Sample
The current study used data from the German Ageing Survey (DEAS), which started in 1996 (first wave). For reasons of data availability, our study was restricted to the second wave, which took place in 2002, as our key independent variable was only assessed in the second wave.
The Federal Ministry for Family Affairs, Senior Citizens, Women and Youth funded the DEAS study. The DEAS study is a representative study of individuals residing in private households (40 years and over). Therefore, the main inclusion criterion was that individuals had to be at least 40 years old. More specifically, first time participants were included when they met the following criteria: [1] born between 1929 and 1974, [2] living in a private household (which means that individuals residing in institutionalized settings were excluded). For panel participants, inclusion criteria were: [1] one or more valid interviews in former waves, [2] willingness to participate in the panel (written consent given by baseline participants), [3] still alive and not living abroad. Various topics (e.g., perception of ageing, social support, health, occupational status, retirement) are covered in the DEAS study.
In the first wave, 4838 individuals were interviewed (50% response rate) and 5194 individuals took part in the second wave (38% response rate). In the first wave, the gross sample included 9613 individuals. Thereof, 3268 individuals (34.0%) refused participation. Other minor reasons for non-response were, for example, that individuals were temporary ill (108 individuals, 1.1%) or permanently ill (383 individuals, 4.0%). Neller showed that the response rate reported in the DEAS study is similar compared with other large survey studies that have taken place in Germany [16]. Klaus et al. provided additional details with regard to the DEAS study [17]. In our analytical sample (i.e., individuals included in regression analysis), n = 3850 individuals were included (with number of specialist visits as outcome measure; n = 3844 with number of GP visits as outcome measure).
Written informed consent was provided by all individuals. An ethical statement for the DEAS study was not needed, as the criteria for it were not met (e.g., examination of patients, risk for the respondents, or the use of invasive methods).
Dependent variable
The use of outpatient physician visits (first dependent variable: number of GP; second dependent variable: number of specialist visits) in the preceding 12 months was assessed in the DEAS study. Several medical specialties were reported in the DEAS study. The number of GP and specialist visits was quantified as: never; once; 2–3 times; 4–6 times; 7–12 times; more often. Following Bock et al. and Flennert et al. [4, 18], it was recoded as: “never” = 0; “once” = 1; “2–3 times” = 2.5; “4–6 times” = 5; “7–12 times” = 9.5; and “more often” = 13.
Independent variables
A single item (based on WHOQOL-BREF [19]) with clear face validity was used to measure meaning in life: “To what extent do you feel your life to be meaningful?” [1 = not at all; 2 = a little; 3 = a moderate amount; 4 = very much; 5 = extremely]. A recent study has provided evidence for the reliability and validity of this single item measure [20].
Based on Andersen’s behavioral model [21], covariates were selected. Namely, predisposing characteristics such as sex, age, marital status (married, and living together with spouse; married, and living separated from spouse; widowed; divorced; single), and occupational status (employed; retired; other: not employed) were controlled for. With regard to enabling factors, income (household net equivalent income) was adjusted for.
With regard to need variables, self-rated health (from 1 = very good to 5 = very bad), physical functioning (subscale physical functioning of the SF-36 [22]; ranging from 0 = worst to 100 = best), depressive symptoms (15 item version of the Center for Epidemiological Studies Depression Scale [23], from 0 to 45, higher values correspond to more depressive symptoms), and the number of chronic illnesses such as diabetes or cancer (ranging from 0 to 11) were adjusted for.
In a sensitivity analysis, other factors that may affect the link between meaning in life and HCU were adjusted for, i.e. religious affiliation (protestant church (not including free churches), roman catholic church, another Christian community, another non-Christian community, no religious group), health locus of control (from 1 = I have practically no influence on my health to 4 = I have strong influence on my health) and network size (number of important people in regular contact, ranging from 0 to 9). In further sensitivity analysis, the continuous outcome measures were replaced by categorical outcome measures (please see the section dependent variables for categories).
Statistical analysis
First, sample characteristics were displayed. Subsequently, negative binomial regression analysis were conducted with GP and specialist visits as outcome measures, respectively [24]. Due to the nature of the data (count data; distribution of visits was positively skewed), we used this type of regression analysis [24,25,26]. Further details with regard to negative binomial regressions are given by Hardin et al. [24]. Meaning in life was our key independent variable. Several potential confounders were adjusted for. The criterion for statistical significance was set at p < .05. Analyses were performed using Stata 15.1 (StataCorp, College Station, Texas, USA).