Abstract
Objectives
Low-grade inflammation is a possible contributing factor in the development and persistence of chronic primary pain syndromes. Related to inflammatory activity is sickness behavior, a set of behavioral responses including increased pain sensitivity, fatigue, malaise, fever, loss of appetite, as well as depressive behavior and anhedonia. To capture these behavioral responses and their relation to longstanding pain, psychometrically sound self-report questionnaires are needed. The Sickness Questionnaire (SicknessQ) was developed to assess self-reported sickness behavior based on studies on acute immune activation while maintaining relevance for persistent conditions. The aim of the current study was to evaluate aspects of the validity and reliability of the SicknessQ in a Swedish sample of persons with longstanding pain.
Methods
Aspects of construct validity were evaluated by means of performing a confirmatory factor analysis (CFA) (testing structural validity) and by relevant hypothesis testing i.e., that ratings of sickness behavior in combination with other related factors (e.g., depression and anxiety) would be significantly related to ratings of avoidance. Reliability was evaluated by means of analyzing the internal consistency of items.
Results
Following the CFA, a non-significant Chi-Square test (χ2 [32, N=190] = 42.95, p=0.094) indicated perfect model fit. Also, the relative fit indices supported adequate model fit (CFI = 0.978; TLI = 0.969; RMSEA = 0.0430). Sickness behavior (p<0.0001), depression (p<0.05) and pain duration (p<0.05) significantly contributed to the regression model, explaining 45% of the total variance in avoidance. Internal consistency was adequate, as indicated by a Cronbach’s α value of 0.82 for the entire questionnaire.
Conclusions
Results indicate that the SicknessQ has adequate structural validity as well as adequate internal consistency, and is significantly associated with avoidance. The SicknessQ appears to have utility as a self-report questionnaire to assess symptoms of sickness behavior for adults with longstanding pain.
Introduction
About one in five adults report that they experience longstanding pain [1, 2] and a substantial proportion of these persons also experience depression, anxiety, insomnia, and pain-related disability [1]. A number of studies focusing on factors central to the development and maintenance of chronic primary pain, pain syndromes best understood as health conditions in their own right [3], point to the possible contributing role of low-grade inflammation for some of these syndromes [4]. For example, several studies argue that inflammatory cytokines play a significant role in understanding fibromyalgia [5].
Related to inflammatory activity is the sickness response and sickness behavior, comprising several responses prompted by inflammatory activity, such as increased pain sensitivity [6], fatigue [7], malaise [8], fever, loss of appetite, as well as depressive behavior and anhedonia [9]. These responses have evolved to enable recovery from an acute illness. However, when symptoms persist they may actually drive behaviors that contribute to longstanding illness [9]. From a behavioral analytical standpoint, symptoms of the sickness response, can from our perspective be viewed as motivational factors antecedent to avoidant behaviors primarily aimed at short term gains, e.g., alleviating pain by means of resting. Related motivational factors include cognitions (e.g., “I can’t stand this pain”) and associated emotional aspects (e.g., low mood), further amplifying the situational motivational state driven by the sickness response [10, 11].
Avoidance, or avoidant behavior, is a central treatment target in cognitive behavior therapy (CBT) and Acceptance and Commitment Therapy (ACT) for longstanding pain, and as a potential process of change in treatment, it appears important to more closely study avoidance in relation to the antecedent motivational factors driving this behavior, such as sickness responses. Tentative results from one recent study supports this notion, indicating that persons with longstanding pain who had higher levels of inflammatory markers at baseline had significantly lower improvements in psychological inflexibility, a construct closely linked to avoidance, following treatment [12]. In order to facilitate research evaluating the role of sickness behavior in relation to longstanding pain and avoidance, as well as in regard to processes of change and efficacy in relation to behavioral treatments, psychometrically sound self-report questionnaires are needed as an adjunct to analyses of inflammatory markers.
The Sickness Questionnaire (SicknessQ) was developed to assess the cardinal symptoms of perceived sickness to be used in both experimental and clinical research. More specifically, the questionnaire aims to capture self-reported sickness experiences such as: The desire to be still, alone and inactive; symptoms of soreness, shakiness, nausea, headache; as well as the experience of feeling depressed and drained [13]. The original validation study illustrated that a one-factor solution with 10 items provided the best fit and that the SicknessQ had adequate internal consistency and was adequately and significantly associated with criteria variables. Two different samples of persons were included in development of the SicknessQ. Blinded healthy participants were injected with endotoxin or placebo to provoke a sickness response to test the items’ sensitivity to inflammatory activation and also an evaluation of the instrument was performed in a larger sample of primary care patients [13]. Two more validation studies of the SicknessQ have been conducted afterwards. A Chinese validation study with clinical and non-clinical participants [14] indicated that a nine-item two-factor-solution of the SicknessQ was a valid and reliable measure of sickness behavior in a Chinese context. Recently, an English validation study with Australian adults with persistent musculoskeletal pain and/or gastrointestinal symptoms displayed that the SicknessQ total score had an adequate model fit [15].
To further facilitate research evaluating the role of aspects of the sickness response in relation to longstanding pain, and assess the SicknessQ’s potential use in this area, evaluating the construct validity and reliability of the SicknessQ for persons with longstanding pain is an important step. In the current study we chose to validate the SicknessQ for patients with longstanding pain given that the questionnaire was developed in Swedish, is brief and aims to capture core symptoms of sickness behavior, both in acute inflammatory states as well as in more persistent conditions.
Aims
The overarching aim in the current study was to evaluate aspects of the construct validity and internal reliability of the SicknessQ in a sample of persons with longstanding pain using the conceptual framework of Mokkink et al. [16]. More specifically we wanted to: (1) Evaluate the structural validity by performing a confirmatory factor analysis (CFA); (2) test the hypothesis that sickness response symptoms, depression, anxiety, self-rated health, pain intensity, and pain disability are significantly related to avoidance (using correlational and regression analyses) [16]; (3) evaluate the internal consistency by means of analyzing Cronbach’s α.
Materials and methods
Participants and procedure
Patients were referred from primary and tertiary care units in Stockholm County to the Behavioral Medicine Pain Treatment Unit at the Karolinska University Hospital between March 2009 and June 2013. The recruitment procedure was consecutive and patients were eligible for study inclusion if they were ≥18 years of age; presented with longstanding pain (≥6 months); could fill out the self-report questionnaires independently in Swedish. The procedure was to ask all patients eligible for study inclusion if they were interested to participate in the study.
Demographics and clinical data
Clinical interviews and self-report questionnaires were used to gather data in the current study. Data on age, gender, pain duration, medication, pain localization, and pain diagnoses were collected at the first visit by a pain physician. Pain physicians diagnosed pain symptoms based on criteria in the International Classification of Diseases, 10th revision (ICD-10). Unspecified pain diagnoses, e.g., ‘Pain or ache, unspecified ’and ‘Chronic pain without known cause’, were classified as Unspecified pain in Table 1. Pain diagnoses that were not possible to categorize retrospectively without more specific knowledge of the factors surrounding the origin of pain, e.g., Lumbago and Myalgia, were classified as Unclassifiable pain in Table 1.
Participant characteristics.
| Characteristics | n (%)/M (SD) | Range (min-max) | n |
|---|---|---|---|
| Demographic | |||
| Female | 149 (78.4%) | 190 | |
| Age, years, M (SD) | 41.0 (13.5) | 68 (18–86) | 190 |
| Pain localizations a | |||
| Neck | 96 (52.2%) | 184 | |
| Lower extremities | 97 (52.7%) | 184 | |
| Upper extremities | 73 (39.7%) | 184 | |
| Thorax and abdominal | 59 (32.2%) | 183 | |
| Back | 88 (47.6%) | 185 | |
| Head | 80 (43.5%) | 184 | |
| Pain type b | |||
| No diagnosis | 5 (2.6%) | 189 | |
| Headaches | 16 (8.5%) | 189 | |
| Nociceptive pain | 20 (10.6%) | 189 | |
| Neuropathic pain | 16 (8.5%) | 189 | |
| Nociplastic pain | 10 (5.3%) | 189 | |
| Unspecified pain | 90 (47.6%) | 189 | |
| Unclassifiable pain | 33 (17.5%) | 189 | |
| Medication | |||
| Opioids | 55 (34.2%) | 161 | |
| Simple or non-opioid analgesics | 105 (65.2%) | 161 |
-
aParticipants could report more than one localization. bParticipants could have more than one pain type.
Sickness behavior
Sickness behavior was assessed using the 35-item version used during the development of the SicknessQ, however in the present study we only included the 10 items part of the final version of the questionnaire [13]. Items comprise the following 10 statements of sickness symptoms: (1) “I want to keep still”; (2) “My body feels sore”; (3) “I wish to be alone”; (4) “I don’t wish to do anything at all”; (5) “I feel depressed”; (6) “I feel drained”; (7) “I feel nauseous”; (8) “I feel shaky”; (9) “I feel tired”; and (10) “I have a headache”. Items are rated on a four-point scale ranging from (0) ‘disagree’ to (3) ‘agree’ and summarized to a total score of 0–30 points. SicknessQ has previous in a sample of patients in Swedish primary care (most commonly seeking consultation for acute infection, muscle and joint pain, or symptoms from airways) been shown to have adequate internal consistency (Cronbach’s α = 0.86), and is adequately and significantly (p<0.001) associated with depression (β=0.41); anxiety (β=0.36); self-rated health (β=0.28), and a single-item of feeling sick (β=0.55) [13].
Avoidance
Avoidance was measured using the avoidance subscale from the Psychological Inflexibility in Pain Scale (PIPS) [17]. The PIPS comprises a total of 12 items, and two subscales labeled ‘Avoidance’ (eight items) and ‘Fusion’ (four items). Items are rated on a seven-point Likert-scale that ranges from (1) ‘never true’ to (7) ‘always true’. Higher scores indicate greater avoidance and psychological inflexibility. Results from two previous Swedish studies illustrate that the questionnaire has adequate validity and reliability, as illustrated by Cronbach’s α of 0.89 [17] and 0.87 [18]. Regarding construct validity, PIPS subscales were significantly correlated (p’s<0.001) with both CPAQ subscales as well as the activity avoidance subscale of the TSK. Regarding concurrent criteria validity PIPS accounted for a significant amount of variance in pain, medication use, work absence, life satisfaction, pain disability, anxiety, depression, kinesiophobia, and acceptance [18].
Pain intensity
Pain intensity during the past week was assessed using a Numeric Rating Scale (NRS), ranging from 0 (‘No pain at all’) to 6 (‘Extremely painful’).
Self-rated health
Self-rated health was evaluated using the first item of the Short Form-12 Health Survey (SF-12), “In general, would you say your health is”, ranging from “Excellent” to “Poor” on a five-point scale [19–22]. Single-item assessments of subjective health perception have continually been found to predict health care use, morbidity and mortality [23].
Anxiety and depression
To assess symptoms of anxiety and depression the Hospital Anxiety and Depression Scale (HADS) was used [24]. HADS consists of two subscales measuring anxiety (HADS-a), and depression (HADS-d) and has 14 items in total, rated on a four-point Likert-scale. Results from previous studies show that the HADS adequately measures anxiety and depression in hospital medical outpatient clinic setting [24], illustrated by Cronbach’s α values of 0.84 for HADS-a, 0.82 for HADS-d and 0.80 for the total scale [25]. Correlations of the subscale scores and assessment of the level of anxiety and depression conducted by researchers were significant (p<0.001) for depression r=0.70, and for anxiety r=0.74, and indicate the use as measures of severity [24]. Regarding construct validity, strong correlation was found in a Swedish sample between total HADS and BDI (Beck Depression Inventory, aiming to evaluate the severity of depression) scores (0.73) [25].
Pain disability
Pain disability, the disabling consequences of longstanding pain on the performance of daily activities, was measured using the Pain Disability Index (PDI) [26]. In total seven items covering day-to-day activities, are rated on a Likert-type scale ranging from (0) ‘No trouble’ to (10) ‘Total disability’. Studies with the questionnaire in English support the reliability and validity of the scale, with adequate internal consistency indicated by a Cronbach’s α value of 0.86 [26, 27]. Regarding concurrent validity, persons with higher scores on the PDI also reported more psychological distress, e.g., on The State-Trait Anxiety Inventory (STAI) and BDI, as well as pain, compared to persons with lower PDI-scores, p<0.001 [27].
Statistical approach
Structural validity and hypothesis testing were used to evaluate aspects of construct validity as described by Mokkink et al. [16], of the SicknessQ. Reliability was evaluated by assessing the internal consistency of the SicknessQ. A 95% confidence interval and a conventional 0.05 alpha level were used as criteria of significance. For hypothesis testing, person-mean imputation was used for imputing single missing questionnaire items, for questionnaires other than the SicknessQ, if the number of missing items were less than 25% for the specific questionnaire. Lastly, the potential differences between included and excluded participants in regards to self-report questionnaires (excluding SicknessQ due to missing data), clinical and demographic variables, were analyzed by using a Chi-square test and Welch’s t-test.
Descriptive statistics
Descriptive statistics were used to broadly characterize the sample by calculating means, standard deviations, ranges, and frequencies of demographic (age, gender) and clinical variables (pain duration, pain type, pain localizations, and pain medication), as well as of the sum scores of the included self-report questionnaires.
Aspects of construct validity and reliability of the SicknessQ
Confirmatory factor analysis
In line with suggestions by Nunnally [28] we aimed to include at least 190 participants in the analysis. Suitability was determined by a significant (<0.05) outcome of Bartlett’s test of sphericity and a Kaiser–Meyer–Olkin (KMO) value >0.7 [29]. Moreover, a KMO value between 0.5 and 0.7 was viewed mediocre, a value between 0.7 and 0.8 good, a value between 0.8 and 0.9 great, and a value >0.9 superb [30]. Assumptions of multivariate normality were examined by calculating Mardia’s multivariate skewness and kurtosis estimates [31], with non-significant results indicating multivariate normality. We considered univariate skewness >2 and kurtosis values >7 as expressive of deviations from univariate normality [32]. Parameter estimation was conducted using maximum likelihood estimation as this approach has been shown to satisfactorily handle non‐normality [33].
When performing the CFA factor loadings above 0.5 were considered adequate [34] and in line with recommendations by Bollen and Long, several fit indices were used to evaluate model fit [35]. The absolute fit of the model was tested using the Chi-Square test in which, in the context of a CFA, a p-value >0.05, indicates a perfect model fit. Additionally, we further evaluated model fit by analyzing approximate fit indices, and we used the Comparative Fit Index (CFI; ideally >0.95), the Tucker-Lewis index (TLI; ideally >0.95) and the Root Mean Square Error of Approximation (RMSEA; ideally <0.05) [36]. Lastly, based on the assessment of statistical output, i.e., residual Chi-Square statistics >10 and on theoretical assumptions that item content was related, we allowed for residual correlations between items within a common factor, if the model fit would substantively improve.
Hypothesis testing
In order to further address construct validity, we performed a number of hypothesis-driven correlational analyses, according to Mokkink et al. [16]. In these analyses assumptions of linearity and normality were inspected using scatterplots, visual inspection in addition to analyses of skewness and kurtosis, as well as by the Shapiro-Wilk test. Absolute skewness values >1 and kurtosis values >1 were used as criteria for non-normally distributed data. In regard to regression analyses, independence and normality of residuals were evaluated using scatterplots, and QQ plots and the homogeneity of residual variance (homoscedasticity) were examined using scatterplots of residuals vs. predicted values. Independence of residuals error terms was examined using scatterplots.
To broadly characterize the relationship between the SicknessQ and demographic variables (age, gender), the clinical variable pain duration, as well as anxiety, depression, self-rated health, pain intensity, pain disability and avoidance, bivariate correlations (Pearson’s r) were calculated. We considered correlations between 0.10 and 0.29 as weak, correlations between 0.30 and 0.49 as moderate, and correlations between 0.50 and 0.89 as strong [37]. As a part of the hypothesis testing we performed a regression analysis in which only variables, in addition to sickness behavior, with significant bivariate correlations with the response variable avoidance were included in the model. This resulted in a model with avoidance as the response variable and with anxiety, depression, self-rated health, pain intensity, pain disability, and sickness behavior included as explanatory variables. Avoidance was chosen as the response variable due to its potential relation to sickness symptoms and responses (antecedent motivational factors) and is also a central treatment target in CBT and ACT.
Internal consistency
To address aspects of reliability we evaluated internal consistency by analyzing Cronbach’s α, i.e.,correlations among items in the scale, and we considered an α value ≥0.80 adequate [38, 39].
Software
All statistical analyses were performed using R version 3.6.2 [40]. To analyze descriptive statistics we used the Tableone package [41]. The MVN package [42] was used for assumption testing prior to performing the CFA, and the REdaS package [43] was used for the KMO. The lavaan package [44] was used for CFA and Maximum likelihood estimation. The Psych package [45] was used for analyzing Cronbach’s α, as well as assumptions before hypothesis testing. We used the Hmisc package [46] for correlation significance levels and the QuantPsyc package [47] for beta values. For the Backward elimination regression analysis, the MASS package [48] was used. Figure 1 was created using the semPlot [49] package.
Confirmatory factor analysis (CFA).
Results
Participants
Of the 215 individuals consenting to study participation, some were excluded from the analyses: Specifically, two had not filled in any questionnaires; data on one participant’s age was missing; and another 22 were excluded due to missing data in the SicknessQ. In total, 190 participants were included in the final dataset. There were no significant differences as regards the rated questionnaires, demographic variables and clinical variables (except pain localization ‘neck’ and pain medication) between the included and excluded participants.
The majority of the 190 included participants were female (78.4%), the mean age was 41.0 years (SD = 13.5) and the mean pain duration was 10.8 (SD = 9.7) years. Fifty-five participants (34.2%) used opioids and 105 (65.2%) used simple or non-opioid analgesics. A large proportion (47.6%) of the participants presented with unspecified pain and further patient characteristics are described in Table 1. Supplementary Table 1 displays the mean sum scores of the included self-report questionnaires, together with standard deviations.
Confirmatory factor analysis and internal consistency
In total, 190 participants completed the SicknessQ and were included in the CFA. The KMO‐test resulted in a value of 0.85 and Bartlett’s test of sphericity was statistically significant (p=0.005), showing that data were suitable for factor analysis. Univariate skewness and kurtosis values ranged from −1.081 to 0.699 and from −1.679 to 0.280 respectively, indicating that they were within the normal range. Mardia’s test of multivariate skewness and kurtosis (p<0.0001) suggested some departure from normality in regards to skewness. To adequately address these issues we performed the CFA using Maximum Likelihood Estimation with robust standard errors and a Satorra–Bentler scaled test statistic [50]. The CFA assessing a one-factor model resulted in a significant Chi-Square (χ2) result (χ2 [35, N=190] = 74.42, p<0.001), which indicated a non-satisfactory fit. Also, the resulting approximate fit indices, i.e., CFI = 0.920; TLI = 0.897 and the RMSEA = 0.077, further illustrated a non-satisfactory model fit. However, model fit improved when allowing for shared residual variance between the following items: Item (1) ”I want to keep still” and item (3) “I wish to be alone”; item (7) “I feel nauseous and item (8) “I feel shaky”; as well as between item (6) “I feel drained” and item (9) “I feel tired”. After the adjustment, the Chi-Square test resulted in non-significant p-value (χ2 [32, N=190] = 42.95, p=0.094), indicating perfect model fit. Also, the relative fit indices all improved, further supporting model fit (CFI = 0.978; TLI = 0.969; RMSEA = 0.0430). Figure 1 displays the results of the CFA. Internal consistency was adequate, as indicated by a Cronbach’s α value of 0.82 for all 10 items. Table 2 provides means and standard deviations for all items of the SicknessQ.
Internal consistency of the SicknessQ.
| Item | M | SD |
|---|---|---|
| I don’t wish to do anything at all | 1.74 | 1.07 |
| I feel depressed | 2.03 | 1.04 |
| I feel nauseous | 1.78 | 1.01 |
| I want to keep still | 1.30 | 1.05 |
| I feel tired | 1.61 | 0.98 |
| I feel drained | 1.99 | 0.99 |
| I feel shaky | 0.96 | 1.09 |
| I wish to be alone | 0.93 | 1.04 |
| My body feels sore | 2.25 | 0.96 |
| I have a headache | 1.55 | 1.28 |
-
Four-level scale ranging from disagree (0) to agree (3) with higher scores corresponding to more sickness behavior. Cronbach’s α for the scale is 0.82. SicknessQ, Sickness questionnaire.
Hypothesis testing
In total, for all included questionnaires, less than two percent of items were missing, and items were missing for all included questionnaires and person-mean imputation of the missing data points was conducted for participants with less than 25% missing data [51]. The number of imputed data points was <0.10%. Visual inspection conveyed that the data was missing at random.
The Shapiro–Wilk test illustrated that variables were normally distributed, and skewness and kurtosis were <1. Visual observation of residual QQ plots indicated normality. Likewise, scale location plots indicated homoscedasticity. There were no outliers exceeding three standard deviations and no participant data was thus removed. There were significant correlations between SicknessQ and: avoidance, anxiety, depression, self-rated health, pain intensity, pain disability, and age. Sex and pain duration were not significantly correlated with sickness behavior. See Table 3 for detailed results.
Correlations between SicknessQ and: avoidance, anxiety, depression, self-rated health, pain intensity, pain disability, age, sex, and pain duration.
| Variables | r |
|---|---|
| Avoidance (PIPS avoidance subscale) | 0.59a |
| Anxiety (HADS anxiety subscale) | 0.55a |
| Depression (HADS depression subscale) | 0.52a |
| Self-rated health (SF-12 item 1) | −0.40a |
| Pain intensity (NRS) | 0.27b |
| Pain disability (PDI) | 0.41a |
| Age | −0.27b |
| Sex | −0.00 |
| Pain duration | −0.08 |
-
n=184–190. aCorrelation is significant at the 0.0001 level (two-tailed). bCorrelation is significant at the 0.001 level (two-tailed). PIPS, Psychological Inflexibility in Pain Scale; HADS, Hospital Anxiety and Depression Scale; PDI, Pain Disability Index.
The response variable avoidance was significantly correlated with sickness behavior (SicknessQ; r=0.59; p<0.0001), anxiety (HADS-a; r=0.43; p<0.0001); depression (HADS-d; r=0.52; p<0.0001); self-rated health (SF12-1; r=−0.34; p<0.0001); pain intensity (r=0.33; p<0.0001); and pain disability (PDI; r=0.62; p<0.0001). Sex, age and pain duration were not significantly correlated with avoidance.
All the significant variables from the correlation analysis with avoidance, except pain disability (as it comprises a response variable), were included in the regression model, with avoidance as the response variable. The demographic control variables sex and age as well as the clinical variable pain duration, were included in the regression irrespective of their correlation to avoidance. In an Enter regression model, sickness behavior (p<0.0001), depression (p<0.05), and pain duration (p<0.05) significantly contributed to the model, explaining 45% of the total variance in avoidance (please see Table 4 for additional test statistics). In a Backward selection analysis, sickness behavior (p<0.0001), depression (p<0.05), and pain duration (p<0.05) significantly contributed to the model.
Regression analyses.
| Predictor variables | Estimate, B | SE | Stand. estimate (beta) | t | p-Value |
|---|---|---|---|---|---|
| Pain intensity | 1.14 | 0.59 | 0.12 | 1.93 | 0.06 |
| Anxiety | 0.10 | 0.16 | 0.05 | 0.63 | 0.53 |
| Depression | 0.43 | 0.19 | 0.18 | 2.29 | 0.02b |
| Sickness behavior | 0.68 | 0.13 | 0.44 | 5.31 | 0.00a |
| Self-rated health | −0.84 | 0.67 | −0.08 | −1.25 | 0.21 |
| Age | 0.08 | 0.05 | 0.11 | 1.60 | 0.11 |
| Sex | −0.22 | 1.55 | −0.01 | −0.14 | 0.89 |
| Pain duration | 0.15 | 0.07 | 0.15 | 2.33 | 0.02b |
-
aCorrelation is significant at the 0.0001 level (two-tailed). bCorrelation is significant at the 0.05 level (two-tailed). Adjusted R2=0.45.
Discussion
This study provides initial support for the structural validity and internal consistency of a Swedish version of the SicknessQ for adults with longstanding pain. Overall, findings show excellent model fit for a one-factor solution and adequate internal consistency, results that are in line with a previous psychometric evaluation based on data collected in a primary care setting [13]. When considering internal consistency, acceptable to very good Cronbach’s α-values range from 0.70 to 0.95, and values near the upper end of the range are recommended for clinical application [39], which, though not providing clear-cut support, indicates adequacy of the SicknessQ for clinical use in the present population. The mean average sickness score in this study (16.1) is higher than in an Australian sample with chronic medically unexplained symptoms, e.g., pain, fatigue, and distress (8.9) [15] and in the original validation article of SicknessQ based on data in a sample of patients from Swedish primary care (10.7) [13]. This was to be expected due to the severity of symptoms of the persons presented in this study. Interestingly, a recent study illustrated that patients with chronic pain (same patients as in the present study), rated levels of sickness behavior on par with those reported by individuals with experimentally induced inflammation via injection of bacterial endotoxin [8].
In this study results from the initial CFA indicated a non-satisfactory fit, but the model fit improved when allowing for shared residual variance between items as presented in the results section. In the validation of the SicknessQ in the Australian sample presented above, an analog approach was used, i.e., similar covariance terms which improved model fit substantively were included in the CFA [15]. In that study no improvement was seen in model fit when comparing the two-factor and three-factor solution with the single factor solution, suggesting that a multidimensional approach offers no advantage, why we did not explore this further in this sample.
The hypothesis testing, as part of the evaluation of construct validity, illustrated adequate correlations between the SicknessQ and other variables common in pain research. Furthermore, findings illustrated that sickness behavior was significantly related to depression, anxiety, self-rated health, pain intensity, pain disability, and avoidance. In line with our assumptions of the role of symptoms of the sickness response, as measured with SicknessQ, as motivational factors in relation to avoidance behaviors [11], these symptoms, correlated with avoidance and explained more variance in avoidance than other emotional symptoms. Pain duration and depression also significantly contributed to the regression model with avoidance as the dependent variable, indicating that these factors under certain circumstances also play a relevant role in avoidance.
Several potential avenues for further research and clinical implications arise from this study. Future studies could evaluate the SicknessQ’s usefulness to predict behavior in a systematic way, by evaluating predictive validity, using longitudinal designs in which baseline levels of sickness behavior are used as predictors of potential alterations in pain symptoms as well as pain‐related functioning over time. Connectedly, it appears relevant to investigate the questionnaire’s test‐retest reliability as well as its sensitivity to change, following for example behavioral treatment for people with longstanding pain. SicknessQ could be utilized in future clinical studies to assess sickness behavior as a potential moderator, predictor or mediator, as well as a possible outcome measure in behavioral treatments for longstanding pain. The SicknessQ was examined in a group of patients experiencing different types of pain, which speaks for the utility of the questionnaire. Since the vast majority of participants in this study are women, future studies including larger samples and more equal proportions of men and women, could based on subgroup analyses evaluate how current results stand under these conditions. Also, since the participants rated a 35-item version used during the development of SicknessQ, future studies should test the current findings using the 10-item version [13].
Some limitations need to be taken into consideration when interpreting the results of the present study. First, slight deviations from a normal distribution of data were found, which could result in misrepresentative estimates of fit indices. However, maximum likelihood estimation has been shown to produce robust estimation also in conditions of non‐normality [33]. In addition, participants were referred to a tertiary care pain clinic, which possibly confines the generalizability of the findings in comparison to patients with longstanding pain in other care settings, e.g., primary care. Also, the scarcity of more specific demographic and clinical information can be cast as a weakness. Furthermore, we did not register patients who declined participation and can therefore not conduct non-responder analyses. However, the sample is representative for patients with longstanding pain, as the distribution of age, sex and pain intensity is similar to other studies [52]. Additionally, as previously mentioned, symptoms of the sickness response were assessed using 10-items from the original 35-item version used during the development of the SicknessQ. This means that we cannot exclude the possibility that, for example, simply the number of items, or the specific order in which items were presented, may have influenced the participants’ scores. Importantly though, all questions related to sickness behavior, as well as the instructions for the test list and the final questionnaire are the same. Lastly, we did not conduct the hypothesis testing in a separate sample of persons than the sample with which we conducted the CFA.
In conclusion, results indicate that the SicknessQ has adequate structural validity and internal consistency, and is significantly associated with avoidance. The SicknessQ appears to have utility as a self-report questionnaire to assess symptoms of sickness behavior in adults with longstanding pain.
Acknowledgments
We are grateful to all the participants who have contributed to the study by answering questionnaires. We are also grateful to the clinicians at Behavioral Medicine, Karolinska University Hospital, for their invaluable assistance in facilitating the current study.
-
Research funding: Financial support was in part provided by grants from Functional Area Medical Psychology, at the Karolinska University Hospital (JÅ), as well as Stockholm County Council ALF (JÅ) (Grant number 20170152) and AFA Insurance (JÅ) (Grant number 140350), Swedish Research Council (BK) (Grant number 2017-00489), Swedish Society of Medicine (BK) (Grant number SLS-691251), Sweden-America Foundation (BK), Fulbright Sweden (BK) as well as Heart-Lung Foundation (BK) (Grant number 20190110).
-
Author contributions: All authors have accepted the entire content of this manuscript and approved its submission.
-
Competing interests: Authors state no conflict of interest.
-
Informed consent: Written informed consent was obtained from participants included in the study.
-
Ethical approval: The study was approved by the Regional Ethical Review Board in Stockholm (Dnr: 2010/662-31/3), and was carried out in accordance with the Declaration of Helsinki (DoH).
References
1. Breivik, H, Collett, B, Ventafridda, V, Cohen, R, Gallacher, D. Survey of chronic pain in Europe: prevalence, impact on daily life, and treatment. Eur J Pain 2006;10:287–333. https://doi.org/10.1016/j.ejpain.2005.06.009.Search in Google Scholar PubMed
2. Johannes, CB, Le, TK, Zhou, X, Johnston, JA, Dworkin, RH. The prevalence of chronic pain in United States adults: results of an internet-based survey. J Pain 2010;11:1230–9. https://doi.org/10.1016/j.jpain.2010.07.002.Search in Google Scholar PubMed
3. Treede, RD, Rief, W, Barke, A, Aziz, Q, Bennett, MI, Benoliel, R, et al.. Chronic pain as a symptom or a disease: the IASP Classification of Chronic Pain for the International Classification of Diseases (ICD-11). Pain 2019;160:19–27. https://doi.org/10.1097/j.pain.0000000000001384.Search in Google Scholar PubMed
4. DeVon, HA, Piano, MR, Rosenfeld, AG, Hoppensteadt, DA. The association of pain with protein inflammatory biomarkers: a review of the literature. Nurs Res 2014;63:51–62. https://doi.org/10.1097/NNR.0000000000000013.Search in Google Scholar PubMed
5. O’Mahony, LF, Srivastava, A, Mehta, P, Ciurtin, C. Is fibromyalgia associated with a unique cytokine profile? A systematic review and meta-analysis. Rheumatology 2021;60:2602–14. https://doi.org/10.1093/rheumatology/keab146.Search in Google Scholar PubMed PubMed Central
6. Karshikoff, B, Lekander, M, Soop, A, Lindstedt, F, Ingvar, M, Kosek, E, et al.. Modality and sex differences in pain sensitivity during human endotoxemia. Brain Behav Immun 2015;46:35–43. https://doi.org/10.1016/j.bbi.2014.11.014.Search in Google Scholar PubMed
7. Lasselin, J, Karshikoff, B, Axelsson, J, Åkerstedt, T, Benson, S, Engler, H, et al.. Fatigue and sleepiness responses to experimental inflammation and exploratory analysis of the effect of baseline inflammation in healthy humans. Brain Behav Immun 2020;83:309–14. https://doi.org/10.1016/j.bbi.2019.10.020.Search in Google Scholar PubMed
8. Jonsjö, MA, Åström, J, Jones, MP, Karshikoff, B, Lodin, K, Holmström, L, et al.. Patients with ME/CFS (Myalgic Encephalomyelitis/Chronic Fatigue Syndrome) and chronic pain report similar level of sickness behavior as individuals injected with bacterial endotoxin at peak inflammation. Brain Behav Immun – Health 2020;2:100028. https://doi.org/10.1016/j.bbih.2019.100028.Search in Google Scholar PubMed PubMed Central
9. Dantzer, R. Cytokine, sickness behavior, and depression. Immunol Allergy Clin 2009;29:247–64. https://doi.org/10.1016/j.iac.2009.02.002.Search in Google Scholar PubMed PubMed Central
10. Michael, J. Implications and refinements of the establishing operation concept. J Appl Behav Anal 2000;33:401–10. https://doi.org/10.1901/jaba.2000.33-401.Search in Google Scholar PubMed PubMed Central
11. Laraway, S, Snycerski, S, Michael, J, Poling, A. Motivating operations and terms to describe them: some further refinements. J Appl Behav Anal 2003;36:407–14. https://doi.org/10.1901/jaba.2003.36-407.Search in Google Scholar PubMed PubMed Central
12. Lasselin, J, Kemani, MK, Kanstrup, M, Olsson, GL, Axelsson, J, Andreasson, A, et al.. Low-grade inflammation may moderate the effect of behavioral treatment for chronic pain in adults. J Behav Med 2016;39:916–24. https://doi.org/10.1007/s10865-016-9769-z.Search in Google Scholar PubMed PubMed Central
13. Andreasson, A, Wicksell, RK, Lodin, K, Karshikoff, B, Axelsson, J, Lekander, M. A global measure of sickness behaviour: development of the Sickness Questionnaire. J Health Psychol 2018;23:1452–63. https://doi.org/10.1177/1359105316659917.Search in Google Scholar PubMed
14. Tang, X, Guan, Q, Duan, W. Sickness Questionnaire: a two-factor instrument reflecting physical and mental symptoms in the Chinese context. J Health Psychol 2020 Jul 16. https://doi.org/10.1177/1359105320942865 [Epub ahead of print].Search in Google Scholar PubMed
15. Andreasson, A, McNaughton, D, Beath, A, Lodin, K, Wicksell, RK, Lekander, M, et al.. Properties of the Sickness Questionnaire in an Australian sample with chronic medically unexplained symptoms. Brain Behav Immun – Health 2020;3:100059. https://doi.org/10.1016/j.bbih.2020.100059.Search in Google Scholar PubMed PubMed Central
16. Mokkink, LB, Terwee, CB, Patrick, DL, Alonso, J, Stratford, PW, Knol, DL, et al.. The COSMIN checklist for assessing the methodological quality of studies on measurement properties of health status measurement instruments: an international Delphi study. Qual Life Res 2010;19:539–49. https://doi.org/10.1007/s11136-010-9606-8.Search in Google Scholar PubMed PubMed Central
17. Wicksell, RK, Renöfält, J, Olsson, GL, Bond, FW, Melin, L. Avoidance and cognitive fusion - central components in pain related disability? Development and preliminary validation of the Psychological Inflexibility in Pain Scale (PIPS). Eur J Pain 2008;12:491–500. https://doi.org/10.1016/j.ejpain.2007.08.003.Search in Google Scholar PubMed
18. Wicksell, RK, Lekander, M, Sorjonen, K, Olsson, GL. The Psychological Inflexibility in Pain Scale (PIPS) - statistical properties and model fit of an instrument to assess change processes in pain related disability. Eur J Pain 2010;14:771.e1–14. https://doi.org/10.1016/j.ejpain.2009.11.015.Search in Google Scholar PubMed
19. Ware, JE, Sherbourne, CD. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care 1992;30:473–83.https://doi.org/10.1097/00005650-199206000-00002.Search in Google Scholar
20. Sullivan, M, Karlsson, J. The Swedish SF-36 Health Survey III. Evaluation of criterion-based validity: results from normative population. J Clin Epidemiol 1998;51:1105–13. https://doi.org/10.1016/s0895-4356(98)00102-4.Search in Google Scholar PubMed
21. Ware, JJ, Kosinski, M, Keller, S. A 12-Item Short-Form Health Survey: construction of scales and preliminary tests of reliability and validity. Med Care 1996;34:220–33. https://doi.org/10.1097/00005650-199603000-00003.Search in Google Scholar PubMed
22. Luo, X, George, ML, Kakouras, I, Edwards, CL, Pietrobon, R, Richardson, W, et al.. Reliability, validity, and responsiveness of the short form 12-item survey (SF-12) in patients with back pain. Spine 2003;28:1739–45. https://doi.org/10.1097/01.BRS.0000083169.58671.96.Search in Google Scholar PubMed
23. DeSalvo, KB, Bloser, N, Reynolds, K, He, J, Muntner, P. Mortality prediction with a single general self-rated health question. A meta-analysis. J Gen Intern Med 2006;21:267–75. https://doi.org/10.1111/j.1525-1497.2005.00291.x.Search in Google Scholar PubMed PubMed Central
24. Zigmond, AS, Snaith, RP. The hospital anxiety and depression scale. Acta Psychiatr Scand 1983;67:361–70. https://doi.org/10.1111/j.1600-0447.1983.tb09716.x.Search in Google Scholar PubMed
25. Lisspers, J, Nygren, A, Söderman, E. Hospital anxiety and depression scale (HAD): some psychometric data for a Swedish sample. Acta Psychiatr Scand 1997;96:281–6. https://doi.org/10.1111/j.1600-0447.1997.tb10164.x.Search in Google Scholar PubMed
26. Tait, R, Pollard, C, Margolis, R, Duckro, P, Krause, S. The Pain Disability Index: psychometric and validity data. Arch Phys Med Rehabil 1987;68:438–41.Search in Google Scholar
27. Tait, RC, Chibnall, JT, Krause, S. The Pain Disability Index: psychometric properties. Pain 1990;40:171–82. https://doi.org/10.1016/0304-3959(90)90068-O.Search in Google Scholar PubMed
28. Nunnally, JC. Psychometric theory. New York, NY: McGraw-Hill; 1978.Search in Google Scholar
29. Babakus, E, Ferguson, CEJr, Jöreskog, K. The sensitivity of confirmatory maximum likelihood factor analysis to violations of measurement scale and distributional assumptions. J Market Res 1987;24:222–8. https://doi.org/10.1177/002224378702400209.Search in Google Scholar
30. Hutcheson, G, Sofroniou, N. The multivariate social scientist: introductory statistics using generalized linear models. Thousand Oaks, CA: SAGE Publications; 1999.Search in Google Scholar
31. Decarlo, L. On the meaning and use of kurtosis. Psychol Methods 1997;2:292–307. https://doi.org/10.1037/1082-989X.2.3.292.Search in Google Scholar
32. Cohen, J, Cohen, P, West, SG, Aiken, LS. Applied multiple regression/correlation analysis for the behavioral sciences. New York: Routledge; 2002.Search in Google Scholar
33. Olsson, UH, Foss, T, Troye, SV, Howell, RD. The performance of ML, GLS, and WLS estimation in structural equation modeling under conditions of misspecification and nonnormality. Struct Equ Model 2000;7:557–95. https://doi.org/10.1207/S15328007SEM0704_3.Search in Google Scholar
34. Hair, JF, Black, WC, Babin, BJ, Anderson, RE, Tatham, RL. Multivariate data analysis. Upper Saddle River, NJ: Prentice-Hall; 2006.Search in Google Scholar
35. Bollen, KA, Long, JS. Testing structural equation models. Thousand Oaks, CA: SAGE Publications; 1993.Search in Google Scholar
36. Schermelleh-Engel, K, Moosbrugger, H, Müller, H. Evaluating the fit of structural equation models: tests of significance and descriptive goodness-of-fit measures. Methods Psychol Res 2003;8:23–74.Search in Google Scholar
37. Cohen, J. Statistical power analysis for the behavioral sciences. Hillsdale, NJ, US: Lawrence Erlbaum Associates; 1977.Search in Google Scholar
38. Cortina, JM. What is coefficient alpha? An examination of theory and applications. J Appl Psychol 1993;78:98–104. https://doi.org/10.1037/0021-9010.78.1.98.Search in Google Scholar
39. Streiner, DL. Starting at the beginning: an introduction to coefficient alpha and internal consistency. J Pers Assess 2003;80:99–103. https://doi.org/10.1207/S15327752JPA8001_18.Search in Google Scholar PubMed
40. R Core Team. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria; 2019. URL https://www.R-project.org/.Search in Google Scholar
41. Yoshida, K. tableone: create “Table 1” to describe baseline characteristics. R package version 0.10.0; 2019. Available from: https://cran.r-project.org/package=tableone.Search in Google Scholar
42. Korkmaz, S, Goksuluk, D, Zararsiz, G. MVN: an R package for assessing multivariate normality. R J 2014;6:151–62.10.32614/RJ-2014-031Search in Google Scholar
43. Maier, MJ. Companion package to the book “R: einführung durch angewandte statistik”. R package version 0.9.3; 2015. Available from: http://cran.r-project.org/package=REdaS.Search in Google Scholar
44. Rosseel, Y. lavaan: an R package for structural equation modeling. J Stat Software 2012;48:1–36. https://doi.org/10.18637/jss.v048.i02.Search in Google Scholar
45. Revelle, W. psych: procedures for personality and psychological research. Evanston, Illinois: Northwestern University; 2020. R package version = 2.0.12. Available from: https://cran.r-project.org/package=psych.Search in Google Scholar
46. Harrell, FEJr. Hmisc: Harrell miscellaneous. Version 4.5-0; 2019.Search in Google Scholar
47. Fletcher, TD. QuantPsyc: quantitative psychology tools. R package version 1.5; 2012. Available from: https://cran.r-project.org/package=QuantPsyc.Search in Google Scholar
48. Venables, WN, Ripley, BD. Modern applied statistics with S, 4th ed. New York: Springer; 2002.10.1007/978-0-387-21706-2Search in Google Scholar
49. Epskamp, S. semPlot: path diagrams and visual analysis of various SEM packages’ output. R package version 1.1.2; 2019. Available from: https://cran.r-project.org/package=semPlot.Search in Google Scholar
50. Cain, M, Zhang, Z, Yuan, K-H. Univariate and multivariate skewness and kurtosis for measuring nonnormality: prevalence, influence and estimation. Behav Res Methods 2017;49:1716–35. https://doi.org/10.3758/s13428-016-0814-1.Search in Google Scholar PubMed
51. Collins, LM, Schafer, JL, Kam, CM. A comparison of inclusive and restrictive strategies in modern missing data procedures. Struct Equ Model 2001;6:330–51. https://doi.org/10.1037/1082-989x.6.4.330.Search in Google Scholar
52. Gerdle, B, Åkerblom, S, Brodda Jansen, G, Enthoven, P, Ernberg, M, Dong, H-J, et al.. Who benefits from multimodal rehabilitation - an exploration of pain, psychological distress, and life impacts in over 35,000 chronic pain patients identified in the Swedish Quality Registry for Pain Rehabilitation. J Pain Res 2019;12:891–908. https://doi.org/10.2147/JPR.S190003.Search in Google Scholar PubMed PubMed Central
Supplementary Material
The online version of this article offers supplementary material (https://doi.org/10.1515/sjpain-2021-0070).
© 2021 Walter de Gruyter GmbH, Berlin/Boston
