Survey data collection.

The quantitative data were collected using a survey administered by trained individuals between 16 May and 7 July 2022. A total of 887 Spanish residents took part in the study. Given the study aims, it was very important to access the general population living in any of the arrangements under consideration. In this vein, participants were a nationally representative sample for the target population, recruited through the services of the “Political and Social Analysis” research group of the University of Santiago de Compostela (Spain) were engaged for this purpose. This group has broad experience in conducting research surveys (those designed for academic purposes). Initial contact was made by telephone in all cases. The survey was applied by telephone call to the community-dwelling participants, except in cases where a face-to-face meeting was requested. For nursing home residents, the surveys were applied in face-to-face meetings. After the initial contact was made with potential participants and their informed consent had been obtained, the surveys were applied under the conditions chosen by those who agreed to participate. Nursing home residents were interviewed at a space in the nursing home that guaranteed the confidentiality of their interview. After eliminating participants who failed to provide information for any of the study variables, the final sample comprised 863 people. The minimum sample size was 108, calculated with G*Power [40] under the assumptions of a two-tailed test, α = .05; β = .01 (95% power); medium effect size (0.5) and 60 predictors.

Survey measurement.

The survey included a block of sociodemographic questions that incorporated the variables of sex, age, having children, education (incomplete primary education, completed primary education, secondary education and university education), living arrangements (living with partner only, living alone and living in a nursing home) and limitations on activities of daily living due to health problems. The latter variable was evaluated using the Global Activity Limitation Indicator (GALI) [41].

A second block of questions incorporated measures of loneliness and social support. Loneliness was assessed using the six-item version of the de Jong Gierveld loneliness scale (DJGLS) [42]. The DJGLS was validated in Spain by Ayala et al. [43] for a sample of older adults. It is a short and commonly used instrument in loneliness research that produces an overall score and two subscores for social and emotional loneliness. This multidimensional approach makes this instrument the most appropriate one to achieve the aims of this research. Three items are used to calculate the emotional loneliness score (e.g., “I miss having people around me”) and another three are used for social loneliness (e.g., “There are plenty of people I can rely on when I have problems”). Each question has three response categories (“yes”, “more or less” and “no”). Items are scored on a scale from 0 to 2, although they are subsequently recodified as dichotomous (0 or 1). In this recodification, following the author’s guidance [43], the middle category (“more or less”) indicates loneliness, together with the category corresponding to the direction in which the question is formulated (e.g., “yes” for the item “I miss having people around me” and “no” for the item “There are plenty of people I can rely on when I have problems”). Higher scores on the scale indicate stronger feelings of loneliness (range from 0 to 6 for the total score, and from 0 to 3 for the emotional and social dimensions). The total scores for both subscales (emotional and social loneliness) were used as outcome variables for the analysis. In addition, the DJGLS allows the use of a cut-off point of 2 or higher to classify participants as experiencing loneliness [44, 45]. In the present study, this cut-off point was used exclusively to provide an estimate of the prevalence of loneliness in the sample of older adults (see Table 1). For the remaining analyses, the scores were used as continuous variables.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Joint display of integrated data collection.

https://doi.org/10.1371/journal.pone.0316751.t001

Functional social support was assessed using the Perceived Social Support Questionnaire (PSSQ) [46]. This scale was originally developed to measure social support among the Spanish-resident population. This instrument provides separate scores for the different sources of social support, offering a notable opportunity to include the specific sources of interest for each study. For each source, two items are used to measure emotional support (e.g., “Could you freely express and share your emotions with this person?”), two for instrumental support (e.g., “If you were sick or needed to be taken to the doctor, would this person be of any help?”), and two for advisory support or guidance (e.g., “Would this person be of any help if you had to make an important decision?”). Scores for the items corresponding to each source and representing the different types of support (emotional, instrumental, advisory) produce a composite functional support score for each source. The responding scale for each item ranges from 0 to 5, and the sum for each source (higher scores representing higher levels of social support provision) was standardised to a 0–10 range. In our case, the scale was adapted to measure support from the seven sources under consideration. As a result, in this study the PSSQ comprised 42 items with questions on the source and level of perceived social support from spouse/partner, sons, daughters, grandchildren, siblings, friends and neighbours. The PSSQ is a widely used instrument in Spanish-speaking populations, having been validated in various contexts (residential, general and hospital) [47, 48].

Quantitative data analysis.

First, descriptive statistics were used to present the respondents’ demographic characteristics. Second, prevalence of risk of loneliness was calculated using the above-mentioned cut-off score and reported as percentages of cases in each subgroup. Chi-square tests were used to determine demographic characteristics associated with risk of loneliness. Both steps offer useful contextual information. Third, the central analysis for this study was conducted: contrasting a multivariate model of the association between loneliness and social support. Structural equation modelling (SEM) was used for this purpose, including both the measurement model (the aforementioned loneliness and social support surveys) and the structural model (association between latent variables). This analysis used the total scores obtained for the relevant variables (emotional/social loneliness and social support from spouse/partner, sons, daughters, grandchildren, siblings, friends and neighbours). Fig 1 summarises the baseline model that was contrasted with the data, including both the measurement model and the structural model. This model was progressively adjusted to the data, using the modification indices. Each decision to modify the baseline model was theoretically informed.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Baseline model contrasted with empirical data via structural equation modelling.

https://doi.org/10.1371/journal.pone.0316751.g001

The significance level was set to p < .05 for all analyses. The fit of the SEM models was evaluated using a combination of relative fit indices and noncentrality-based indices. From the first group, Bollen’s Incremental Fit Index (IFI), the Tucker-Lewis Index (TLI), and the Bentler-Bonett Norme Fit Index (NFI) were used. From the second group of indices, the Root Mean Square Error of Approximation (RMSEA) and Bentler’s Comparative Fit Index (CFI) were used. For IFI, TLI, NFI and CFI, values exceeding or approaching .95 indicate a good fit of the model to the data [49]. More specifically, cut-off values of .99, .95, .92 and .90 are established as respectively corresponding to models achieving excellent, close, fair and poor fit. For RMSEA, values below .05 indicate a good fit, while values of .08 or lower suggest an acceptable fit [50]. It is also worth bearing in mind that large model size tends to be penalised in the calculation of indicators, whose values may deteriorate in models with a high number of parameters (observed variables) [51]. All statistical analyses were performed using IBM-SPSS (V29) and IBM-AMOS (V29).

In-depth interview procedures.

Semi-structured interviews were used to collect information in the qualitative phase, with 20 women and 10 men interviewed of ages ranging from 65 to 99 years (M = 76.8), of whom 13 were living alone (one-person households), 11 were only with a partner (couple without children living together) and 6 were in a nursing home. In this way, it was possible to include a sufficient number of participants, as defined by the inclusion criteria, particularly the criterion regarding living arrangements. The interviews were conducted between 12 September and 20 December 2022. Two participant identification and selection strategies were used. First, for the selection of community-dwelling older adults, a similar approach to that used for the quantitative phase sample was applied, involving engaging the services of the Foundation for Advanced Social Research. This not-for-profit foundation is dedicated to supporting research work and offers extensive experience in the extraction of samples for qualitative studies. It works with research groups in university settings and hence has detailed knowledge of academic research practices and standards. Second, in the case of nursing home residents, the research team directly contacted the nursing homes and provided information on the aims and ethical aspects of the study. Nursing home staff identified potential participants and established initial contact, after which arrangements were made to visit the nursing home and conduct the interview.

The inclusion criteria used for the study (see “Study design and target population” subsection) formed the starting point for participant selection. The identification of potential participants also incorporated an age-based criterion to ensure that the sample included a range of ages and avoided focusing on a narrow range (such as 65 to 75 year olds). It was also established that a significant number of participants had to be male. In this regard, the living arrangements included in the study implied a risk that the sample might be almost exclusively female, given their greater life expectancy and the feminine structure of the population aged over 65 years in Spain. The initial contact was used to confirm that the potential interviewee was part of the target population. Following confirmation that an individual was prepared to participate, a member of the research team then attended the interview, which was arranged at university offices, the interviewee’s home or another location meeting the requirements of being a comfortable and appropriate space for an interview. In the case of nursing home residents, a social worker at the nursing home provided a space that ensured the interview would be private and confidential. The opening minutes of the interview were used to reconfirm that the person fit the study inclusion criteria in all cases.

The interviews were conducted in Spanish, took between 40 and 75 minutes and involved the application of a semi-structured script including questions related to the study aims, such as: who do you have relations with in your daily life? Who is the most important? If you have any needs, who do you turn to? How valuable is their support? What kind of support is most important for you? The main focus was the experience of loneliness in old age, however, with questions including: do you feel alone? In which situations do you feel most lonely? Do you think there is a higher chance of feeling alone during this stage of life? Which circumstances can affect feelings of loneliness in old age? The script was designed to be flexible and relatively unstructured, making it possible to incorporate significant emerging topics into the dialogue. Table 1 presents a joint display illustrating the integration of quantitative and qualitative data collection instruments. The final column outlines the rationale for combining these methods. The qualitative questions included in the table are a subset of those included in the full script (see S1 File).

Qualitative material analysis.

All interviews were audio-recorded, transcribed and imported in Spanish into Atlas.ti (V.23) for analysis. First, the interviewees’ statements were used to identify the most emotionally close social networks. The in-depth interview script included an open-ended question designed to elicit participants’ perceptions of their most significant social relationships. This approach allowed for a more spontaneous and less structured identification of key social ties than a closed-ended questionnaire. The resulting transcripts were used to construct a database where the columns represented the 30 interviewees and the rows listed all the significant individuals mentioned. This information was processed using UciNet 6 software [52] to graphically represent the interviewees’ significant social relations. Second, a thematic analysis of the thirty interviews was conducted [53, 54]. Our approach to thematic analysis was influenced by the sequential explanatory design of our research. As noted in the “Study design and target population” subsection, the qualitative data collection phase was informed by the quantitative data collected in the first phase of the research. Furthermore, our thematic analysis aimed to deepen and expand our understanding of the model of association between social support and loneliness identified in the quantitative analysis. Finally and notably, our research design was based on a theoretical elaboration of the available empirical evidence, as analysed in the first section of this paper. As a result, our approach to coding and identifying themes was primarily deductive [53], driven by our analytic and theoretical focus on the complex relation between sources of social support and the dimensions of loneliness. The thematic analysis was completed in the following steps. The analysis started with familiarisation with the data (reading and re-reading). As part of this first step, the second author produced the initial codes in Atlas.ti linked to the research questions, focusing on the qualitative aspects (degree of emotional closeness, satisfaction with ties) and functional aspects (type of social support received and reciprocity) of the social relations and the experience of loneliness (dimensions and feelings of loneliness). Reflective notes were used throughout the process [55], a very useful tool for organizing information into meaningful groups. This initial set of codes was examined and refined by the first two authors in a close review of transcripts and interview summaries. The second step addressed the construction of themes [55]. The results of the quantitative analysis theoretically and analytically oriented both the search for themes and the relationship between them. The research team worked as a group and collaboratively to organize codes by giving a common meaning to those groups of codes that collapsed into emerging themes. These themes were classified into main themes and sub-themes, generating a provisional thematic map. The research team then identified and developed broader patterns and combinations of themes until they reached a consensus regarding the key themes of the interviews linked to the features of social support (by source) and their association with feelings of loneliness. Subsequently, the second author reviewed this collection of potential themes, contrasting it with the qualitative material; that is, with the excerpts and codes identified as an expression and part of each theme. This work produced the final thematic map, which the team subsequently analysed in an integrated manner together with the results of the quantitative phase.

Ethical considerations.

All study procedures were approved by the Universidad Complutense de Madrid ethical committee (report reference CE_20220217–14_SOC). Emphasis was placed on protecting participant confidentiality in both phases of the study (quantitative and qualitative). Participants were individually informed about the nature of the study and its risks and benefits. Notable benefits included the contribution to generating knowledge in a field of study (loneliness) that is particularly significant for society in general and for older adults in particular. In terms of risks, participants were warned that the interview included themes of family and social relations, which might entail discussing unpleasant or sad situations. Each participant had the opportunity to read their interview transcript prior to codification and analysis. Only one person requested to do so. In the quantitative phase, verbal consent was obtained. The record of this consent constituted the first item of the survey, which could not be continued unless the corresponding box was checked. Written consent was obtained in the qualitative phase, and all signed documents have been archived and are currently in the custody of the lead researcher. Participants in the quantitative phase were not paid for participating in the study, but qualitative-phase participants did receive payment.

Social support from family and loneliness.

In Figs 3 and 4, support from spouse/partner stands out as being related to lower loneliness scores. However, it is striking that the regression coefficients show its significance to be greater in the case of social loneliness (β = -.293; p < .001) than for emotional loneliness (β = -.074; p < .001). The analysis of qualitative material is aligned with these results. In all the interviews with participants living with their partner (without cohabiting children in all cases), the spouse emerged as a significant source of day-to-day support, helping participants to cope with the problems inherent to old age and acting as an emotional refuge to avoid loneliness. It is worth noting that in practice, spousal support has a combined impact in terms of reducing both social and emotional loneliness:

She really is for me [referring to her being a supporting person], she’s always looking after me, anything, she’s doing my drops, she’s reminded me twice today (…) And then at four this morning I was trying not to wake her up, she noticed me getting up and woke up because I was going to the bathroom to do my drops and there she is behind me, she’s woken up, so you can see how much my wife supports me (…) She looks after me like a boy, like a child, I mean to say she’s supporting me 100% and she’s always with me. (OP1, male, living with partner, 72 years)

Well, my husband [referring to closest support network], I don’t see my children every day (…). With my husband, we’re really close, he’s had an operation too and we’ve both got our ailments but right now we support and help each other (…) Like I’m telling you, as I wasn’t well, my husband even put my socks on. (OP12, female, living with partner, 76 years)‬‬‬‬‬

In the marriage, we both try not to bother others if we can do it ourselves. So if it’s been necessary to go to the doctor, we go, if we have to go by metro or take a taxi we go to the doctor, I mean to a consultant or to the clinic, we always go on our own, there’s no need because we can look after ourselves fine for the time being, thank God (…) But there are day-to-day things that we don’t bother about. (OP18, male, living with partner, 81 years)

The association pattern between social support and loneliness is significantly more complex in the case of sons and daughters. Support from the latter was related to lower scores for emotional loneliness (β = -.047; p < .05), although there was a smaller association in comparison with other sources (spouse, siblings, grandchildren and neighbours). It is worth emphasising that there is no significant association between emotional loneliness and support from sons. But there was a positive association between social loneliness and both sons (β = .753; p < .001) and daughters (β = .940; p < .001), such that higher scores for the support measure were associated with higher scores for social loneliness. The material collected in the interviews offered highly useful context to analyse the complexity of the relationship between emotional and social loneliness, on one hand, and social support from sons and daughters, on the other. In fact, the interviews revealed a clear distinction between the two. Specifically, the interviewees described daughters as a key source of emotional support in their lives, more easily creating a relationship of attachment based on communication, intimacy and mutual care. However, this relationship may be affected by difficulties maintaining frequent social contact:

My daughter, my daughter is by far the most important person or thing in my life (…) we talk a lot in our relationship, yes. We’re lucky to live fairly close by but she’s really busy with work too (…) so she’s quite pushed for time, and what we do is talk on the phone for the hour it takes her to go to work, and that’s an hour every day, whether or not we see each other, so in that hour you can imagine, we talk about everything that has happened to us during the day, everything we’ve seen, everything we’ve been told, everything under the sun, we talk about ourselves and everything around us, so there really is lots of communication in our relationship. Regardless of when we need to be there for everything (…) it’s a bond we have and I hope that all mothers and daughters are like that. (OP5, female, living alone, 67 years)

Sons and daughters were frequently compared in the interviews. It was a significant theme for the participants:

Yes, whenever I’ve needed it, because I’ve already needed them to take care of me, it’s been my daughter who’s been there unconditionally, and two friends of mine who’ve even brought me food, they’ve gone with me to the doctor, and all that. If I’ve occasionally asked my son to do something or other, he hasn’t refused, but he hasn’t volunteered or checked if I’ve needed help. (OP4, female, living with partner, 73 years)

More specifically, the interviews indicated a tendency for sons to specialise in providing instrumental support, compatible with an emotional distancing that helps provide context for the data in Figs 1 and 2. In this regard, we found descriptions of support from sons as being strictly instrumental and task-focused, with far less emphasis on emotions:

My son, I can’t do anything without him, I don’t know anything. He’s the one who goes to the bank, the doctor, he does everything, he’s everything to me; I’m nothing, nothing without him. (OP10, female, living alone, 91 years)

My son has a more independent nature. So as he’s got older I think that he loves me, because he shows me, but he doesn’t need me. He knows I’m there even though he doesn’t need me. So the time he can give me or find out how I’m doing or whatever, he might prefer to chat with a friend, who thinks more like he does, or maybe he has more of a laugh with them than me, I don’t know. Because he does love me, he’s shown me, but I don’t think he feels the need for the closeness that my daughter wants with me. (OP5, female, living alone, 67 years)

In any event, it is worth noting that social support from sons and daughters significantly correlates with social support from grandchildren. This is significant: this support was one of the most important sources both in the SEM analyses and in the analysis of the in-depth interviews. In fact, perceived support from grandchildren is related to lower scores for emotional loneliness (β = -.145; p < .001), but the intensity of the association is particularly significant in the case of social loneliness, owing to its magnitude (β = -3.773; p < .001). This latter case represents the most intense association in the model set out in Fig 3, showing the importance of support from grandchildren in the social loneliness experienced by older adults. The analysis of the in-depth interviews contributes to describing this process, which is largely modulated by the age of the grandchildren. The emotional closeness of the relationship is the most significant element when the grandchildren are younger, but interaction with grandchildren generally creates specific wellbeing that is largely related to the maintenance of significant social relations:

(…) my grandchildren are indispensable, it really is a totally different kind of love, now I’m the grandfather, now they adore you. One’s eight, another is six. I give them four years, probably they won’t come and see me once they’re 12, but now they come to grandad’s house and they have a great time. (OP9, male, living alone, 68 years)

My grandchildren have gone on holidays, one of them because another is studying in the US on a grant, we’re not rich. And we call each other, I call them, they call me, they don’t come and see me as much, I’m always missing them like always. Although they’re women, one is 19 and the other 17, they’re doing really well with their studies, they get great marks and that keeps me happier too, it keeps me going. (OP8, male, nursing home, 78 years)

Family sources of support can be observed to be basic elements in understanding processes related to loneliness. As we have seen, the interviewees tend to locate their social ties involving emotional closeness and social interaction within the family setting, with widespread recognition that the family is a fundamental source of support. This is the context for the association between support from siblings and emotional loneliness (β = -.111; p < .001) and social loneliness (β = -.058; p < .01). Both cases involve negative regression coefficients. This is reinforced by the correlation with support from friends (a particularly important source of support for older adults, as discussed in the “Non-family sources of support: friends and neighbours” subsection). The importance of sibling support in reducing loneliness and the functional equivalence of support from siblings and friends are clearly described in the qualitative context:

I think so, having the sense of family with friends, siblings, nephews, who you can call at any time and say “hey, this has happened, it’s just that this has happened to me” and they say “fine, don’t worry, I’m coming”. That’s essential for me. You feel a bit sheltered, and if you need anything you know they’ll come. They might not come as quickly as you want, but they’ll come. (OP18, male, living with partner, 81 years).

My siblings, I’ve got siblings I’m very close to (…) we don’t call each other much, but I know I’ve got their love and friendship. My siblings have been really crucial for me (…). (OP26, female, living alone, 78 years)

Non-family sources of support: Friends and neighbours.

The results of our study show a significant negative association between support from friends, on one hand, and the emotional (β = -.397; p < .001) and social (β = -.265; p < .001) dimensions of loneliness, on the other. This association is more significant in the case of emotional loneliness. Friendships are a key category in the in-depth interviews. In the qualitative material, the support generated in friendships is highly important in emotional terms, since it is linked to the existence of close bonds of trust:

I don’t know, at this time of life you can get low for any reason, I don’t know, you can have slightly depressed moments (…) and that feeling of loneliness, which I haven’t had, that’s bad. It’s one thing for me to be alone because I want to, which I normally look for (…) I value the emotional support more than money, and more than anything knowing that you have someone there who’s bothered about you, I don’t know, that’s something that I do value. Because I’ve done it, with friends, and knowing that they’re there, it’s what I was telling you, even if you don’t see them, even then… but knowing they’re there, that really gives you strength. (OP9, male, living alone, 68 years)

A particularly high number of interviewees (20 people, see Fig 2) state that friends are networks for emotional support, available on a day-to-day basis. This implies two interrelated elements that stand out in the analysis of the interviews. First, friendships create interactions with positive effects for both emotional loneliness and social loneliness. Second, the association very frequently takes the form of a day-to-day relationship:

I see my friends almost every day. We’re a group of friends who all live here in the same street, all of us are older now, and we meet up for coffee, then we go shopping. We do it every day, with or without aches and pains, some not so good, others well, but we get out (…) we tell each other our news and then we all go home, just that bit of time in the morning, we don’t go out later in the afternoon. (OP12, female, living with partner, 76 years)

The huge importance of the intimacy and trust that characterises the friendships described in the interviews means that when necessary, support from friends is functionally equivalent to the support received from certain family relationships:

My sister hardly gives me any support (…) but right now I get lots more support from my friends and my nursing home companion than the others, than my family, it’s as clear as that. I haven’t been a very family-focused person either, I’ve always had a bit of a rootless life. (OP3, male, nursing home, 67 years)

In our study, support from friends is the only non-family support with a significant role in reducing loneliness. In this regard, a significant association was not found between support from neighbours and social loneliness (Fig 4), and there was a positive association in the case of emotional loneliness (Fig 3), with higher scores for support related to higher scores for loneliness (β = 5.745; p < .001). Although these findings will be examined in the discussion section, it is worth noting that in the discourses collected in the in-depth interviews, relations with neighbours only emerge as a significant factor in reducing loneliness when they converge into friendship with those living in the same neighbourhood. This involves spending time together, supporting each other and developing a space for emotional cooperation and closeness, which forms the basis for constructing patterns of frequent social interaction:

Well, there are a few neighbours, a young woman who lives opposite my apartment, who always offers me help. I don’t have many friends here in Madrid, but I do have two, three, four or five. We have a group too, and I think if someone needs any help they can get it there. Meanwhile we’re always in touch, mostly through the chat group, through WhatsApp (…) (OP14, female, living with partner, 81 years)

I’ve got a friend who’s 86, but I don’t know. She lives on this street, just at the end, on the same side and everything, and I get on so well with her, I can talk about more with her than with others my own age, you know? I don’t know, because I can talk about everything, we go out for lunch, sometimes we go to the cinema (…) I’m just like a young kid when I’m out with her. I don’t know, she’s got a very young spirit, she lived in France for a long time and I don’t know, she gets everything, we understand each other very well (OP16, female, living alone, 65 years)