We investigated two main questions in this prospective cohort study, conducted in a middle-aged Danish population (aged 49–61 at baseline). First, we investigated how perceived emotional and perceived instrumental social support is associated with GP contact tendency, and the numbers of GP contacts and secondly, whether there is differential vulnerability according to sex, number of morbidities, and according to occupational social class; that is, whether there is differential vulnerability. Overall, we found that perceived emotional and instrumental social support dimensions of social relations were not significantly associated with GP contact in this adult population and that there were no strong indications of differential vulnerability.
To our knowledge, there are no previous studies specifically measuring the perceived emotional and instrumental dimensions of functional aspects of social relations in relation to GP utilisation, and generally, we found few studies investigating how functional aspects of social relations are associated with GP utilisation. In the following, we will therefore relate our findings to those from studies applying other functional measures of social relations, such as social integration and social anchorage among middle-aged and older people [8, 14,15,16,17]. Two previous studies on social integration and social anchorage found no association with GP utilisation in individuals aged 50 years and older, nor in individuals aged 60–78 years, respectively [15, 26]. On the other hand, the latter study also found that a sense of community cohesion and belonging was associated with a higher frequency of GP use. However, with only a small effect, leading the authors to conclude that social- and psychological factors only influence GP use marginally; and that comorbidity was the strongest predictor of frequent GP use . Moreover, in contrast with our findings that instrumental social support is not associated with GP contact, a study found that receiving and providing financial support increased both the likelihood and number of GP visits, among people of 60 years and older. However, in the same study, emotional social support was not associated with GP utilisation . In line with results published by Bremer et al. 2018, we find no strong indications of interactions between emotional and instrumental social support and comorbidity. Bremer et al. did not find any interaction effect between comorbidity and the functional aspect of social integration on GP contact. However, they did find differential effects of social contact frequency and number of emotionally close relationships according to health status (measured as self-rated health). Among people with good health status, high social contact frequency was associated with more GP visits. Finally, among people with poor health, a higher number of emotionally close relationships were associated with more GP visits . Moreover, our findings are in line with those of Korten et al. who did not find that social support was associated with the volume of GP contacts for either men or women. Moreover, they found that the strongest predictor for healthcare utilisation was physical health . On the other hand, Korten et al. find indications of low (relative to high) social support being associated with a lower tendency to contact a GP among men compared to women .
Strength and limitations
The large study population, and the prospective nature of the study design, are two major strengths of this study, giving substantial weight to the results and diminishing the risk of reverse causality . Another strength of this study is the application of detailed register data of high validity to measure contacts with GPs. Applying these objective measures of GP contacts throughout the two-year study period eliminates the risk of recall bias as opposed to if a questionnaire based measure of GP contacts had been applied .
The measures of social support included in the analyses have been validated for content and face validity as well as for reliability, and it has been argued that they are suitable for measuring functional aspects of social relations, specifically among middle-aged individuals . To our knowledge, there is no validated measure combining the emotional and instrumental social support measures, and therefore we believe it is appropriate to conduct separate analyses for the two types of social support respectively. Moreover, the social support aspects of social relations may play a larger role in an individual’s decision-making in regard to contacting the GP than in the decision-making regarding more specialised healthcare that – as in the Danish healthcare system – may need a referral. Hence, GP contact is a reasonable outcome measure when studying how social support affects healthcare utilisation. A further strength is the inclusion of the Danish Occupational Social Class Measure (DOSCM) as a measure of socioeconomic position. The DOSCM is based on an assessment of the occupational skills and competencies necessary for a particular job, and the influence and control associated with the position. The measure has been demonstrated to be suitable specifically to the late middle-aged, as this group might be transitioning from working age to old age. Moreover, the use of DOSCM – rather than separate measures of income, education and occupation – enables analysis of possible effects of material resources, combined with skills and knowledge linked to social standing in society .
The study also has, nevertheless, potential limitations that warrant attention. Although the study population is relatively large, we cannot completely rule out that the insignificant findings were a result of lack of statistical power. Moreover, as described elsewhere, the CAMB population is a selective population. Compared to non-participants, a larger proportion of participants were employed, male, and had Danish origin. Furthermore, only people living in a specific geographic area (defined as ‘the eastern parts of Denmark’) were invited to participate [28, 29]. Moreover, non-respondents showed a higher all-cause mortality than respondents throughout the follow-up [28, 29]. Yet, participants did not differ substantially from non-participants, either regarding number of contacts with GP or educational level, during the first year of follow-up [28, 29]. We conducted a sensitivity analysis including the 278 individuals excluded due to missing values in the full population, in which the results did not differ substantially. In the analyses, it was not possible to account for possible changes in level of emotional and instrumental social support during the two-year follow-up period. To limit the bias possibly arising from this and which might diminish or erase an existing association, it might have been appropriate to choose a shorter follow-up, namely one year or six months. Moreover, we included a self-reported measure of health status: number of (self-reported) morbidities. This might have biased the results, particularly if the effect of health on GP contact is either stronger or weaker than measured through these self-reported measures, or if individuals either forget or leave out health information when filling out the questionnaire. However, as we find no reasons to assume that this possible measurement error relates to the perceived emotional social support or perceived instrumental social support, it is unlikely that it has had any particular effect on the results. In relation to this, an ideal alternative to the self-reported measure of health status would have been objectively registered diagnoses and conditions related to each of the registered GP contacts. However, as the DNHSR is established mainly for administrative purposes it only contains information on type of consultation (physical, telephone, email etc.) and no information on the reason for contact. It is possible that the pattern found in this study is somewhat blurred by not knowing whether the contacts reflect routine consultations or conditions requiring more intense medical treatment. If for instance, the majority of health conditions in this population sample were of a serious nature requiring intense treatment, it might be that the seriousness of the condition, rather than the perceived level of social support, were the main driver of GP contact. In the Danish primary care system, each citizen is registered to a single dedicated general practice. As such, it is rare that a person contacts another general practitioner or visits different practices. Hence, for practical purposes in a statistical analysis, persons are nested within practices. It is likely that some practice characteristics may make their patients consult their GP more/less. Without specifying these characteristics, heterogeneity between practices with respect to these will make observations within practices dependent. This could have been adjusted for in the analysis employing the GEE approach covering whole practices, not just the individual patients (as implemented now). Such additional adjustment would not affect the point estimates, only their standard errors and thereby the width of their confidence intervals. In the current study, we did not implemented these adjustments. This is primarily because we did not have access to these data and gaining access would have involved considerable time and administrative efforts. Moreover, using a proxy for practice, e.g. municipality, would render an adjustment with little effect. Furthermore, the dependences between patients within practices are unlikely to be strong: predictors for health care seeking behavior are typically characteristics of the patient rather than the practice, e.g. cohabitation, geographical distance, and there are guidelines in place to regulate and standardize follow-up for specific groups of patients. Therefore, we argue that the dependence within practices is not substantial. Another potential limitation is the question of novelty as our survey data were collected in 2009-2011, and the follow-up data on GP contacts refer to the period 2012-2013. However, as the healthcare system in Denmark has not changed markedly in terms of how it is organized and structured during the last ten years (and more), we do believe that the data applied in this paper are still relevant. Finally, but still very important to consider, is the age of the sample, which is 49-61 years at baseline. Middle-aged adults may be much less dependent on others to seek healthcare than for example older adults and altogether this might be part of the reason why we found no associations between perceived emotional and instrumental social support and GP contact. In sum, it would indeed be interesting to replicate the analyses in a sample of older people and including objective measures of health status.
Implications and future research
Generally, our findings do not support the argument that perceived social support as a functional aspect of social relations, independent from health-related needs, are associated with general practitioner contact among middle-aged people. Hence, this study does not give weight to the promotion of social interventions to strengthen social relations and social support, reducing social isolation and loneliness, among middle-aged people with low social support. However, in the future it would be valuable to replicate the study with several GP follow-up points, and to include more of the functional aspects of social relations such as relational strain and social anchorage. In this way, it might be possible to explore how changes in social support over time affect contact with GPs, and how different functional aspects might be a stronger or lesser predictor of GP contact over time. Furthermore, it would be valuable to conduct in-depth qualitative studies to better understand the mechanisms between perceived social support and GP contact among middle-aged people.