The current survey drew on data from a nationally representative online survey of Germans aged 18 to 74 (where 3,091 respondents participated). A key aim of this study was to clarify the determinants of healthcare use and sick leave days. In this current study, we restricted our sample to full-time employed individuals aged 18 to 64 years (n = 1,342). The age restriction was made because individuals aged 65 years and over are commonly retired in Germany. Moreover, we focused on full-time employed individuals to ease the comparison. In our view, it is rather difficult to compare the number of part-time employed individuals (e.g., when one individual is working 5 h a week and another individual is working 30 h a week).

The survey was carried out in mid-March 2022. The market research firm Bilendi & respondi – an ISO 26,362 certified online sample provider – recruited participants using its own actively managed online access panel. The participants were rewarded by bilendi & respondi based on their Mingle points system. The points awarded were nominal in value, as a small compensation for the time it took to complete the survey.

Respondents were drawn from an online sample in such a way that their age, gender, and federal state distribution were representative of the entire German adult population (quota sampling) [14]. About 11,900 individuals were invited to participate. A sample selection bias could not be calculated for reasons of data availability.

All individuals provided informed consent. This study was approved by the University Medical Center Hamburg-Center Eppendorf’s Local Psychological Ethics Committee (LPEK-0412).

Respondents self-reported the number of sick leave days in the preceding 12 months. The individuals were instructed as follows: “Please indicate all days, not only those for which you have received a doctor’s certificate of incapacity for work”.

This is a common assessment of sick leave days. For example, it is in accordance with the assessment used in the German Socio-Economic Panel (GSOEP) [15] – a well-known and long-running household panel.

With regard to personality-related factors, we included these factors: The 10-item Big Five Inventory (BFI-10) [16] was used to quantify personality (i.e., agreeableness, conscientiousness, extraversion, neuroticism and openness to experience). It is an established tool to quantify the key personality characteristics (two items per dimension; each dimension goes from 1 to 7, higher values reflect a more pronounced personality factor). Moreover, altruism was quantified using the subscale ‘altruism’ of the International Personality Item Pool (IPIP-5F30F-R1 [17]) consisting of six items (ranging from 1 to 5 in each case). By averaging the recoded items, a score was calculated (1 to 5; higher scores indicate higher altruism). Based on the Interpersonality Reactivity Index (IRI [9]; German version: Saarbrucken personality questionnaire, SPF [18]; called SPF-K ([19]), empathy was measured. It has four items (5 levels in each case). Following Paulus [19], a sum score was created (ranging from 4 to 20, higher values reflect higher empathy). Further details are provided by Paulus [19].

With regard to psychosocial factors, we included these factors: Loneliness was quantified using the 6-item De Jong Gierveld loneliness tool) [20]. It consists of six items. By averaging the items, a loneliness score was computed (from 1 to 4; higher values reflect higher levels of loneliness). Perceived social isolation was quantified based on the Bude and Lantermann [21] tool which has four items. A score was created by averaging the items (from 1 to 4, with higher values reflecting higher perceived social isolation). Coronavirus anxiety was measured using the coronavirus anxiety scale [22,23,24]. It has five items. A sum score was computed (from 0 to 20, higher values reflect higher coronavirus anxiety). Moreover, the Patient Health Questionnaire-9 (PHQ-9) was used to assess depressive symptoms. It consists of nine items (sum score ranges from 0 to 27, with higher values corresponding to more depressive symptoms) [25]. To assess anxiety symptoms, The Generalized Anxiety Disorder-7 (GAD-7) [26] was used. It has seven items (sum score ranges from 0 to 21, with higher values reflecting more anxiety symptoms).

Based on prior research (e.g., [27,28,29] and based on theoretical considerations, covariates were selected. More precisely, as covariates, we included several sociodemographic and health-related factors in regression analysis: Age, sex (three categories, reference category: men; women; diverse), one or more children in own household (reference category: no; yes), family situation (married, living together with spouse; married, not living together with spouse; single; widowed; divorced; dichotomized into: married, living together with spouse; other including all other categories (as reference category)), and school education (reference category: Upper secondary school; Qualification for applied upper secondary school; Polytechnic Secondary School; Intermediate Secondary School; Lower Secondary School; Currently in school training/education; Without school-leaving qualification). Additionally, we included vaccination against Covid-19 (reference category: no; yes), self-rated health (single-item measure ranging from 1 to 5; higher values reflect better self-rated health) and the presence of one or more chronic conditions (no; yes) in regression analysis.

In additional analysis, and in accordance with prior research (e.g., [30, 31]), it was also adjusted for some lifestyle-factors. More precisely, it was adjusted for alcohol consumption (reference category: daily; several times per week; once a week; 1–3 times per month; less often; never), smoking behavior (reference category: yes, daily; yes, sometimes; no, not anymore; never smoker), and frequency of sports activities (reference category: no sports activity; less than one hour a week; regularly, 1–2 h a week; regularly, 2–4 h a week; regularly, more than 4 h a week).

Firstly, sample characteristics are shown. Thereafter, multiple negative binomial regressions were used to examine the personality-related and psychosocial correlates of sick leave days. For example, compared to a Poisson model, a negative binomial regression had much smaller BIC values (Poisson model, BIC: 35,484.5; negative binomial model, BIC: 7,196.3). This shows that the negative binomial model fits our data much better.

The significance level was set at p < 0.05. Stata 16.1 (Stata Corp., College Station, Texas) was used for performing statistical analyses.