Measures

The dependent variable psychological distress was measured by the Hopkins Symptom Checklist (HSCL-25) (Derogatis, Lipman, Rickels, Uhlenhuth & Covi, 1974) which was included in the postal questionnaires. The HSCL-25 is commonly used as a self-report tool for identifying common mental health problems. HSCL-25 contains 25 items, including 10 items related to anxiety (HSCL-a) and 15 items related to depression (HSCL-d) which may be used as separate subscales. Responses are given on a scale from 1 = “not troubled” to 4 = “extremely troubled”. According to convention, a mean score equal to or greater than 1.75 indicates elevated risk for clinical anxiety or depression (Sandanger et al., 1998). For both subscales (i.e., HSCL-a and HSCL-d), we used this standard cut-off value to indicate high symptom levels (Winokur, Winokur, Rickels & Cox, 1984). As anxiety and depression tend to co-occur (Pollack, 2000), we combined the scores into four categories, reflecting 0 = no symptoms of anxiety or depression, 1 = only anxiety symptoms ( ≥  1.75), 2 = only depressive symptoms ( ≥  1.75), and finally 3 = comorbid anxiety and depressive symptoms (CAD; both subscales  ≥  1.75). Cronbach’s alpha for the full scale was 0.92, and for the HSCL-a and HSCL-d subscales, alphas were estimated to 0.82 and 0.89, respectively.

The exposure variable was physical functional limitations, restricted due to data availability, to the difficulties in functioning approach. In accordance with recommendations from the WHO (2011), the “difficulties in functioning approach” (“A restriction or inability to perform an activity in the manner or within the range considered normal”) is favored over an “impairment approach” (“Any temporary or permanent loss or abnormality of a body structure or function”) to measure the extent of functional limitations (Barbotte et al., 2001, WHO, 2011). Data on functional limitations were retrieved from the interview section of the surveys and assessed by two separate items. The first item dealt with whether or not the respondent had difficulties climbing up and down one floor of stairs without any rest. The second item asked whether the respondent had difficulties carrying a 5 kg object for more than 10 m. The difficulties should have lasted six months or more, and be present even with the use of assistive devices. The response alternatives were “yes” and “no” for both items. We combined responses to the two items into one variable with the categories 0 = “no difficulties”, 1 = “difficulties walking stairs”, 2 = “difficulties carrying”, and 3 = “difficulties with both walking stairs and carrying”. For the mediation analysis, this variable was recoded to include three groups: 0 = “no difficulties”, 1 = “one difficulty”, and 2 = “both difficulties present simultaneously”.

Pain was indicated by two questions: i) “Do you have bodily pain” (yes/no) and ii) “During the past four weeks, how often have you used the following drugs?” In the latter, the use of pain medication either with or without prescription were among the choice of drugs (response alternatives: 0 = “not used over the last four weeks”; 1 = “weekly, but not daily”; 2 = “at least every week”; and 3 = “daily”. The two alternatives of using pain medication with, or without prescription were combined. We constructed a new variable based on these two sources of information (0 = neither pain nor use of pain medication, 1 = either pain, but no medication use; or no pain, but pain medication use, and 2 = pain and use of medication).

Information on sense of mastery was obtained from the postal questionnaire. Sense of mastery was measured with a five item version of Pearlin and Schooler’s (1978) Mastery Scale (i.e., the perceived constraint facet of the scale). The five statements were: “I have little control over the things that happen to me”, “There is really no way I can solve some of the problems I have”, “There is little I can do to change important things in my life”, “I often feel helpless in dealing the with problems of my life” and “I sometimes feel like a piece in the game of life” (five response alternatives: 1 = “strongly agree” to 5 = “strongly disagree”). The sum scores were used in the analyses (range 5–25). Cronbach’s alpha was estimated to 0.84 in the current sample.

Information concerning social support was a part of the interview section of the survey. Social support was measured with the Oslo Social Support Scale (OSS-3) (Meltzer, 2003). This measure contains three items reflecting different dimensions of perceived social support; “How many people are so close to you that you can count on them if you have serious personal problems?” (1 = “none”, 2 = “1 or 2”; 3 = “3–5”; 4 = “5+”), “How much concern do people show in what you are doing?” (1 = “much interest”, 2 = “some interest”, 3 = “neither much or little interest”, 4 = “little interest”, and 5 = “no interest”) and “How easy is it to get practical help from neighbors if you should need it?” (1 = “quite easy”, 2 = “easy”, 3 = “neither easy or difficult”, 4 = “difficult”, and 5 = “very difficult”). The items were summed, giving a total score ranging between 3 and 14 (Dalgard et al., 2006; Lavikainen et al., 2006).

Information on age and gender was retrieved from national registers. Age was treated as a continuous variable when included as a confounder. To explore differences over periods in the life span, the age variable was grouped into three time periods, young adult (16–44 years), middle age adult (45–66 years), and elderly (≥ 67 years). When dichotomized for the mediation analyses, to have a relatively balanced sample size in each group, the groups were defined as 16–44 and ≥ 45 years. Information on level of education was obtained from the Norwegian population register and categorized into three educational levels; low (i.e., elementary and middle school/up to 9 years of education), medium (i.e., completed secondary high school or one year of university studies) and high (i.e., holding a bachelor or higher university degree). Family size was also obtained from the Norwegian population register and had two classifications. The classifications were, either having only 2 family members or having a family size that differed from 2 members. Self-rated health was selected as a proxy measure to adjust for chronic disease. It was used as a confounder and measured by a single question: “How do you evaluate your health in general?” with five response alternatives from very good to very bad. The variable was simplified to 3 categories: 1 = “very good/good”, 2 = “neither good nor bad”, and 3 = “bad or very bad”.