Older Adults’ Social Relationships and Health Care Utilization: A Systematic Review

Abstract

Background. Deficiencies in older people’s social relationships (including loneliness, social isolation, and low social support) have been implicated as a cause of premature mortality and increased morbidity. Whether they affect service use is unclear.

Objectives. To determine whether social relationships are associated with older adults’ use of health services, independently of health-related needs.

Search Methods. We searched 8 electronic databases (MEDLINE, Embase, CINAHL, Web of Science, PsycINFO, Scopus, the Cochrane Library, and the Centre for Reviews and Dissemination) for data published between 1983 and 2016. We also identified relevant sources from scanning the reference lists of included studies and review articles, contacting authors to identify additional studies, and searching the tables of contents of key journals.

Selection Criteria. Studies met inclusion criteria if more than 50% of participants were older than 60 years or mean age was older than 60 years; they included a measure of social networks, received social support, or perceived support; and they reported quantitative data on the association between social relationships and older adults’ health service utilization.

Data Collection and Analysis. Two researchers independently screened studies for inclusion. They extracted data and appraised study quality by using standardized forms. In a narrative synthesis, we grouped the studies according to the outcome of interest (physician visits, hospital admissions, hospital readmissions, emergency department use, hospital length of stay, utilization of home- and community-based services, contact with general health services, and mental health service use) and the domain of social relationships covered (social networks, received social support, or perceived support). For each service type and social relationship domain, we assessed the strength of the evidence across studies according to the quantity and quality of studies and consistency of findings.

Main Results. The literature search retrieved 26 077 citations, 126 of which met inclusion criteria. Data were reported across 226 678 participants from 19 countries. We identified strong evidence of an association between weaker social relationships and increased rates of readmission to hospital (75% of high-quality studies reported evidence of an association in the same direction). In evidence of moderate strength, according to 2 high-quality and 3 medium-quality studies, smaller social networks were associated with longer hospital stays. When we considered received and perceived social support separately, they were not linked to health care use. Overall, the evidence did not indicate that older patients with weaker social relationships place greater demands on ambulatory care (including physician visits and community- or home-based services) than warranted by their needs.

Authors’ Conclusions. Current evidence does not support the view that, independently of health status, older patients with lower levels of social support place greater demands on ambulatory care. Future research on social relationships would benefit from a consensus on clinically relevant concepts to measure.

Public Health Implications. Our findings are important for public health because they challenge the notion that lonely older adults are a burden on all health and social care services. In high-income countries, interventions aimed at reducing social isolation and loneliness are promoted as a means of preventing inappropriate service use. Our review cautions against assuming that reductions in care utilization can be achieved by intervening to strengthen social relationships.

PLAIN-LANGUAGE SUMMARY

We looked at published research on the link between older adults’ use of health services and their social relationships (including loneliness, isolation, and social support). We searched 8 electronic databases for studies published between 1983 and 2016. We identified more than 26 000 articles. We included 126 studies in our review, providing information on 226 678 people from 19 countries. We found strong evidence of a link between weaker social relationships and readmissions to hospital but no evidence on the use of community services or emergency departments. Evidence of moderate strength linked larger social networks to shorter hospital stays. We conclude that current evidence does not support the idea that older patients with low levels of social support use more ambulatory care than they need. Our study suggests that campaigns to reduce isolation or loneliness may not reduce utilization of ambulatory care services, though strengthening social support could shorten hospital stays and prevent some readmissions.

Over the past decade, the quantity and quality of social relationships in later life have become a major societal concern. Across Europe and North America, national campaigns have been set up to “end loneliness in older age” (Coalitie Erbij, Netherlands; Campaign to End Loneliness, United Kingdom), “mobilize society against isolation in the elderly” (MONALISA, France), strengthen older adults’ social connections (“Connect2affect,” United States), and “Reach Isolated Seniors Everywhere” (RISE, Canada). One of the drivers of these campaigns is the growing body of research associating social relationships with mortality and morbidity: recent meta-analyses have shown that individuals with weaker social ties are on average 30% more likely to die early or to develop cardiovascular disease.^1,2

The implications of older adults’ relationships for health care utilization are unclear. It is often claimed that dissatisfaction with the quality and quantity of one’s relationships leads to inappropriate consultations with health professionals, and that interventions aimed at strengthening social relationships could help to prevent avoidable health care utilization (e.g., The Campaign to End Loneliness, 2015).³ Older adults who lack social relationships and feel isolated report lower levels of self-rated physical health,⁴ which can prompt them to consult a physician more frequently than socially connected individuals.⁵ Seeking medical assistance might also be a way for them to satisfy an unmet need for interpersonal interaction and stimulation.⁶ Conversely, it may be that older adults who have limited access to social relationships and the support they provide (e.g., transportation, financial aid, health-related advice)⁷ place fewer demands on health services; if so, their infrequent contacts would be a precious opportunity for health monitoring or preventive care.⁸

To date, individual studies have reported conflicting findings on the association between social relationships and family doctor visits, hospital admissions, or emergency department attendances.^9–13 Some of this variation may reflect the range of ways in which social relationships have been operationalized in the literature, from subjective appraisals of relationship quality to more objective measures of social contact.¹⁴ In their behavioral model of health care utilization, Andersen et al. suggest that both subjective aspects, such as the perceived availability of support, and objective characteristics such as social network composition and size, are important for accessing services.¹⁵ This model does not, however, indicate which of these aspects of relationships might be more influential; nor does it specify whether they are barriers to, or facilitators of, care utilization. To determine the nature and direction of effect between different dimensions of social relationships and older adults’ use of health services, we set out to systematically review the empirical evidence.

METHODS

Our study followed the Centre for Reviews and Dissemination’s Guidance for undertaking reviews in health care.¹⁶ We registered a protocol with the International Prospective Register of Systematic Reviews (registration number: CRD42016045682).

Search Strategy

Using a combination of index and free-text terms, we searched 8 databases (MEDLINE, Embase, CINAHL, Web of Science, PsycINFO, Scopus, the Cochrane Library, and the Centre for Reviews and Dissemination) for literature published up until July 2016 (see Appendix A, available as a supplement to the online version of this article at http://www.ajph.org, for an example of the search strategy used in MEDLINE). We also searched the reference lists of all included studies and review articles, contacted authors to identify additional studies, and scanned tables of contents of key journals for additional reports of relevance.

Selection Criteria

Researchers commonly distinguish between 3 major dimensions of social relationships: (1) the structure of a person’s social network (e.g., frequency of contact with family and friends), (2) the objective availability of social support (e.g., whether someone receives emotional, informational, tangible, or belonging support), and (3) subjective perceptions of social support (e.g., the perceived adequacy of the quantity or quality of one’s relationships, or feelings of loneliness).¹⁷ The evidence to date does not suggest that one dimension may be more problematic for service use than the others; we therefore included studies if they investigated any one of these dimensions and reported quantitative data on health service utilization. Our aim was to study the implications of both the quantity and the quality of social relationships; because marital status, living arrangement, or number of relatives provide limited information on relationship quality or quantity (e.g., people may be married or parents but not see their spouse or child), these measures qualified for inclusion only when combined with an assessment of contact frequency or appraisal of relationship quality.

Because the focus for this review was on older people, we excluded studies in which less than 50% of the sample was older than 60 years or in which mean age was younger than 60 years. To ensure that our findings were relevant to the current context of service delivery and disease burden in high-income settings, we excluded work conducted in low- and middle-income countries as defined by the World Bank and we restricted our search to studies published in the past 25 years. We excluded reports written in languages other than those read by the research team (English, French, Spanish, or Italian) because of resource constraints.

Title and Abstract Review, Data Extraction

Two researchers independently checked the titles and abstracts of the articles identified through our search. They then extracted data from the studies that met inclusion criteria into a standardized form that captured study design, data collection methods, participant characteristics, predictor and outcome variables, methods of statistical analysis, and findings relevant to the review. We extracted results of statistical tests (95% confidence intervals for relative risk, odds or hazard ratios, and P values) where reported; when data were missing or unclear, we contacted authors.

Quality Assessment

We appraised the quality of individual studies by using the standardized scale developed by Gomes and Higginson.¹⁸ We graded studies as high quality if they had used multivariate analysis and scored 70% or higher; medium quality if they included multivariate analysis but scored less than 70%, or did not include multivariate analysis but scored at least 60%; and poor quality if no multivariate analysis was presented and it scored less than 60%. Three reviewers assessed the studies and disagreements were resolved by discussion.

Data Synthesis

We did not attempt meta-analysis because of heterogeneity in study design, population, social relationship measure, and outcome assessment (see Appendix B, available as a supplement to the online version of this article at http://www.ajph.org, for details of each study). Instead, we conducted a narrative synthesis according to the stages outlined in the Centre for Reviews and Dissemination guidance: preliminary synthesis, exploration of relationships within and between studies, and robustness assessment.¹⁶ After grouping studies according to the outcomes and dimensions of social relationships on which they reported, we systematically assessed the strength of the evidence for each type of service and social relationship domain by using the algorithm presented in Figure 1. We then compared the strength of the evidence across different measures of relationships for each service type and across service types. When writing up our findings, we gave more weight and space to studies with a lower risk of bias; results from medium- and low-quality studies are discussed only when they provided additional insights into the literature.

FIGURE 1—

Algorithm Used to Grade Strength of Evidence

Note. Example of how we applied this to the evidence on perceived social support and family physician visits: we identified 4 studies, 3 of high quality and 1 of medium quality. Less than 70% of high-quality studies (2 out of 3) but more than 50% of high- and medium-quality studies (3 out of 4) reported no evidence of effect; we therefore rated the evidence as being of moderate strength.

Source. Adapted from Gomes and Higginson.¹⁸

RESULTS

Of the 26 074 citations identified via our electronic search, 123 references met our inclusion criteria; we found an additional 3 articles through hand searching and contacting authors (see Figure 2 for a flow diagram of the study selection process). These 126 articles were based on 104 data sources and reported information from 226 678 participants across 19 different high-income countries, primarily from the United States (59 studies) and Canada (18 studies). Sample sizes ranged from 38 to 100 894 and more than a third of articles (39.6%) contained secondary analyses of previously collected data. Study designs included 1 randomized controlled trial,¹⁹ 38 (30.2%) longitudinal cohort studies with follow-up times ranging from 14 days to 6 years,^{10–12,20–54} 76 (60.3%) cross-sectional studies,^{9,13,55–128} and 11 (8.7%) case–control studies.^129–139

FIGURE 2—

PRISMA Flowchart of Study Selection for Systematic Review of Social Relationships and Health Care Service Use by Older Adults

The main outcomes investigated were physician visits (26 studies), hospital admissions (26 studies), hospital readmissions (15 studies), emergency department visits (10 studies), hospital length of stay (15 studies), and use of home- and community-based services such as home health aides, visiting nurses, congregate meals, Meals on Wheels, or transportation (53 studies). Social network was the most commonly measured aspect of social relationships (59 studies); 27 studies assessed the objective availability of supportive relationships, 47 studies included a measure of perceptions of relationships, and 29 studies used a multidimensional measure (i.e., a measure that combined 2 or more of these 3 dimensions). Quality scores varied widely, ranging from 31% to 90%; we classed 61 studies as high quality, 53 as medium, and 12 as low. The strength of the evidence for each outcome and measure of social relationships is summarized in Table 1.

TABLE 1—

Social Relationships and Health Care Service Use by Older Adults: Strength of, and Message From, the Evidence From a Systematic Review of the Literature: 1983 to 2016

		High-Strength Evidence
Dependent Variable	Measure of Social Relationships	Relationship With Service Use	Consistencya	No. of Participants	Moderate-Strength Evidenceb: Relationship With Service Use
Family physician visits	Social network				No evidence of association9,12,78
	Perceived support				No evidence of association12,32,78
Physician visits (across specialties)	Combined measure				No evidence of association66–68, 81,113
Hospital admission	Social network				No evidence of association10,20,25,27,29,103,105,106,108,125,127
	Received support				No evidence of association29,56
	Perceived support				No evidence of association13,25,105,108,110
	Combined				No evidence of association42,45,67,68,81
Readmission to hospital	Perceived support				No evidence of association26,31,33
	Combined	Weaker social relationships were associated with greater likelihood of readmission34,46,48	75 (3/4)	1 176
Length of hospital stay	Social network				Smaller social networks were associated with spending more days in hospital19,98,106,108,127
	Perceived support				No evidence of association33,108,110
	Combined				No effect23,69,97
Emergency department visit	Social network	No evidence of association11,77,125	75 (3/4)	3754
	Perceived support				No evidence of association11,77,104
General home- and community-based services	Social network	No evidence of association37,52,54,82,83,92,124,127	89 (8/9)	10 029c
	Received support				No evidence of association50,51,72,82,83,92,94,107,121,129
	Perceived support	No evidence of association21,37,82,110,120	83 (5/6)	4 049
Senior and day center use	Perceived support				No evidence of association59,62
General health service use	Social network				No evidence of association52,112

^aPercentage of high-quality studies reporting an effect in the same direction (number of high-quality studies reporting the same effect or number of high-quality studies included).

^bDomains for which the evidence was of low strength (e.g., received support and family physician visits), either because there were fewer than 3 medium-quality studies on the topic or because fewer than 50% of medium- or high-quality studies agreed, are not included in this table.

^cTwo studies used data from the 1988 National Survey of Hispanic Elderly People; when totaling number of participants, we only entered the biggest sample number (Tran¹²⁴) to avoid counting individuals multiple times.

Physician Contact

Contacts with family physicians were the subject of 5 studies, and visits to physicians in general (i.e., where specialty was not specified) were analyzed in 21 studies.

Contacts with family physicians.

Of the 5 studies that examined the association between social relationships and visits to family physicians, 3 were of high and 2 of medium quality. The evidence on social networks and perceived support was mixed, though in both cases more than 50% of medium- and high-quality studies reported no evidence of association. In 1 high-quality study, loneliness (i.e., the negative feeling associated with people judging that the quantity or quality of relationships is inadequate) was weakly but significantly associated with general practitioner contact in Denmark, whereas having access to a social network was not.⁹ Conversely, another high-quality study found no association between the perceived availability of help or close relationships and number of visits to a general practitioner in Australia, but reported that men with larger social networks were more likely to have at least some contact with their general practitioner.¹² In the third high-quality study, perceived availability of support was unrelated to service use among Dutch patients with Parkinson’s disease.³² Social network, meanwhile, did not explain variation among frequent users of family physician services in Canada.⁷⁸ We identified no study on received support and only 1 medium-quality study with a multidimensional measure of relationships.

Contact with all physicians.

Nine high-quality, 10 medium-quality, and 3 low-quality studies reported data on visits to physicians in general. Overall, they did not clearly identify social relationships as either enabling or preventing service use. Two high-quality studies on perceived support produced mixed results.^35,38 A medium-quality study found no evidence of association once potential confounders had been adjusted for, and another suggested that loneliness may be associated with greater physician utilization among older US women.^63,108 Evidence for the effect of received support and access to social networks was similarly mixed.^{27,29,53,56,61,101,106,108,125,127} When assessed with a multidimensional measure, social relationships did not tend to predict physician use.^66,68,81,113

Hospital Use

Hospital admission.

Twenty-six studies considered the influence of social relationships on hospital admissions. Fifteen of these were of high quality, 9 were scored as medium, and 1 as low quality. A majority found no evidence for an association between social relationships and admission to hospital. Eleven out of 16 studies with a measure of social network—5 of which used data from the American Longitudinal Study of Aging—reported that contact with friends or relatives was not significantly associated with hospitalization.^{10,20,25,27,29,103,105,106,108,125,127} The remaining studies identified limited social contact as a predictor of hospital admission,^{13,96,134,136} with a further study suggesting that this was only the case among recently widowed women.³⁹ Although Canadians who received more assistance with activities of daily living were more likely to be hospitalized,¹¹⁰ there was no evidence of association between received support and hospitalization among US community dwellers.^29,56 Evidence on perceived support was mixed, with 5 studies reporting no association,^{13,25,105,108,110} 2 suggesting that feeling more supported was associated with lesser likelihood of admission,^35,103 and 2 reporting mixed results.^10,30 When relationships were assessed with a multidimensional measure, they did not predict hospital admission.^42,45,66,81

Hospital readmission.

Readmission to hospital was assessed in 11 high- and 4 medium-quality studies. Having a smaller social network was associated with increased risk of readmission in 2 high-quality studies,^49,96 though secondary analysis of data from American Longitudinal Study of Aging found no association.¹²⁷ Having an informal carer predicted greater likelihood of rehospitalization in 1 high-quality study,²⁴ but a second found no evidence of association.²⁸ Although perceived support was identified as a risk factor for readmission in 2 medium-quality studies,^43,108 evidence from high-quality studies did not support this.^31,33 Three out of the 4 high-quality studies that used a multidimensional measure found that weaker social relationships were associated with greater likelihood of readmission.^34,46,48

Length of time spent in hospital.

Our search retrieved 15 studies with data on social relationships and hospital length of stay. Six were of high quality, 8 were of medium quality, and 1 scored as low quality. The 5 studies with a measure of social network all reported findings that showed an inverse relationship between social network size and length of time spent in hospital, though in the case of the 2 studies that used American Longitudinal Study of Aging data, it was contact with nonkin, rather than family members, that made a difference.^{19,98,106,108,127} Whereas findings from 1 high- and 1 medium-quality study suggested that lower perceived support was associated with more days spent in hospital for cardiac patients in North America⁷⁰ and chronic heart failure patients in Sweden,⁴³ 3 other studies found no relationship with length of hospital stay.^33,108,110 We identified moderate-strength evidence suggesting that social relationships measured with a multidimensional tool were not significantly associated with time spent in hospital, on the basis of null findings from 3 medium-quality studies set in Australia, Spain, and Denmark.^23,69,97

Emergency department use.

Of the 10 studies that considered emergency department attendance in their analyses, 5 were of high quality, 4 were of medium quality, and 1 was of low quality. In 2 high-quality cross-sectional studies from Israel and 1 prospective cohort from the United States, amount of social interaction did not affect emergency department usage.^11,77,125 In a fourth study from Ireland, having a smaller social network was associated with greater emergency department utilization, as was loneliness—but not the perceived availability of support.¹³ The other studies on perceived support did not report evidence of an association.^11,77,104 High-quality studies found no association between emergency department visits and receipt of support or social relationships as measured with a multidimensional tool.^11,75

Other Services

Home- and community-based service use.

Forty-four studies examined the association between social relationships and services provided in the home or community (as opposed to within an institution or other isolated setting). There were 19 high-quality studies, 21 medium-quality studies, and 4 low-quality studies. When social network and received or perceived support were measured separately, most studies reported no evidence of association: of the 9 high-quality studies on social network, 8 found no association with service utilization,^{37,52,54,82,83,92,124,127} and perceived support did not predict service use in 5 out of 6 high-quality studies.^{21,37,82,110,120} Seven high- and medium-quality studies found no association between received support and service use,^{21,76,82,120,129,131,137} with a further 3 reporting null results for specific services (rehabilitative care), populations (people living alone), or type of support (help with activities of daily living).^50,83,92 In 3 of the 8 studies that used a multidimensional measure of relationships, stronger social ties were associated with greater service use,^75,88,126 but the remainder of the evidence was mixed.^{32,65,86,99,120}

Senior center and day center use.

Ten studies—5 of high, 3 of medium, and 2 of low quality—presented evidence on the association between social relationships and senior or day center use. Two of the 4 high- and medium-quality studies with a measure of social networks found that older adults who frequently interacted with family and friends were more likely to attend senior centers,^62,87 whereas 1 reported mixed findings⁸³ and 1 no effect.⁸⁰ High-quality studies identified no relationship between lower received⁸³ or perceived^59,62 social support and center use. Only 1 study, of high quality, found evidence of a positive association between a multidimensional measure of social relationships and center use.⁹⁰

General health service use.

Eight high-quality and 2 medium-quality studies reported data on multiple health service use. Social network size was unrelated to service utilization in 2 high-quality studies,^52,112 with a further 2 reports suggesting variation according to whether networks involved kin or nonkin members.^114,139 The 1 study on received support found no evidence of association with health care utilization.¹¹⁰ Although 1 high-quality study from the United States found that the perceived availability of support was associated with lesser likelihood of medical service use,²¹ another reported no evidence of association.¹¹⁰ A third study from Canada found that among people seeking help for psychological distress, community dwellers who reported minimal social support used health services less.¹¹¹ The evidence from studies with a multidimensional measure of relationships was also mixed, with 1 high-quality study reporting an inverse association with health service utilization in Canada⁷⁵ and another from Israel identifying participants with a diversified network as making greatest and most recent use of services.⁹⁵

Mental health service use.

All 5 studies on social relationships and mental health service use were of medium quality. Two studies reported that having access to social relationships (received support or relationships as measured by a multidimensional tool) was associated with lesser use of services,^32,132 and 3 studies reported no association with service utilization.^57,115,133

Medication, preventive services, cancer support, ambulance, and hospital social work services.

Small numbers of studies with different measures of relationships on medication use (2 studies^58,81), preventive services (1 study⁸⁴), cancer support services (2 studies^71,74), ambulance use (1 study⁴¹), and hospital social work services (1 study⁵⁵) did not allow us to gauge the effect of social relationships on these outcomes in this review.

DISCUSSION

We found strong evidence of an association between social relationships measured by using a multidimensional measure and the likelihood of early hospital readmission (75% of high-quality studies reported evidence of an association in the same direction). In evidence of moderate strength, on the basis of 2 high-quality and 3 medium-quality studies, smaller social networks were associated with longer hospital stays. When received and perceived social support were measured separately, they were not linked to health care use. Overall, the evidence did not indicate that older patients with weaker social relationships place greater demands on ambulatory care (including physician visits and community- or home-based services) than warranted by their needs.

Of the 126 articles we reviewed, 91 adjusted for health status (e.g., physical or mental health diagnosis or functional health) in multivariate analyses, 16 focused on a sample with similar health-related needs (e.g., a group of stroke patients or a population of frail older adults), and 1 was a randomized controlled trial. Our review does not therefore invalidate the message from the literature that links social relationships to worse physical and mental health outcomes,^1,2,140,141 potentially increasing the burden on services. Rather, it suggests that, independently of health-related needs and with the exception of hospital readmission, there is limited evidence that social networks and support are associated with variation in health and social care utilization.

The epidemiological literature on morbidity and mortality suggests that relationships may play a greater role in shaping people’s experiences once they are ill, rather than in disease etiology.¹⁴² Stronger social networks and support are associated with higher levels of patient adherence to medical treatment,¹⁴³ and individuals who perceive their relationships to be unsatisfactory are less likely to use active coping methods.¹⁴⁴ The fact that we found comparatively consistent evidence linking social relationships to hospital readmission and length of stay (i.e., measures of volume of contact, which imply that participants are already using or have already used the service) may be a further indication of the importance of relationships for prognosis. This interpretation is consistent with the stress-buffering model of relationships and health-related outcomes, according to which relationships primarily affect outcomes among people who are under stress.⁷ This could explain why a majority of studies we reviewed found no evidence of association with service use, as these studies did not stratify their analyses according to health status. Future work is needed to investigate the potentially interactive effects of health and social relationships on care usage and to test whether the stress-buffering model might be a useful framework for conceptualizing the link between social relationships and service use.

Strengths and Limitations

To our knowledge, this is the first systematic review of the evidence on social relationships and older people’s use of health services. Our review included all settings across the care system that were studied in the research literature and was not shaped by the authors’ previous assumptions. Most of the included studies were observational, and there was considerable heterogeneity in the study populations, measurement of social relationships, and outcomes. In common with other reviews of observational studies, confounding by unmeasured common causes cannot be excluded; nor can the possibility of reverse association be eliminated, particularly in the case of senior center and day center activities, which provide older adults with the opportunity to create new relationships and reduce isolation. Observational studies were the most appropriate design for examining the association between social relationships and service utilization, but combining such data risks magnifying any bias in the individual studies. We therefore made no attempt to pool the data. We took care to select a tool for quality appraisal that was appropriate for observational studies and assessed both study design and reporting quality. We differentiated between admissions to hospital and readmissions within 30 days. This is an important measure for health services performance, but relies on accurate identification of an index admission. If readmissions were misclassified as index admissions, it is possible that we may have underestimated the impact of social relationships on readmissions.

Previous reviews of the evidence on social relationships and health have highlighted variation in effect sizes according to the social relationship domain studied. For example, in their meta-analysis of studies on mortality, Holt-Lunstad et al. found a stronger relationship with composite measures of social integration compared with binary measures of living alone.¹⁷ Although our findings also varied with the measure of social relationships (Table 1), no clear pattern emerged to suggest whether perceptions of social support, or more objective characteristics such as social network size, might be more important for service use. Aside from the studies that employed a multidimensional tool, only 5 articles included a measure of each of the 3 dimensions of relationships covered in this review^{11,46,76,82,129}; the heterogeneity of tools they used, and the range of outcomes they assessed, may explain their conflicting findings.

Public Health Implications

Our findings are important for public health because they challenge the claim commonly voiced in the media and by campaigners and policymakers that lonely older adults are a burden on all health and social care services. In high-income countries, interventions aimed at improving the quality and quantity of older adults’ social relationships are currently being promoted as a means of preventing inappropriate service use. Our review cautions policymakers, practitioners, and researchers against assuming that reductions in care utilization can be achieved by intervening to strengthen social relationships.

Future research on social relationships would benefit from a consensus on clinically relevant concepts to measure. At present, because of the fragmented nature of the evidence, it is not clear that social relationships, independently of their implications for health and well-being, affect the service use of older people. A commitment to taking social relationships into account when one is studying service use and agreement among researchers and stakeholders on what is important to measure will enable us to build a body of work that is credible and answers important questions for designing future preventive interventions.