The role of stigma in health and functioning in chronic pain: Not just catastrophizing

People with chronic pain are frequently exposed to stigma, which is typically distressing and may lead to internal stigmatizing thoughts. The thought content associated with stigma has similarities to pain catastrophizing, although these concepts differ in that stigma is arguably more social in origin. Stigma can be measured by the Stigma Scale for Chronic Illness – 8‐item version (SSCI‐8). In this study, we first demonstrate the validity of this measure in Swedish. We then examine the role of stigma in the health and functioning of people with chronic pain, particularly beyond the role played by pain catastrophizing.


| INTRODUCTION
"It's not just all in my head" -a phrase reflecting the need to dispute doubtful responses that people with chronic pain often face.Indeed, doubt and suspicion about the legitimacy of a person's pain are common, particularly when there are perceived inconsistencies or when pain cannot be medically explained (De Ruddere & Craig, 2016).Enacted stigma, where people in one's social context react negatively to certain characteristics one might have (Molina et al., 2013), is thus part of the chronic pain experience -both from the public and from health care professionals (De Ruddere & Craig, 2016).In some cases, this enacted stigma becomes internalized, so that a person comes to stigmatize themselves (Corrigan & Watson, 2006;Molina et al., 2013).Stigma has been shown to be related to depression, reduced functioning (Bean et al., 2022;Scott et al., 2019), low social support and increased activity avoidance (Bean et al., 2022).Internalized stigma may also lead to feeling less worthy (Corrigan & Rao, 2012).
More than a third of people with chronic pain may experience internalized stigma, and such stigma correlates with pain catastrophizing (Waugh et al., 2014), a psychological process important in the maintenance of chronic pain (Vlaeyen & Linton, 2000).Theoretically, this correlation makes sense since both concepts include reference to negative evaluation, such as the pain being terrible in the case of catastrophizing or the evaluation of being a terrible person in the case of stigma.However, catastrophizing, arguably, does not explicitly take social aspects into account, whereas the role of social context is naturally implied in the case of stigma.While catastrophizing is an evaluation a person makes in relation to their pain, stigma is an evaluation a person makes in relation to another person.This difference is important if one wishes to address the insensitive treatment that people with chronic pain receive, to understand how this might lead to greater pain impact and to find ways to mitigate this.
To assess stigma, validated measures are needed.The Neuro-QoL Short Form v1.0 Stigma is a self-report measure of enacted and internalized stigma.It is better known as the Stigma Scale for Chronic Illness -8-item version (SSCI-8), as it was originally named, and this name will be used in this paper.The original version in English has been deemed psychometrically sound (Molina et al., 2013;Scott et al., 2019).A Swedish translation has not been psychometrically validated before now.
The first aim of this study is to validate the Swedish SSCI-8.Factor structure will be examined as well as reliability in terms of internal consistency and temporal stability.Convergent construct validity will be assessed through correlations between stigma on the one hand, and pain interference, work and social adjustment and depression on the other hand.The second aim is to investigate if stigma is able to account for variance, beyond variance accounted for by pain catastrophizing, in the pain-related outcomes including pain interference, work and social adjustment and depression.

| Procedure
In this cross-sectional study, adults with chronic pain, fluent in Swedish and with access to an internet-enabled device were enrolled after recruitment via social media and patient organizations.People with all pain conditions were included, although recruitment especially focused on low back pain, endometriosis, vulvodynia and fibromyalgia in order to ensure inclusion of commonly studied conditions as well as less commonly studied conditions.Data collection was conducted between October and December 2021, and all measures were administered on two occasions, 2 weeks apart.This time interval was chosen in order to have an interval long enough to create an independent assessment while minimizing the risk of experiences and behaviour patterns changing.For this particular study, we are interested in the second time point results for the SSCI-8 only, in order to establish temporal stability, whereas the second time point results for the other measures will be presented in a future paper with a different aim.In-depth descriptions of the procedures can be found in Sundström et al. (2023).The study has received ethical approval from the Swedish Ethical Review Authority (DNR: 2021-02656) and all participants provided informed consent.

| Measures
2.2.1 | Information on demographics and pain Demographic information including gender, age, minority group status, family origin, relationship status, education level, work status and financial situation was collected.Pain-related information including pain conditions, pain intensity during the past week, current pain intensity, duration of pain, frequency of health care visits, number of days with pain in a month, pain treatments, confirmation of diagnosis and pain sites was also provided by participants.
2.2.2 | The SSCI-8   The SSCI-8 consists of eight items scored on a scale from 1 to 5, with higher scores indicating higher levels of stigma (Molina et al., 2013).This short version is derived from a longer 24-item version that was designed on the basis of focus groups regarding aspects affecting the quality of life in people with neurological conditions, as well as reviewing the literature and existing instruments when drafting the items (Rao et al., 2009).In the original study, using cognitive interviews, participants were asked about the initial item phrasings as well as item relevance.A team of experts analysed the interview content and revised the items accordingly.Psychometric evaluation indicated high internal consistency, and a bifactor solution accounting for internalized and enacted stigma (Rao et al., 2009).The 8-item version was created based on the need for a brief measure, and items measuring enacted as well as internalized stigma were selected based on item response theory parameters from earlier studies (Molina et al., 2013).This short version was then validated in a population with neurological conditions (Molina et al., 2013).With a Cronbach's alpha of 0.89, the internal consistency was found to be good.Validity was established by looking at correlations between the SSCI-8 and problems in conducting usual activities as well as between the SSCI-8 and psychological distress in terms of depression or anxiety (Molina et al., 2013).Further validation in chronic pain provided evidence for good internal consistency, and validity was established by examining correlations with measures of injustice experiences, pain intensity, functioning and depression (Scott et al., 2019).
In terms of factor structure, a one-factor solution generally appears to best fit the data, based on exploratory factor analysis and parallel analysis (Molina et al., 2013;Scott et al., 2019).However, in the study by Molina et al. (2013), two of the items appeared to specifically capture internalized stigma, and conducting EFA with two factors in mind did indicate factor loadings in line with proposed categorizations of enacted and internalized stigma, even though the one-factor solution fit the data best (Molina et al., 2013).Specifically, the items "I felt embarrassed about my illness" and "I felt embarrassed because of my physical limitations" were the items measuring internalized stigma in the study by Molina et al. (2013), whereas the rest measured enacted stigma.
The Swedish translation used in the current study was made available through licence agreement with researchers at Northwestern University in the United States.The translation followed principles of good practice as defined by the Professional Society for Health Economics and Outcomes Research, aiming for conceptual equivalence between languages as well as a culturally relevant translation (Wild et al., 2005).The translation process consisted of two forward translations by native Swedish speakers which were synthesized by a third native-speaking person.This synthesized Swedish version was then back-translated by a native English speaker also fluent in Swedish.The final translated version was examined regarding suitability for a Swedish population.
2.2.3 | The catastrophizing subscale from the coping strategy questionnaire A six-item subscale assessing pain catastrophizing (PC) was used.It was derived from the Coping Strategies Questionnaire (CSQ; Rosenstiel & Keefe, 1983).Each item of this measure is scored on a scale from 0 to 6 where higher scores reflect higher catastrophizing.The Swedish version of the CSQ is deemed adequate (Jensen & Linton, 1993).

| Additional measures
Besides the demographic and pain-related information, the SSCI-8 and the PC subscale, additional established standardized measures were completed.These included pain interference items from the Brief Pain Inventory (BPI; Cleeland, 2009) adjusted to reflect pain interference during the past week, the Work and Social Adjustment Scale (WSAS; Jansson-Fröjmark, 2014;Mundt et al., 2002) and the Patient Health Questionnaire-9 (PHQ-9; Hansson et al., 2009;Kroenke et al., 2001) for measuring depression.Higher scores on the pain interference items and the PHQ-9 indicate higher levels of pain interference and depression, whereas higher scores on the WSAS indicate lower levels of adjustment.In-depth descriptions of these measures are available in Sundström et al. (2023).In general, measures were chosen as they are established measures of the constructs of interest and relatively short, in order to minimize the burden for participants.

| Statistical analyses
The statistical analyses were in general conducted in SPSS 28.0.1.0.However, checking for multivariate normality, computing the reliability composite and conducting a confirmatory factor analysis (CFA) were done in R 4.3.0,and a network analysis of measure items was conducted in Jasp 0.17.
2.3.1 | Normality, outliers and missing data Univariate normality was assessed through analyses of skew and kurtosis and visually inspecting histograms.Two background variables, namely days of pain per month and frequency of health care visits, reflected very little variation and will not be included in the analyses.In item-level analyses of the SSCI-8, non-normality was found for item 3.As multivariate non-normality assessed by Mardia's test was also found for the SSCI-8 items, factor analysis was conducted with methods robust to non-normality.
One univariate outlier was identified, defined as a score above or below 3 standard deviations (SD).Eight multivariate outliers were identified when examining the Mahalanobis distances for all the SSCI-8 items, and one multivariate outlier was detected when examining Mahalanobis distances between the SSCI-8 summary score and summary scores of the pain-related outcomes.In sensitivity testing, the outliers did not influence the results in any meaningful way and were thus retained in the analyses presented in this paper.
Pain duration was not included in the analyses due to excessive missing data.For variables with minimal missing data, the variables were retained but the data points missing were excluded from analyses.As the level of missing data was progressive from one instrument to the next in the survey, the number of participants analysed differs on this basis.

| Background variables
Pearson correlations and point-biserial correlations on recoded dichotomized variables were used to assess the relationship between background variables and the SSCI-8 as well as between background variables and the pain-related outcomes of pain interference, work and social adjustment and depression.A one-way ANOVA was carried out to examine differences in SSCI-8 scores based on pain type.

| Factor structure
A one-factor solution for the SSCI-8 has been proposed by previous studies (Molina et al., 2013;Scott et al., 2019).However, there are clearly two items measuring internalized stigma (Molina et al., 2013), therefore both one-factor and two-factor solutions were examined in CFA.Due to multivariate non-normality, a Satorra-Bentler correction was used for maximum-likelihood estimation with robust standard errors.
Apart from the Chi-Square (χ2), goodness of fit was evaluated based on whether the root mean square error of approximation (RMSEA) was 0.08 or less, whether the standardized root mean squared residual (SRMR) was 0.09 or less and whether the Tucker-Lewis index (TLI) and the comparative fit index (CFI) were above 0.90 (Marsh et al., 2005).The threshold for salient item loadings was >0.32 (Tabachnick & Fidell, 2012).

| Network analysis
As an alternative approach to estimating the factor structure, where latent variables and their connections to each item are sought, a network analysis was also conducted in order to see how the items directly relate to each other -without trying to capture a latent variable.This was estimated by using graphical least absolute shrinkage and selection operator regularization with extended Bayesian information criterion model selection (EBICglasso; Foygel & Drton, 2010).

| Reliability
First, internal consistency was assessed.Traditionally, Cronbach's alpha is often used for estimating this, and an alpha between 0.7 and 0.95 is often considered desirable (Bland & Altman, 1997).However, Cronbach's alpha is often criticized for having known limitations that are not taken into account when it is used (Hayes & Coutts, 2020;McNeish, 2018;Sijtsma, 2009;Trizano-Hermosilla & Alvarado, 2016).In this paper, we do follow the convention of presenting Cronbach's alpha, but we also report a composite reliability statistic.This statistic is calculated based on one's factor model using structural equation modelling (Green & Yang, 2009;Jorgensen et al., 2018).
Reliability was further examined by looking at temporal stability with intraclass correlation coefficients (ICC) between results on each item across two time points.This was done by looking at absolute agreement for single ratings in a two-way random-effect model.Results below 0.5 were deemed as poor, results between 0.5 and 0.75 were moderate and results between 0.75 and 0.9 were good (Koo & Li, 2016).

| Convergent construct validity
Pearson correlations including pain interference, work and social adjustment and depression were examined.Small correlations were defined as 0.1 or higher, medium as 0.3 or higher and strong as 0.5 or higher (Cohen, 1988).In order to examine the generality of these potential correlations across pain types, the correlations were also examined separately for each pain type and compared using z-scores.

| Explained variance in pain-related outcomes
The assumptions for multiple regression were examined.Specifically, relationships between predictors and outcomes were linear, multicollinearity was ruled out, as was the presence of influential outliers, and residual values were independent.The data also met the assumptions of homoscedasticity and normally distributed residuals.
Three hierarchical multiple regression analyses were carried out, one for each of the three outcomes of pain interference, work and social adjustment and depression.In the first steps, age, education level, financial situation, work status and pain intensity were included.As pain intensity was measured with two items, only the pain intensity item correlating most strongly to the outcomes, namely pain intensity during the last week, was included.In the last steps, pain catastrophizing was first entered, followed by stigma.

| Sample characteristics
The sample of 404 participants consisted mostly of women (93.8%), with family origins from Sweden (83.2%) and a mean age of 47.75 years.Most were married or in a relationship (69%).About half the sample were employed or self-employed, about half the sample had a good or very good financial situation, and about half the sample had completed a university or college education.The most frequent pain sites were the lower back or spine, the pelvic region and the neck region, not being mutually exclusive.Low back or spine pain was reported by 68.8% of the sample, pelvic pain by 50.2% and neck pain by 45.5%.Most participants had been prescribed some sort of medication (78%), and around half the sample had generalized pain.See Table 1 for additional details on sample characteristics.On the second occasion of data gathering, 298 participants (73.8%) completed the SSCI-8 again.

| Background variables in relation to the SSCI-8
Scores on the SSCI-8 had a small positive correlation with pain intensity currently as well as during the last week.Being younger, having generalized pain, having been prescribed pain medications in general as well as opioids in particular, having gone through psychological treatment and having had the pain diagnosis confirmed by a medical doctor were all positively correlated with SSCI-8 scores.Furthermore, being part-or full-time employed or self-employed, and having a good or very good financial situation, correlated negatively with scores on the SSCI-8.See Table 2 for details.ANOVA revealed a difference between pain types in regards to SSCI-8 scores (F (df = 3) = 2.922; p = 0.034).Tukey HSD post hoc test further revealed this difference to lie between low back pain and endometriosis as main diagnoses, where those with endometriosis experienced significantly more stigma in comparison (p = 0.036).Descriptive statistics for each pain type can be seen in Table 3.

| Background variables in relation to pain-related outcomes
Age had a small negative correlation with BPI pain interference scores and medium negative correlation with  depression symptoms on the PHQ-9, meaning functioning appeared worse for younger people.Pain intensity during the last week had medium correlations to WSAS and PHQ-9 and strong correlation with the BPI pain interference, whereas current pain intensity showed a medium correlation with all three outcomes.Having generalized pain, having had opioids prescribed, and having gone through psychological treatment correlated positively, to a small degree, with all three outcomes.Having had generally pain medication prescribed, as well as having had the pain diagnosis confirmed by a medical doctor, correlated positively, in a small way, with scores on WSAS and BPI pain interference, while having had psychopharmacological drugs prescribed had small correlations positively with the PHQ-9 and BPI pain interference.Identifying as part of a minority group showed a small correlation with depression.Being employed or self-employed, having a good or very good financial situation and having a university education correlated negatively, to a small degree, with scores on all three outcome measures.See Table 2 for details.

| Factor structure
Neither a two-factor solution nor a one-factor solution initially met criteria for a good fit.Modification indices were examined for both solutions.For the two-factor solution, modification indices suggested covarying residuals for items across the two latent factors which would not have made theoretical sense.This, in combination with previous EFAs and parallel analyses suggesting a single-factor solution led us to pursue adjustments for a one-factor model.As modification indices suggested allowing for covarying residuals, this was done until a good model fit was reached.Due to risks of overfitting, it was checked at each step whether items considered for covarying residuals were in fact similar in content.Each successive adjustment was guided by examining modification indices again.Two modifications were sufficient for a good model fit on the CFI, the TLI and the SRMR (χ 2 (df = 18) = 98.836; p < 0.001; robust CFI = 0.937; robust TLI = 0.903; SRMR = 0.061; robust RMSEA = 0.118), whereas in total eight modifications would have been needed for the RMSEA to also indicate a good fit.Factor loadings for the modified model with two adjustments were all adequate, ranging between 0.51 and 0.82.See Table 4 for details.
T A B L E 3 Subgroup descriptive statistics on SSCI-8 scores.

T A B L E 4
Standardized factor loadings and intraclass correlations between two time points for items in the SSCI-8.

| Network analysis
The graphical network displayed in Figure 1 shows similarities to the results from the factor analysis, where the items that needed to be allowed to have covarying residuals in the factor analysis in order to prompt a good model fit are also strongly connected in the network analysis.
In addition, the items exerting the most expected influence on the remaining items in the network were items 1 (z = 1.34), 4 (z = 0.8) and 7 (z = 1.18).

| Reliability
The internal consistency in terms of Cronbach's alpha was good with a value of 0.87, and it was found that internal consistency could not improve by removing any items.
The composite reliability value based on the estimated factor model parameters was also deemed as good at 0.78.Temporal stability assessed with ICC was moderate for all items.See Table 4 for details.

| Convergent construct validity
SSCI-8 scores demonstrated a medium positive correlation with BPI pain interference when looking at the whole sample.Looking at the different pain types, correlations ranged from small for the endometriosis sample to medium for the fibromyalgia sample and large for low back pain and those with other conditions.There was also a medium positive correlation between SSCI-8 scores and WSAS scores in the total pain sample, medium correlations in the fibromyalgia subsample and those with other conditions, and large correlations in the low back pain and endometriosis samples.Lastly, SSCI-8 scores demonstrated a large positive correlation with PHQ-9 in the entire pain sample, as well as in all subsamples except for endometriosis where the correlation was medium.See Table 5 for details on these correlations.
When comparing how the correlations between the SSCI-8 scores and the three outcomes differed between pain types, two statistically significant differences emerged.Specifically, for participants with low back pain, the correlation between SSCI-8 and BPI pain interference was stronger than for participants with endometriosis (z = 2.41; p < 0.05).This was also the case for participants in the other category, where the correlation between SSCI-8 and BPI pain interference was stronger than for participants with endometriosis (z = 2.22; p < 0.05).
In the final model, only pain intensity, pain catastrophizing and stigma significantly contributed to the explained variance in pain interference, with pain intensity carrying the largest increment, followed by pain catastrophizing and stigma.In total, this final model explained 45.2% of the variance in pain interference.See Table 6.The final model explained 37.4% of the variance in work and social adjustment.In the final model, age, education level and financial situation no longer significantly contributed to explained variance in work and social adjustment.Stigma explained the largest increment in the variance, followed by pain intensity, work status and lastly, pain catastrophizing.See Table 6.
In the final model, education level, financial status, work status and pain intensity no longer significantly contributed to explained variance in depression.The variable most strongly contributing to explained variance in the final model was pain catastrophizing, followed by stigma and then age.The final model accounted for 44.4% of the variance in depression.See Table 6.

CONCLUSIONS
The first aim of this study was to demonstrate adequate validity and reliability of the Swedish version of the SSCI-8.A solution was found through CFA, though two adjustments, allowing covarying residuals, were needed to achieve overall goodness of fit as indicated by the CFI, TLI and SRMR.However, additional adjustments would be needed for the RMSEA to also indicate a good fit.The covarying residuals indicate that some items in the SSCI-8 may be redundant.This is also supported by the network analysis, showing high correlations between these same items.At the same time, it is somewhat unfortunate from a theoretical perspective for a one-factor solution to emerge as there could be significant value in a measure able to capture internalized and enacted stigma as separate facets.Stigma can be regarded as an environmental variable, such as in the case of enacted stigma, and it can be an experiential variable from the recipient's point of view.Lumping these two together makes it difficult to define the stigma measured by the SSCI-8 and to consider relevant interventions based on their scores.
For the time being, the factor structure of the current SSCI-8 is deemed adequate.However, there is a 14-item version of the SSCI that seems to fit a two-factor solution (Rao et al., 2016).This, in combination with the potential redundancy of the current 8-item measure, might indicate that it could be possible that the SSCI-8 could benefit from refinements leading to a different set of eight items that could reflect both aspects.Potential refinements should consider that some items in the measure exert more influence within the measure network, meaning that agreeing to these particular statements indicates a higher probability of agreeing to additional items in this measure.These items might be considered a priority to preserve.Results further indicate a reliable instrument both in terms of internal consistency and temporal stability.As for convergent construct validity, the SSCI-8 correlated with pain interference, work and social adjustment and depression.
Looking at correlations with background variables, SSCI-8 scores seemed, not unexpectedly, to be higher for people who had more pain-related problems.Specifically, people with greater pain intensity, people with generalized pain, people prescribed medication such as opioids and people who have gone through psychological treatment all scored higher on the measure.Factors not related to the pain experience that might influence scores on the SSCI-8 were age, where an older age was related to lower scores on the SSCI-8, and financial situation, where a better financial situation was related to lower scores on the SSCI-8.This means that certain demographic factors might influence the amount of stigma that one is exposed to or how such exposure is perceived.
An interesting finding is that participants with endometriosis scored higher on the SSCI-8 compared to those with low back pain.This might not be surprising as it has been argued that endometriosis has been largely delegitimized at a structural level due to its involvement of the female reproductive organ and that there is an attitude that women are somehow to blame for illness (Hudson, 2021).It should be mentioned, however, that endometriosis subsample is made up of individuals who are on average younger than the ones in the other groups, and older age seems to correlate with lower scores on the SSCI-8.Thus, age could be an important variable in understanding these results.Still, and of additional interest, even though those with endometriosis scored higher on the SSCI-8 than those with low back pain, stigma scores for the low back pain participants correlated more strongly with pain interference compared to those with endometriosis.Although the degree of stigma scores did not differ significantly between endometriosis and the other category in the sample, the correlation between stigma scores and pain interference did differ in strength between these two pain groups as well.Similar to the comparison of endometriosis and low back pain, stigma scores correlated less strongly with pain interference for the endometriosis group compared to the other category.Although correlations between stigma scores and the pain-related outcomes were consistently significant for all pain types, indicating the general importance of stigma as potentially affecting pain-related outcomes across pain conditions, the degree to which stigma may impact certain pain-related outcomes may not be uniform across pain types or across individuals.
The second aim of this study was to see whether variance in reported stigma is able to account for explained variance in pain interference, work and social adjustment and depression beyond the variance accounted for by pain catastrophizing.The results show that this is the case.While pain catastrophizing once again looks like an important variable when it comes to pain-related outcomes, stigma added unique variance to the model that is not entirely captured by pain catastrophizing in itself.This is especially true for work and social adjustment, where stigma contributed to more explained variance than pain catastrophizing when looking at the standardized regression coefficients.Furthermore, in regards to both work and social adjustment and depression, stigma experiences may influence these outcomes to a larger degree than the actual intensity of the pain.This may be an important finding, as few studies have brought stigma forward as a potentially important factor affecting chronic pain outcomes.With the concept of stigma comes a social dimension that is lacking when only highlighting pain catastrophizing or other similar variables of psychological individual difference.This means that stigma brings something else to the table in understanding pain-related outcomes.Furthermore, treating pain through the targeting of pain catastrophizing could, if presented insensitively, in itself be stigmatizing or blaming.The possible message received could be that one would be in less pain if one could stop catastrophizing.This could lead to a breakdown in trust or collaboration, leading to counterproductive interventions.
Moving forward with chronic pain treatment, there could be value in routinely assessing stigma and intervening when needed.For example, since acceptance and commitment therapy (ACT) have been shown to improve functioning and reduce anxiety and depression in people with chronic pain (Hughes et al., 2017;Lai et al., 2023), this therapeutic approach could potentially help people experiencing stigma in relation to their chronic pain.For instance, feelings embarrassment, guilt, fear of exclusion or negative self-judgements could easily lead to avoidance of otherwise valued activities.Helping the individual to pursue what gives them meaning even though they might encounter stigma from the environment and in turn experience difficult emotions and thoughts could be a way to reduce suffering from perceived stigma and its impacts.Other therapeutic approaches could also be helpful at an individual level.However, there is also a need for change at a societal level to decrease enacted stigma.This includes enacted stigma that can take place within health care.
Here, there is a potential training need, perhaps specifically around reactions to pain experiences that might not adhere to a clear picture of physical pathology or a conventional medical point of view.
Limitations to this study include predominance of women in the data and excessive missing data for pain duration.This means we are unable to adequately examine the role of these factors in relation to results.Still, the sample does include a variety of pain conditions and many participants have had their diagnosis confirmed by the health care providers.This suggests that many participants are experiencing clinically significant pain.
An additional limitation is the lack of cut-off scores to guide interpretation of the prevalence of stigma in this sample.Looking at Table 3, the mean stigma scores seem to be relatively low in each subgroup, but being guided by clear cut-off scores would have eased such an interpretation.However, even if the mean stigma scores in this sample could potentially indicate low levels of stigma, the correlations to pain-related outcomes do indicate that those who experience higher degrees of stigma are also in general worse off in terms of the pain-related outcomes.There are also the usual limitations here as will apply in survey studies, from participants recruited online, relying on self-report measures, and not including any experimental manipulation so as to demonstrate cause and effect.Reliability, generality and direction of relations remain to be seen.Additional studies with methods that address these limitations are recommended.
To conclude, the Swedish SSCI-8 is deemed an adequate measure to capture stigma as a single dimension.However, refinements could lead to an even better instrument in terms of factor structure and could expand what is studied so that it includes a separate perspective on internalized stigma.Using the measure to assess the role of stigma compared to the role of pain catastrophizing suggests that stigma is an important factor that contributes significantly to explained variance in pain-related outcomes.As stigma appears to add to the burden of people in pain, it is worthy of further study and as a potential target in treatment development.

T
sample M = 47.75, SD = 13.02Age in low back pain subsample M = 53.29,SD = 12.08 Age in endometriosis subsample M = 36.73,SD = 9.70 Age in fibromyalgia subsample M = 50.53,SD = 11.00Age in "other" subsample M = 48.58,SD = 13.23 Health care visits due to pain in last 6 months M = 4.73, SD = 9.19 Days with pain per month M = 28.14,SD = 4.05 Correlation analyses between background variables and the SSCI-8 as well as the outcomes.

F
I G U R E 1 A network illustration of how the items of the SSCI-8 are interconnected.The circles illustrate the item variables, and the thickness of edges between the circles illustrates the strength of the correlation between variables.Blue edges indicate positive correlations and red edges indicate negative correlations.

Percentage or mean and standard deviation
Sample characteristics.
T A B L E 1 Hierarchical multiple regression for predicting pain-related outcomes.
T A B L E 6