The correlation between illness perception, pain intensity and quality of life in elderly with low back pain in Denmark: a cross-sectional study

Hypotheses

We hypothesised that IP, pain intensity and QoL would be associated, i.e., people perceiving their LBP as a greater threat to them would report more pain and lower QoL and would tend to use more treatments and more often use pharmacological treatments for LBP (as opposed to non-pharmacological treatments).

Participants

This cross-sectional survey study is based on an inception cohort from the county of Elsinore, Denmark of 640 children (330 females and 310 males) who, in 1965, as 14-year-old pupils had their history of LBP registered by the local school doctor (Harreby et al., 1995). The follow-up study was conducted in September–October 2019. Participants now had a mean age of 68.5 (SD: 0.5) years.

Online survey

The survey was adaptive based on how each respondent answered. As around 30 questions were added specifically to this survey (and not part of the original cohort survey), comprehension of the questions was evaluated on five citizens with LBP, by means of cognitive interviewing using the think-aloud technique (Willis, 2004). Afterwards, to test functionality and completeness of the questionnaire, all questions as well as information material were pilot tested in 25 citizens between 60–70 years with 60% having LBP.

Data were collected via REDCap, as previously described in Ginnerup-Nielsen et al. (2021). To minimize the risk of incorrect use and false (double) responses, the survey was sent through the public “Digital Post” system (electronic mailbox for letters from Danish authorities) administered by an online platform “e-Boks” (e Boks, 2021), linked to the individual’s personal identification number. Access to e-Boks is mandatory for people who do not have special needs. Around 90% of 60–70 year old Danes have access to e-Boks (Danmarks Statistik, 2017).

Participants were sent an e-letter, with a link to an online survey that was managed via REDCap (Research Electronic Data Capture) electronic data capture tools hosted at the capital region of Denmark. REDCap is a secure, web-based software platform designed for data collection in research studies (Harris et al., 2019; Harris et al., 2009). The letter included information about the study and questionnaire as well as the possibility to withdraw from the study at any time. Participants were encouraged to respond to the questionnaire whether they had LBP or not. All subjects gave informed consent when initiating the survey. In case of incomplete or non-response, two reminders were sent within six weeks.

Ethics approval

The Regional Health Research Ethics Committee of the Capital Region of Denmark reviewed the outline of this cohort study. The committee deemed the study exempt from approval (Reference no.: H-20012468) as it was only based on questionnaires. Such studies can be implemented without permission from the Health Research Ethics Committee according to Danish legislation (Committee Act §1, paragraph 1).

Variables and outcome measures

The questionnaire included a maximum of 52 questions for people reporting no LBP and 105 questions for people reporting LBP based on the two initial triage questions: “Have you ever had low back pain”—With response options: “yes” and “no” and “How many days have you had low back pain within the last 12 months?” with response options: “0 days”, “1–7 days”, “8-30 days”, “more than 30 days, but not daily” or “daily”. Analyses in this study are based on people who reported LBP at least 1–7 days within the last year.

The rest of the questions concerned LBP intensity—with response options: “mild”, “moderate”, “severe” and “fluctuating” and use of medications and treatments for LBP. It furthermore concerned earlier back pain related illnesses and examinations or surgeries as well as IP (related to LBP), health related QoL, musculoskeletal health, fitness and physical function, lifestyle, and demographics.

The brief illness perception questionnaire

The Danish version of The Brief illness perception questionnaire (B-IPQ) was used to assess illness perceptions (Broadbent et al., 2006a). The B-IPQ is a generic questionnaire developed as a short version of the original 84-item revised illness perception questionnaire (IPQ-R) (Moss-Morris et al., 2002) and assesses illness perceptions based on the five dimensions: Identity, Cause, Timeline, Consequences and Cure-Control (Broadbent et al., 2006a).

The B-IPQ contains nine items. The first eight items are scored on a numeric rating scale (1–10) with 1 being no perceived threat (e.g., no affect at all) in items 1, 2, 5, 6 and 8 and highest perceived threat (e.g., no control at all) in items 3, 4 and 7. Item 9 is a free text field in which the respondent can formulate their beliefs about their condition. Both the English and Danish versions of the B-IPQ are available online (Broadbent et al., 2006b) and the wording of the B-IPQ version used in this study is presented in Table 1. We only present analyses based on the questions 1-8.

We replaced the word “illness” with “low back pain” in all items which is recommended when using the B-IPQ in specific conditions (Broadbent et al., 2006a). Furthermore, we changed the wording in item 4 from “How much do you think your treatment can help your low back pain?” to “How much do you think treatment can help your low back pain?” as our respondents were not necessarily being treated for their LBP.

Although an overall score can be calculated from the B-IPQ questionnaire (Løchting et al., 2013), items are usually presented separately. Accordingly, analyses in this study were based on each item score.

EQ-5D (3L) and EQ-VAS

We used the EuroQoL 5 Domain (EQ-5D) questionnaire with three response options to assess health related QoL. The EQ-5D is a standardised measure of health status that provides a simple, generic health measure. It is simple to use and has been validated in a wide range of health conditions and languages (Rabin & Charro, 2001). The EQ-5D covers questions within five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. Response options are given via three Likert boxes ranging from being in perfect health (e.g., I have no problems in walking about) to being seriously affected (e.g., I am confined to bed). From the responses, an EQ-5D index is calculated ranging from −0.624 (worst rated health) to 1.000 (best rated health) (Lauritsen, 2007).

The EQ-5D also contains an EQ visual analogue scale (0-100) called EQ VAS. The scale is used to record the respondent’s self-rated health with endpoints labelled “the best health you can imagine” and “the worst health you can imagine” (EuroQol Group, 1990). In 2009 the EQ-5D was translated into Danish and a Danish valuation set for reference was developed (Wittrup-Jensen et al., 2009).

Self-management strategies

Self-management strategies included the use of different predefined types of non-pharmacological or pharmacological treatments for LBP the last 12 months. Users were described based on treatment type: they could be users of pharmacological treatments and users of non-pharmacological treatments. Furthermore, the number of users of any kind of treatment (at least one type of either non-pharmacological or pharmacological treatment) was calculated.

A user of both pharmacological and non- pharmacological treatments was included in both groups.

Use of pharmacological treatments within the last 12 months was defined by a positive response to the question: “Have you taken painkillers because of low back pain within the last 12 months (Tablets, capsules, suppositories, oral solution, injections)?”.

Use of non-pharmacological treatments included receiving either: physiotherapy and/or chiropractic’s specifically for LBP within the last 12 months and/or doing low back pain specific exercise at least weekly.

As a reference from the general practitioner (GP) is needed to access physiotherapy or receive prescription drugs we decided not to include visit to the GP as a self-management strategy.

Statistical analyses

This study was designed as a descriptive cross-sectional study.

The included population was described regarding to sex, age, LBP intensity, disease management (pharmacological/non-pharmacological), B-IPQ scores and EQ-5D-scores. As data generally did not follow a normal distribution, medians, and interquartile ranges (IQR) were presented.

Correlations between IP (B-IPQ) and QoL (EQ-5D and EQ-VAS) were explored using Spearman’s correlation. Given that there is no consistent interpretation of the correlation coefficient, interpretation was based on rule of thumb: Correlations below 0.3, between 0.3 to 0.6 and greater than 0.6 were considered low, moderate, and high, respectively (Andresen, 2000; Overholser & Sowinski, 2008).

Associations between pain intensity and IP were explored via Kruskal-Wallis one way analysis of variance (ANOVA) and presented in boxplots.

Cluster analysis

To investigate how eventual IP based profiles would differ in terms of LBP intensity, QoL and use of treatments, a cluster analysis (CA) based on the B-IPQ-item scores was performed and the distribution of users of pharmacological or non-pharmacological treatments as well as the levels of pain intensity and QoL (EQ-5D and EQ-VAS) in the resulting clusters were explored. The CA was performed using the 2-step procedure recommended by Clatworthy et al. (2007) and described in Ginnerup-Nielsen et al. (2021). First, we performed a hierarchical analysis using Ward’s method with squared Euclidean Distance based on a reformed agglomeration schedule to determine the optimal number of clusters, then, we used the centroids and the number of clusters from the analysis as a starting point in a K-means CA.

χ test (with Fisher’s exact test for expected counts <5) was used to assess differences in cluster distributions of males/females and users of pharmacological/non-pharmacological treatments. Kruskal-Wallis test for non-parametric independent samples was used to assess cluster differences in age, QoL and pain intensity.

All analyses were performed in SPSS (version 3.3). All P-values and 95% confidence intervals were two-sided. Statistical significance was set at an α level of 0.05.

Statistical tests were reported with P-values for standard hypothesis tests and claims of statistical significance were only intended for exploratory purposes.

Brief illness perception scores

At item level, the lowest scores were seen on item 8 (emotional representation): “How much does your low back pain affect you emotionally?” with a median of 2 (IQR 1.0−3.5) and the highest scores were seen on item 2 (Timeline): “How long do you think your low back pain will continue?” with a median of 9.0 (IQR 3.5-10), indicating that, despite heterogenic responses, participants generally expected their LBP to continue for very long but were not considerably emotionally affected by it.

On the item 3 (personal control): “How much control do you feel you have over your low back pain?”, participants scored a median of 7.0 (IQR 5.0−9.0) while the median score on item 4 (treatment control): “How much do you think treatment can help your low back pain?” was 5 (IQR 3.0−8.0), indicating that participants generally felt in control of their pain, but did not necessarily trust in treatment to have an effect although scores also demonstrated heterogeneity between respondents (Table 2).

Quality of life (EQ-5D and EQ-VAS)

The median EQ-5D score was 0.824 (IQR: 0.773−1.000) while the median EQ-VAS was 80 (IQR: 66–88.5) indicating that participants’ QoL was generally mildly affected according to age adjusted Danish population norms (Sørensen et al., 2009).

Correlations between illness perception items

Analyses of the interrelationships between the B-IPQ items showed moderate to high statistically significant correlation coefficients between the perceived severity of LBP symptoms (item 5, Identity) and consequences of LBP (item 1, Consequences) (r: 0.76). Hence, participants with more severe symptoms perceived their LBP to have more consequences for their lives. Furthermore, participants who were more concerned about their LBP (item 6, Concern) were more emotionally affected (item 8, Emotions) (r: 0.72), they also perceived more consequences to their LBP (item 1, Consequences) (r: 0.69) and had lower QoL (EQ-5D) (r: −0.62).

Moderate statistically significant correlations with coefficients ranging from −0.32 to 0.59 were seen for the rest of the B-IPQ items, except from a low correlation of (r: 0.15) between item 2 (Timeline) and item 7 (Coherence) and (r: 0.23), and item 2 (Timeline) and item 8 (Emotions). Generally, low and mostly non-statistically significant correlations were found between item 4 (Treatment control) and all other B-IPQ items, indicating that participants’ perceptions about possible effects of treatment were not related to other perceptions (Table 3).

Correlations between illness perception and EQ-5D

High, statistically significant correlations were found between the B-IPQ item 1 (Consequences) and the EQ-5D (r: −0.62). Low to no correlation was seen between B-IPQ item 4 (Treatment control) and any other health related outcome measure and low to moderate correlations were found between all B-IPQ items and EQ-5D and EQ-VAS (rs: 0.27 to −0.53). Correlations indicated that the more threatening participants perceived their LBP, the lower QoL they reported (Table 3).

Associations between illness perception and pain intensity

We found all IP items, except from item 4 (treatment control) (P = 0.06), to be associated with pain levels (P < 0.01), with higher pain levels being associated with perceiving LBP as a greater threat. Participants reporting fluctuating pain patterns perceived their LBP almost as threatening as participants reporting severe pain (Fig. 2).

Cluster analysis based on illness perception scores

Based on the change between coefficients in the agglomeration schedule from the hierarchical CA and the subsequent K-Means analysis, we found that two illness perception clusters would be the optimal clustering solution.

In the two-cluster solution we found statistically significant cluster differences in all B-IPQ item scores, except item 4, as well as in EQ-5D and EQ-VAS scores (Table 4).

Cluster 1 contained 146 participants while Cluster 2 contained 67 participants. In Cluster 1, 41 (28.1%) had experienced LBP 1-7 days the last year while 38 (26.0%) had daily pain, 48 (32.9%) characterized their pain as mild, while 9 (6.2%) characterized their pain as severe. Cluster 1 reported lower quality of life than Cluster 2 with a difference between medians of −0.176 (95% CI [−0.233–−0.119]) on the EQ-5D.

Cluster 1 generally perceived LBP as a greater threat than Cluster 2. Cluster 1 was more concerned (item 6) (difference between medians: 2 (95% CI [1.45–2.55])) and perceived LBP to cause more symptoms (item 5) and have more consequences (item 1) than Cluster 2 with differences between medians of respectively: 3 (95% CI [2.37–3.63]) and 2 (95% CI [1.66–2.34]). Respondents in Cluster 1 also expected their LBP to last much longer than respondents in Cluster 2 (item 2) (difference between medians: 8 (95% CI [7.18–8.82])) and perceived themselves to be in less control of their pain than respondents in Cluster 2 (difference between medians: −3 (95% CI [−3.63–−2.37])). No statistically significant difference was found between clusters concerning their confidence in the effect of treatment for their LBP (item 4).

In Cluster 2, 48 (71.6%) had experienced LBP 1-7 days the last year while none reported daily pain, 38 (56.7%) characterized their pain as mild while 3 (4.5%) characterized their pain as severe. Hence, the population in Cluster 2 had experienced fewer days with LBP within last year and markedly fewer respondents reported severe or fluctuating pain patterns than in Cluster 1.

There was a statistically significant difference between clusters concerning use of pharmacological treatments, as 55 (37.7%) in Cluster 1 had used pharmacological treatment within the last 12 months and 20 (13.7%) reported daily use. In Cluster 2, 10 (14.9%) had taken pharmacological treatment within the last 12 months (P = 0.001) and only one (1.5%) reported daily use. No statistically significant difference was found between clusters in the use of non-pharmacological treatments or in the proportions of users of any kind of treatment. The proportions of women in the 2 clusters were 83 (56.8%) in Cluster 1 and 35 (52.2%) in Cluster 2 (Table 4).