Men and loneliness in the ‘west’: A critical interpretive synthesis

Loneliness considerable in recent but has seldom investigated through a gendered lens. This review aims to critically summarise research substantively related to men and loneliness. A critical interpretive synthesis, incorporating a pre-defined search strategy, was employed to analyse a broad variety of data. The searches were conducted in July 2019, across seven databases: MEDLINE; PsycINFO; Scopus, ASSIA, SSCI; Sociological Abstracts; and Social Policy and Practice. Included studies were written in English, conducted in high income nations in Western Europe, North America, and Australasia, with no date limitations. Qualitative studies required a substantive focus on sex/men or gender/masculinities and loneliness, and quantitative studies either an explicit focus on sex differences in loneliness, or data substantially relevant to loneliness and men. Seventy-nine studies met the inclusion criteria. sex men likely to engage in risky/unhealthy vi) feeling/being ‘insufficiently masculine’ can result in loneliness; and vii) intersections of identity facilitate different gendered results.

relationships. It also appeared to be characterised either by the construction of a masculine persona, or a fear of the repercussions for initiating intimate conversation. When attempting to ascertain whether a man is lonely, acknowledging gendered variation in survey responses is required. However, it is unclear how generalisable these 'synthetic constructs' are, and their potential inter-relatedness requires further research.

Background
Loneliness has become widely recognised as a significant and growing public health concern.
Although many definitions of loneliness exist, the predominant consensus is that it is a negative emotion representing a feeling of a lack, or loss, of meaningful social relationships (1)(2)(3)(4)(5). In the United Kingdom (UK), research has suggested over nine million adults are either 'often' or 'always' lonely (6).
Loneliness has been associated with depression (7), compared to smoking 15 cigarettes a day (8), identified as potentially increasing the risk of cardio-vascular (9), and found to increase the risk of premature death (10).
Recognising this, in 2018, the UK government published a strategy for tackling loneliness (11).
Despite personalised approaches that recognise the complex and subjective nature of loneliness being one of its guiding principles, a consideration of the distinct needs of men was absent. A growing body of evidence highlighting how gender intersects with other social, economic, environmental, political and cultural determinants of health has driven an increasing focus on men's and women's health (12). The importance of gender (masculinities) in men's experience of, and help-seeking for, mental health is particularly well documented (13)(14)(15). Furthermore, failure to recognize and/or seek help for mental health problems has been associated with men's higher rate of suicide (12). It has also been argued that gendered cultures and modes of social organisation mean men and women do not have the same conceptualisations of loneliness, nor of what is desired to prevent and alleviate it (16). To date, there has been a paucity of literature specifically focused on the influence of sex or gender on men's constructions and/or experiences of loneliness, and its implications for policy and practice. Our review aimed to bridge this gap in the knowledge by critically examining the current literature to address the following question: What is the influence of sex or gender on men's constructions and/or experiences of loneliness?

Methods
The review followed the principles of critical interpretive synthesis (CIS) (17), adapted to incorporate a systematic search strategy. As this was a secondary synthesis of data, ethical approvals were not required.

Methodology
The goal of CIS is to critically examine relevant literature and data, then transform the underlying data into theoretical and conceptual ideas named 'synthetic constructs' (17). The approach was well suited to the current review question as it facilitates the integration of findings from diverse studies and disciplines into a 'single, coherent framework' (18, p2). CIS has also been recognised as a useful approach for reviewing 'challenging' areas of health research, such as mental health, in which people may not be as forthcoming (19).
Three deviations from Dixon-Woods et al.'s description of CIS were employed. The first was a 'resultsbased convergent synthesis' design (20), meaning that the statistical and qualitative studies were initially examined separately. The second was the use of thematic synthesis methods to identify 'data-driven' themes arising directly from the evidence base (21). In combination, these two adaptations allowed the review to summarise a particularly varied evidence base prior to attempting to form 'synthetic constructs'. The final adaptation was the use of an 'a priori' search strategy. Dixonwoods et al. (17) considered an 'iterative' and 'flexible' search strategy necessary for a truly 'critical' analysis. However, as summarising the existing evidence was an important element of this review, a 'pre-defined sequence' (17, p9) was considered more appropriate.

Search strategy
Electronic database searches were used to search for literature published anytime up to 30 June 2019. All searches followed a base strategy of (1) 'loneliness or social isolation' AND (2) 'gender or men or masculinities (and related terms, see additional file 1)', followed by (3) NOT 'children or neurological or biochemical or animal (and/or related terms, see additional file 1)'. Only 'loneliness' and 'social isolation', and not related terms such as 'social connectedness', 'social support', and 'solitude', were employed (pilot searches revealed excessive numbers of irrelevant studies when using additional terminology, and the review question requires the investigation of constructions of 'loneliness', thus a focus on the word was considered useful). To focus on studies with likely relevance to the UK, only English language studies, from high-income nations in Western Europe, North America, and Australasia, were considered for inclusion. Pilot searches included studies from all OECD nations, but this resulted in excessive numbers of results, among which some studies emphasised their difference from 'western' settings (22,23). An interpretive method for identifying 'adults' was specified as precise age ranges resulted in less meaningful inclusions/exclusions. Searches were conducted in MEDLINE (OVID), PsycINFO (OVID), Scopus, ASSIA (proquest), SSCI (Web of Science), sociological abstracts (proquest), and social policy and practice (OVID). A combination of Medical Subject Headings (MeSH) terms, subject categories, and free text searches were used. Additional file 1 details the exact strings for each database.

Inclusion criteria
Seven points for inclusion were constructed. Separate criteria for quantitative and qualitative studies were applied when necessary. No study was excluded according to quality. 6b. Quantitative studies -sample includes adult men and women, or men and a measure of gender roles (as comparators were required).
7a. Qualitative studies -reports substantive findings on sex/gender and loneliness.
7b. Quantitative studies -has an explicit focus on sex/gender and loneliness (i.e. the study states it focuses on sex/gender and loneliness, using any variables), or compares different types of loneliness by sex/gender, or includes a test of sex/gender, loneliness, and a third variable that is not a demographic characteristic, physical health, or mental health concern (including variables related to drug/alcohol use, exercise, or nutrition). Screening 1.
All records uploaded to Endnote, duplications removed.

2.
Title and abstract screening. Articles that did not include participants from specified nations, specified a focus on children or adolescents, only featured women, are neurological or bio-chemical studies, or which clearly did not investigate 'loneliness' and sex/gender in accordance with the inclusion criteria (including studies focusing on physical or mental health without a specified focus on gender and loneliness), were excluded at this stage. This stage was conducted by one reviewer, then two separate random samples of 2.5% were checked by a supervisor each.
Disagreements were resolved through discussion, and the remaining 95% of articles re-assessed by the lead researcher.

3.
Full text retrieval. Articles were included according to the seven-point criteria. This stage was conducted by one reviewer, then the final preferred reporting items for systematic reviews and meta-analyses (PRISMA) chart, and Endnote files of included/excluded articles, were checked by 2 supervisors.

Data extraction and analysis
Findings were analysed via a three-stage process. Each stage was primarily conducted by the lead researcher, with advice and input from all listed authors.

1.
To begin the 'results-based convergent synthesis' (20), separate tables summarising the characteristics and findings of the quantitative and qualitative studies were created. For the quantitative studies, this relayed the article's title, the details of its sample, and summaries of its relevant variables, statistical methods, and results (additional file 2). For the qualitative studies, this presents the article's title, the context of the study, the details of the sample, and a summary of the relevant findings as relayed by the author/s (additional file 3).

2.
Two thematic analyses, aimed at summarising the content of the quantitative and qualitative separately, were conducted. These are presented as two sets of tables, in which each study with information related to that theme was displayed in the corresponding table (additional files 4 and 5). This stage aimed to identify 'datadriven' themes (21). It was not considered necessary to assign all articles to a theme as all articles were to be re-examined at stage 3. Sixty-four of the 79 studies were related to at least one theme.

3.
The final stage involved two components. These were conducted alternately and iteratively to allow for the reinterpretation of the data into theoretical concepts not identified by the original authors. The first component was a contrasting of the themes and data from the quantitative and qualitative thematic analyses, and the second was a revisiting of all 79 articles to re-examine whether they had theoretical relevance to the synthetic constructs that were being built. Whilst conducting the above tasks, tables with all data relevant to each synthetic construct were built, which are presented as the final results. Seventeen of the 79 included articles are not cited in the main manuscript as, after going through the above steps, they were not deemed to have provided sufficiently relevant data.

Study characteristics
Seventy-nine studies, published between 1978 and 2018, were included in the final review ( Figure 1).
Sixty-three (80%) were quantitative (additional file 2), and 16 qualitative (additional file 3 were excluded due to a lack of access. Figure 2 details the process undertaken for these 141 articles.

Synthetic constructs
Seven 'synthetic constructs' were identified. These are listed table 1, then summarized narratively in the following sub-sections. The results from stage 2 of the analysis are available in additional files 4 and 5. 1. Men's loneliness appears to be more closely associated with a perception they possess poor quality social networks The evidence for this construct was built from 18 studies (table 2). Four were primarily related to social network size. Stokes and Levin (24) found social network size and type to be more strongly associated with loneliness in men, and Bell and Gonzalez (25) found 'social integration' a predictor of loneliness in men but not women. Bell (26), however, found no sex difference in the association between 'network density' and loneliness, and Dykstra and de Jong-Gierveld (27) found that women showed a stronger association between social network size and loneliness.
Research investigating sex differences in people's perceptions of their social networks offers clearer support for this construct. Patulney and Wong (28) found that, even after controlling for people's preferences, men considered themselves to be more 'socially disconnected' than women. In the four studies using the Rokach 'causes of loneliness' scale (29)(30)(31)(32), three found men were more likely to state 'social marginality' as a cause of loneliness (30)(31)(32). Sundberg (33) found that feeling 'alone or alienated from positive persons, places, or things' was a more commonly selected survey item by men (although there was no difference in feelings of 'isolation'). Lastly, Stevens and Westerhof (34) found that 'emotional support from friends' mediated sex differences in loneliness, in that men had less emotional support, thus were lonelier.
Several studies less clearly focused on social networks further underlines these findings. Tornstam (35) surveyed people's views of what caused them loneliness, and men were more likely to select items related to a physical distance from people, particularly family, whereas women were more likely to select items focused on emotions, such as being 'misunderstood'. Similarly, Juntilla et al. (36) recorded that men were more likely to select items such as having fewer 'good friends' or 'nobody to talk to', whereas women were more likely to say they were 'shut out' or 'excluded by others'. Lastly, men were often found to be 'socially' lonelier, and women 'emotionally' lonelier (27,(37)(38)(39). As social loneliness refers to a perception that one does not possess adequate networks, and emotional loneliness a lack of intimacy, this seems to further place women's loneliness as denoting an emotional experience, and men's as a result of (perceived to be inadequate) social networks.   (41, p1209).
On the other side of this 'pair' were findings indicating that this reluctance may result in loneliness: 'the imperative of "separateness" allied to masculine self-identity appears to hamper the establishment of new relationships' (42, p39 Indeed, among four studies surveying people's opinions of lonely men and women, three suggested lonely men are particularly unlikely to be socially 'accepted' (47)(48)(49), and the other showed no significant difference (50). While the construction of a dominant masculine identity may characterise some men's emotional reticence, they may also fear the ramifications of openly discussing emotional vulnerabilities. Twenty-eight studies contributed evidence related to this construct (Table 4). Ten found a significantly larger difference in the prevalence of loneliness between single and married men than they did between single and married women (27,28,34,37,38,(51)(52)(53)(54)(55), and two also suggested that single men are lonelier than those in unmarried romantic relationships (51,56). The findings from seven qualitative studies further emphasised the particular importance of romantic relationships to men (40,42,46,(57)(58)(59)(60). As one man stated after the death of his wife, "all that intimacy goes and there is nothing to replace it" (59, p426). Three (61-63) also found men to be more 'romantically' lonely, though only one of these (62) found this to be statistically significant, and a fourth study measuring this concept found no significant difference (64). However, the survey instruments used in these included items that did not separate having a romantic partner with desiring one. For instance, 'I have a romantic partner with whom I share my most intimate thoughts and feelings' was an item in three (61)(62)(63). As such, these studies are less effective for investigating this construct than the term 'romantic loneliness' implies.
Four studies refuted this construct. Spahni et al. (65) found widowhood affected men and women similarly, and Woodward et al. (66) found no significant sex differences in loneliness after divorce.
Collins (60) found that only two of seven males widowers interviewed expressed a desire for a new partner, albeit only one specifically stated they did not desire one. Lastly, Dahlberg et al. (67) found that long term widowhood (7+ years) only predicted loneliness in women. However, they also found that recent widowhood (<7 years) was a much stronger predictor of loneliness in men, and they did not record whether respondents entered a new relationship in those seven years.
Two important sub-constructs were identified. Dykstra and Fokkema (37) found 'partnercenteredness' to be a statistically significant explanation of loneliness in divorced men, yet this was not significant among women, indicating men's attitudes explain the findings related to this construct.
Tornstam (35), however, noted that married women were significantly lonelier than married men, suggesting that marriage may also be less protective for women. Lastly, two perspectives on why men may be more greatly affected by partner status were identifiable. Firstly, men did seem to consider a romantic partner the first, and primary, person for meaningful social interaction: 'we [men] do not talk like I'd talk with my wife' (46, p1252). Secondly, men's social networks were often found to be limited and dependent on their spouse (42,59,60). For instance, when discussing services aiming to alleviate loneliness, one man stated: a group focused on men is self-perpetuating because then the people participating get more practice in being responsible for their own social network, their own social life, instead of that doing, that thing through the female spouse (59, p804). Four studies compared sex differences in the prevalence of loneliness using multiple tools (table 5).
Three included both a direct question asking how often the respondent felt lonely, and a scale representing loneliness (39,52,68). Men reported less loneliness in response to the direct question than on the indirect scale in all three. The final study found no overall sex difference according to the University of California loneliness (UCLA) scale, but that men appeared much lonelier when using the loneliness deprivation scale, and slightly lonelier using the 'emotional/social loneliness inventory' (69). half of the studies found no significant sex difference. Six out of seven studies both investigating nonstudent samples and using a direct question found women to be lonelier, and no studies using a direct question found men to be lonelier.  (table 7). Three asked men how they coped with loneliness, in which the most consistent finding was a reference to risky/unhealthy behaviours such as drinking alcohol (35,40), substance abuse (36), visiting sex workers (40), and gambling (36). However, Botterill et al. (53), who found that loneliness was equally likely to result in problem gambling among men and women, offered some contrary evidence. Additionally, it is notable that Munoz-Laboy et al. (40) concluded the men in their study were more likely to enact these behaviours in lieu of discussing emotional issues with people who were not a romantic partner. may be critical to men's loneliness, as participants in this study implied they felt lonely despite remaining present at many of the same team events.  (29)(30)(31)(32). As noted above, gay men discussed a sense of loneliness arising from not fulfilling gender norms (71,72). In sporting arenas, heroic ideals of loneliness, such as 'suffering alone' to avoid 'letting the team down' (73), or being a 'lonely hero' in the individual sport of running (74), were constructed by male participants. Female runners, however, lamented this 'lone wolf' mentality as a source of loneliness (74). Additionally, male students appeared to be lonelier than female students across the included studies (table 6), whereas older women were lonelier than older men in some studies (10,52), possibly because of the greater number of female widows (10).
First, older men prefer services with a constructive goal (60,75,76,78), or, as Collins (60) termed it, services that are 'task-focused'. Secondly, several accounts suggested that older men's lives are viewed from, or placed within, a feminine context (42,60,77), resulting in services that are unsuitable to older men's wants and needs. Indeed, the assumption that the above 'isolating events' induced loneliness was considered part of this feminine context by some (42,58). Lastly, several studies highlighted the importance of a supportive and non-competitive atmosphere -'there's nothing to prove, and people come here in relaxed fashion doing that which they are able to do' (75, p142).

Discussion
Loneliness is a pressing public health concern, and an area which may require gender-sensitive approaches and intervention. Our review has bridged an important gap in the literature by synthesising the current evidence on men and loneliness. Men's reluctance to seek help, particularly for emotional issues, is well recognised (14), and the notion that men are hesitant to discuss loneliness would appear a parallel concept. Notably, though, a second dimension to this was manifested in the current review; that a disinclination to discussing emotional issues may be a cause of loneliness, and not just a barrier to seeking help for it. Though this review did not find a significant body of evidence for it, the notion that feelings of loneliness are particularly likely to result in risky/unhealthy behaviour among is also consistent with a large body of research. Courtenay (13), for example, link men's disinclination to seek help for mental health issues with higher rates of alcohol and drug abuse, Lee and Hanson (79) found loneliness to be a predictor of recidivism for sexual offences, and Hubach et al. (80) suggested that young men self-medicated loneliness with drugs and sex. Furthermore, inclusion point 7b specified that in statistical studies not explicitly focused on sex and loneliness, but including tests of sex, loneliness, and a third variable, the third variable should not be related to unhealthy behaviours such as drug or alcohol abuse. As such, this review may have missed evidence for this construct.
Both this review and some narrative literature has suggested the possibility that men are less inclined to state they are lonely in response to a direct survey question (81,82). However, this may not mean an indirect scale is 'better'-a recent Age UK report found many people are lonelier using a direct question, suggesting the UCLA scale also misses some people's loneliness (83). Indeed, for Jylhä (84), an indirect scale forces the respondent to answer according to the researcher's definition of loneliness, thus inherently lacks validity. The De jong-Gierveld scale often found men to be slightly lonelier than women (table 6), something that was largely a result of their greater 'social' loneliness, and Oshagan and Allen (69, p402) proposed that the loneliness deprivation scale's focus on 'deeper, more existential feelings of sorrow and aloneness' may explain men's greater loneliness on this.
Though the UCLA scale appears least likely to find a sex difference in the prevalence of loneliness, Juntilla et al. (36) found many of the individual items displayed statistically significant sex differences.
They even find that men seem to be emotionally lonelier, and women socially lonelier, suggesting different ways of measuring forms of loneliness may also provide markedly different results.
Sensitivity to how gender (or other factors) affect the findings, and the use of multiple tools if possible, would therefore seem more appropriate than the pursuit of a 'best' tool.
Franklin et al. (16) define loneliness as 'belongingness', and trace men's belongingness as historically constructed through participation in public realms, whereas women act/acted as 'kin-keepers', taking responsibility for family and friendship networks. In turn, men's construction of loneliness as related to a perception of poor social networks may represent poor engagement with the public realm, and their greater reliance on romantic relationships could originate from the tendency for women to act as 'kin-keepers'. Connell's (85,86) concept of 'hegemonic' masculinities, the theoretical notion that certain masculinities are privileged for their reification of unequal gender relations, offers a useful framework for understanding the concept of a 'loneliness of outsiderness'. Indeed, Connell's (87, p40) list of insults denoting femininity and/or homosexuality strikingly exemplifies how this may be culturally constructed. It may also offer insight on why men may not wish to discuss emotional issues, or even admit to loneliness on a survey, given that to do so may undermine a powerful re.
At times, the synthetic constructs identified in the current review appeared to be intertwined. Indeed, men's reluctance to discuss emotional issues has been proposed to be the cause of men's poorer quality social networks (46), their greater dependency on romantic relationships (46), their disinclination to state they are lonely (39,40,81), and their increased propensity towards risky or unhealthy behaviours (40). The notion of being/feeling insufficiently masculine as a cause of loneliness, though, would appear to suggest a complex situation where being emotionally candid can be a cause of loneliness too. Nevertheless, some studies, albeit all of which were focused on older men, have suggested this can be overcome, particularly in social settings where men can take part in 'constructive' activities inflected with a 'supportive' atmosphere (46,75,76,78,89).

Strengths and limitations
Synthetic constructs are identified inductively by examining a broad array of data, thus reviews focused specifically on researching one of these constructs would likely provide more robust investigations of their accuracy, truthfulness, and generalisability. This, however, would require several entire reviews, is only possible after the identification of the constructs, and would be unable to present an overall summary of key concepts related to men and loneliness. The use of a predefined search strategy, by necessity of limited sensitivity, meant that this review may have missed relevant perspectives an iterative or more sensitive strategy would allow the inclusion of. An iterative search strategy, though, would be unable to systematically summarise the evidence base.
Additionally, though the articles that could not be accessed were deemed unlikely to add important perspectives, it cannot be guaranteed.
37% of the included studies were published before the year 2000, and many were conducted with students or older populations, or in different countries that, at times, portrayed markedly different results. Though occasionally commented on, the potential inter-relatedness of these synthetic constructs was neither widely nor thoroughly investigated. Methodological issues were present in some, such as those measuring 'romantic' loneliness (61)(62)(63)(64), or those using sex-role scales, which have been noted as inherently problematic (86,90). The use of loneliness scales may also be problematic given the findings above. The decision to not exclude articles on the basis of quality may further limit the confidence with which the constructs represent genuine trends. Lastly, if intersections of identity affect men's constructions and experiences of loneliness, these synthetic constructs may not be equally true for all groups of men. Whilst there was evidence for the importance of age, sexuality, student status, and even involvement in sport, areas such as ethnicity and social class received little attention. All this suggests the extent to which the synthetic constructs represent a meaningful perspective of modern Britain is unclear. Conversely, the focus on a small number of 'western' countries severely constrains their international relevance. Nevertheless, they are useful as a summary of important perspectives in the field of men and loneliness.
Implications for future research, policy and practice The seven constructs, and their sub-constructs, can be used as a concise, if not exhaustive, summary of what may be important in research, policy, and practice related to men and loneliness. Partner status appeared to be a more powerful 'risk-factor' for men than women, suggesting that policy and practice can benefit from acknowledging this risk-factor, but also that actions to facilitate quality social networks, regardless of partner status, may work to reduce dependency on romantic relationships. Working to overcome a masculine reticence to discuss emotional issues appears central to preventing and alleviating loneliness in men. Moreover, doing so appeared to require both challenging masculine ideals of invulnerability, and facilitating an atmosphere in which men do not fear being emotionally candid. In studies of older men, environments with a constructive focus appeared to provide meaning to a social activity, within which supportive atmospheres could be built.
The synthetic constructs also provided important information on identifying whether a man is lonely. Authors' contributions JR offer major contributions to the study design, conducted and analysed the research, and had a major role in writing the manuscript. PG and MK offered major contributions to the design, checked the data and analysis, and offered edits and advice on the final manuscript. All authors have read and approved the manuscript.