Abstract Background and Objectives Social network typologies have been used to classify the general population but have not previously been applied to the stroke population. This study investigated whether social network types remain stable following a stroke, and if not, why some people shift network type. Research Design and Methods We used a mixed methods design. Participants were recruited from two acute stroke units. They completed the Stroke Social Network Scale (SSNS) two weeks and six months post stroke and in-depth interviews 8–15 months following the stroke. Qualitative data was analysed using Framework Analysis; k-means cluster analysis was applied to the six-month data set. Results Eighty-seven participants were recruited, 71 were followed up at six months, and 29 completed in-depth interviews. It was possible to classify all 29 participants into one of the following network types both prestroke and post stroke: diverse; friends-based; family-based; restricted-supported; restricted-unsupported. The main shift that took place post stroke was participants moving out of a diverse network into a family-based one. The friends-based network type was relatively stable. Two network types became more populated post stroke: restricted-unsupported and family-based. Triangulatory evidence was provided by k-means cluster analysis, which produced a cluster solution (for n = 71) with comparable characteristics to the network types derived from qualitative analysis. Discussion and Implications Following a stroke, a person’s social network is vulnerable to change. Explanatory factors for shifting network type included the physical and also psychological impact of having a stroke, as well as the tendency to lose contact with friends rather than family. Social networks, Social isolation, Analysis—mixed methods, Friendship, Chronic illness Stroke is a leading cause of complex disability in older adults (Adamson, Beswick, & Ebrahim, 2004); around 75% of people who have a stroke are aged over 65 (National Audit Office, 2010). Following a stroke, a person is at risk of losing contact with friends and their wider social network (Northcott, Moss, Harrison, & Hilari, 2016; Vickers, 2010). They also take part in fewer social activities (Cruice, Worrall, & Hickson, 2006; Fotiadou, Northcott, Chatzidaki, & Hilari, 2014), and the family unit is placed under strain (Northcott, et al., 2016; Winkler, Bedford, Northcott, & Hilari, 2014). Yet, the social impact of having a stroke varies: a subset appear relatively able to preserve important elements of their kin and non-kin social contact (Northcott & Hilari, 2011). The current project aimed to develop a social network typology in order to provide an explanatory framework in which to explore why some people’s social networks are more vulnerable than others post stroke, and whether there are protective factors. The social relationships a person has can profoundly impact on mood, life satisfaction, and physical health. A meta-analysis examined the impact of social relationships on illness-related mortality (Holt-Lunstad & Smith, 2012). The analysis included 148 studies of 308,849 participants who were followed for an average of 7.5 years. On average, people with stronger social networks had a 50% increased likelihood of survival; the effect was strongest in those studies that used complex measures of social integration (e.g., measuring frequency and density of social contact) rather than those studies using binary measures (e.g., whether living alone). Furthermore, the quality of social relationships has been consistently associated with life satisfaction (Pinquart & Sorensen, 2000), with those who perceive themselves poorly supported at greater risk of developing depression (Teo, Choi, & Valenstein, 2013). These patterns have also been observed in the stroke population. A recent systematic review (n = 4,816) found that both low-perceived social support and reduced social network were consistently associated with depression, worse quality of life, and worse physical recovery (Northcott, et al., 2016). It is therefore of concern that those who have a stroke are at risk of becoming isolated. Given the adverse consequences of becoming isolated post stroke, it is important to understand the reasons why some people experience a contraction of their social network, whereas others are able to maintain contact with diverse sources of support (e.g., family, friends, local community). Developing a social network typology may provide a useful framework for understanding these variations in response. Further, the type of network a person belongs to has been shown to be an important factor in how a person responds to aging and ill-health, and the patterns of support they both seek and receive (Machielse, 2015; Wenger, 1994). Developing a clearer understanding of how social network patterns change following a stroke may therefore enable services to support the social well-being of people post stroke more effectively. Social network research has distinguished major network “types” through examining key elements, such as frequency of contact with kin and non-kin network members. Despite the heterogeneous methodologies, for example, Framework Analysis applied to qualitative interviews (Spencer & Pahl, 2006) or k-means Cluster Analysis applied to population-based survey data (Litwin, 2001), previous research has consistently found the following broad network types: a network type characterized by frequent contact with diverse sources, including family, friends, neighbors, and often also community involvement; a network where friends are the key source of support and contact; a family-based network, where a person has more contact with family than friends; and a restricted network, distinguished by there being few social contacts (Fiori, Antonucci, & Cortina, 2006; Li & Zhang, 2015; Litwin, 2001; Litwin & Shiovitz-Ezra, 2011; Spencer & Pahl, 2006; Wenger, 1994). Depending on the variables used to define the typology, additional network types have also been found, for example, a network type characterized by close contact with neighbors with few friends and family (Spencer & Pahl, 2006; Wenger, 1994). Those in “diverse” networks, thus having frequent contact with both family and friends, consistently have higher morale and subjective well-being than those in more restricted network types (Fiori et al., 2006; Li & Zhang, 2015; Litwin, 2001; Litwin & Shiovitz-Ezra, 2011). Most social network typology research is cross-sectional. However, both Wenger (1994) and Li and Zhang (2015) followed a cohort of older people thus providing insight into how social networks change over time. Both these studies found that people most commonly shift into either family-based or restricted networks. In terms of predicting who will shift network type, Li and Zhang (2015) found that those with worse health were at a higher risk of shifting network type; while Wenger (1994) documented that increasing frailty and old age were factors. However, Wenger (1994) also noted that most network types remain stable and as such represent a life-long adaptation with only a small percentage shifting each year. Social network typology work has not yet been applied to the stroke population. A stroke potentially represents a challenge to the functioning of a person’s social network (Northcott, et al., 2016). Tracking people with stroke over the first year post onset therefore provides an opportunity to explore in more detail the mechanisms by which people potentially shift network, and what factors enable some people to maintain their prestroke network type. In order to determine whether the developed network typology was a reasonable sectoring of the social world, we were also interested in triangulating our findings using mixed methods. Triangulation, where two different methods lead to similar findings, potentially increases confidence in the findings (Singleton, Straits, & Straits, 1993). Specifically, if both qualitative and quantitative methodologies produced social network types with similar characteristics it would provide reassurance as to the validity of the typology. This study addresses the following three research questions: Which network types are stable post stroke and which are vulnerable to change? What factors explain why a person changes network type? Will a similar network typology be found when using qualitative and quantitative methodologies? Design and Methods This study is primarily a qualitative study: qualitative methods were used to develop a typology, explore network shifts, and explanatory factors for change. Quantitative methods (cluster analysis) were used to provide triangulatory evidence for the qualitative typology. The study formed one part of a larger study exploring quality of life and social relationships post stroke (Hilari et al., 2009; Northcott & Hilari, 2011). Ethical approval was gained by the appropriate National Health Service local ethics committees; pseudonyms, replacement terms, and vaguer descriptors are used throughout this paper to preserve anonymity. Participants Participants were recruited from two acute stroke units in metropolitan teaching hospitals. People were eligible to take part if they were over 18 years old; had a first stroke; and were admitted to hospital for at least three days. Exclusion criteria included: not living at home prior to the stroke; severe co-morbidity, for example, advanced cancer; unable to give informed consent; history of mental health problems or cognitive decline; non-fluent English speaker premorbidly. Those with any severity of expressive aphasia and mild-moderate receptive aphasia could self-report on the measures used. However, those with very severe receptive aphasia [scoring <7/15 on the receptive domains of the Frenchay Aphasia Screening Test (Enderby, Wood, Wade, & Hewer, 1987)] were asked to nominate a proxy. Proxy responses are analysed elsewhere. A subset was purposively selected to take part in in-depth qualitative interviews. When selecting who to invite to the qualitative arm of the project, a sampling matrix was used in order to systematically optimize the range and diversity of relevant characteristics (Ritchie, Lewis, & Elam, 2003). Primary criteria included: stroke severity, age (whether aged over 65 or not; secondary target of recruiting a minimum of five people aged over 80), and perceived social support. Secondary criteria included additional social factors (living arrangements, size of network, number of close friends), gender, and ethnic background. Those with aphasia were preferentially included to ensure they were adequately represented (see Supplementary Appendix A for further details). Procedures and Methods Participants’ social networks were assessed within two weeks of having the stroke (baseline), and again six months later. The outcome measure used was the Stroke Social Network Scale (SSNS) developed as part of this project and validated on stroke survivors with and without aphasia (Northcott & Hilari, 2013). During the baseline assessment, participants were asked to reflect on the month prior to their stroke. The SSNS consists of five subdomains: children; relatives; friends; groups; satisfaction with social network. There is good evidence for the scale’s internal consistency (α = 0.85), acceptability, validity, and sensitivity to change. There are 19 items, and questions focus on frequency of contact (either face to face or remote, e.g., via telephone, E-mail, letter), proximity, quantity (e.g., number of close friends, number of close relatives), and satisfaction (e.g., with overall social network). Overall scores range from 0 to 100, with higher scores indicative of a better functioning social network (Northcott & Hilari, 2013). At approximately 12 months post stroke a subset of participants then took part in in-depth interviews. A topic guide was used (see Supplementary Appendix B). The order in which topics were covered varied from participant to participant, following in an organic way from participant responses. The interviews explored participants’ social networks and how these had changed since the stroke; experiences of receiving and giving support; and probing of the impact of the stroke on family relationships and friendships. Interviews took on average 65 minutes and were audio recorded. Interviews were carried out by the first author (S. Northcott) who is a speech and language therapist with experience of working with people who have aphasia. Two early interviews were listened to by a senior researcher who gave feedback helping to ensure the interviews were nonbiased and enabled participants to explore topics fully. The researcher also kept field notes which included contextual information as well as reflections on the influence of her own background on the research process. Qualitative Data Analysis All the interviews were transcribed verbatim, and analysed using the Framework method (Ritchie & Spencer, 1994). Several steps were followed in using this approach (Spencer, Ritchie, O’Connor, Morrell, & Ormston, 2014). Initial themes and concepts were identified through reviewing the data. These were then used to construct a thematic framework. The framework contained eight main themes (e.g., Theme 3: Friendship post stroke), under which more detailed subthemes were nested (e.g., Subtheme 3.2: changes to friendship post stroke). The transcribed material was then indexed, thus a decision was made for each phrase or sentence as to which section of the framework it belonged to. Having indexed the material, thematic matrices were constructed. Each main theme was a separate matrix, and each subtheme a separate column, while participants were assigned a particular row. The indexed data was then summarized and synthesized, and entered into the appropriate cell in the matrices. The advantage of this matrix-based system is that it enabled systematic cross-case and within case analysis, facilitating descriptive accounts of the range and diversity of experiences. Further, since the matrices link back to the transcripts, it allows the analyst to “move back and forth between different levels of abstraction without losing sight of the raw data” (Spencer et al., 2014, p. 283). In order to avoid bias, a senior researcher was involved in all stages of analysis, for example, reviewing charted material in order to reflect on the emerging analytic themes. We used this matrix-based analytic system to facilitate the development of a social network typology. We aimed to create a multifactorial typology (i.e. a person was assigned to a network category as determined by several variables), where the network categories were discrete and independent of one another, thus an individual belonged only to one category. Our aim was that the emergent typology should be meaningful, easily recognizable, and capture important patterns. In determining relevant variables with which to classify participants, we were initially guided by the literature, research aims, and by a systematic analysis of the different elements identified within relevant subthemes. However, the process of establishing whether the typology “fitted” with the data set was iterative, requiring interrogation of the whole data set, to ensure all participants could be assigned to a network category and that no participant in fact belonged to more than one network category. Where this was not the case, the variables or categories were refined, and the typology reworked. For example, we initially considered proximity as a variable as used in previous typologies (Litwin, 2001; Wenger, 1994). Yet in the present data set it was rare that adult children lived close by, and this variable did not facilitate differentiating the network categories. A further consideration was that the typology needed to work with both prestroke and post stroke data (i.e. it would be possible to categorize all participants at both time points). In analysing the reasons why people shift network type, we explored explicit explanations (i.e. those provided by the participants), as well as searching for “linkages” within the data (where a particular phenomenon co-occurs in the data set with another phenomenon), inferring possible explanations from the identified patterns of association. In order to classify participants into social network categories, the primary source of information was the in-depth qualitative interview carried out at approximately one year post stroke and which included reflection on both their current and premorbid social network. However, we also referred to the SSNS as administered at both two weeks and six months. The process of classification relied on the qualitative techniques described above. Quantitative Data Analysis: Cluster Analysis Cluster analysis is an exploratory data analysis tool which classifies multivariate data into clusters or subgroups (Burns & Burns, 2008; Hair & Black, 2000). The variables entered into cluster analysis were four of the five factors (six-month data set) which made up the SSNS (Northcott & Hilari, 2013): Children, Friends, Relatives, and Groups. We did not include the Satisfaction domain in order to better match the qualitative typology where satisfaction was considered only in so far as it differentiated between different network members (i.e. satisfied with contact with friends versus children). Hierarchical cluster analysis was used initially to determine the optimum number of clusters (Burns & Burns, 2008), at which point k-means cluster analysis was used. k-means clustering produces the number of clusters requested which are “of the greatest possible distinction” (Burns & Burns, 2008). Since the variables employed (social network factors) did not all have the same variance, they were standardized prior to entry into cluster analysis (Hair & Black, 2000). In terms of interpreting results, analysis of variance was used to determine the factors on which the clusters were differentiated. Further information on cluster analysis methodology is supplied in Supplementary Appendix C. Results Participants Eighty-seven participants were recruited into the main study (completing the social network assessment), of whom 71 were followed up at six months. Of these 71 participants, 32 were selected to take part in in-depth interviews about a year post stroke (range 8–15 months): 29 consented, one declined, and two were no longer contactable. Participant characteristics are displayed in Table 1. At six months, the majority were White (80%), male (56%), and married/had partner (53%). Eleven (16%) had aphasia. For the in-depth interviews, again, the majority were White (72%), male (59%) and married/had partner (55%). Ten (34%) had aphasia. Supplementary Appendix A shows how the participants fitted into the sampling matrix. Table 1. Participant Characteristics Variable Respondent n (%) 6 months 8–15 months n = 71 n = 29 Gender Female 31 (44%) 12 (41%) Male 40 (56%) 17 (59%) Age Mean (SD) 69.3 (14.1) 68.0 (14.0) Range 18–91 18–90 Ethnic group Asian 9 (13%) 2 (7%) Black 5 (7%) 6 (21%) White British 52 (73%) 15 (51%) White non-British 5 (7%) 6 (21%) Marital status Married/has partner 38 (54%) 16 (55%) Single, divorced, or widowed 33 (46%) 13 (45%) Living arrangements Living alone/in-hospital or institution 32 (45%) 11 (38%) Living at home with someone 39 (55%) 18 (62%) Employment statusa Not employed 60 (86%) 25 (86%) Working part time/voluntary work 4 (6%) 2 (7%) Working full time/full time education 6 (8 %) 2 (7%) Stroke type Ischemic 62 (87%) 21 (72%) Hemorrhagic 9 (13%) 8 (28%) Communication disability None 58 (82%) 18 (62%) Dysarthria 2 (3%) 1 (3%) Aphasia 11 (15%) 10 (35%) Variable Respondent n (%) 6 months 8–15 months n = 71 n = 29 Gender Female 31 (44%) 12 (41%) Male 40 (56%) 17 (59%) Age Mean (SD) 69.3 (14.1) 68.0 (14.0) Range 18–91 18–90 Ethnic group Asian 9 (13%) 2 (7%) Black 5 (7%) 6 (21%) White British 52 (73%) 15 (51%) White non-British 5 (7%) 6 (21%) Marital status Married/has partner 38 (54%) 16 (55%) Single, divorced, or widowed 33 (46%) 13 (45%) Living arrangements Living alone/in-hospital or institution 32 (45%) 11 (38%) Living at home with someone 39 (55%) 18 (62%) Employment statusa Not employed 60 (86%) 25 (86%) Working part time/voluntary work 4 (6%) 2 (7%) Working full time/full time education 6 (8 %) 2 (7%) Stroke type Ischemic 62 (87%) 21 (72%) Hemorrhagic 9 (13%) 8 (28%) Communication disability None 58 (82%) 18 (62%) Dysarthria 2 (3%) 1 (3%) Aphasia 11 (15%) 10 (35%) aMissing data for n = 1 at 6 month data point for employment status. View Large Table 1. Participant Characteristics Variable Respondent n (%) 6 months 8–15 months n = 71 n = 29 Gender Female 31 (44%) 12 (41%) Male 40 (56%) 17 (59%) Age Mean (SD) 69.3 (14.1) 68.0 (14.0) Range 18–91 18–90 Ethnic group Asian 9 (13%) 2 (7%) Black 5 (7%) 6 (21%) White British 52 (73%) 15 (51%) White non-British 5 (7%) 6 (21%) Marital status Married/has partner 38 (54%) 16 (55%) Single, divorced, or widowed 33 (46%) 13 (45%) Living arrangements Living alone/in-hospital or institution 32 (45%) 11 (38%) Living at home with someone 39 (55%) 18 (62%) Employment statusa Not employed 60 (86%) 25 (86%) Working part time/voluntary work 4 (6%) 2 (7%) Working full time/full time education 6 (8 %) 2 (7%) Stroke type Ischemic 62 (87%) 21 (72%) Hemorrhagic 9 (13%) 8 (28%) Communication disability None 58 (82%) 18 (62%) Dysarthria 2 (3%) 1 (3%) Aphasia 11 (15%) 10 (35%) Variable Respondent n (%) 6 months 8–15 months n = 71 n = 29 Gender Female 31 (44%) 12 (41%) Male 40 (56%) 17 (59%) Age Mean (SD) 69.3 (14.1) 68.0 (14.0) Range 18–91 18–90 Ethnic group Asian 9 (13%) 2 (7%) Black 5 (7%) 6 (21%) White British 52 (73%) 15 (51%) White non-British 5 (7%) 6 (21%) Marital status Married/has partner 38 (54%) 16 (55%) Single, divorced, or widowed 33 (46%) 13 (45%) Living arrangements Living alone/in-hospital or institution 32 (45%) 11 (38%) Living at home with someone 39 (55%) 18 (62%) Employment statusa Not employed 60 (86%) 25 (86%) Working part time/voluntary work 4 (6%) 2 (7%) Working full time/full time education 6 (8 %) 2 (7%) Stroke type Ischemic 62 (87%) 21 (72%) Hemorrhagic 9 (13%) 8 (28%) Communication disability None 58 (82%) 18 (62%) Dysarthria 2 (3%) 1 (3%) Aphasia 11 (15%) 10 (35%) aMissing data for n = 1 at 6 month data point for employment status. View Large Table 2 displays descriptive statistics for the SSNS. The overall score reduced from mean (SD) = 60.69 (15.22) at baseline to mean (SD) = 56.78 (15.44) by six months post stroke [t(70) = 3.89, p < 0.001]. The only domain where there was significant change was the Friends domain, t(70) = 4.25, p < 0.001, whereas the Children domain [t(70) = 1.56, n.s.] and Relatives domain [t(70) = 0.89, n.s.] were the most stable. Table 2. Stroke Social Network Scale: Descriptive Statistics Baselinea 6 months n = 87 n = 71 Overall score Mean (SD) 60.7 (15.2) 56.8 (15.4) Range 11.3–91.7 10.3–85.1 Satisfaction Mean (SD) 85.2 (15.6) 82.6 (19.2) Median (IQR) 88.3 (78.3–96.7) 86.7 (80.0–93.3) Range 35.8–100 6.7–100 Children Mean (SD) 57.6 (35.5) 58.8 (34.2) Range 0–100 0–100 Relatives Mean 37.8 (28.5) 36.8 (29.2) Range 0–88.9 0–93.3 Friends Mean (SD) 57.0 (25.0) 44.0 (28.1) Range 0–95 0–95 Groups Mean (SD) 35.1 (37.1) 31.0 (34.2) Range 0–100 0–100 Baselinea 6 months n = 87 n = 71 Overall score Mean (SD) 60.7 (15.2) 56.8 (15.4) Range 11.3–91.7 10.3–85.1 Satisfaction Mean (SD) 85.2 (15.6) 82.6 (19.2) Median (IQR) 88.3 (78.3–96.7) 86.7 (80.0–93.3) Range 35.8–100 6.7–100 Children Mean (SD) 57.6 (35.5) 58.8 (34.2) Range 0–100 0–100 Relatives Mean 37.8 (28.5) 36.8 (29.2) Range 0–88.9 0–93.3 Friends Mean (SD) 57.0 (25.0) 44.0 (28.1) Range 0–95 0–95 Groups Mean (SD) 35.1 (37.1) 31.0 (34.2) Range 0–100 0–100 Note: IQR, interquartile range; SD, standard deviation. aCollected two weeks post stroke, questions relate to the month prior to the stroke. View Large Table 2. Stroke Social Network Scale: Descriptive Statistics Baselinea 6 months n = 87 n = 71 Overall score Mean (SD) 60.7 (15.2) 56.8 (15.4) Range 11.3–91.7 10.3–85.1 Satisfaction Mean (SD) 85.2 (15.6) 82.6 (19.2) Median (IQR) 88.3 (78.3–96.7) 86.7 (80.0–93.3) Range 35.8–100 6.7–100 Children Mean (SD) 57.6 (35.5) 58.8 (34.2) Range 0–100 0–100 Relatives Mean 37.8 (28.5) 36.8 (29.2) Range 0–88.9 0–93.3 Friends Mean (SD) 57.0 (25.0) 44.0 (28.1) Range 0–95 0–95 Groups Mean (SD) 35.1 (37.1) 31.0 (34.2) Range 0–100 0–100 Baselinea 6 months n = 87 n = 71 Overall score Mean (SD) 60.7 (15.2) 56.8 (15.4) Range 11.3–91.7 10.3–85.1 Satisfaction Mean (SD) 85.2 (15.6) 82.6 (19.2) Median (IQR) 88.3 (78.3–96.7) 86.7 (80.0–93.3) Range 35.8–100 6.7–100 Children Mean (SD) 57.6 (35.5) 58.8 (34.2) Range 0–100 0–100 Relatives Mean 37.8 (28.5) 36.8 (29.2) Range 0–88.9 0–93.3 Friends Mean (SD) 57.0 (25.0) 44.0 (28.1) Range 0–95 0–95 Groups Mean (SD) 35.1 (37.1) 31.0 (34.2) Range 0–100 0–100 Note: IQR, interquartile range; SD, standard deviation. aCollected two weeks post stroke, questions relate to the month prior to the stroke. View Large Qualitative Analysis to Develop a Social Network Typology (n = 29) The variables used to classify participants to a social network category were as follows: Perceived amount of contact with children, relatives and close friends, and satisfaction with that contact; Composition of the network (relative number of kin versus non-kin, and close friends versus more casual social contacts) and perceived importance of different network members; Which network members were most likely to provide different types of functional support, for example, emotional or practical support, and the meaning this support held for participants. Five network categories were derived using these variables. It was possible to classify all 29 participants in the qualitative arm of the project into one of these categories both prestroke and post stroke. The categories were defined as follows: “Diverse network.” These participants had the most extensive social networks, comprising both kin and non-kin. They had close relationships with their immediate families, whom they saw frequently, as well as strong friendships. “Friends-based network.” Friends occupied a central role in this network type. Friends were likely to be the main source of emotional support. “Family-based network.” Family were the main source of support for these participants, and they had close relationships with several family members. Friends were not considered as important as family ties, and participants were unlikely to be in frequent contact with friends. “Restricted-supported network.” These participants had limited social ties, and few or no close friends. Despite their relative isolation they felt well-supported by one or two family members, for example, a spouse or daughter. “Restricted-unsupported network.” As with the above category, these participants had limited social ties. They either had no children or did not live near a child; had few or no close intimate friends; and received very limited functional support from any source. Which Network Types Are Stable Post Stroke and Which Are Vulnerable to Change? Figure 1 displays the patterns of change that occurred post stroke. The three network types found to be most stable were: friends-based, family-based, and restricted-unsupported. Around one third of participants shifted network type. The most common pattern of change was to move from a diverse network into a family-based one. Categories which became more numerous post stroke were family-based and restricted-unsupported. Nobody moved out of the restricted network types into a more supportive category following a stroke, nor did anyone acquire a diverse or friends-based network. Figure 1. View largeDownload slide Patterns of change in social network type. Figure 1. View largeDownload slide Patterns of change in social network type. Figure 2 details the category membership of each participant before and after the stroke, as well as giving information on each participant’s living arrangements post stroke, age, and presence of aphasia. However, since the sample was purposively selected to emphasize diversity it is therefore not representative of the parent stroke population and prevalence rates should be interpreted cautiously. Figure 2. View largeDownload slide Network type of all participants (n = 29) before and after stroke. Figure 2. View largeDownload slide Network type of all participants (n = 29) before and after stroke. What Factors Explain Why a Person Changes Network Type? Retaining or Losing a Diverse Network What determined whether a person retained their diverse network was the extent to which they maintained non-kin contact: the family element of their network remained constant and supportive. Factors which enabled people to keep their non-kin contact post stroke included: being able to leave the house (and therefore more able to access social activities and meet with diverse network members); not feeling withdrawn or depressed; the quality of their prestroke friendships; having friends who lived locally; and regular supportive groups including church membership. This is illustrated by Winnifred, a 65-year old. She had lived in the same house for 30 years, and had many local friends. Extended family also lived nearby. She described her house as “full… a lot of people here, always.” She had recovered sufficiently to be able to leave the house, enabling her to attend church regularly, where she “know everybody”, and to go for short walks. She described adopting a positive outlook: “you have a smile on your face, they too have a smile”. In terms of those with a diverse network who shifted into a family-based network, there were two subgroups. One subgroup had severe physical disability and could not leave their house without considerable assistance. This group still hoped for further recovery, and to resume social activities. An example is Tomasz, aged 66. Although his friends were important to him (“I don’t have 10% friends, they are 100% friends”) his friendships had been altered by the stroke: he could not leave his flat without his son assisting him, and he found it “rather impossible” to be in receipt of his friends’ goodwill rather than helping them. He now saw his friends “rarely”. He conceptualized changes to his social network as temporary, and lived in “hope that tomorrow might be better.” The other subgroup who shifted from diverse to a family-based network, as distinct from the above pattern, were not housebound. Instead, the primary factor appeared to be changing social desires, for example, feeling withdrawn, depressed, or hesitant about leaving the house. There was also a new selectivity about who they preferred to spend time with: seeing family and small intimate gatherings were preferred to large, noisy social occasions which appeared to hold less meaning for them post stroke. Case Example: Moving From a Diverse Social Network to a Family-Based One Prior to the stroke “Peter” was a successful business man, had an active social life as well as a close relationship with his family. One year post stroke he had made a good recovery and was able to walk, although he still experienced extreme fatigue. He had handed over the family business to his son which he described as “very traumatic”. He rarely saw any of his old friends, preferring to see his close family. He explained “I can’t be bothered” to socialize outside the family, fearing others might perceive him as “the weaker member of the pack”. He described feeling vulnerable and disinclined to leave the house. Maintaining a Friends-Based Network All those who belonged to a “friends-based” network prestroke retained this network type post stroke, with the exception of one participant. In some cases this was despite various risk factors for losing friends such as feeling withdrawn, having severe aphasia, losing shared activities including paid work, physical disability, and exhaustion. Although many did experience reduced social activity, friends were still the main source of support. In terms of what enabled them to maintain this network type, one factor was that they had well-developed friendships prior to the stroke: they spoke about how long they had known their close friends, the experiences and interests they shared, that they felt understood, and the bond this created (“As close as I will ever be to anyone. She totally understands absolutely everything… we’re old horses together”, Patricia, aged 62, speaking about a friend she had known for 25 years). Several participants also spoke of honesty and frankness within their most important friendships. Participants with a friends-based network all described the support and concern they received from their friends post stroke. This is illustrated by John, aged 76. He explained that the stroke hadn’t changed his friendships (“I’m lucky there, they’re very good.”). When asked to elaborate, he explained: “Well, they’re supportive, it’s hard to explain what supportive means, I suppose, but it’s this showing concern… it’s the tone of voice, and how they ask me about things.” A further factor is that these participants had fewer family resources to fall back on: they either had no grown up children or troubled relationships with their children. The majority lived on their own. As such, they had more reason to maintain friendships. As noted by Leonisa, aged 74, “When you are living alone, you cannot, you have nobody to talk, to laugh together, so when you are with the friends, it’s nice.” The only person to shift out of this category had a distinctive profile: prior to the stroke she was in less good health, and had a smaller friendship base in part because good friends had either moved away or died. Post stroke she had become housebound and had significant discourse difficulties associated with her right hemisphere stroke. She developed a restricted-unsupported network and her main social contact post stroke was with her daily carer. Case Example: Maintaining a Friends-Based Network “Steve” was in his 40s, working, living alone with no children. His primary sources of support were his close friends. He had a severe stroke leaving him with long-term mobility difficulties and severe expressive aphasia. Although he lost touch with many acquaintances, he retained his most important friends, and described feeling close to them. When asked about his friendships, it took him over six minutes to write: “They are the only think [thing] I have.” Maintaining a Family-Based Network Participants with a family-based network prior to the stroke remained in this network type post stroke. Both before and after the stroke, family were the core of their network (“My children are everything to me”, Edward, aged 58). Although there were changes within the family unit (e.g., a child helping with the shopping; participant more likely to receive than make visits), the network structure remained unchanged, and it was rare to move out of this group. Although a frequently observed pattern was that people with a diverse network developed a family-based network post stroke, no-one with a friends-based or restricted network developed a family-based network. Restricted Network Types The restricted-supported network type was a vulnerable structure as it was dependent on one or two people for all functional support needs. Reasons for participants shifting out of this category were the ill-health of the relative and a relationship break-up. For older people, it was perhaps particularly vulnerable as their key supporters were more likely to be older, and therefore more likely to experience ill-health themselves. An example is Ivy, 82 years old and single. She had lived all her life with her sister. She reported that in the last few years “most of my friends have died.” She described her sister as “very kind, and she gave me courage.” After her stroke, Ivy became housebound and more reliant on her sister. A few months later, her sister was taken ill, and it was unlikely she would be able to return home. As a result Ivy saw only her daily carer. In terms of the restricted-unsupported network type, prior to the stroke only one participant belonged to this category; more participants developed this network type post stroke. All those classified as having a restricted-unsupported network described symptoms of depression post stroke. Case Example: Moving From Restricted-Supported to Restricted-Unsupported Prior to the stroke, “Chris” was working full-time as a manager and was involved with a number of sporting activities at which he excelled. For emotional support, he relied on his partner. His two sons did not live locally. At the age of 58, he had a severe stroke leaving him with severe aphasia (verbal output limited to “yes”, “no” and swear words) and difficulty walking. His partner left him and he lost contact with the people he knew through work and sport. One year post stroke the only people he saw regularly were his sons (about once a fortnight). He indicated that his life was awful. Quantitative Analysis to Develop a Social Network Typology (n = 71) Cluster analysis was also used to develop a network typology, using data from the main study collected six months post stroke (n = 71). Hierarchical cluster analysis found that a four cluster solution was optimal (see Supplementary Appendix C for more detail, including plots of the agglomeration coefficients). Four clusters were therefore requested using k-means cluster analysis. The final cluster centers are presented in Table 3. Means approximately half a standard deviation above or below the overall mean for the sample represent defining peaks of the clusters. Table 3. Factor Means for the Different Clusters Delineating characteristics Network type Children factor Friends factor Relatives factor Groups factor Frequency (%) Diverse .45 0.73 0.44 −0.49 n = 21 (30%) Friends-based −.62 0.48 −0.08 1.09 n = 23 (32%) Family-based .88 −1.26 1.12 −0.47 n = 9 (13%) Restricted −.18 −.82 −0.97 −0.58 n = 18 (25%) Delineating characteristics Network type Children factor Friends factor Relatives factor Groups factor Frequency (%) Diverse .45 0.73 0.44 −0.49 n = 21 (30%) Friends-based −.62 0.48 −0.08 1.09 n = 23 (32%) Family-based .88 −1.26 1.12 −0.47 n = 9 (13%) Restricted −.18 −.82 −0.97 −0.58 n = 18 (25%) Note: Numbers in bold represent means approximately half a standard deviation above or below overall mean for the sample, and thus define the character of each cluster. View Large Table 3. Factor Means for the Different Clusters Delineating characteristics Network type Children factor Friends factor Relatives factor Groups factor Frequency (%) Diverse .45 0.73 0.44 −0.49 n = 21 (30%) Friends-based −.62 0.48 −0.08 1.09 n = 23 (32%) Family-based .88 −1.26 1.12 −0.47 n = 9 (13%) Restricted −.18 −.82 −0.97 −0.58 n = 18 (25%) Delineating characteristics Network type Children factor Friends factor Relatives factor Groups factor Frequency (%) Diverse .45 0.73 0.44 −0.49 n = 21 (30%) Friends-based −.62 0.48 −0.08 1.09 n = 23 (32%) Family-based .88 −1.26 1.12 −0.47 n = 9 (13%) Restricted −.18 −.82 −0.97 −0.58 n = 18 (25%) Note: Numbers in bold represent means approximately half a standard deviation above or below overall mean for the sample, and thus define the character of each cluster. View Large All four delineating variables contributed to differentiating between the clusters, as indicated by their significant F values (see Supplementary Appendix D). The four clusters contained between 9 and 23 participants. The clusters that emerged could be matched to the social network typology developed in the qualitative data. They were characterized as follows: 1. Cluster One: Diverse. Above average Children, Relatives, and Friends factors, although below average on the Groups factor. 2. Cluster Two: Friends-based. Above average non-kin (Friends and Groups); below average Children and Relatives factors. 3. Cluster Three: Family-based. Strong in both Children and Relatives factors; non-kin contact below average. 4. Cluster Four: Restricted. Low scores in all domains. There is only one restricted network type revealed. In fact, it would not have been possible to replicate the restricted supported/unsupported distinction as information on which network members provided functional support was not collected and so could not be entered into cluster analysis. The network types found through cluster analysis are represented graphically in Figure 3. Figure 3. View largeDownload slide Mean scores of the social network factors by cluster type. Figure 3. View largeDownload slide Mean scores of the social network factors by cluster type. Discussion This study developed a social network typology in order to better understand potential alterations to social networks following a stroke. Seventy-one participants were assessed using the SSNS at two weeks and six months post stroke, and a subset of 29 participants took part in in-depth interviews 8–15 months post stroke. Based on the qualitative material, we developed a typology comprising the following network types: diverse; friends-based; family-based; restricted-supported; and restricted-unsupported. It was possible to assign all 29 participants from the qualitative arm of the project to a network type both before and after the stroke. The most populated network type prior to the stroke was the “diverse” network (plentiful contact with kin and non-kin); post stroke the family-based network type was most populated. The main shift that took place was participants moving out of a diverse network into a family-based one, explained by the tendency for people to lose friends but keep in contact with family. Yet despite the general trend for friendship loss, the friends-based network type appeared relatively stable. Another trend was that the restricted-unsupported network type became more populated post stroke. Triangulatory evidence for the validity of the typology was provided by k-means cluster analysis. The factors of the SSNS were used as delineating variables (n = 71) to produce clusters (or network “types”) with comparable characteristics to the qualitative typology, providing supporting evidence that the sample can be meaningfully sectored into these broad network groups. Although the variables used to classify participants were specific to this project, emerging from the qualitative data set, the resulting network types match previous social network typology research. In common with previous literature, we discovered the following network types: diverse, friends-based, family-based, restricted (Fiori et al., 2006; Litwin, 2001; Litwin & Shiovitz-Ezra, 2011; Spencer & Pahl, 2006; Wenger, 1994). We further subdivided the restricted category. Wenger (1994) similarly found her “private restricted” category could be subcategorized into “independent married couples” and “dependent elderly”; likewise Spencer and Pahl (2006) found two restricted categories: “partner-based” and “professional based” (i.e. reliance on professionals as no close kin). Our “restricted-supported” category reflects that in our sample, the close supportive relationship was sometimes a sibling or child rather than a partner. Cross-sectional research has found that those with restricted and family-based networks tend to be older and more disabled (Fiori et al., 2006; Litwin, 2001); and longitudinal studies document frailty and old age (Wenger, 1994) and poor health (Li and Zhang, 2015) as risk factors for moving network type. Our research similarly found that physical disability, particularly where a person becomes housebound, is an obstacle to maintaining a diverse network. However, other factors also appeared to be important, such as the availability of locally based friends and local supportive groups, as well as the psychological impact of illness and whether a person felt vulnerable and withdrawn. A longitudinal stroke study found that having few social contacts outside the house was a significant predictor of depression. By three years post stroke 66% of nondepressed participants had met with a friend or relative in the previous week compared to only 7% of depressed participants (Astrom, Adolfsson, & Asplund, 1993). Friendship loss post stroke is well-documented (Northcott, et al., 2016), thus it might be anticipated that a friends-based network would be vulnerable to network change post stroke. In fact, this did not happen. It seems likely that this reflects the perceived quality of the friendships, and also that these participants had fewer family resources to fall back on. Labi, Phillips, and Greshman (1980) also documented that those who named a friend rather than a spouse as their significant other, and those who lived alone, were less likely to reduce out-of-house socializing following a mild stroke. The longitudinal social network study conducted by Wenger (1994) concluded that most network types in fact remain stable over time, with only a small percentage shifting each year. By contrast, over one third of the sample in our qualitative study shifted network types. Although the sampling was purposive and prevalence rates need to be interpreted cautiously, still, it is likely that following a stroke network shifts are relatively common. Of interest, people with restricted networks prior to the stroke did not develop a more supportive network type post stroke, despite increased support needs. Similarly, people with a friends-based network did not develop a family-based or diverse network. These patterns suggest that close family ties were not commonly built up following the stroke, but rather, where family provided increased support generally this reflected close family relationships prior to the stroke. Certainly, the participants who felt closer to their adult children post stroke tended to be those who had high-quality relationships with their children prior to the stroke. Other research has found that how adult children support their elderly parents can be predicted by patterns set up earlier in life (Régnier-Loilier, 2006). One factor which we anticipated might cause a person to shift network type was aphasia. There is evidence that people with aphasia take part in fewer social activities, are at risk of losing friends, and can feel excluded from social participation (Cruice et al., 2006; Parr, 2007; Vickers, 2010). People with aphasia have been found to have significantly fewer contacts with friends than healthy older adults (Hilari & Northcott, 2016). It is therefore of interest that aphasia did not preclude a person retaining a friends-based network. In common with preexisting literature (Northcott, et al., 2016; Parr, 2007), these participants sometimes encountered stigma within their wider social networks, and their aphasia could make conversations more difficult or effortful. Nonetheless, participants who belonged to this network type prestroke appeared able to retain their most important friends and so this network type post stroke, despite having aphasia. We also anticipated that age might be a factor in what network type a person belonged to. The “old-old” (over 80) tend to maintain fewer elective ties and lose touch with those on the periphery of their networks (Fingerman, 2004). Previous social network typology research has found old age is predictive of belonging to a restricted network type (Li & Zhang, 2015). Survey research (n = 26,784) exploring the relationship between old age, mental health, and social networks has found that smaller social network structures (number of meaningful contacts) were significantly associated with worse mental health, and that this association was strongest amongst those aged over 80 (Litwin, Stoeckel, & Schwartz, 2015). The authors suggest that this finding “emphasizes the importance of having a sufficient number of meaningful people with whom to interact in the latest stage of life” (p307). In the present study, the restricted networks were predominantly made up of those aged over 75 years old: as anticipated, it was common for these participants to describe how friends and family members had died, become disabled or unwell, which, combined with their own frailty or disability, had limited contact. Although those in restricted-supported networks benefitted from the emotional closeness of a key supporter, it was a network structure vulnerable to change. Those in the restricted-unsupported networks described poor mental health. There were, however, also participants aged over 75 represented in the diverse and friends-based categories both prestroke and post stroke. The social network typology developed through cluster analysis matched the qualitative typology well. The only surprising result was that Cluster One (Diverse) scored below average on the Groups factor. In fact, group membership was not a delineating variable for the qualitative typology, so this result does not contradict the qualitative typology. Nonetheless, it is a surprising finding, as the literature suggests those in diverse networks are typically involved in multiple groups/community activities (Fiori et al., 2006; Litwin, 2001; Wenger, 1994). This result may reflect the fact that group membership for this sample was low even before the stroke. Thus premorbidly, the mean (SD) number of groups that participants belonged to was 0.88 (1.03). Approximately 50% of participants did not belong to any group. Six months post stroke, the mean number of groups had reduced only slightly to 0.79 (1.0). Thus for the majority of participants in this sample group membership was not a key factor in their social network. The only participants for whom group membership appeared to contribute a significant component six months post stroke were those in Cluster Two (“friends-based”). Strengths/Limitations A strength of the study was the inclusion of people with aphasia, with interviews being conducted by experienced speech and language therapists. Another strength of the study was the use of cluster analysis to provide triangulatory evidence as to the validity of the qualitative typology. Nonetheless, it would have provided stronger evidence had the timescales matched (quantitative data collected at six months; qualitative at roughly one year), and the delineating variables been more similar. For example, a key consideration in assigning participants to the qualitative typology was determining how much support they received from different network members: the SSNS does not assess this, and therefore this could not be entered into cluster analysis. A limitation of the study is the timescale: at one year post stroke, many were anticipating further recovery and as such, their social networks were arguably still in the process of transition. In terms of generalizability, this study explores the experiences of those recruited through London teaching hospitals, and may not transfer to other contexts. Clinical Implications There is broad consensus that health service provision should consider the social consequences of health states (Intercollegiate Stroke Working Party, 2016). It is of concern, therefore, that having a stroke increased the risk of a person developing a restricted social network. In our qualitative typology, 3% belonged to a restricted unsupported-network prior to the stroke: this rose to 17% by six months post stroke. Given the close links between social isolation and other adverse outcomes, such as worse recovery and poorer quality of life (Northcott, et al., 2016), stroke services should consider how best to support a person in maintaining social contacts or even developing new contacts. This may involve health professionals paying greater attention to a client’s social context during the rehabilitation process, and potentially working more closely with social services and the voluntary sector. Preliminary evidence suggests that therapy delivered in a group context, and on-going peer support, may be helpful in alleviating isolation (van der Gaag et al., 2005; Vickers, 2010). Caution is perhaps necessary, however, before prescribing increased social participation for all who have reduced social networks post stroke. Machielse (2015) argues that the socially isolated are not a homogenous group, and that “interventions should fit with the ambitions and strategies of the clients involved” (p. 350). Their study explored social work interventions, interviewing socially isolated individuals, social workers, and examining log books detailing case studies. For individuals who had been isolated for many years it took time to build trust, and brief “one size fits all” social interventions were less successful. There is also arguably a need to respect that some participants preferred spending time with close family and friends following the stroke rather than seeking to participate more widely: Machielse (2015) also describes a subset who she termed “secures” who sought protective “safe” family environments. Our research found that it is not only physical disability that causes a person to shift network type, but also the psychological impact of stroke, causing a withdrawal from social interaction. There is arguably a need for further research exploring how best to provide psychological support post stroke in a way that enables people to re-engage socially, for example, through approaches such as solution focused brief therapy (Northcott, Burns, Simpson, & Hilari, 2015). Another implication from our research is that for those reporting friends-based networks prior to the stroke, friends were their primary source of support rather than family, and this remained true even after the stroke. Other research has found that friends can play a key role in enabling a person to “live successfully” following stroke and aphasia (Brown, Davidson, Worrall, & Howe, 2013). Although there is rightly recognition of the importance of including spouses and primary carers during rehabilitation in stroke guidelines (Intercollegiate Stroke Working Party, 2016), there is less emphasis on considering a person’s friendship circle. It may be that supporting friends when they first visit hospital to communicate more successfully with a person who has aphasia, or considering maintaining friendships as a legitimate goal of therapy, would be valuable. Conclusion/Summary We interviewed people two weeks, six months, and one year post stroke in order to develop a social network typology and explore patterns of social network change post stroke. Around one third of participants shifted from one social network “type” to another post stroke, most commonly from a “diverse” network (plentiful contact with both family and friends) to a family-based network. The friends-based network appeared relatively stable. More people belonged to restricted network types following the stroke. Triangulatory evidence for the validity of the typology was provided by k-means cluster analysis. Stroke services should consider the social impact of stroke, and how best to support those in restricted network types. Supplementary Material Supplementary data is available at The Gerontologist online. Funding This study was supported by a grant from the Consortium for Healthcare Research of the Health Foundation. Conflicts of Interest None Acknowledgments We are very grateful to Jane Ritchie, who was closely involved at all stages of data collection and analysis. We would also like to acknowledge the stroke unit teams of St Mary’s Hospital in London and the Royal Free Hospital, as well as Alice Lamb, and, of course, the participants themselves. References Adamson J. Beswick A. , & Ebrahim S . ( 2004 ). Is stroke the most common cause of disability ? Journal of Stroke and Cerebrovascular Diseases , 13 , 171 – 177 . doi: 10.1016/j.jstrokecerebrovasdis.2004.06.003 Google Scholar CrossRef Search ADS PubMed Astrom M. Adolfsson R. , & Asplund K . ( 1993 ). Major depression in stroke patients. A 3-year longitudinal study . Stroke , 24 , 976 – 982 . doi: org/10.1161/01.str.24.7.976 Google Scholar CrossRef Search ADS PubMed Brown K. Davidson B. Worrall L. E. , & Howe T . ( 2013 ). “Making a good time”: the role of friendship in living successfully with aphasia . International Journal of Speech-Language Pathology , 15 , 165 – 175 . doi: 10.3109/17549507.2012.692814 Google Scholar CrossRef Search ADS PubMed Burns R. P. , & Burns R . ( 2008 ). Business research methods and statistics using SPSS . London : Sage . Cruice M. Worrall L. , & Hickson L . ( 2006 ). Quantifying aphasic people’s social lives in the context of non-aphasic peers . Aphasiology , 17 , 333 – 353 . doi: org/10.1080/02687030600790136 Google Scholar CrossRef Search ADS Enderby P. M. Wood V. A. Wade D. T. , & Hewer R. L . ( 1987 ). The Frenchay Aphasia Screening Test: a short, simple test for aphasia appropriate for non-specialists . International Rehabilitation Medicine , 8 , 166 – 170 . doi: org/10.3109/03790798709166209 Google Scholar CrossRef Search ADS PubMed Fingerman K. L . ( 2004 ). The consequential stranger: Peripheral relationships across the life span . In F. R. Lang K. L. Fingerman (Eds.), Growing together: Personal relationships across the life span (pp. 183 – 209 ). New York : Cambridge University Press . Google Scholar CrossRef Search ADS Fiori K. L. Antonucci T. C. , & Cortina K. S . ( 2006 ). Social network typologies and mental health among older adults . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 61 , P25 – P32 . doi: org/10.1093/geronb/61.1.p25 Google Scholar CrossRef Search ADS Fotiadou D. Northcott S. Chatzidaki A. , & Hilari K . ( 2014 ). Aphasia blog talk: How does stroke and aphasia affect a person’s social relationships ? Aphasiology , 28 , 1281 – 1300 . doi: org/10.1080/02687038.2014.928664 Google Scholar CrossRef Search ADS Hair J. H. , & Black W. C . ( 2000 ). Cluster analysis . In L. G. Grimm P. R. Yarnold (Eds.), Reading and understanding more multivariate statistics (pp. 147 – 205 ). Washington, DC : American Psychological Association . Hilari K. Lamping D. L. Smith S. C. Northcott S. Lamb A. , & Marshall J . ( 2009 ). Psychometric properties of the Stroke and Aphasia Quality of Life Scale (SAQOL-39) in a generic stroke population . Clinical Rehabilitation , 23 , 544 – 557 . doi: 10.1177/0269215508101729 Google Scholar CrossRef Search ADS PubMed Hilari K. , & Northcott S . ( 2016 ). “Struggling to stay connected”: comparing the social relationships of healthy older people and people with stroke and aphasia . Aphasiology . doi: doi.org/10.1080/02687038.2016.1218436 Holt-Lunstad J. , & Smith T . ( 2012 ). Social relationships and mortality . Social and Personality Psychology , 6 , 41 – 53 . doi: org/10.1111/j.1751-9004.2011.00406.x Intercollegiate Stroke Working Party . ( 2016 ). National clinical guideline for stroke . London : Royal College of Physicians . Labi M. L. Phillips T. F. Gresham G. E . ( 1980 ). Psychosocial disability in physically restored long-term stroke survivors . Archives of Physical Medicine and Rehabilitation , 61 , 561 – 565 . Google Scholar PubMed Li T. Zhang Y . ( 2015 ). Social network types and the health of older adults: exploring reciprocal associations . Social Science and Medicine , 130 , 59 – 68 . doi: 10.1016/j.socscimed.2015.02.007 Google Scholar CrossRef Search ADS PubMed Litwin H . ( 2001 ). Social network type and morale in old age . Gerontologist , 41 , 516 – 524 . doi: org/10.1093/geront/41.4.516 Google Scholar CrossRef Search ADS PubMed Litwin H. Shiovitz-Ezra S . ( 2011 ). Social network type and subjective well-being in a national sample of older Americans . The Gerontologist , 51 , 379 – 388 . doi: 10.1093/geront/gnq094 Google Scholar CrossRef Search ADS PubMed Litwin H. Stoeckel K. J. , & Schwartz E . ( 2015 ). Social networks and mental health among older Europeans: are there age effects ? European Journal of Ageing , 12 , 299 – 309 . doi: doi.org/10.1007/s10433-015-0347-y Google Scholar CrossRef Search ADS PubMed Machielse A . ( 2015 ). The heterogeneity of socially isolated older adults: a social isolation typology . Journal of Gerontological Social Work , 58 , 338 – 356 . doi: 10.1080/01634372.2015.1007258 Google Scholar CrossRef Search ADS PubMed Northcott S. Burns K. Simpson A. , & Hilari K . ( 2015 ). ‘Living with aphasia the best way I can’: A feasibility study exploring Solution-Focused Brief Therapy for people with aphasia . Folia Phoniatrica et Logopaedica , 67 , 156 – 167 . doi: 10.1159/000439217 Google Scholar CrossRef Search ADS PubMed Northcott S. , & Hilari K . ( 2011 ). Why do people lose their friends after a stroke ? International Journal of Language and Communication Disorders , 46 , 524 – 534 . doi: 10.1111/j.1460-6984.2011.00079.x Google Scholar CrossRef Search ADS PubMed Northcott S. Hilari K . ( 2013 ). Stroke Social Network Scale: development and psychometric evaluation of a new patient-reported measure . Clinical Rehabilitation , 27 , 823 – 833 . doi: 10.1177/0269215513479388 Google Scholar CrossRef Search ADS PubMed Northcott S. Moss B. Harrison K. Hilari K . ( 2016 ). A systematic review of the impact of stroke on social support and social networks: associated factors and patterns of change . Clinical Rehabilitation , 30 , 811 – 831 . doi: 10.1177/0269215515602136 Google Scholar CrossRef Search ADS PubMed Parr S . ( 2007 ). Living with severe aphasia: tracking social exclusion . Aphasiology , 21 , 98 – 123 . doi: org/10.1080/02687030600798337 Google Scholar CrossRef Search ADS Pinquart M. , & Sorensen S . ( 2000 ). Influences of socioeconomic status, social network, and competence on subjective well-being in later life: a meta-analysis . Psychology and Aging , 15 , 187 – 224 . doi: org/10.1037//0882-79188.8.131.52 Google Scholar CrossRef Search ADS PubMed Régnier-Loilier A . ( 2006 ). How often do adult children see their parents ? Population and Societies , 427 , 2 – 5 . Ritchie J. Lewis J. , & Elam G . ( 2003 ). Designing and selecting samples . In J. Ritchie J. Lewis (Eds.), Qualitative research practice: A guide for social science students and researchers . London : Sage . Ritchie J. , & Spencer L . ( 1994 ). Qualitative data analysis for applied policy research . In A. Bryman R. Burgess (Eds.), Analysing qualitative data . London : Routledge . Google Scholar CrossRef Search ADS Singleton R. A. Jr Straits B. C. , & Straits M. M . ( 1993 ). Approaches to social research . Oxford: Oxford University Press . Spencer L. , & Pahl R . ( 2006 ). Rethinking friendship: Hidden solidarities today . New Jersery : Princeton University Press . Spencer L. Ritchie J. O’Connor W. Morrell G. , & Ormston R . ( 2014 ). Analysis in Practice . In J. Ritchie J. Lewis C. McNaughton Nicholls R. Ormston (Eds.), Qualitative research practice: A guide for social science students and researchers ( 2nd ed.). London : Sage . Teo A. R. Choi H. Valenstein M . ( 2013 ). Social relationships and depression: ten-year follow-up from a nationally representative study . PLoS One , 8 , e62396 . doi: 10.1371/journal.pone.0062396 Google Scholar CrossRef Search ADS PubMed van der Gaag A. Smith L. Davis S. Moss B. Cornelius V. Laing S. , & Mowles C . ( 2005 ). Therapy and support services for people with long-term stroke and aphasia and their relatives: A six-month follow-up study . Clinical Rehabilitation , 19 , 372 – 380 . doi: org/10.1191/0269215505cr785oa Google Scholar CrossRef Search ADS PubMed Vickers C . ( 2010 ). Social networks after the onset of aphasia: The impact of aphasia group attendance . Aphasiology , 24 , 902 – 913 . doi: doi.org/10.1080/02687030903438532 Google Scholar CrossRef Search ADS Wenger G. C . ( 1994 ). Support networks of older people: A guide for practitioners . Bangor : Centre for Social Policy Research and Development, University of Wales . Winkler M. Bedford V. Northcott S. , & Hilari K . ( 2014 ). Aphasia blog talk: How does stroke and aphasia affect the carer and their relationship with the person with aphasia ? Aphasiology , 28 , 1301 – 1319 . doi: org/10.1080/02687038.2014.928665 Google Scholar CrossRef Search ADS © The Author(s) 2017. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: firstname.lastname@example.org. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)
The Gerontologist – Oxford University Press
Published: Mar 14, 2017
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera