and

Branch

(1978)

Gottlieb

[W

(1980)

(1981)

Hirsch

Duncan-Jones

[59]

(651

[46]

1721

el ol

(1979)

Henderson

and

[45]

and Bloom

1361

Gottlieb

(1980)

[35]

[73]

Carveth

(1982)

Caldwell

Jette (1983)

(1982)

Blazer

(1981) 1641 Barrerra (1981)

Barrerra

(1977) WI

Andrews

long-term

-

helping

network

emotional

support

aid;

social reinforcement;

socializing;

efforts:

positively

guidance;

that

affect success of coping

of interaction

sense of

nurturance;

by:

with

or

of

type of

amount

negatively

Five forms

help and guidance

worth;

of personal

reassurance

reliability;

social integration;

attachment;

are provided

members

of contact

problem-centered

are suggested:

importance

contact;

relations

feedback;

Social

adequacy

size;

of contact;

network

frequency

definitions

network

Three

support;

accessibility;

source;

that six aspects should

or neighbors

considered:

Indicated

friends,

all types by family,

be

care assistance:

provided

that

provided

behaviors

definition

receive when

of help, paid or unpaid,

Informal

assistance

actually

individuals

with

of natural

Diversity

_

Conceolual

I items

Eleven

For

activities

if assistance

interactions

supportive

of daily

of

interaction

of support

degree

degree of satisfaction

of

with

for

of

personal

classes:

and in five areas of that occurred

activity,

their

members

amount

specify

Respondents with

network

adequacy

relationships

and perceived

each of six social

action

problem-solving;

four

categories into

are used to specify

environmental

two questions

availability

Fifty

influence:

emotionally-sustaining;

are organized

six empirically-generated behaviors

number

of days in

II

of the contact

birth);

number

importance

helping

since event;

Twenty

discussions

since event (e.g. a child’s

in

living

of support

of daily

is provided

interaction;

in

are used received

to: roles and

areas

members:

in the six conceptual

five core network

contact

relate

behaviors

the literature

support

Face

Face

Face residents

Face

support.

interviewer

with

r = 0.53

ratings

subject of satisfaction

between

with

with

and

responses

Eysenck

relationships

had significantly

fewer

raters on of behaviors

correlations informants

(2) Correlation

(I)

network

(4) Significant

Scale

correlations Personality

(3) Low

size

previously-validated

and definition

Family

Environment

Moos’

by three of four

and less adequate

(2) New

(I)

category

(2) Agreement

(I)

with

of Family

and network

selected from

instrument

Items

Validity

correlations subscale

Scale (FES),

Cohesion

(2) Positive

(I)

Face

of social

stress: crisis supporl; social participation

of social

that

of studies

the presence of

items are used to reflect degree

received

Forty

basic and instrumental

specify

support

Respondents

perceived

frequency

items selected

attachments;

from

of helping

month

amount

the preceding

under

neighborhood;

items generated

from

to specify

Forty

support

definition

components

are used to measure

Operational

Methodological

three types of support

Sixteen

Table

of

Test&retest,

0.81

0.22 4.51

Intercorrelat~ons.

0.37

r = 0.5lLO.79 consistency,

Test-retest, (2) Internal

0.86

for: code

consistency

0.59-0.73

coefficients.

units, 0.76; assignments,

scoring

Inter-rater

(I)

-

clusters:

(2) Alphas,

coefficients, 0.93-0.94

Correlation 0.25-0.43

(I)

Within

(3) Alphas,

0.4‘&0.91

(2) Correlation

(I)

r = 0.88

Reliability study,

no results given

Pilot

Support available IO an individual through social ties to other individuals, groups and the community

Lin er ol. (1981) [57] Dean CI al. (1981) 1221

Continuedoverleaf

Norbeck et cd. (1981) 1661

Miller t-f al. (1976) (261

McFarlane et of. (1981) (62,63]

Interpersonal transactions that include: expression of affection; affirmation of another’s behaviors; and/or giving aid

-

-

Information leading individual IO believe s/he is cared for, loved, and a part of a network of mutual obligation

Jenkins CI al. (1981) I’4

Marmot (1982) [SS]

In work/nonwork relationships, I2 questions specify degree of: reliability; willingness to listen; helpfulness; and from supeyvisors, concern and competency

Relationship with one or more persons that is characterized by relatively frequent interactions, strong, positive feelings. perceived ability IO lend aid

House and Wells (1978) [47]

After respondent identifies up to 20 network members and domain of relationship, nine questions are used to rate members on three functional areas of support and on thra network components

Questionnaire includes two series of questions that specify: For confidants; number; frequency of contact; and quality, availability and reciprocity of relationship; For nonconfidants: number; frequency of contact; domain of relationship and social participation

In six categories of potential life stress: work; money/linana; home/family; personal/social; personal health; society, respondents specify with whom discussions held, helpfulness, and if relationship reciprocal

Series of questions covering IO areas including confidants; social contacts; social/religious participation; and network loss during previous Year

Four indices used IO specify the degree of: (I) Family problems (4 items); (2) Satisfaction with neighborhood (2 items); (3) Instrumental/expressive support (26 items); (4) identification/ characterization of relationship with confidant(s) (II items)

A series of open-ended questions are used to probe structural position, satisfaction, problems and support in five life domains: work; finance; housing; social life; marriage

Three indices are used: (I) Family Environment Scale (FES) consisting of three subscales. each with nine true-false questions on quality of social relations in family; (2) Work Environment Index (WEI) consisting of three subscales of nine trutfalse questions on social climate at work; (3) A Traditional Social Support Index asking respondents about network ties and religious/social participation

Holahan and Moos (1981) [ZS]

(I) Significant correlations with Cohen and Lazarus scak for affirmation and a&t but not for aid (2) No significant correlations with measures of mood states nor life experiences

-

(I) Instrument evaluated by four judges and found acceptable (2) Helpfulness of spouse in all categories significantly different between parent-therapist couples and couples in therapy

-

(I) Tes-retest for support and network items, r = 0.8M.92 (2) Internal consistency for items, 0.69-0.98

-

Tes-retest: (I) Individuals named in each category, I = 0.62-0.99 (2) Helpfulness indicated in each category, 0.54094

-

Inler-ilem correlation for four indices: (I) 0.5%0.76 (2) 0.28-0.82 (3) Zero order, 0.67; (4) Five factors loaded high

Correlations: (I) Interrater, 0.754K~O; (2) Subjectinformant: Initial, 0.62-0.72; Follow-up, 0.52-0.73

FaCe

(I) Indices l-2 taken from other sources (2) Family problem index, money problems, family income, life demands, neighborhood satisfaction and reciprocal, durable, confidential relationships significantly related to depression scale (CES-D)

Alphas, 0.7H.92

Alphas: (I) FES, 0.89; (2) WEI, 0.88

Questions adapted from other sources

(I) FES and WEI part of previouslyvalidated instruments (2) Predictive validity of FES and WEI inferred from unpublished papers

1 ;! ET

8 vr

g

E 8 g

*Turner R. J. Measures of social support:

Williams er al (1981) [68] Donald et al. (1978) [69] ware PI al. (1980) [70]

Vachon er al. (1982) 1341

some instruments

-

Two definitions: (I) Information leading to belief one is loved, esteemed. and a member of a network (2) Weiss’ provisions of social relations

Turner ( I98 I) [49] Turner (1981)’

ties; confidants:

care;

and their properties.

Health Care Research

Unit, University

Nine items measure social contacts and resources now and during previous month to year

List of questions on network social participation; and aid

1981 (unpublished

paper).

(I) Face (2) Factor analysis identified social health (i.e. social support) as separate factor (3) Significant correlations with physical. mental, general health indices

(I) For vignetles, correlations with traditional support variables, r = 0.24-0.44 (2) For all scales, significant correlations with five measures of distress

from other sources

Alphas for three indices (I) Vignettes, 0.87; (2) Love/esteem, 0.724.75; Social relations, 0.73

(3)

(I) Test-retest, r = 0.56-0.66 (2) Alphas, 0.314.95

Alpha. 0.56

Indices adapted

-

from other sources

(I) Test retest: Number scale, I = 0.90; Satisfaction scale, I = 0.83 (2) Alphas: Number scale, 0.97; Satisfaction scale, 0.94

adapted

Both scales associated negatively with tests of personality and positively with life attitudes and self esteem. Number scale associated negatively with external control and cognitive interference and positively with positive life events. Satisfaction scale positively associated with negative life events. No relationship between scales and social desireability.

Questions

of Western Ontario,

of

Two-part questionnaire is used lo measure: (I) Tangible support: respondents identify up to five people who would provide help in nine situations (2) Emotional and informational support: after listing network members, respondents rate each as to information provided; reliability; boosts spirits; is caring; can be confided in

Appraisal of whether and to what extent an interaction, pattern of interaction or relationship is helpful. Three types of support are identified: tangible; emotional; informational

Schaefer er 01. (1981) [48]

Three indices are used: (I) Responses to nine vignettes about love, esteem and network: (2) Eight-item scale of reflected love and self-esteem; (3) Eighteen-item scale of provisions social relations

In 27 common situations, respondents list up lo nine people who can be counted on (number scale) and specify degree of satisfaction with support (satisfaction scale)

Existence or availability of people upon whom we can rely, who let us know they care about, value and love us

al.

sarason 9,

(1983) [67]

Access to and use of individuals, groups or organizations in dealing with life

Respondents were asked if there was anyone to whom they could tell anything; married respondents were also asked if they could talk to spouse about important matters

Eight-item scale is used to measure: assistance; concern; trust; value/interest similarity among up to IO importanl nonfamily members

Pearlin el ol. (1981) 152)

t-1 al. (1981) [71]

Oxley

Table I-continued

B

? ;6

Table 2. Methodological Studv

Conceptual definition

components of studies of social networks

Operational definition

Validitv

Rcliabilitv

Barnra (1980) [81]

Individuals who provide the functions that define support

(I) In six areas of social support. respondents identify individuals who typically supply such support and who actually supplied it in the previous month (2) Respondents identify those persons to be encountered in a social conflict and with whom they actually conflicted during the previous month

-

(I) Test-retest for: Support categories r = 0. I g-0.87; Perceived network size, r = 0.88; Actual network size, r = 0.88; Individuals named both times, r = 0.48-0.73 (2) Alphas: Perceived, 0.78; Actual, 0.74

Berkanovic er ol. (1981)

-

Respondents identify up to seven persons to whom they talk about health-related matters and for each, give: age; proximity; frequency of visits. Respondents also are asked: whether they know one another; show concern; are consulted on health matters

Questions drawn from other sources

-

Branch and Jette (1983) [36]

Those significant others with whom elders have close contact

Respondents identify children, other relatives and friends who are seen or talked to often and characterize: proximity; frequency of contact; closeness; health (of children); duration of friendships

-

-

Froland et al. (1979) [23]

Social ties that have a potential for providing social support defined as accessible and important

Extensive series of questions characterize networks in terms of: overall structure (size, proximity, density domain of relation); patterns of interaction (durability, frequency of contact); supportive functions (help provided, strength of ties); stability of network (changes, losses, deaths)

-

-

Gallo (1982) (831

Set of interpersonal links from which dependable others gratify a person’s psychosocial needs

Respondents identify up to four persons with whom they talk on matters of concern or importance and for each, specify: frequency of contact; if they know one another; proximity; content; intensity; homogeneity; duration and directedness of relationship

Questions adapted from other sources

-

Hirsch (1980) [72]

Natural support system: significant others who are members of one’s social network or unaffiliated nonmental-health professionals

(I) Respondent lists up to 20 significant others with whom s/he is likely to interact during specified time period and uses matrix to indicate where one is a friend of another (2) Interviewer rates social network variables: feedback; sex differences in relationships; frequency of contact; preferences of interactions; multidimensionality

Fats

From list of up to six individuals who are most important to respondent, or to whom s/he feels closest, respondents identify extent to which each is relied on for material assistance, emotional support, companionship and information

Correlation between respondent and family members on number of close friends, I = 0.62

-

-

-

1821

Mitchell (1982) [3l]

-

Peru& and Targ (1982) [32]

A set of direct and indirect ties among a defined group of individuals or organizations

Wentowski (1981) (801

From person who committed patient to hospital, interviewer identified and then interviewed patient’s network members on network relationships to determine: network size; density; openness; pattern of ties; and role structure Structured interviews and participant observations are used to identify providers of support and reciprocal nature of helping behaviors

867

-

868

PATRICK O'REILLY

The first obvious but not very original conclusion is the need for increased agreement on a conceptual definition of social support. Particularly critical in this effort to clarify the definition of support is ending the confusion resulting from including components of social networks in the conception of social support. Social network is an analytic concept, used to describe the structure of linkages between individuals or groups of individuals [50,51]. Such networks have a variety of functions of which the provision of social support is but one. Social support is provided through the behaviors or actions of members of a network and communicated through the network’s structure. Clearly, network analysis can provide an effective approach toward understanding and explaining how support is offered and received. However, the network is not, and should not be, confused with the support its members transmit to one another. While we continue to seek to achieve a broader agreement on a conceptual definition of support, investigators must also realize the inappropriateness of operationalizing this concept without benefit of a clear definition. The continuing lack of a linkage between conceptual definition and operationalization results in very likely the often ambiguous and inconclusive findings obtained in social support research. Operational definitions of social support Following from conceptual definitions are those issues related to operationalization of social support. Operationalized definitions were either stated by the investigators or reflected in the elements included in the research instruments that they used. Returning to Table 1, the measures of social support that are the least appropriate appear to be those measures that were based on a few items, such as the presence of a confidant [52] or social participation [53], which investigators selected from a more extensive questionnaire. The use of such indices as measures of social support often appears to have been carried out retrospectively, using concepts originally designated as social health or social relationships. The use of questionnaire items in these ways could indicate the interchangeability of multiple concepts, conceptual confusion, or the inappropriate extension of one concept (i.e. social relationship) to cover a second (i.e. social support). Also in these studies the connection between the investigators’ conceptual definitions of social support and their operationalized measures of this concept is often weak or nonexistent. As a result, a variety of indirect indices of ‘social support’ were found to Ix associated with different aspects of health status, but it is not at all clear what the investigators meant by social support. Of the remaining support studies, researchers appear to have made operational decisions in three areas: (a) specificity of questions; (b) type of format; and (c) specificity of dimensions of support. Speczjiciry of questions. The first distinction among the social support instruments was whether the questions measuring support were constructed for use in a general [35,48,54-711 or specific population. By including population-specific questions, investigators were able to address issues related to type and

provision of support that were germane to that population, such as assistance provided after divorce [45,72], death [53], to new mothers [46], to the elderly [36,73], or in a work setting [25,47]. Type of format. Following from Broadhead et al.‘s [7] work, it appears that investigators used two broad formats to elicit information on support. They were: Network format: Who provides support? Investigators using this format posed questions in one of two ways. In one group of studies [48, 59, 60, 62, 63, 671. investigators described one or more real or hypothetical situations and asked respondents if support or help would be provided in this situation, and if so by whom. A second group of studies [25,45,47.66,72] used either previouslyelicited names of network members or simply a type of relationship (i.e. wife, friend, coworker) and asked if support had been or would be provided by such individuals in a particular situation. Clearly it was important to these investigators to know not only if support were provided, but who provided the support. Behavioral format: Is support being provided? Investigators using this approach listed specific helping or supportive actions or behaviors and obtained information on whether or not respondents received or would expect to receive any of these behaviors, often limiting a positive response to behaviors that of time period within a specific occurred [56.57, 64,65,68,69]. The providers of the supportive behaviors were not identified; what these investigators considered critical was the provision of a particular behavior or the perception of support, regardless of the provider. It should be noted that the above format distinctions should not be interpreted as being mutually exclusive in terms of the overall design of an instrument. In fact, some investigators utilized more than one format in different sections of the same instrument [25,26,48, 56, 571. Specificity of dimensions of support. Finally, investigators chose whether or not to measure support as one broad, single variable [36, 52, 53, 5867,681 or to measure its conceptualized components separately, such as tangible aid, emotional support, or the ability to confide in someone [48,49,56,57,59,66]. Based upon the above, the decision tree for selecting an approach to the measurement of social support appears (at the top of facing page). When the multiple options for instrument design are combined with the wide range of interests among researchers, it is difficult to imagine any single instrument being developed that would accurately measure the perceptions of support or supportive behaviors in the variety of research that continues to be conducted. Finally, when operationalizing support, researchers should consider the probability that there are two basic types of support [38,74, 751; everyday support leading to a general sense of well-being, similar to what Weiss [6] has proposed, and support at critical times, as Cobb [4] has proposed. There are difficulties in measuring both types of support. Everyday support is difficult to measure as it is so much a part of our ‘taken for granted world’ that often we are not conscious of its importance until it is no longer

Social support and social network research Conceptual

869

Definition

What questions?

What component? I

Multicomponent

2 What dimension?

Specific dimension

available. This support may not lend itself readily to standardized measures and survey methodology, and may be more accurately described with observational or longitudinal studies using semi-structured interview formats [76]. As for support at critical times, it is not readily measured prior to the crisis, and when measured retrospectively it is subject to all of the biases of this type of research. So again, use of longitudinal data would be more appropriate than the use of cross-sectional data [75]. Specific indicators of social support

After deciding which approach will be used for the instrument, the investigator must still decide on the specific indicators to be selected to measure support. As Table 1 illustrates, not only were the indicators used quite broad, but a number of researchers included questions related to participation in clubs or religious services, satisfaction with neighborhood, recent losses from the network, feelings about oneself or society, and network dimensions, such as size, frequency of contact, and reciprocal nature of relationships [54,56,58,61,62,71]. The use of such diverse indicators may be a reflection of the lack of clarity in how some investigators conceptualized social support, or it may simply reflect the investigators’ use of available data. However, while some of these items might be considered indirect indicators of support, particularly of its availability, others, while cited as measures of social support, are

actually measures of social participation, social isolation, state of personal well-being, and most often, components of social networks. As Pearlin et al. [52] remind us, with social support “. . . we confront a notion that is sufficiently unspecified that it potentially embraces virtually all social relationships, even the most intermittent and contractual.” And, it may be added, something that measures everything ends up measuring nothing. An additional issue related to the choice of indicators for measuring social support is the question of the appropriateness of relying primarily on subjective perceptions of support, or alternatively, of using only actual helping behaviors as the indicators of support. A perception of social support could be classified as a coping or adaptive mechanism based on some sense of past experiences. Subjective perceptions of support can also be influenced by the personality, mood, and illness state of an individual [771. In addition, such indicators do not provide much in sight into the specific behavioral interactions that resulted in these perceptions of support. That is, they usually do not inform us as to who did/would do what for whom in what circumstance that was/would be deemed helpful or supportive. Alternatively, when the measurement of support is based primarily on having individuals indicate if others have provided them with specific behaviors or actions, then there is no information on whether or to what extent such actions were considered supportive. The probability exists that behaviors

870

PATRICK O'REILLY

intended to be supportive are interpreted differently, even negatively, by different people. In summary, if research findings are intended to help us better understand the social processes related to social support and ultimately to develop interventions aimed at improving health outcomes, then the indices used to measure support need to be consistent with conceptual definitions, differentiate the multidimensionality of support, specify the perceived adequacy of support, and where appropriate, identify not only general supportive behavior but the specific supportive behaviors and interactions that are presumed to affect the health outcomes being measured. Under these conditions, the development of one instrument that can accurately and precisely measure social support in all potential situations is highly unlikely. It is probable that a widely-accepted, valid, and reliable instrument can be developed for studies in which the population and health outcomes are fairly general, such as in studies of the relationship of social support to rates of mortality or depression in a random population. But, the more specific the aims of a study, the less likely that a single valid and reliable instrument, sensitive to the provision of support relevant to the study’s outcome can be developed. In these circumstances, it is likely that over time a series of instruments tailored to particular populations and/or health outcomes could be developed. As part of this process, and in lieu of always developing a totally new instrument, investigators should attempt to identify instruments that: (1) use a format that has been shown to be convenient and reliable; (2) reflect the multifaceted nature of social support; and (3) can be adapted to obtain responses that are specific to the provision of social support relevant to the aims of their particular study. Validity and reliability Although specifically requested, few of the responses to our inquiry provided any additional information about validity and reliability of instruments beyond what was stated in published reports. Seven reports did not even mention reliability, four reported on validity, two reported only on reliability, and 11 reported on both validity and reliability. While the latter proportion may appear to be better than expected, the methods used and, in many instances, the results obtained, do not provide overwhelming confidence in the data collection instruments. Various approaches were used to establish validity. Face validity, the weakest of the available approaches, was frequently used to provide evidence of the accuracy of the operational indices. Moreover, the results of studies in which more exacting methods of validation were utilized, such as discriminate, predictive, or convergent validity, generally were only modestly supportive, and often poorly to nonsupportive, of the validity of the instrument [49, 56, 59,61,64,66]. In addition, four investigators [25,47, 71, 731 used questions selected from other questionnaires and inappropriately noted the tests of validity for the original questionnaire as an indicator of the validity of the selected items. As for reliability, nine of the 24 investigators reported that they had carried out a test of reliability.

Six of these investigators and seven others who included no evidence of a test of reliability did report the correlation coefficients for their instruments, which ranged from 0.22 to 0.98, and/or alphas, which ranged from 0.31 to 0.97. Of the nine checks on reliability, six investigators [48,59, 62,64, 66,671 reported generally favorable results using a test-retest approach, and two [61,65] reported favorable interrater correlations. One study [61] contacted informants from the social network in an attempt to establish the reliability of the subjects’ selfassessments of support, and reported positive results. The difficulty with interpreting this latter result is that a second study [59] also used a network informant for verification of subject responses, but claimed the positive findings to be a measure of the validity of their instrument. The use of an informant from a subject’s social network is a particularly powerful measure, but it can establish either validity or reliability, not both. The degree of correspondence between the responses of a subject and an informant informs the researcher of the accuracy of the subject’s responses; that is, of their validity. However, it should also be remembered that the accuracy of the respondents’ answers should not be interpreted as meaning that the questions included in the instrument are valid indicators of the concept being measured. In the absence of an objective standard or criterion, complete validity of a measure cannot be established. Reliability, on the other hand, is more easily determined, but was tested in only a minority of the studies reviewed here. It appears that most investigators either moved rapidly from instrument design into data collection, without paying sufficient attention to establishing the validity and reliability of their instruments, or they inappropriately extracted ‘support’ items from a longer validated measure without validating the selected questions. These major methodological shortcomings in the development of measures of support need to be addressed before research findings can be confidently accepted and interventions planned. SOCIAL NETWORKS

Conceptual

and operational

definitions

The remaining nine reports reviewed were designed to measure social networks. They are found on Table 2. The initial research on social networks was carried out by social anthropologists (78,791 who attempted to clarify the relationship between total or extended social networks and social behavior. More recently, the use of this analytic concept, particularly in the health field, has generally been limited to obtaining information from a focal individual about his/her personal social network; that is, those individuals to whom one has direct links (501. In the network studies reviewed here some investigators were concerned with these personal social networks [23, 32, 36,801 but the majority focused on the even more restrictive social support network (31, 72,8 l-831, what researchers defined as the significant or important ties that provide support to a focal individual. Unlike the contrasting definitions used for social support, there appears to be strong agreement on the definitions of

Social support and social network research the term ‘network’ in both the broad and narrow scope in which it has been applied. However, as with social support research, a review of the operationalization of social network or social support network reveals limited correspondence among the different approaches used. Three levels of decisionmaking again appear to be involved in this operationalization process: (a) specificity of questions; (b) specificity of network; and (c) specificity of components to be measured. Specificity of questions. Three of the investigators [32,36,80] worded their instruments such that they were applicable only for a specific population, such as the elderly or the hospitalized. The other instruments were designed to be applicable to a general population. Specificity ofnetwork. A seemingly critical decision was made by investigators as to whether or not to limit the number of network members identified by the respondents. Four investigators [23,32,80,8 l] asked for a list of significant others. The others collected information on a predetermined number of network members, ranging from 4 to 20. It is not clear how the specific number for network members was selected or if all investigators addressed the issue of what, if any, effect such a limitation would have on the conclusions drawn from the data. It would appear that the reasons for this decision may have had more to do with the logistics of data collection than with the logic of scientific investigation. Specificity of components. The network components identified in the reports can be divided into those that are structural and those that are interactional. Among the structural dimensions, the most frequently measured were: relationship, size, density, and proximity. Among the interactive dimensions, durability, frequency of contact, and intensity of relationship were used most often by the investigators. After size of the network, frequency of contact was used more often than any other component, and it was used in only five of the nine network studies reviewed. Some of the studies [23,31,82] also included supportive functions such as help provided, instrumental role, type of activity, and concern shown. The inclusion of indicators of support in the network instruments again points out the confusion and/or lack of specificity among investigators about these two concepts. Related to the issue of social network versus support network as the unit of study is the question of whether studying only supportive ties may create a distorted picture of a social network‘ in which some members can be both supportive and nonsupportive [84,85]. Focusing only on the supportive aspects of these ties will not accurately account for the stresses and tensions that lead to nonsupportive or possibly destructive actions within the network. Such actions could result from situations in which the focal individual is reluctant to request assistance from members of his network, or a network member is reluctant or unable to provide support or acts contrary to the interests of the focal individual. It is also important for researchers using social network analysis to appreciate the fact that individuals, particularly in urban settings, have affiliations with individuals in

871

different groups. As such, some members of that network may be called upon to be supportive in some situations and not in others. The identification of a support system without reference to specific situations could, as a result, exclude key individuals from the analysis [40, 861. Validity and reliability As for validity and reliability of the instruments, it can be seen on Table 2 that most of the blocks of information are missing. Particularly distressing is the almost total lack of information on the reliability of these methods (Orth-Gomer and Unden [44] noted a similar situation in the instruments they reviewed). The criticisms discussed above need not be repeated, and should be considered as equally applicable to this group of studies. Additional concern has also been voiced about the reported low to moderate validity of subjects’ responses to questions about their social networks [87,88]. In view of these questions it would be reasonable for any investigator studying personal social networks to consider the value of using network informants (i.e. key members of the network) as a check on the accuracy of the responses obtained from subjects. CONCLUSION

In summary, the criticism of the lack of critical assessments of social support and social network research as it relates to health status appears quite justified. Before the utility of these concepts is accepted, it is necessary that increased attention be paid to clarify what the essential elements of social support and social networks are. It must be understood that social support and the effects of its provision are a function of an individual’s social network, and while the network has structural and interactive dimensions which can affect the provision of support, they are not themselves supportive nor necessarily indicators of supportive behaviors. Aspects of either social support or social networks can be investigated independent of each other. However, because of their interrelationship, it is potentially more productive to describe both the functional (and dysfunctional) and structural properties of the social support system. Maintaining the distinction between these concepts will be important in terms of identifying in what circumstances or situations behavioral (functional) or social (structural) interventions would be more effective in altering health status. These more clearly specified conceptual definitions should also increase agreement as to the operational components used in instrument design. Investigators should also address the issue of whether or not one questionnaire can be designed that is equally applicable in all situations, for all populations and health outcomes being observed. More likely we should be looking to identify the instrument(s) or formats that are most effective in measuring support or networks in specific populations and for particular health outcomes. Finally, investigators need to be more rigorous in the standards they use to establish the validity and reliability of their instruments. However, the most critical issue remains the lack of agreement on what are the conceptual and

PATRICKO’REILLY

872

operational definitions of social support and social network. As long as conceptual and operational confusion remains, the predictive utility of the concepts will not progress to the stage where logical and meaningful interventions can be developed. Acknowledgemenrs-The author is indebted to John B. McKinlay for his suggesting the subject of this paper, and his early guidance in its formative stages, and to Kirsten Levy for her assistance in the preparation of the manuscript. This research was supported in part by a grant from the National Heart. Luna and Blood Institute, NIH (HL 18318). REFERENCES

1. Cassel J. The contribution of the social environment to host resistance. Am. J. Epidem. 104, 107, 1976. 2. Antonovsky A. Health, Stress and Coping. Jossey-Bass, San Francisco, Calif., 1981. 3. Caplan G. Support Systems and Community Mental Health. Behavioral Publications, New York, 1974. 4. Cobb S. Social support as a moderator of life stress. Psychosom.

Med. 38, 300, 1976.

5. Kahn R. Aging and social support. In Aging from Birth IO Death, A.A.A.S. (Edited by Riley M. W.), Selected Symposia Series 30. American Association for the Advancement of Science, Washington, D.C., 1979. 6. Weiss R. W. The provisions of social relationships. In Doing Unto Others (Edited by Rubin Z.). Prentice-Hall, Englewood Cliffs, N.J., 1974. 7. Broadhead W. E., Kaplan B. H., James S. A. er al. The epidemiologic evidence for a relationship between social support and health. Am. J. Epidem. 117, 521, 1983. 8. Hamburg B. H. and Killia M. Relationship of social support, stress, illness and use of health services in healthy people. In Healthy People, The Surgeon General’s Report on Health Promorion and Disease Prevention, Backaround Pawrs. DHEW Pub. No. 79-55071A. G&t Printing- Of&., Washington, D.C., 1979. 9. Kaplan B. H., Cassel J. C. and Gore S. Social support and health. Med. Care 15, 47, 1977. 10. McKinlay J. B. Social network influences on morbid . . _. eptsodes and the career ot help seeking. In The Relevance of Social Science for Medicine (Edited by Eisenberg L. and Kleinman A.). Reidel, Dondrecht, 1980. 11. Cohen S. and Svme S. L. (Eds) Social Suooorl and Health. Academic Press, New York, 1985. ‘* 12. 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. Epidem. 109, 186, 1979.

13. 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. Epidem. 116, 123, 1982. 14. De Araujo G., Van Arsdel P. P. Jr, Holmes T. H. and Dudley D. L. Life change, coping ability, and chronic intrinsic asthma. J. psychosom. Res. 17, 359, 1973. 15. Dimond M. Social support and adaptation to chronic illness: the case of maintenance hemodialysis. Res. Nurs. Hhh 2, 101, 1979.

16. Earp J. L., Ory M. G. and Strogatz D. S. The effects of family involvement and practitioner home visits on the control of hypertension. Am. J. publ. Hhh 72, 1146, 1982.

17. Matsumoto Y. S. Social stress and coronary heart disease in Japan: a hypothesis. Milbank Meml. Fund Q. 48, 9, 1970.

18. Medalie J. H. and Goldbourt U. Angina pectoris among 10,000 men. II. Psychosocial and other risk factors as

evidenced by a multivariate analysis of a five-year incidence study. Am. J. Med. 60, 910, 1976. 19. Morisky D. E., Levine D. M.. Green L. W. er al. Five-year blood pressure control and mortality following health education for hypertensive patients. Am. J. p&l. Hlth 73, 153, 1983. 20. Andrews G.. Tennant C.. Hewson D. M. and Vaillant

G. L. Life events, stress, social support, coping style and risk of psychological impairment. J. nerv. ment. Dis. 166, 307, 1978.

2I Brown G. W. and Harris T. Social Origins of Depression: A Study of Psychiarric Disorder in Women. The Free Press, New York, 1978. 22. Dean A., Lin N. and Ensell W. M. The epidemiological significance of social support systems in depression. In Research in Community Mental Health (Edited by Simmons R. G.), Vol. II. JAI Press, New York, 1981. 23. Froland C., Brodsky G., Olsen M. and Stewart L. Social support and social adjustment: implications for mental health professionals. Commun. menu. Hhh J. 15, 82, 1979. 24. Henderson S. The social network, support, and neurosis: the function of attachment in adult life. Br. J. Psychiaf.

131, 185, 1977.

25. Holahan C. J. and Moos R. H. Social support and psychological distress: a longitudinal analysis. J. abnorm. Psycho/. 90, 365. 1981: 26. Miller P. McC. and Ingham J. G. Friends, confidants, and symptoms. Sot. Psychiat. 11, 51, 1976. 21. Pavkel E. G.. Emms E. M.. Fletcher J. and Rassabv E. ‘s. Life events and social support in puerperal depression. Br. J. Psychiaf. 136, 339, 1980. 28. Cohan C. I. and Sokolovsky J. Schizophrenia and social networks: ex-patients in the inner-city. Schizophr. Bull. 4, 546, 1978. 29. Frydman M. I. Social support, life events and psychiatric symptoms: a study of direct, conditional and interaction effects. Sot. Psychiar. 16, 69, 1981. 30. Hammer M., Makiesky-Barrows S. and Gutwirth L. Social networks and schizophrenia. Schizophr. Bull. 4, 522, 1978.

31. Mitchell R. Social networks of psychiatric clients: the personal and environmental context. Am. J. commun. Psychol. 10, 387, 1982. 32. Perrucci R. and Targ D. Network structure and reactions to primary deviance of mental patients. J. Hlfh sac. Behav. 23, 2, 1982.

33. Parkes C. M., Benjamin B. and Fitzgerald R. G. Broken heart: a statistical study of increased mortality among widowers. Br. med. J. 1, 740, 1969. 34. Vachon M. L. S., Rogers J., Lyall W. A. et al. Predictors and correlates of high distress in adaption to conjugal bereavement. Am. J. Psychiat. 139, 998, 1982. 35. Barrera M. Social support in the adjustment of pregnant adolescents: assessment issues. In Social Networks and Social Support (Edited by Gottlieb B. H.). Sage, Beverly Hills, Calif., 198I. 36. Branch L. G. and Jette A. M. Elders’ use of informal long-term care assistance. Geronfologisl 23, 51, 1983. 37. Caulan R. D.. Cobb S. and French J. R. P. Jr. Relati’onships of cessation of smoking with job stress, personality, and social support. J. appl. Psycho/. 60, 211, 1975. 38. House J. Work Stress and Social Support (Edited by Reading M. H.). Addison-Wesley, Reading, Mass., 1981. 39. Gore S. The effect of social support in moderating the health consequences of unemployment. J. Hhh sot. Behav. 19, 157, 1978. 40. McKinlay J. B. Social networks, lay consultation and help-seeking behavior. Sot. Forces 51, 275, 1973. 41. Nuckolls K. B., Cassel J. and Kaplan B. H. Psycho-

social assets, life crisis, and the prognosis of pregnancy. Am. J. Epidem. 95, 431, 1972.

Social support and social network research 42. Tolsdorf C. C. Social networks, support, and coping: an exploratory study. Fem. Process. 15, 407, 1976. 43. Wilcox B.- L. Social support in adjusting to marital disruption: a network analysis. In Social Nerworks and So&l Support (Edited by Gottlieb II. H.). Sage, Beverly Hills, Calif., 1981. 44. Orth-Gomer K. and Uden A. L. The measurement of social support in population surveys. Sot. Sci. Med. 24, 83, 1987. 45. Caldwell R. A. and Bloom B. L. Social support: its structure and impact on marital disruption. Am. J. commun. Psychol. 10, 647, 1982. 46. Carveth W. B. and Gottlieb B. H. Social support and stress. Can. J. Behav. Sri. 11, 179. 1979. 47. House J. S. and Wells J. A. Occupational stress, social support and health. In Reducing Occupational Stress (Edited by McLean A., Black G. and_Colligan M.), Conference Proceedines. DHEW Pub. No. 78-140. Govt Printing Office, Washington, D.C., 1978. 48. Schaefer C., Coyne J. C. and Lazarus R. S. The health-related functions of social support. J. behov. Med. 4, 381, 1981.

49. Turner R. J. Social support as a contingency in psychological well-being. J. Hlrh sot. Behuv. 22, 357. 1981. 50. Mitchell J. C. The concept and use of social networks. In Social Nerworks in Urban Siruotions (Edited by Mitchell J. C.). University of Manchester Press, 1969. 51. Craven P. and Wellman B. The network city. Social. Inquiry 43, 57, 1973.

52. Pearlin L. I., Lieberman M. A., Menaghan E. G. and Mullan J. T. The stress process. J. Hlth sot. Behav. 22, 337, 1981. 53. Vachon M. L. S., Sheldon A. R., Lancer W. J. er al. Correlates of enduring distress patterns following bereavement: social network, life situations, and personality. Am. J. Psychior. 139, 998, 1982. 54. Andrews G., Schonel M. and Tennent C. The relationship between physical, psychological, and social morbidity in a suburban community. Am. J. Epidem. 105, 324, 1977.

55. Andrews G., Tennant C., Hewson D. and Schonell M. The relationship of social factors to physical and psychiatric illness. Am. J. Epidem. 108, 27, 1978. 56. Lia N., Simeoper R., Ensel W. M. and Kuo W. Social support, stressful life events and illness: a model and an emperical test. J. Hlth sot. Behau. 20, 108, 1979. 57. Lin N., Dean A. and Ensel W. M. Social support scales: a methodological note. Schizophr. Bull. 7, 73. 1981. 58. Marmot M. G. Socio-economic and cultural factors in ischaemic heart disease. Ado. Curdiol. 29, 68, 1982. 59. Henderson S., Duncan-Jones P., Byrne D. G. and Scott R. Measuring social relationships: the interview schedule for social interaction. Psychoion. _ - Med. 10, 723, 1980. 60 Duncan-Jones P. The structure of social relationships: analysis of a survey instrument, parts l-2. Sec. Psychiaf. 16, 143, 551, 1981. 61 Jenkins R., Mann H. H. and Belsey E. The background, design, and use of a short interview to assess and support in research and clinical settings. Sot. Sci. Med. 15E, 195, 1981. 62. McFarlane A. H., Norman G., Streiner D. L. e( al. A longitudinal study of the influence of the psychosocial environment on health status: a preliminary report. J. Hlth sot. Behov. 21, 124, 1981. 63. McFarlane H. H., Neale K. A., Norman G. A. er 01. Methodological issues in developing a scale to measure social support. Schizophr. Bull. 7, 90, 1981. 64. Barrera M. Jr, Sandler I. N. and Ransay T. B. Preliminary development of a scale of social support: studies on college students. Am. J. commun. Psychol. 9, 435, 1981.

873

65. Gottlieb B. The development and application of a classification scheme of informal helping behaviors. Can. J. behuu. Sci. 10, 105, 1978. 66. Norbeck J. S., Lindsey A. and Carrieri V. L. The development of an instrument to measure social sup port. Nurs. Res. 30, 264, 1981. 67. Sarason I. G., Levine H. M., Bashom R. B. and Samson B. R. Assessing social support: the social support questionnaire. JT Person. so>: Psychol. 44, 127, 1983. 68. Williams A. W.. Ware J. E. and Donald C. A. A model of mental health, life events, and social supports applicable to general populations. J. Hlth sot. Behav. 22,324, 1981. 69. Donald C. H., Ware J. E., Brook R. H. and DaviesAvery A. Conceptualization and Measurement of Health in the Health Insurance Study, Vol IV, Social Health, R-1987/4-HEW. Rand Corp, Santa Monica, Calif., 1978. 70. Ware J. E., Davies-Avery A. and Brook R. H. Conceptualization and Measuremenl of Health for Adults in the Health Insurance Study, Vol. VI, Analysis of Relationships Among Health Srarus Measures, R-198716

HEW. Rand Corp, Santa Monica, Calif., 1980. 71. Oxley D., Barrera M. Jr and Sadella E. K. Relationships among community size, mediators, and social support variables: a path analysis. Am. J. commun. Psychol. 9, 637, 1981.

72 Hirsch B. J. Natural support systems and coping with major life changes. Am. J. commun. Psychol. 8, 159, 1980.

73. Blazer D. G. Social support and mortality in an elderly community population. Am. J. Epidem. 115, 684, 1982. 74. Veiel H. 0. Dimensions of social support: a conceptual frame work for research. Sot. Psych&. 28, 156, 1985. 75. Brown G. W.. Andrews B.. Harris T.. Adler Z. and Bridge L. Social support. self-esteem and depression. Psycholog. Med. 16; 813, 1986. 76. Sokolovskv J. and Cohen C. I. Toward a resolution of methodological dilemmas in network mapping. Schizophr. Bull. 7, 109, 1981. 77. Alloway R. and Bebbington P. The buffer theory of social support-a review of the literature. Psycholog. Med. 17, 91, 1987.

78. Barnes J. A. Class and communities in a Norwegian island parish. Hum. Relations 7, 39, 1954. 79. Bott E. Family and Social Network: Norms and External Relationships in Ordinary Urban Families’ (Revised Edition). Tavistock, London, 1971. 80. Wentowski G. J. Reciprocity and the coping strategies of older people: cultural dimensions of network building. Geronrologisr 21, 600, 1981. 81. Barrera M. A method for the assessment of social support networks in community survey research. Connections 3, 8. 1980.

82. Berkanovic E., Telesky C. and Reader S. Structural and social psychological factors in the decision to seek medical care for symptoms. Med. Core 19, 693, 1981. 83. Gallo F. The effects of social support networks on the health of the elderly. Sot. Work Hlfh Care 8, 65, 1982. 84. Wellman B. Applying network analysis to the study of support. In Social Networks and Social Support (Edited by Gottlieb B. H.). Sage, Beverly Hills, Calif., 1981. 85. Fischer C. S., Jackson S. R. M., Stueve C. H. et al. Networks and Places: Social Relations Serring. Free Press, New York, 1977.

in the Urban

86. McCallister L. and Fischer C. S. A procedure for surveying personal networks. Sociolog. Merh. Res. 7, 131, 1978. 87. Killworth P. and Bernard H. Informant accuracy in social network data. Hum. Org. 35, 269, 1976. 88. Hammer M. Some comments on the validity of network data. Connections 3, 13, 1980.