Frequency of stressful life events and associations with mental health and general subjective health in the general population

Socioeconomic status was shown to have an impact on life events and other stressors (Lantz et al. 2005; Hobson et al. 1998). Higher financial stress was associated with moderate-to-severe functional limitations and poor-to-fair self-rated health. Differential exposure to stress and negative life events based on socioeconomic disparities may be one determinant for socioeconomic inequalities in health. Gender also determines the risk of exposure to stress (Landrine et al. 1995; Thomas et al. 2008). Gender has been found to influence the predictive role of stressful life events (SLEs) for the risk of major depressive episodes over a long-term period of 25 years (Assari and Lankarani 2016). The predictive role of SLEs for depression was stronger in men than women. However, these findings may be due to gender differences in the threshold of reporting stress, perceiving risks or use of coping strategies (Harris et al. 2006).

Health is defined as “a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity,” according to the World Health Organization (WHO). A meta-analysis by Kraaij, Arensman and Spinhoven (Kraaij et al. 2002) revealed that almost all negative life events appeared to have a modest but significant relationship with depression. For instance, the total numbers of negative life events and of daily hassles had the strongest relationships with depression. Furthermore, longitudinal data from a 10-year follow-up health survey were analyzed including life events as stressors and morbidity years of ten chronic conditions (Hsu 2011). Accordingly, loss of offspring increased the risk for impaired physical health, psychological health and life satisfaction. Being newly widowed was related to depressive symptoms, whereas being newly divorced increased the risk of poor social support. There is also longitudinal evidence for life events related to the development of anxiety disorders (Jonker et al. 2017). With reference to morbidity years, increased failure of successful aging was observed, especially after medical illness such as stroke (Kornerup et al. 2010), respiratory system disease (Lietzén et al. 2011), cancer (Lillberg et al. 2003), diabetes and visceral adiposity (Mooy et al. 2000). Chronic inflammatory diseases of the central nervous system such as multiple sclerosis have also been linked to SLEs (Mohr et al. 2004). There is evidence for gene-environment correlations with regard to SLEs and psychotic experiences (Shakoor et al. 2016). The nature and extent of the association between life event exposure and stress reactivity in young adults depended on the valence of the events together with the sex of the individual and their social network size (Phillips et al. 2005). To provide a theoretical framework, Schwarzer and Schulz (2003) proposed a model suggesting different mediating pathways with regard to the association of stressful life events and illness indicators. The potential mediators were health-compromising behaviors (e.g., smoking, alcohol consumption, physical inactivity), physiological changes and negative affect.

Originally, according to Holmes and Rahe, critical life events were supposed to be associated with stress experience regardless of the valence of a certain life event (Holmes and Rahe 1967). This means that important life changes, whether negative, such as the death of a close friend, or positive, such as marriage, all induce stress. This theoretical framework implies that life events require the individual to adapt and regain stability. The Social Readjustment Rating Scale [SRRS; (Holmes and Rahe 1967)] is a widely used checklist for the occurrence of stressful life events (SLEs) (Scully et al. 2000). The authors suggested scores reflecting the severity of the impact of the life event for each item, which refer to the past year. The scores called “life-changing units” range from 11 to 100. Stress is operationalized as a cumulative psychological construct and estimated from the summed frequency of life events. Higher sum scores correspond to higher stress levels. There are modified English versions of the SRRS (Hobson et al. 1998; Hobson and Delunas 2001; Miller and Rahe 1997); some are based on specific populations (Goldberg and Comstock 1980; Lüdtke et al. 2011). In summary, checklists are still the most popular assessment methods for stressful life events (Hobson and Delunas 2001; Cleary 1980) although interview methods have been introduced [e.g., (Paykel 1997; Wittchen et al. 1989)]. As criticized by some researchers (Hobson and Delunas 2001), the life event items contain both discrete events (e.g., sustained injury or disease) and possible symptoms or consequences of stress (e.g., change of sleep patterns). Additionally, critical life events vary in their valence, ranging from positive to negative, e.g., marriage and death of a spouse. Also, it is not clear whether and how operationalization of a stress score improves the predictive value of life events with regard to mental or general health (Cooper et al. 1985).

First, we aimed to present the frequency of stressful life events for a large representative sample of the German population. Thus, we wanted to provide data on the frequency of life events depending on age, sex and socioeconomic status. This allows for cross-national comparison and for the investigation of cultural differences. Second, we further wanted to determine which of the SLEs are of interest for epidemiological research in terms of psychological distress (depression and anxiety) and subjective health. Highly stressful life events (based on the SRRS guidelines) are assumed to have a negative impact on health resulting in higher distress (depression, anxiety) and lower self-reported general health. Finally, we tested whether weighted, as suggested by the original authors (Holmes and Rahe 1967), or unweighted SLE scores of life events predict distress and subjective health taking sociodemographic variables into account. We expected stronger covariations of SLE with indicators of psychological distress compared with the more abstract variable of subjective general health in terms of convergent and discriminant validity.

Prior to enrollment, participants signed written informed consent. The study and its procedure were approved by the ethics committee of the Statutory Physician Board of the State of Rhineland-Palatinate and controlled by the local and federal data safety commissioners. The GHS sample was drawn randomly from the local registry of the city of Mainz and the district of Mainz-Bingen. The sample was stratified 1:1 for gender and residence and in equal strata for decades of age. Participants between the ages of 35 to 74 years were included. Individuals with insufficient knowledge of the German language, or physical or mental inability to participate, were excluded; 5.2% were excluded based on these exclusion criteria. The response rate (defined as the recruitment efficacy proportion, i.e., the number of persons participating in the baseline examination divided by the sum of the number of persons participating in the baseline examination plus those who refused and those who could not be contacted) was 55.5%. The sample is described in detail in the supplement material (Table S2a and Table Sb) separately for men and women. Mean age was 55.0 (SD = 11.1) years; 49.4% were female.

Data were collected during a 5-h examination in the study center. They included evaluations of classic cardiovascular risk factors, laboratory examinations, a computer-assisted personal interview and anthropometric measurements. In the computer-assisted personal interview, participants were asked questions about sociodemographic variables and completed questionnaires. All examinations were performed according to standard operating procedures by certified medical technical assistants.

The Social Readjustment Rating Scale (SRRS) (Holmes and Rahe 1967) is a widely used checklist for stressful life events (Scully et al. 2000; Hobson and Delunas 2001). It provides a standardized measure of the impact of a wide range of common stressors. The original version includes 43 items. An adapted German version with 36 items was administered in the GHS (Cronbach’s alpha = 0.81). Seven items were excluded because of partially overlapping content and time constraints. In line with the original scale, the GHS version of the SRRS used four response categories (0 = never occurred; 1 = occurred in the last year; 2 = occurred 2 to 5 years ago; 3 = occurred > 5 years ago). Response categories 2 and 3 were combined as “occurred > 1 year” in the descriptive statistics based on single SLEs stratified for sociodemographic variables (see supplement material Table S1, Table S2, Table S2a and Table S2b). Since weighted scores indicating the severity of stress for each life event were only available for SLE regarding the past year (Miller and Rahe 1997), we accordingly used only the categories 0 and 1 in multiple regression analyses with SLE sum scores as dependent variables to compare unweighted and weighted sum scores of the SRRS items. Death of the spouse or a close family member, divorce or marital separation, and jail terms were regarded as the most severe SLEs.

Depression was measured by the Patient Health Questionnaire (PHQ-9) (Kroenke and Spitzer 2002), which quantifies the frequency of being bothered by each of the nine diagnostic criteria of major depression over the past 2 weeks. The items were rated on a 4-point Likert scale, 0 = not at all to 3 = nearly every day. Responses are summed to create a score between 0 and 27 points.

Generalized anxiety (GAD) was identified by the GAD-2 screener derived from the anxiety module of the PHQ (Kroenke et al. 2007). Participants answered the two GAD items “Feeling nervous, anxious or on edge” and “Not being able to stop or control worrying” on a 4-point Likert scale, 0 = not at all to 3 = nearly every day.

General subjective health was measured by the item “How do you rate your health in general on a 5-point Likert scale ranging from 1 = “very good” to 5 = “bad”? (Brähler et al. 1979). To ease the interpretation, we recoded this variable so that a high score indicated better health.

Sociodemographic variables used in this study were: sex, age, socioeconomic status (SES), partnership and income. Sex and partnership were dichotomous variables. Age was measured as a continuous variable for descriptive statistics, and subsequent analysis of variance categories were calculated based on 10-year intervals. SES was defined according to Lampert and Kroll’s scores of SES with a range from 3 to 21: 3 indicates the lowest SES and 21 the highest. Subsequently, we generated three SES categories (low, moderate, high) based on Lampert et al.'s (Lampert and Kroll 2009) recommendation for descriptive analyses and analyses of variance. Following the Organisation for Economic Co-operation and Development [OECD (OECD 2011)], we estimated equivalence income by dividing net household income by the square root of the number of people living permanently in a household.

Descriptive statistics were calculated for each SRRS item. Group-specific statistics were provided, such as sex, age and socioeconomic status. To test group differences, univariate and multivariate analyses of variance were subsequently performed. Spearman correlation analyses (SLE at the item and aggregated level) and multiple hierarchical regression analyses (SLE at the aggregated level) with the enter method were used to investigate the association of stressful life events with health-related variables. We did not estimate missing data for inference statistical analyses because they can be systematic, for instance, because of sex (e.g., pregnancy), age (e.g., school-related questions, retirement) and other determinants. For multivariate analyses (i.e., variance and regression analyses), listwise deletion was used. Multiple regression models with SLE on the item level were not conducted because of the detrimental reduction of the sample size and lack of variance for many single SLE item predictors. In the first block of the regression analyses at the aggregated level, sociodemographic variables were controlled. The second block included the aggregated SLE sum score (unweighted or weighted). Due to the big baseline sample, p values should be treated with caution and are primarily given for descriptive reasons; they correspond to two-tailed tests. Statistical analyses were conducted using SPSS for Windows 23. Correlation and regression coefficients were interpreted based on Cohen (Cohen 1988).