Positive psychological well-being predicts lower severe pain in the general population: a 2-year follow-up study of the SwePain cohort

Based on a large Swedish population-based cohort study, [26, 27] this study assesses biopsychosocial aspects of pain in a sampling frame gathered from the Swedish Total Population Register. No longitudinal analyses have been reported for this cohort. Baseline data (T0) were collected using a representative stratified random sample of 34,000 individuals from the general population in south-eastern Sweden. The random sampling was stratified by gender and municipality to reach individuals living in urban and rural areas [26]. Data were collected by Statistics Sweden. The selected individuals received a postal questionnaire in March 2013, which could be returned either by post or electronically. A reminder was sent to non-responders after 2 weeks; if necessary, another reminder was sent 2 weeks later. The collection of questionnaires ended in May 2013. Follow-up data (T1) were collected 2 years later. Only individuals who completed and returned the first questionnaire were eligible to participate in the follow-up assessment. Eligible individuals received a postal survey in March 2015, which could be returned by post or electronically. Two reminders were sent. The collection of the follow-up data ended in May 2015. Both surveys at T0 and T1 included the same questions. The survey for this study is described below. Completion of the postal survey was deemed to be the agreement of participant’s informed consent. The study was approved by the local ethics committee of Linköping University, Sweden (Dnr: 2011 72/31), and this research was conducted in accordance with the Declaration of the World Medical Association.

For the present longitudinal study, inclusion criteria were age between 18 and 85 years and individuals who answered at both T0 and T1 regardless of their pain status (Fig. 1). Chronic pain (CP) was defined by a single question regarding the presence and duration of pain: “Do you usually have pain?” Three response options were available: (1) no; (2) yes, with less duration than 3 months; and (3) yes, with duration of more than 3 months. Individuals who answered 3 were classified as CP, and individuals who answered 1 or 2 were classified as no CP. Hence, this study had two sub-cohorts: sub-cohort 1—individuals without CP at baseline irrespective of their CP status at follow-up; and sub-cohort 2—individuals with CP at baseline irrespective of their CP status at follow-up (Fig. 1). This strategy allowed us to include those individuals who shift from no pain at T0 to CP at T1 and vice versa and to examine whether the associations for these two sub-cohorts differ.

Reporting of the results from this study was done in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement (see Additional file 1) [28].

Two indicators of psychological well-being within the hedonic perspective were considered—positive well-being and life satisfaction.

All statistical analyses were performed using IBM SPSS Statistics (version 23.0; IBM Inc., New York, USA). Two-sided statistical tests were used and p < 0.05 was regarded as significant. All analyses were stratified by the two sub-cohorts—participants without CP at T0 (CPG = 0) and participants with CP at T0 (CPG > 0). Distributions and descriptive statistics were examined for all variables at both T0 and T1. Means and standard deviations (SDs) for continuous variables and frequencies with percentages (n; %) for categorical variables were calculated. The associations of the baseline predictor variables (positive well-being and life satisfaction) with the outcome variable (pain severity as derived by the five ordered CPGs) were analysed using ordinal regression under a Generalized Linear Model (GLZ), which allows the outcome variable to have an ordinal distribution [43, 44]. GLZ is a flexible generalization of ordinary linear regression and can be used to analyse data with binary, discrete, or continuous outcomes [44].

We used GLZ with an ordinal distribution and a cumulative logit link function. The statistical significance of the model was examined by the Wald test. [44] Multicollinearity among covariates was estimated through variance inflation factor (VIF) with a cut-off score of > 2 as an indicator of multicollinearity [45]. Because the VIF was < 2, all covariates were included in the analysis. We present three models: one unadjusted in which crude odds ratios (ORs) with corresponding 95% confidence intervals (CIs) were calculated for both baseline predictors studied (i.e. well-being and life satisfaction) in a simultaneous model; one adjusted in which both baseline predictors studied were adjusted for the nine covariates, as previously described, excluding the baseline CPGs; and one fully adjusted model including CPGs at baseline in which both baseline predictors studied were adjusted for the nine covariates and the baseline CPGs as well. However, the fully adjusted model cannot be applied to sub-cohort 1 since everyone had a CPG = 0.

Finally, when significant associations between the baseline covariates and pain severity as measured by CPGs were found, a multiple post hoc sensitivity analysis was additionally performed to show how the interaction between well-being and life satisfaction depended on these selected participant characteristics (age, sex, county of birth, marital status, education, employment, household income, multimorbidity, and sleep problems in the association with pain severity, i.e. weaker vs. stronger).

The flowchart of the sample selection is depicted in Fig. 1. At T0, 15,563 individuals (46% men, 54% women) completed and returned the questionnaire, a 46% response. Of these, 11,386 (74%) completed the survey at T1. The dropout analysis showed that the response rate at T0 was lower among men, singles, and participants born abroad, and the response rate at T1 declined for younger ages, men, singles, secondary educated, unemployed, having lower household income, multimorbidity, higher functional impairment, being severely distressed, and born abroad (Table 1). Of the 11,386 individuals who participated at both T0 and T1, 2025 were further excluded because of missing data on the CPG questionnaires at either T0 or T1. Hence, 9361 individuals were included in this longitudinal study. The sub-cohort 1 and the sub-cohort 2 consisted of 5693 and 3668 participants, respectively (Fig. 1).

The total sample consisted of 4266 men (45.6%) and 5095 women (54.4%) and the mean age was 52.5 (SD = 17.5) years. In sub-cohort 1, the mean age was 51.0 (SD = 18.3) years and 49.6% were women. In sub-cohort 2, the mean age was 54.7 years (SD = 8) years and 62.0% were women. In both sub-cohorts, women were younger than men. Table 2 illustrates the baseline characteristics for the two sub-cohorts stratified by the three nominal categories of the well-being (severe distress, moderate distress, and positive well-being) and by the two nominal categories of the life satisfaction (satisfied and not satisfied).

Percentages of well-being, life satisfaction, and CPGs were fairly stable over time for the two sub-cohorts (see Additional files 2 and 3). In sub-cohort 1 without CP at baseline, the overall proportions of positive well-being and life satisfaction were highest among CPG-0 at T1 (see Additional file 2). In sub-cohort 2 with CP at baseline, the report of severe distress and not satisfied was more noticeable for those with CPG I and II at both time points (see Additional file 3).

The unadjusted and adjusted ordinal regression analyses for both sub-cohorts are presented in Table 3. In sub-cohort 1, the adjusted analysis (i.e. corrected for the nine baseline potential confounders) showed that participants with positive well-being had significantly lower cumulative ordered odds of being in a higher CPG grade. Life satisfaction, conversely, was not statistically associated with CPGs in sub-cohort 1 (Table 3). Furthermore, in the adjusted analysis, we found that being a woman, multimorbidity, and sleep problems were significantly associated with CPGs at T1 (see Additional file 4; Sub-cohort 1). The sensitivity analyses revealed that the association between positive well-being and pain severity is moderated by being a woman (p = 0.029), multimorbidity (p < 0.001), and sleep problems (p = 0.004).

In sub-cohort 2, the adjusted analysis (i.e. corrected for the nine baseline potential confounders) showed that participants with positive well-being and moderate distress, with severe distress as a reference, had significantly lower cumulative ordered odds of being in a higher CPG grade. In this model, life satisfaction was non-significant (Table 3). The fully adjusted model including CPGs at baseline showed that participants with both positive well-being and life satisfaction had significantly lower cumulative ordered odds of being in a higher CPG grade (Table 3). Moreover, in the fully adjusted mode, we found that CPGs at T0 were associated with changes in CPGs at T1. The strongest association was found for CPG-IV compared to CPG-I (see Additional file 5).

In the fully adjusted model, we also found that being a woman, born abroad, married, university educated, multimorbidity, and sleep problems were significantly associated with CPGs at T1 (see Additional file 4; Sub-cohort 2). The sensitivity analyses revealed that the association between positive well-being and pain severity is moderated by multimorbidity (p < 0.040) and sleep problems (p = 0.016). The same analysis between life satisfaction and the above-mentioned covariates showed that none emerged as a moderator of our findings (p = 0.059) for the interaction for all covariates.

This is the first study to reveal a predictive value of positive well-being and life satisfaction in a large cohort of the general population with and without chronic pain, alongside with its prospective nature, detailed exposure data, and confounder information. Furthermore, two internationally accepted and commonly used measures of well-being [38, 39] within the hedonic perspective were chosen to address our study hypothesis [1, 2]. However, our study has some shortcomings. One limitation is that all the assessments reported here are based on self-report measurements, so it is possible that our results are overestimated. It has been reported that self-reported health measures are more likely to exhibit biased associations because they exhibit stronger associations between well-being and health outcomes than health assessments made by physicians [50]. Although we had sufficient sample sizes per nominal category of well-being and life satisfaction, the proportion of participants reported higher well-being and life satisfaction was substantially larger. Thus, a ceiling effect might be presented [51]. Furthermore, we observed that the response rate at follow-up weakened among individuals with stronger indicators of biopsychological stress, indicating a response bias or a selection bias. Finally, we did not adjust for baseline-specific chronic pain conditions (i.e. traumatic or non-traumatic neck pain, low back pain, fibromyalgia). These adjustments should be considered in future studies.

Acknowledging these caveats, our study highlights the importance of well-being not only as an additional important factor worthy to be assessed in clinical daily practice, but also as a possible new treatment option in pain populations. Although the pathways connecting well-being to positive outcomes remain unknown, there is evidence that well-being may receive its protective effect through three mechanisms [7, 16, 52‐54]. It has been proposed that it can effectively reinforce one’s physiological and biological functioning, and the effective functioning, in turn, retains the health state status in healthy populations or may benefit recovery in diseased populations [7, 16, 52‐54]. It can further promote individuals to embrace healthy habits and practices such as cessation of smoking, physical exercise, weight control, and healthy sleep habits [7, 9, 10]. Lastly, some theoretical models proposed that well-being can buffer the negative effects of stress by helping individuals embrace adaptive coping skills for stressful life events [19, 55]. However, it is not known if these three mechanisms function independently, are inter-correlated, or are simply reversed associations [1]. For this reason, future experimental studies should examine more thoroughly these pathways of positive well-being in pain populations.