Article Text
Abstract
Background Sense of coherence (SOC) is a health-promoting resource within the salutogenic theory that reflects an individual's coping ability. The association between SOC and mental health has been confirmed, but its association with mortality is less clear. We examined the association between SOC and all-cause mortality in an adult Dutch population.
Methods Between 1996 and 1998, a postal questionnaire, including the three-item SOC scale, was completed by 12 024 men and women aged 20–65 years, who had participated in a health examination (MORGEN project) 6 months to 3 years earlier. Vital status was recorded up to November 2011; in total, 603 deaths were registered (5%). The participants were divided into three groups with a weak (21.1%), intermediate (60.3%) or strong (18.6%) SOC. Cox proportional hazard models were used with an intermediate SOC as the reference group. Adjustments were made for sex, age, socioeconomic factors, indicators of health status and lifestyle.
Results A weak SOC, as compared with an intermediate SOC, was associated with a higher all-cause mortality risk after, on average, 13.5 years of follow-up and adjusted for sex and age (HR=1.40, 95% CI 1.14 to 1.70). After additional adjustments, the higher all-cause mortality risk remained statistically significant (HR=1.27, 95% CI 1.01 to 1.59). Mortality risk for the strong SOC group did not differ from that for the intermediate group.
Conclusions A weak SOC was associated with a higher risk of all-cause mortality. Health promotion focusing on strengthening SOC may be a promising new strategy, potentially affecting not only mental health but also mortality.
- Mortality
- Mental Health
- Psychosocial Factors
- Cognition
Statistics from Altmetric.com
Introduction
Sense of coherence (SOC) reflects an individual's ability to cope with difficult or stressful situations. It is a core construct of the salutogenic theory, which focuses on the origins of health and wellbeing rather than on disease. An individual's SOC consists of three components: the idea that the stimuli from one's environment are structured, predictable and explicable (comprehensibility), the feeling that sufficient resources are available to deal with the stressors (manageability) and the feeling that the challenges are worthy of investment and engagement (meaningfulness).1 People are constantly confronted with stressors in the course of their life that can push them to the dis-ease end of the ease/dis-ease continuum.1 ,2 A strong SOC is hypothesised to facilitate adaptive coping with the inherent stressors of everyday life (mental pathway) and as such is strongly related to personality factors such as neuroticism.3 ,4 A strong SOC is also hypothesised to promote the beneficial use of the resources at one's disposal (behaviour pathway) in order to maintain, or move to, a healthy state. A third pathway underlying the hypothesised influence of SOC on physical health is the physiological pathway, via changes in the neuroimmune and endocrine system.
In a review of more than 50 articles on the association between SOC and different health measures, Flensborg-Madsen et al5 conclude that the evidence for this association is relatively strong for health measures that include psychological or mental aspects. For example, several studies have found that SOC is negatively related to depressive symptoms and anxiety.6 ,7 However, the evidence is less convincing for physical health, disease incidence and mortality.5 Nonetheless, some studies seem to suggest that a strong SOC is associated with a lower incidence of diabetes, cancer and stroke.8–10 Research investigating the association between SOC and mortality is limited and has produced inconsistent results.11–13 Surtees et al11 compared the all-cause mortality risk between a weak and a high SOC group (measured with SOC-3) and found that a strong SOC was associated with a 30% reduction in all-cause mortality. However, Poppius et al12 found that a strong SOC was only associated with a lower mortality risk among white-collar workers when they compared three SOC-tertiles based on the original 29-item SOC scale. In the Umeå study, using the continuous SOC-13 measure, SOC was only associated with 1-year mortality and not with 4-year mortality.13 Haukkala et al6 found that the differences in all-cause mortality risk between three tertiles were non-significant using the SOC-13.
In addition, Lundberg observed that the health of people with a strong SOC was not better than the health of people with an intermediate SOC.14 In light of this, he wondered whether a stronger SOC was health-promoting or whether a weaker SOC was damaging to health. In the Helsinki Heart Study, these patterns were also observed when the authors compared the risk of coronary heart disease across tertiles of SOC scores (SOC-29). The researchers found that people with a strong SOC had an equal relative risk of coronary heart disease as people with an intermediate SOC.15 Later, Kouvonen et al8 noted that Finnish male employees with a medium and strong SOC had similar risks of diabetes incidence (using SOC-13). The idea that a strong SOC is not necessarily better for health than an intermediate SOC, but that a weak SOC is especially detrimental, has received limited attention in research.
We were able to study the relation between SOC and all-cause mortality in the Netherlands, as SOC was part of early measurements in a large-scale multipurpose prospective cohort study on health and chronic diseases. In order to explore further the observations made by Lundberg, we analysed the association of both a weak and a strong SOC with mortality risk as compared with an intermediate SOC.
Methods
Study design and population
The sampling frame for the prospective cohort study was the study population of the Monitoring Project on Chronic Risk Factors (MORGEN project), initially designed to study the impact of lifestyle and biological risk factors on disease.16 During the MORGEN project, data were collected on health status and risk factors from a random sample of 20–65-year-old inhabitants of three cities in the Netherlands (Amsterdam, Doetinchem and Maastricht) from 1993 to 1997. In total, 23 100 participants completed a general questionnaire on risk factors and health and a food frequency questionnaire. In addition, these persons participated in several physical measurements (such as blood pressure and height/weight measurements). Participants, who gave consent for further research (94%), were invited to complete the Health and Lifestyle Questionnaire (HLEQ) designed to study the relationship between psychosocial factors and disease (see figure 1). A detailed description of the HLEQ and its use within the European Prospective Investigation into Cancer and Nutrition can be found elsewhere.17 ,18 The HLEQ was administered either 6 months or 3 years after the participants had participated in the MORGEN project (between 1996 and 1998). In total, 12 024 people completed this questionnaire, of which 5339 were men and 6685 were women (response rate=56%).
Measures
All-cause mortality
Municipal population registrations were used to obtain vital status for all the participants. The registrations were completed up to November 2011.
Sense of coherence
The HLEQ included a three-item version of the SOC scale with an item for each of the three SOC components19 ,20 : (1) Do you usually see a solution to problems and difficulties that other people find hopeless? (meaningfulness) (2) Do you usually feel that the things that happen to you in your daily life are hard to understand? (comprehensibility) (3) Do you usually feel that your daily life is a source of personal satisfaction? (manageability). Participants could score the items as 1 (yes, usually), 2 (yes, sometimes) or 3 (no). The sum of these three items (after reverse coding of the comprehensibility item) reflected the total SOC, with higher scores indicating a weaker SOC. In line with previous studies,20 ,21 participants were divided into three SOC groups: weak (scores 6–9), intermediate (scores 4–5) and strong (score 3).
Sociodemographic variables
In the MORGEN study, age was divided into five categories with 10-year bands. Marital status was dichotomised into living with or without a partner. Educational level was categorised into low (primary school; lower vocational education; lower general secondary education), intermediate (community college; secondary education for at least 3 years) or high (higher vocation education; university).
Health status
Health status was assessed in the MORGEN study and included physiological measures and self-reported health. Participants who reported having (had) cancer, stroke, myocardial infarction or diabetes were classified as having prevalent disease(s). Self-reported health was measured by asking the respondents to rate their own health as excellent, good, adequate, moderate and poor (with higher scores indicating poorer subjective health). Neuroticism scores (scores between 0 and 12) were obtained by administering the neuroticism subscale of Eysenck's Personality Scale (α=0.86).22 Body mass index (BMI) was calculated by dividing the participant's weight by his/her height in metres squared.23 Participants’ systolic blood pressure was measured (mm Hg) while they were sitting down using a random zero sphygmomanometer. Cholesterol levels are reported in mmol/L.16
Lifestyle
Measures of lifestyle in the MORGEN study included smoking, alcohol consumption and physical activity. Smoking behaviour was categorised into three levels: never, past and current smokers. Alcohol behaviour was measured by asking the participants how many glasses of alcohol they drink on average per week. The classification of sport engagement was based on a single question, ‘Do you engage in sport?’
Statistical analysis
Baseline characteristics across SOC levels were analysed using χ2 (for categorical data) or ANOVA (for continuous data). Survival analyses were carried out to investigate the relationship between SOC and all-cause mortality after, on average, 13.5 years of follow-up. The association between SOC and all-cause mortality was assessed with Cox proportional hazard models. The intermediate SOC group was chosen as the reference group to explore the idea posited by Lundberg that a strong SOC is not necessarily better for our health than an intermediate SOC, but that a weak SOC is especially detrimental.14
The person-days of follow-up were calculated as the number of days from the completion of the HLEQ to the date of death, the date of migration or the end of the study (November 2011) depending on which occurred first. The average follow-up was 13.5 years (total of 154 262 person-years). The HRs are presented with the 95% CIs adjusted for sex and age (model 1), and additionally adjusted for marital status and educational level (model 2), health status (model 3) and lifestyle (model 4) to explore their role as potential confounders. The proportional hazards assumption was justified for all variables (p>0.05). Sex and educational level were considered as effect modifiers, but the interaction terms of sex or educational level and SOC were not significant (p>0.05). The analyses were conducted in SAS V.9.3 statistical programme package (SAS institute, Inc, Cary, North Carolina, USA).
Results
Complete data on the three-item SOC scale were available for 11 738 participants (97.6%), of which the baseline characteristics are presented in table 1. The mean age of the study population was 42.1 years (SD=11.1) and 44% of the participants was male.
Table 2 presents the differences in baseline characteristics by SOC level. Those with a strong SOC were more likely to be male, to have a higher educational level and to live with a steady partner. Additionally, those with a strong SOC more often engaged in sports, were less often current smokers and had fewer prevalent diseases. Participants with a stronger SOC also reported more positive neuroticism scores, they had a lower average blood pressure, but they drank on average more alcohol at baseline. Participants with a weak SOC reported a poorer self-rated health at baseline. No association was observed between SOC and BMI or having a high cholesterol level. Most of the significant differences between SOC groups on the covariates occur between the weakest versus the intermediate or strong SOC groups (except for BMI and cholesterol level). Some covariates also show significant differences between the intermediate and strong SOC group (ie, sex, age, educational level, alcohol consumption, self-rated health and neuroticism).
In total, 603 participants (5%) died after a median of 8.5 years of follow-up (352 men and 251 women). Figure 2 shows the survival probabilities for all participants across SOC levels. The association between SOC and all-cause mortality risk was significant for all participants.
Table 3 shows the HRs for all-cause mortality by SOC level. After adjustment for sex and age, a weak SOC was associated with a 40% higher risk of all-cause mortality compared with the intermediate SOC level. A strong SOC was not associated with a lower all-cause mortality risk. Further adjustments for educational level, marital status, indicators of health status (obesity, prevalent diseases, blood pressure, cholesterol level, subjective health and neuroticism) and lifestyle (smoking, sport and alcohol consumption) slightly attenuated the association between a weak SOC and all-cause mortality risk (HR=1.27, 95% CI 1.01 to 1.59).
The HRs were also examined for a model in which a strong SOC was chosen as the reference category (instead of the intermediate SOC). A weak SOC was associated with a higher all-cause mortality risk (HR=1.33, 95% CI 1.03 to 1.73) as compared with the strong SOC group, when adjusted for sex and age. This association lost its significance when further adjusted for health status and lifestyle (HR=1.12, 95% CI 0.83 to 1.51). The HR of the intermediate group did not differ significantly from the HR of the strong SOC group (HR=0.96, 95% CI 0.76 to 1.20).
The HRs were also examined for continuous SOC scores. After adjusting for sex, age, marital status and educational level, the HR for SOC was 1.07 (95% CI 1.01 to 1.15), indicating that a weaker SOC (ie, a higher SOC score) is associated with a higher all-cause mortality risk. This latter association lost its significance when further adjusted for health status (HR = 1.06, 95% CI 0.99 to 1.14).
Discussion
This study showed that a weak SOC was associated with a higher risk of all-cause mortality as compared with an intermediate SOC. This finding was only slightly attenuated by adjustments for age, sex, educational level, marital status, indicators of health status and lifestyle. A strong SOC was not associated with a lower all-cause mortality risk as compared with an intermediate SOC. The analysis based on the continuous SOC scores also indicated that a weaker SOC is associated with higher all-cause mortality risk, but this association lost its significance when it was adjusted for health status. When a strong SOC was chosen as the reference category, only the weak SOC group had a significantly higher all-cause mortality risk, although this association lost its significance when further adjusted for health status and lifestyle.
Strengths of the current study include the size of the sample and the opportunity to adjust for multiple risk factors, including sociodemographic variables, indicators of health status (obesity, blood pressure, cholesterol level, prevalent diseases, subjective health and neuroticism) and lifestyle (smoking, sport engagement and alcohol consumption). Together with the reliable ascertainment of deaths from municipal registrations, this study allowed us to prospectively examine the association between SOC and all-cause mortality. The observed effect size was substantial, with a 27% higher all-cause mortality risk for people with a weak SOC after adjusting for several risk factors, as compared with the intermediate SOC group.
In this study, persons with a strong SOC did not differ in their all-cause mortality risk from those with an intermediate SOC, but those with a weak SOC did have a higher all-cause mortality risk. This supports the idea posited by Lundberg that a strong SOC is not necessarily better than an intermediate SOC, but that a weak SOC is especially harmful. Supporting our results is a study conducted by Konttinen et al24 demonstrating that “the variation in the lowest SOC tertile was more strongly associated with health variables than in the highest tertile” (p. 2401). In addition, Kivimäki et al25 showed that only a weak SOC predicted health. These findings seem to suggest that a strong SOC is not necessarily leading to good health. Other studies have observed the same patterns in health-related outcomes between different levels of SOC,8 ,9 ,14 ,15 but in these studies no explicit comparison was made with an intermediate SOC to see whether a strong SOC also had a positive influence on health. The reader should, however, be aware that these studies used different measurements of SOC (with 3, 13 or 29 items) and divided participants differently into groups based on the SOC scores. This makes it difficult to compare the results of this study with other studies. The difference in all-cause mortality risk should be explored further using different SOC scales and comparing different group divisions.
Three possible pathways have been proposed to explain the relationship between SOC and mortality: the behavioural, the physiological and the mental/psychological pathways.8 ,9 ,26 The behavioural pathway assumes that a stronger SOC enables people to deal with stressful situations and that this will allow them to select appropriate resources and behaviours to deal with this situation. The relationship between SOC and healthy lifestyle choices has been confirmed in several studies,9 ,27–30 and this may lead to better health. The second pathway postulates that the level of SOC has direct physiological consequences via the central pathways of the neuroimmune and endocrine system.21 ,31 These physiological differences between people with a strong and a weak SOC may lead to differences in disease incidence and mortality. The mental pathway is grounded in the observation that several studies have found a strong positive association between SOC and mental health, and a strong negative association between SOC and stress.5 ,32 By reducing levels of stress, SOC is hypothesised to promote good mental health and to reduce the mortality and morbidity of stress-related diseases. In our study, SOC was associated with lifestyle (smoking, sport engagement and alcohol consumption) and with blood pressure (a physiological measure), but not with cholesterol level. The differences between SOC groups on these variables were especially visible when we compared the low SOC group with the intermediate and strong SOC groups. The differences were less often significant when we compared the intermediate with the strong SOC group. The association between SOC and all-cause mortality was partly explained by these factors, as the associations attenuated slightly when the variables were added to the analysis. However, to draw firm conclusions on the underlying mechanisms, further research is required that specifically tests for these mechanisms.
Several limitations should be taken into account when the results are being interpreted. First, although the SOC-3 has been found to show satisfactory test–retest reliability and to have a strong relationship with the original scale,20 it might not be appropriate to indicate those with a really strong SOC compared with the intermediate group.33 This may explain the lack of difference in the mortality risk between the intermediate and the strong SOC group. Future studies should test the claim made by Lundberg using the SOC-29 or SOC-13 version,14 which show a greater variety in strong SOC.33
The response rate for the HLEQ was 56%; this is reasonably good for a postal survey. Nonetheless, general non-response and item-non-response may have influenced our results. Therefore, we compared the participants who completed the HLEQ with the participants who did not complete the HLEQ on important study variables (measured in the general questionnaire of the MORGEN study). The participants completing the HLEQ were more often female, were younger, were higher educated, had less prevalent diseases at baseline and reported better self-rated health than those participants that did not complete the HLEQ. The selection bias may have reduced the contrast in the study population, especially because those people with a low SOC score were likely to be non-responders. This reduced contrast may have led to an underestimation of the associations. Differences in item-non-response were limited; of all the participants completing the HLEQ, only 2.4% had no SOC score. Because of this small number of missing SOC scores, any attenuation of the results is likely to be minimal.
The time period between the measurement of the covariates (in MORGEN) and SOC (in HLEQ), up to 3 years, is also a study limitation because covariates may have changed during this period. However, a stratified analysis of persons followed up 6 months or 3 years after completion of the MORGEN measurement showed that the association between SOC and all-cause mortality did not differ between the two groups.
Another limitation of this study is that the measurement of SOC is restricted to one time point. In contradiction to Antonovsky's hypothesised stability of SOC,1 several studies have found that SOC levels can deteriorate (ie, when confronted with stressful events) or improve during the course of life.34–36 Because of the single-point measurement of SOC, we were unable to consider these possible fluctuations in SOC. However, these fluctuations may be very relevant for mortality risk, for example, when SOC levels deteriorate after the SOC measurement due to negative life experiences of the participants. Moreover, the causal relationship between SOC and health could not be established. According to Antonovsky,1 health can also influence the level of SOC. However, in a study among Finnish employees, health was not found to predict SOC.25 More research is needed on the stability of SOC and on the causal relationship between SOC and physical health, requiring large follow-up studies with multiple SOC measurements.
Several studies have found that SOC is strongly related to personality traits such as neuroticism or extraversion,3 and to negative affect scales such as depression and anxiety.24 Therefore, it can be questioned whether SOC has divergent validity in relation to these factors. Feldt et al,3 for example, suggest that SOC and neuroticism could be considered opposite constructs. However, this suggestion is contradicted by Olson et al,37 who found that, in three different samples, SOC could only be partially explained by psychopathological variables. In a large Swedish study, Höchwalder found that a large part of the variation in SOC could be attributed to personality traits.4 “Nevertheless, SOC describes additional aspects in people that are not captured by the big five personality factors” (p. 594). With regard to the high correlations of SOC with negative affect scales, the same uncertainty regarding the divergent validity of SOC can be found. For example, Konttinen et al24 found very high inverse correlations of SOC with both anxiety and depression, leading them to question whether it is possible to discriminate between the three measures. Yet, Schnyder et al7 conclude, after studying the relationship between SOC and depression and anxiety, that “SOC is not merely a proxy measure of psychopathology, but rather a partially independent, general measure of a person's world view” (p. 296). These findings indicate that there is still uncertainty about the incremental validity of SOC and personality factors (such as neuroticism) and negative affect (such as depression) in their relationship to disease, mortality and health. In this study, neuroticism was moderately correlated with SOC (r=−0.49), which is somewhat lower than the correlations reported in other studies.3 ,4 Adjusting for neuroticism slightly attenuated the relationship between SOC and mortality, but it remained significant, which is in line with the study conducted by Surtees et al.11 More research is needed to get a better understanding of the relationship between SOC and mortality and the added value of the SOC concept in relation to other (psychological or negative affect) concepts in explaining and predicting health.
The findings suggest that the association between SOC and all-cause mortality can only partly be explained by lifestyle and physiological factors. Therefore, health promotion efforts may benefit from strengthening SOC. There have been a limited number of studies examining the effect of interventions on the development of SOC, for example, among unemployed Finnish individuals.38 In this intervention, unemployed individuals followed a re-employment programme that included group counselling and activities that improved their job-search and coping skills. The SOC level was increased at the end of the intervention, especially among employees over 30 years of age.38 In a study among Finnish employees suffering from burnout symptoms, the effect of two different interventions on the SOC of the employees was measured.39 Both interventions incorporated a reflection on personal norms, values, attitudes and patterns of behaviour. The intervention groups showed a significant increase in SOC as compared with the control group. Future research should investigate further whether interventions can be specifically targeted at strengthening SOC, especially because Antonovsky hypothesised that SOC was relatively stable, at least from around the age of 30 and for people with a strong SOC.1
SOC might be a relevant concept to explore further in research and health promotion. To our knowledge, this study was the first to test the idea that a weak SOC is especially detrimental to our health, whereas a strong SOC is not necessarily better than an intermediate SOC. Future studies on SOC should explore this hypothesis further, for example, in relation to quality of life, subjective health, disease incidence and mortality.
What is already known
-
Sense of coherence is positively related to mental health, health-promoting lifestyle behaviours and quality of life.
What this study adds
-
A weak sense of coherence is associated with a higher all-cause mortality risk in a large Dutch prospective cohort, whereas the all-cause mortality risk for people with a strong and an intermediate sense of coherence does not differ.
References
Footnotes
-
Contributors All authors have contributed to the conception and design of the study and have been substantially involved in the writing of the article by revising the manuscript critically. WMMV and HSJP were mainly responsible for the data collection and management. SS was mainly responsible for the data analysis. The final article is approved for publishing by all authors. HSJP is the guarantor.
-
Competing interests None.
-
Patient consent Obtained.
-
Ethics approval Ethical Board of the Dutch Organisation of Applied Science.
-
Provenance and peer review Not commissioned; externally peer reviewed.