This study showed that unfavourable levels of positive and negative experiences of social support were associated with poor mental health and with changes from a good to a poor mental health. Low levels of positive experiences of social support were additionally associated with a low fruit and vegetable intake. High levels of negative experiences of social support on the other hand, were additionally associated with smoking, physical inactivity, overweight and poor self-perceived health, and with unhealthy changes in alcohol consumption, physical activity and self-perceived health. Neither positive nor negative experiences of social support were associated with hypertension or hypercholesterolemia.
Although the relationship between social support and mental health is already well established in existing literature, results regarding other lifestyle and health outcomes are however inconclusive [14
]. Investigating the unhealthy changes over a 10-year period enabled us to estimate the associations between social support and the onset of different lifestyle and health indicators. This analytical strategy was also applied by a few other studies [12
]. Glass et al. (2006) reported that social engagement, defined as social and productive activity, was associated with prevalent and incident depressive symptoms, whereas Melchior et al. (2003) found a significant association between inadequate social support and incident self-reported poor health in men, but not in women [15
]. Green et al. (2008) only found significant associations between social network characteristics and cognitive and functional status but not decline [12
]. The researchers suggested that cognitive and functional decline might therefore be the cause rather than the consequence of social network characteristics [12
]. This might also be partly explaining why we have found an association between social support and current smoking and fruit and vegetable intake, but not for future smoking or intake of fruits and vegetables. Regarding physical activity, our results showed that negative experiences of social support related to current and future physical inactivity. This corroborated with the findings of Kouvonen et al. (2011) that practical support was associated with the recommended amount of leisure time physical activity in cross-sectional and longitudinal analysis. However, they did not find a longitudinal association concerning confiding/emotional support and incident physical activity [18
]. In our study we used positive and negative experiences of social support as separate predictors of health. This led to the result that smoking, physical inactivity, overweight and self-perceived health appeared to be more affected by negative experiences of support than by positive experiences of support. Newsom et al. (2005), who studied the relative importance of positive and negative exchanges, found that positive exchanges only related to well-being, whereas negative exchanges additionally related to psychological distress [39
]. This could indicate that different mechanisms for positive and negative experiences of support may play a role in how they influence health.
Strengths and limitations
We studied the relationship between social support and a variety of health outcomes. Outcomes with different levels of subjectivity were included ranging from subjective outcomes (mental health and self-perceived health) to less subjective outcomes (lifestyle factors) to objective outcomes (clinically measured factors), in order to overcome the problem of correlated measurement error. Correlated measurement error is likely to occur when both the exposure and the outcome measurement have the same source of error [40
]. This is particularly of concern when both measurements have high levels of subjectivity, like with social support and self-reported health. Furthermore, we tried to diminish the bias caused by correlated measurement error by using repeated measurements to estimate the prevalence of the outcomes. Additionally, we studied the association with incident outcomes, hereby excluding the prevalent cases of poor health at round 2. Nevertheless, correlated measurement error still could be a methodological explanation for both the found association between social support and self-perceived health or mental health, and the lack of association between social support and hypertension or hypercholesterolemia. Also, correlated measurement error could have played a role in the association between 1) positive experiences of social support and low fruit and vegetable intake, and 2) negative experiences of social support and current smoking. Concerning both outcomes, a statistically significant association was found for the prevalent outcomes but not the incident outcomes. Another limitation of this study was that although we used a prospective analytical approach, we cannot provide evidence that improving the quality of social support leads to a more beneficial lifestyle and improves health status. This is because we had no information about changes in social support prior to our measurement, and were therefore not able to investigate the effects of changes in social support on lifestyle and health.
Another strength of our study was the low number of non-response regarding the Social Experiences Checklist. Participants were excluded if they had one or more missing items on the Social Experiences Checklist. Because each questionnaire was checked for item non-response at the community health service by a trained research assistant, the number of people with missing items for both the positive and negative experiences of social support was low (2.9%). People with missing values for the social support assessment were likely to be older, lower educated and/or more often unemployed than the analytical sample. However, the influence of age, educational level and employment status on the risk estimates was not larger than the effects of social support on the different health indicators. Therefore, we believe that any attenuation of the effects of social support on the different health outcomes caused by this selection bias is minimal.
Social support as determinant of health is complex and closely interwoven with other factors, including socio-economic and cultural factors [41
]. The participants included in our study lived in a rural area and almost all of them originated from the Netherlands. Research in other populations and countries is needed to gain insight in the universal impact of social support on health and health-related factors. For the analysis in this study, the second, third and fourth round of the Doetinchem Cohort Study were used. Although the response rates were generally good throughout these three subsequent rounds (79%, 75% and 78% respectively), we cannot rule out the possibility that attrition due to drop-outs or intermittent missing data in the outcome variables might have influenced our results. Nevertheless, this possible selection bias would mainly affect the prevalence estimates, and much less the estimated magnitudes of the associations [25
The results of this study imply that both positive and negative aspects of social support are related to some, but not all, indicators of lifestyle and health. More research that differentiates between positive and negative experiences of social support is necessary to confirm our findings, and to give insight into how they may operate differently on health. In our analysis, we assessed social support at a single time point. Longitudinal studies using repeated measurements of social support are needed to investigate the influence of changes in social support on lifestyle and health, before any recommendations regarding public health interventions can be given.