Social relationships and physician utilization among older adults—A systematic review

Background In older age health needs and demand for health services utilization increase. Individual’s social relationships can play a decisive role regarding the utilization of outpatient health care services. This systematic review examines the associations of structural and functional dimensions of social relationships with outpatient health services use of older adults. Methods The databases PubMed, CINAHL, SocINDEX, PsycINFO, International Bibliography of the Social Sciences (IBSS), Sociological Abstracts, and Applied Social Sciences Index and Abstracts (ASSIA) were searched in February 2016. The methodological and reporting quality of the articles was assessed and the results were synthesized descriptively and systematically. Results Out of 1.392 hits, 36 articles (35 studies) were included in the systematic review. The methodological and reporting quality of the included articles was reasonable. Various structural and functional characteristics of social relationships were associated with the use (yes/no) and the frequency of using outpatient care among older adults. The majority of the associations between structural dimensions of social relationships and the use of physicians were positive and moderate in strength. The associations between functional dimensions of social relationships and the probability of using physician services were inconsistent and varied in strength. For the most part, social relationship variables assigned to the structural dimension were positively and weakly to moderately associated with the frequency of physician visits. Functional aspects of social relationships also tended to have positive associations with the frequency of physician utilization. The associations were weak to moderate in strength. Conclusions Measuring social relationships and their influence on health services use is a challenging methodological endeavor indicated by the inconclusive results. The results suggest that the outpatient care utilization behavior of older individuals being structurally and functionally integrated in social relationships is different to older adults being socially isolated or having no social support. All in all, the current status of quantitative data was insufficient. Future health services research should accentuate social ties in more detail, especially according to quality aspects of social relationships.


Results
Out of 1.392 hits, 36 articles (35 studies) were included in the systematic review. The methodological and reporting quality of the included articles was reasonable. Various structural and functional characteristics of social relationships were associated with the use (yes/no) and the frequency of using outpatient care among older adults. The majority of the associations between structural dimensions of social relationships and the use of physicians were positive and moderate in strength. The associations between functional dimensions of social relationships and the probability of using physician services were inconsistent and varied in strength. For the most part, social relationship variables assigned to the structural dimension were positively and weakly to moderately associated with the frequency of physician visits. Functional aspects of social relationships also tended to have positive associations with the frequency of physician utilization. The associations were weak to moderate in strength. PLOS  utilization among older adults associated with various dimensions of social relationships. The second aim was to evaluate magnitude and consistency of the associations between social ties and health services use.

Method
A systematic review on studies dealing with social relationships and the utilization of outpatient care physicians among older adults was conducted. The performance of this review was based on the PRISMA checklist [19] and a study protocol including all preliminary specifications published on PROSPERO, registration number CRD42016036004 (S1 File, S1 Table).

Search strategy and inclusion criteria
After developing the research question and performing a pilot run of literature search, seven databases were used (February 11 th 2016). The databases PubMed, CINAHL, SocINDEX, Psy-cINFO, International Bibliography of the Social Sciences, Sociological Abstracts, and Applied Social Sciences Index and Abstracts were searched for the keywords and various synonyms "social relationships", "utilization", "outpatient care" and "aged" in title and abstract (S1 Text). MeSH-terms and limiters were adapted to each electronic database. In addition, references of relevant articles were searched for further matching studies. At first, one reviewer (DB) screened the titles and abstracts of all articles identified by electronic and reference search. In a second step, two independent reviewers (DB and LI) applied a predefined set of inclusion criteria on all relevant articles by performing a full text screening. In case of disagreement between the reviewers, a third investigator (OK) was consulted and the study was discussed until consensus was accomplished.
Within the full text screening, articles had to pass five predefined inclusion criteria. Firstly, records were controlled for the criterion "peer-reviewed journal articles in German or English". Peer-reviewed journal articles represent good scientific practice to secure quality, to foster objectivity and to provide transparency. Due to language skills and a reasonable use of resources of the reviewers, German and English articles were screened. Secondly, records were checked for three different study designs: quantitative observational 1) cross-sectional, 2) casecontrol and 3) cohort studies. Thirdly, full texts were inspected for the criterion "communitydwelling or noninstitutionalized individuals fifty years and older". The rationale behind this population was to extract a reference group still active on the labor market, and to expand the number of potentially relevant studies. Compared with younger cohorts, individuals within their fifties or older show more chronic illnesses and increased rates of health care use [2]. The fourth inclusion criterion was the accounting for utilization or frequency of use of outpatient care services as the dependent or outcome variable. These measures of use are solidly established in health services research and increase the chance of comparability. Finally, studies had to include and analyze social relationship variables. To gather information on the full spectrum of social relationships including structural and functional aspects this broad term was implemented.

Data extraction and quality assessment
The data was extracted using a standardized form including information about the author, year, country, research design, study year (follow-up if applicable), sample size, response rate, age, gender, outcome, social relationship variables, and confounders in the fully adjusted model.
The quality assessment, including the methodological and reporting quality, was based on a checklist following the Newcastle-Ottawa-Scale [20] and its adaptation of Herzog, Alvarez-Pasquin [21]. The checklist included the three sections "selection", "comparability and confounders" and "outcome". It consisted of ten (cohort studies) respectively eight items (crosssectional studies) which could be answered by "yes", "no" or "unclear". Instead of reporting a sum score, a global rating was preferred [22]. The quality of cross-sectional studies which met three or less criteria were ranked as "low", four or five as "medium" and six or more as "high". Cohort studies with four or less fulfilled criteria were rated as "low" quality, five to seven as "medium" and eight or more as "high".

Analysis strategy
The results were descriptively and systematically synthesized. All associations between social relationships and utilization of physicians were extracted and categorized. Each social relationship variable was assigned to a social relationship category and dimension. For a better overview, closely related indicators were aggregated within categories (e.g., marital status or social support). Moreover, social relationship variables were classified as "structural" or "functional" [15]. The functional dimension was split into "received support" and "provided support". To answer our two research questions, we looked comprehensively at all associations between social relationships and physician use. For the sake of clarity and presentation, we focused on the statistically significant associations in our following tables (p<0.05). Due to the heterogeneity of the included studies a meta-analysis was not performed. Instead, we decided to complement our descriptive analysis by assessing the quality of the studies and by presenting a full description of the relevant quantitative data to maximize transparency and to enable rating the certainty of the results [23]. Since use (yes/no) and frequency of practitioner visits show a distinct level of information and have different meanings, the results are reported separately.

Literature search
A total of 1,392 publications were identified through database search. After removing 158 duplications, 1,234 articles remained for title and abstract screening (Fig 1). 1,176 publications were excluded based on title and abstract screening. Fifty-eight full-text articles were assessed for eligibility (S3 Table). Thirty-four were eliminated due to various reasons (deviant age group, deviant outcome, no social variable, none relevant data shown or analyzed). Twelve records were identified through reference search of included articles. In the full text screening inter-rater agreement on study inclusion was 88%. In the end, thirty-six publications based on thirty-five studies were included in the review and the synthesis. Though two articles [8,24] were based on the same study, their methodological and reporting quality was evaluated separately and their results were analyzed independently due to differing samples and data sets.
The methodological and reporting quality of 47.2% of the records was categorized as "high", 44.4% as "medium" and 8.3% as "low" ( Table 2). Apart from criterion two (nonrespondents and response rate) and criterion seven (independent assessment of outcome), the majority of the articles met the criteria of methodological and reporting quality ( Table 2, Table). Criterion two (non-respondents and response rate) was met only by five articles [6,12,13,27,37] and criterion seven (independent assessment of outcome) by eight records [6,29,31,34,35,37,41,47].

Associations between social relationships and physician utilization (yes vs. no)
Fourty associations between social relationships and the use of physicians were found in thirteen articles (S4 Table). In seven articles, fourteen associations were statistically significant ( Table 3). In other words, two thirds of the associations were statistically insignificant.
Seven out of these fourteen associations included variables of the structural dimension of social ties [12,33,36,38,44]. Suominen-Taipale, Koskinen [33] found consistent and relatively strong negative associations between being single, widowed, divorced or separated and the probability of physician utilization compared to older adults who are married and cohabiting. Systematic review of social ties and physician visits in old age Li and Chi [12] reported a strong positive association between living with at least one child and the physician use. Regarding the social network size, Park [38] observed a moderate positive association between having social network members and the use of physicians, while Liao, Chang [44] found a weak negative association between the household size and the probability of visiting a physician. Seven out of fourteen associations included variables of the functional dimension of social relationships [10,12,50]. Wolinsky and Johnson [50] found consistently positive, but weak associations between nonkin or kin social support and physician consultations. Fritel, Panjo [10] showed a higher probability of using outpatient care doctors for older people with weak social support. Otherwise, discussing health with friends or close relatives was associated strongly and positively with using health services [10]. Li and Chi [12] analyzed specific forms of social support in their study. For older people receiving or providing financial support or providing instrumental support they observed consistent and strong negative links to the utilization of physicians [12].

Associations between social relationships and frequency of physician utilization
Ninety-two associations between social relationships and the frequency of physician use were found in twenty-eight articles (S5 Table). In seventeen articles, thirty-seven associations were statistically significant (Table 4). Consequently, more than half of the associations were statistically insignificant.
Twenty-two out of these thirty-seven associations included variables of the structural dimension of social ties. Three studies found positive associations between being married and the frequency of physician visits [9,45,49] and one article reported a positive association between being widowed and the frequency of physician consultations [45]. Furthermore, living alone was positively and weakly associated with a higher frequency of using outpatient health services in three records [26,27,45]. Dalsgaard, Vedsted [29] found no, positive and negative differences for older adults who are living alone depending on their age and gender. Living with others (e.g., child or others except spouse) was associated negatively with the frequency of utilizing physicians in three studies [12,27,48]. Foreman, Yu [9] reported a strong and positive association between living with children and the frequency of health services use.   Systematic review of social ties and physician visits in old age The size of the social network was positively and strongly associated with a higher frequency of physician visits [32]. Counting friends or neighbors amongst their social networks, older adults reported a higher number of physician consultations [39]. Coulton and Frost [41] found out that socially isolated older people showed a lower number of physician contacts than socially integrated older adults. Moreover, Harris, Cook [34] and Militades and Wu [49] observed positive associations between higher contact frequencies in social networks and the frequency of physician use. Fifteen out of thirty-seven associations included variables of the functional dimension of social relationships. Two studies showed positive and weak associations between received social support and the frequency of physician utilization [6,8]. Financial, instrumental or informational support was associated weakly with more physician visits [11,12]. Emotional support was associated with less consultations [12]. Schafer [7] reported moderate to strong and positive associations between the likelihood of discussing health and the frequency of physician use taking several social ties into consideration (partner, children, non-kin). Harmonious social relationships decreased the frequency of physician visits [9] and respectful social ties increased the use rate [39].
Li and Chi [12] investigated the association between providing social support and the frequency of using physicians. Providing instrumental support was associated negatively and weakly. The provision of financial support was linked positively and weakly. Systematic review of social ties and physician visits in old age Systematic review of social ties and physician visits in old age

Summary of findings
This review provides a comprehensive overview and furthers the understanding of the association between social relationships and health services use among older adults (50 years and older). The first objective of this study was to systematically review social relationships associated with the utilization of outpatient care services of older people. The second aim was to evaluate magnitude and consistency of the associations between social ties and health services use. We included thirty-six records on thirty-five different studies reporting structural and functional dimensions of social relationships linked to the utilization of health services into our analyses. In most cases empirical evidence was insufficient and for several of the social tie variables inconsistent results were found. Taking into account the fully adjusted model, associations between use measures and social relationship variables were for the most part weak and statistically insignificant. Potentially, associations were underestimated by that strict criterion.
Overall, most of the studies focused on associations between social ties and frequency of physician use. The structural dimension of social relationships and its association with physician visits (use and frequency of use) was investigated far more often than the functional dimension. Though a substantial number of social relationship dimensions were explored until now, none of the included studies included a holistic approach of social tie measures (degree of integration, received and perceived social support) [15] and theirs links to health services utilization.
The majority of the associations between structural dimensions of social relationships and the use of physicians were positive and moderate in strength. The associations between functional dimensions of social relationships and the probability of using physician services were inconsistent and varied in strength. For the most part, social relationship variables assigned to the structural dimension were positively and weakly to moderately associated with the frequency of physician visits. Functional aspects of social relationships also tended to have positive associations with the frequency of physician utilization. The associations were weak or moderate in strength. All in all, the current status of quantitative data was insufficient to draw precise and generalizable conclusions.
Our review reveals that the link between various social relationship indicators and health care use as well as frequency of use have been investigated in few studies. This clearly indicates that further research is needed. Limitations Including a broad range of seven medical and sociological databases, we were able to minimize the risk of missing relevant articles. Nevertheless, the risk of publication bias is still existent. More than half of the studies (64%) were performed in North America, and therefore, findings cannot be generalized. Since the majority of included studies (61%) had cross-sectional design, conclusions concerning causal relations are not possible. Due to the fact that ten studies did not (four studies) or did not clearly meet (six studies) the quality criterion of representativeness and thirty articles did not (twelve studies) or did not clearly report (eighteen studies) information on non-respondents and response rate, the results were moderately robust. Overall, the methodological and reporting quality of the studies was mostly categorized as medium or high (92%).
Most of the studies referred to one year of physician use. Still, the range of the utilization variable was substantial between the studies (from 15 days to two years). As the time span was quite long in some studies, and considering the older age of the interviewed individuals, risk of memory bias was existent, particularly, if the information on consultations was not compared to medical records (twenty-seven studies).
Since there were no consistent measures of predictors (social relationships) and outcome variables (use and frequency of outpatient care visits), data was analyzed systematically, but descriptively. A prerequisite of meta-analyses is a high level of accordance across the included studies regarding independent and dependent variable measures and data analysis approaches [52]. Due to the heterogeneity of the included studies (e.g. study designs, sampling procedures, data collection methods, definition of outcome and exposure variables, confounders, quality of studies, statistical analysis and reporting) a meta-analysis was not conducted. In most cases the associations were small and statistically not significant. The current status of evidence is insufficient and partly inconsistent.
Unfortunately, analyses of group-differences concerning age, gender, and chronic conditions could not conducted on the basis of the review material.

Conclusions
Social relationships can increase or decrease the probability to consult a physician, and they can influence the frequency of visits. All in all, older people who are structurally integrated by social relationships are more likely to consult a physician at all and to contact a physician more often. Functional aspects of social relationships, depending on the form of social support, can increase or decrease the probability of physician use. Older adults who are experiencing social support tend to have a higher rate of physician visits than older people without any or less support.
On the one hand, this could be read as good news, since structural and functional aspects of social relationships tend to enable the utilization of health services, and thereby potentially foster older adults' health. Social relationships could offer informational, instrumental and emotional resources with regard to health, health care services and treatments. On the other hand, considering increasing numbers of single-person-households and an increasing risk of loneliness and social isolation in older age [53], this could be interpreted as a cause of concern, since older individuals who are not socially integrated may not find their way to health care services. The results do not include information about the adequacy of health care regarding access to health services, extent of health treatment, and quality of health care.
Social ties have an impact on the patient's motives for a consultation and on the patient's compliance regarding future visits for treatment, prevention or rehabilitation [54,55]. Consequently, health care practitioners should consider information on patient's social environments into their clinical routine. By default, physicians should assess social networks among the elderly screening for social resources or social needs of support. Furthermore, relevant stakeholders (e.g., physicians, public health institutions and health insurance companies) need to find ways to ensure that older adults can use outpatient care services regardless of their structural and functional level of social integration.
The variety of dimensions of social relationships presented in this review illustrates that utilization of outpatient health care services is a complex social process. Besides methodological challenges, the complex picture of social tie's impact on health care utilization bases on the fact that relationships are not always of positive virtue [56,57]. In contrary, "some of the most powerful impacts on health [and health services use] that social relationships may have, are through acts of abuse, violence, and trauma" [14]. This fact may represent a possible explanation for the inconsistent pattern of social relationships on health services use among older adults.
Furthermore, the inconclusive results demonstrate that measuring social relationships and their influence on health services use is a challenging methodological endeavor. Future health services research should accentuate social relationship variables more in detail, and not only in terms of structure and quantity, but also according to functional and quality aspects of social relationships.
The relatively low number of included studies indicates a deficit of elaborated observational studies dealing with the role of social relationships for the utilization of health services among older populations. The majority of the identified studies have a cross-sectional design investigating a number of possible social relationships of health services use. It is crucial to determine social ties for health services use more clearly and to identify causal relations, especially in the form of prospective cohort studies.
Methodologically, it can be constructive to directly connect the question of social relationships and health care utilization to the scientific debate of health care inequalities [58][59][60][61][62] by conducting mediator or moderator analyses to create further clarity. This may complement the identification and understanding of social inequalities in health services utilization. In the future, this can be directed into new approaches to reduce social inequalities in health services utilization and to offer needs-based access to health care and adequate levels of treatment.