Associations of health-related quality of life with sociodemographic characteristics, health, pain, and lifestyle factors, and motivation for changing lifestyle in adults living with chronic pain: a cross-sectional exploratory study

Study design

We conducted a cross-sectional exploratory study among adults with chronic pain reported according to the STROBE extension for cross-sectional studies [17, 18].

Setting

The Multidisciplinary Pain Center (MPC) at Naestved Hospital (Region Zealand, Denmark) is a specialised regional pain centre that consults around 1,000 outpatients with complex chronic pain conditions annually and provides multidisciplinary pain treatment delivered by physicians, nurses, psychologists, physiotherapists and a social worker.

Participants

A convenience sample of 144 outpatients referred to the MPC, who were either in-treatment or recently completed the treatment at the time of the investigation, was contacted by e-mail between December 2018 and March 2019. All participants were aged ≥18 years, had ≥3 months of pain experience before the investigation and had complex health challenges that no longer could be covered by general practice, demanding a specialised multidisciplinary intervention.

Data collection

An online questionnaire of 101 questions was developed for the study, using SurveyXact by Ramboell (Aarhus, Denmark) software. The questionnaire was pilot tested for possible inconsistencies and misinterpretations among a group of 11 outpatients and health professionals and adjusted in terms of the ease of completing the questionnaire. The questionnaire was primarily available for completion online. Two rounds of online distributions were planned and conducted. The procedure was extended with the third round of reminders to increase the response rate. Three reminders were sent by e-mail at 10 days intervals to increase the response rate. Additionally, the outpatients having an appointment at the MPC were asked if they wished to participate. The participants could complete the questionnaire either on paper or iPad, depending on their preferences. Practical support was available from trained instructors (graduated occupational therapists), both in-person in the waiting room at the MPC, by phone or by e-mail.

Variables

Analysis

We conducted the descriptive statistical analysis, estimating means and standard deviations (SD) for continuous variables and frequency distributions for categorical variables. We described and compared the mean of the EQ-5D Index and the proportions of specific HRQoL-related problems in the participants with the Danish normative population means [31] using a two-sample t-test.

We applied the following rules to the tests of the association between HRQoL and health, pain and lifestyle factors, as well as motivation for changing lifestyle: a) For all categorical variables, categories with observations n<5 were collapsed with the next category, e.g. extreme and severe levels; b) ‘Do not know’-answers were recoded as indecisive answers to the middle position on the five-item scales (‘neither good/nor poor’) and included in the analysis.

We included all the independent variables in a multiple linear regression model because of their sufficient clinical relevance for HRQoL in people living with chronic pain and practical relevance for the research objectives of this study [32]. In the analysis, dichotomised EQ-VAS score (VAS 0–50/51–100) was used for ‘poor’ or ‘good’ health status (independent variable, binary). All the regression models contained covariates gender (male-female) and age (continuous). Finally, we explored possible significant associations between the specific EQ-5D domain scores, i.e. HRQoL-related problems in the particular life areas (categorical) and health (EQ-VAS as the dependent variable, continuous) using multiple linear regression analysis. We tested the assumptions underpinning multiple linear regression analysis visually by scatterplots that showed a linear relationship between HRQoL and the independent variables, and Q–Q plots, and numerically by Shapiro–Wilk W test that showed normal distribution of the residuals. Inspection of Variance Inflation Factor (VIF) values for the independent variables was conducted to check their multicollinearity. The residuals’ variance homogeneity (homoscedasticity) was inspected by plotting the residuals vs. fitted (predicted) values. We reported the multiple linear regression analysis results as poorer or better health, based on the beta (β) coefficients.

We analysed the association of motivation (a binary variable ‘very highly to highly motivated’ or ‘other’, e.g. ‘moderately motivated’ to ‘very low motivated’ or ‘no opinion’) with gender and age (grouped) using a logistic regression model. We inspected the logistic regression model using post-estimation tools, such as calculating predicted residuals, estimating predicted margins and the Pearson χ2 test for goodness of fit. The logistic regression analysis results were reported in odds ratios as predicted difference in motivation level, accounted for gender and age.

We evaluated statistical significance at the 95%-confidence level (95% CI), p<0.05, for all the tests. We used Stata/IC 16.0 (StataCorp, College Station, TX, USA) for the data analysis.

Descriptive analysis

Lifestyle factors and motivation for changing lifestyle

Sixty-four percent of the participants were overweight, obese or severely obese (Table 2). Forty-seven percent were current smokers, while none had a high-risk alcohol consumption and 73% reported no alcohol consumption at all. Forty-three percentof the participants did not reach the World Health Organisation’s recommended minimum of 150 min of moderate physical activity per week for adults. The observed frequencies of sedentary lifestyle (43%) and obesity (42%) were higher compared to the general adult Danish population, comparable with the study cohort at gender and age distribution (17% obese and 29% inactive in 2017) [20, 33, 34]. Many participants evaluated their physical fitness (66%) and sleep quality (68%) ‘poor to very poor’, while most placed their eating habits at ‘neither good/nor poor’ (58%) and stress at ‘neither low/nor high’-level (41%). However, the majority of the participants (72%) considered lifestyle as an important or very important factor for good quality of life, independently of gender (p=0.12) or age group (F=0.20). Ninety-two percent expressed moderate to very high motivation for changing lifestyle. Multiple (n≥2) metabolic risks in various combinations such as BMI≥30, sedentary lifestyle, unhealthy eating, poor physical fitness, poor sleep quality, high stress level and cigarette smoking were found in 58% of the participants.

Health and HRQoL

Most participants (62%) experienced rather poor health (EQ-VAS score 0–50) and moderate-to-extreme levels of the problems (scores≥3) in pain/discomfort (94%), usual activities (81%) and mobility (61%), see Figure 1.

Figure 1:

The proportions of HRQoL-related problems in mobility, self-care, usual activities, pain/discomfort and anxiety/depression in the study sample.

The two-sample t-test showed that the mean of EQ-5D Index in the study sample differed from the general Danish adult population norms calculated on the sample, comparable with the participants in this survey (n=15.700; age mean 47; SD 16; range 20–79). Thus, the sample had significantly lower HRQoL values than the general population (p=0.000; diff. −0.492 (95% CI −0.520; −0.464)). Maximal cumulative HRQoL-scores observed in the study cohort (EQ-5D Index score max.=0.824) were below the mean value for the general population (EQ-5D Index score mean=0.889; SD 0.154). As in this study sample, the most frequent problem reported by the general Danish population was pain/discomfort experienced at the moderate-to-extreme level by 19–52%, while amongst Danes in general, only 7–35% reported the moderate-to-extreme level of problems in usual activities, and 2–25% – in mobility [31].

The observed associations

Associations between specific HRQoL-related problems and health

The multiple linear regression model (p=0.002; 121 observations), including HRQoL-related problems in mobility, self-care, usual activities, pain/discomfort and anxiety/depression adjusted for gender and age, explained 24% of the variances in health. The health score was significantly poorer in those with serious-to-extreme HRQoL-related problems related to usual activities (β=−0.328, p=0.030), holding other variables constant (Table 4).

The residual plots revealed linear relationships between the dependent and independent variables. Q–Q plot and the Shapiro–Wilk W test (p=0.56) delivered no reason to reject the assumption of normally distributed residuals could. The residual-vs.-fitted plot illustrated the homogeneity of the variance. No concern of multicollinearity arose from the tolerance values for the model variables >0.1 and VIF=2.16.

Implications

• – Pain intensity and sleep quality are important factors to consider when assessing the quality of life in adults living with chronic pain;

• – Addressing problems in usual activities, such as work, study, housework, family and leisure, are essential for the improvement in the health status of the chronic pain population;

• – Further research is needed to clarify if a comprehensive multidisciplinary lifestyle-oriented intervention program would improve health and quality of life in adults living with chronic pain.

Limitations

A larger study sample might have demonstrated further associations between HRQoL and other independent variables than pain intensity and sleep quality. Despite the assistance offered by the researchers, only a few participants took contact and received support. More supportive outreach research activities and fewer questions might potentially have helped to avoid missing data. Selection bias towards the participants capable of completing the survey regarding internet access, time and sufficient Danish language, might have been present. Furthermore, a greater number of men might also have affected the results of the analyses of differences between men and women. Therefore, these results need to be cautiously interpreted. The cross-sectional study design and self-reported retrospective data may also contain recall bias, misreporting and social desirability bias, particularly in lifestyle-related data. Anyhow, the support delivered by the research team during the survey distribution and conduct allowed us to reach out to the outpatients who may generally find participating in research activities difficult, which may have strengthened the generalisability of the current investigation.