The data used, were provided from the Health Survey of Nord-Trøndelag (HUNT), Norway. HUNT is considered one of the largest health surveys in the world, and is well suited to epidemiological research because of the stabile and homogenous population . The third health survey in Nord-Trøndelag (HUNT 3) was conducted from October 2006 to June 2008. Approximately 105,000 inhabitants were invited by a self-administered questionnaire, sent through the mail. The participation rate was 50 807 (49%). The HUNT study and this study were approved by the Regional Committee for Medical and Health Research Ethics (REK), Norway.
Participants and settings
For this study, participants included women and men (20–69 years) in the working age population reporting MSP over the last year (n=6702). The inclusion criteria were that participants have had pain or stiffness in muscles or joints that lasted at least 3 consecutive months, had a job, and answered “moderate”, “strong” or “very strong” on the question: “How strong has your physical pain been during the last 4 weeks?”. All participants who answered “Yes” or “No” to the question “Have you been on sick leave in the past 12 months”, were included to represent outcome variable as the work group and the sick leave group (with/without certified sick leave from doctor) in the study. The total data material consisted of N = 50 807 with a prevalence of musculoskeletal pain (MSP) or stiffness in muscles or joints that had lasted at least 3 consecutive months of 39.5% (n=20 051). The final sample included only those participants who reported “moderate” to “very strong” pain, n=6702, and those who had answered the question about sick leave. Missing were 1.6%.
In spite of the gross measure of pain, both the work group and sick leave group showed homogeneity according to reporting moderate, strong, or very strong pain. The work group reported a 78.1% occurrence rate of moderate pain, 20.1% of strong pain, and 1.8% of very strong pain, respectively 78.8%, 19.3% and 1.9% in the sick leave group. The nature of MSP, and the fact that the work group and sick leave group were homogeneous, would counteract the gross measure in sick leave.
The independent variables were selected based on the salutogenic theory , and supported empirically from the resilience research within the dimensions of personal, social, and functional resources . Personal resources were measured by a 12 items short form of the Eysenck Personality Questionnaire (EPQ) scale . The items were as follows: 1)“Are you a life of the party type of person;” 2)“Are you mostly quiet and reserved when you are around other people;” 3)“Describe yourself as you normally are;” 4)“Do you like meeting new people;” 5)“Do you like to have a lot of life and excitement around you;” 6)“Are you a relatively lively person;” 7)“Do you usually take the first step to make new friends;” 8)“Are you often worried;” 8)“Are your feelings easily hurt;” 9)“Do you often feel that you lose interest;”10)“Do you have nervous problems;” 11)“Do you often feel tired and indifferent/unmotivated without reason” and 12)“Do you worry that terrible things might happen.” The response options were “No” and “Yes.”In short, six items represented extroversion, known as positive affects, and six items represented neuroticism, known as negative emotions. The meaning variable was measured by a single item: “When something bad happens in my life, I think that is happen for a purpose,” with response options “No,” “Yes,” and “Don’t know.”
Social resources were measured by a single question of social support: “Do you have friends that can help you when you need them,” with the response options “No,” and “Yes.” Social cohesion was measured by, “Do you have friends that you can speak to confidentially,” with the response options “No” and “Yes.” Social activities were measured by six items, namely, “How many times in the last 6 months have you participated in an association or club meeting/activity – in music, singing or theatre – in parish work – in outdoor activities – in dance and in sports or worked out”. Each item had four response options from “more than “1×/week” to “never.”
Functional resources were measured as physical exercise and self-rated health (SRH). Present health status was measured by a single question health indicator, which was “How is your health now,” with four response options from “poor” to “very good?” Physical exercise was measured by one question: “How often do you exercise?” with the response options on a scale from 1–5, from “never” to “nearly every day.” Feeling strong was measured by, “Do you feel for the most part, strong and fit or tired or worn out?” The response options were on a scale from 1–7, from “very strong and fit” to “very tired and worn out.” The question was reversed so high scores indicated strength and a feeling of being fit.
Work characteristics were measured with 12 items containing different personal, social, and functional resources and were reduced by factor analysis . The items were as follows:
(1) “There is a good collegiality at work;” (2) My co-workers are there for me (support me);” (3) I get along well with my co-workers;” (4) “Does your job require you to work very fast;” (5) “Does your job require you to work very hard;” (6) “Does your job require too great a work effort;” (7) “Do you have the possibility to decide for yourself how to carry out your work;” (8) “Do you have the possibility to decide for yourself what should be done in your work;” (9) “Is your work so physically demanding that you are often physically worn out after a long day’s wor7k;” (10) “Are you bullied/harassed at work;” and (11) “Does your job require creativity.” The last item 12) was a single question only for the age range of 20–29 years “All things considered, how much do you enjoy your work.” For all questions the response options were divided into four point likert scale indicating agreement or disagreement.
The data used in this study was collected by a self administered mail questionnaire, and reported sick leave was the only question in this study administered as an interview during the clinical examinations administered by the HUNT research centre, and registered in the HUNT data bank.
The statistical analysis was first used testing for assumptions of normally distributed data to meet criteria for parametric tests. Factor and reliability analysis was used to determine the suitability of constructing scales, and composite scores of means were made when appropriate. Gender was included in the analysis.
Factor analysis for the Eysenck Personality Questionnaire (EPQ) obtained a factor solution through (direct oblimin rotation) a structure of 6 items in an extroversion (EPQ- E) scale, and 6 items in a neuroticism (EPQ-N) scale. The KMO and Bartlett’s test was .809. Both EPQ-E and EPQ-N achieved alpha coefficients well in excess of .73 and .74, respectively.
Work characteristics (12 items) were analyzed by factor analysis and obtained three component loadings (with direct oblimin rotation), which explained 62% of the variance; work support explained 26.5% , work load 19.5% and work control 16%. The Cronbach’s alpha were .869, .817, and .747, respectively. The first component work support loaded on item 1), 2), 3), and 10) explained 26.5% of the variance of work, and the second component work load loaded on question 4), 5), 6), and 9) explained 19% of the variance of work. The last work control component loaded on question 7) and 8) explained 16% of the variance of work.
The variable work support was reversed so that the high score was “strongly agree,” and the variable work control was reversed so that a high score was “yes, often.” The items 10), 11), and 9) were excluded in factor analysis through reliability analysis because Cronbach’s alpha increased from .489 to .869, from .697 to .817 and from .721 to .747 respectively, which indicate that these items are measuring something else. The item 12) had too few cases to be included in the analysis. The work support variable and work control variable were reversed so that high scores indicated greater support and greater level of work control, to ease the following analysis. All three variables were used as mean scores of multiple items. In summary, the work load component included working “hard” and “fast,” work control included deciding “what” and “how” work should be done, and work support included “well-being” and “support.”
Further, ordinal variables with more than four levels were treated as continuous, due to the large sample size. Bivariate analysis was obtained with different types of Pearson’s correlation coefficient relevant to the present level of measurement. Mann–Whitney tests were used for group comparisons (work and sick leave) because of unequal group size and violation of the homogeneity of variance assumption.
Crosstabs and chi-square statistics (Phi and Pearson) were used to analyze categorical variables relationships and differences between groups. We included only covariates that were significantly associated with SRH, and the statistical significant variables from bivariate analysis in the multivariable logistic regression model. Finally, logistic regression was used to formulate a model about health promoting resources that might determine whether a person with pain is working or being sick listed. For all analysis, a significance level of p=.01 was selected to evaluate the significance of the results. Data was analyzed using SPSS version 19.