Items, response options, and scoring on the SSS-8.
The model was estimated using robust weighted least-squares estimation with mean and variance adjustment based on polychoric correlations. Loadings are standardized. Tucker-Lewis index, 0.95; comparative fit index, 0.97; root mean square error of approximation, 0.08 (90% CI, 0.08-0.09).
The number of health care visits was estimated using a negative binomial regression model that adjusted for depression (Patient Health Questionnaire–2) and anxiety (Generalized Anxiety Disorder–2) scores; error bars show 95% CIs.
eTable. Analysis of factorial invariance for age and sex using multigroup confirmatory factor analyses
Gierk B, Kohlmann S, Kroenke K, Spangenberg L, Zenger M, Brähler E, Löwe B. The Somatic Symptom Scale–8 (SSS-8)A Brief Measure of Somatic Symptom Burden. JAMA Intern Med. 2014;174(3):399-407. doi:10.1001/jamainternmed.2013.12179
Somatic symptoms are the core features of many medical diseases, and they are used to evaluate the severity and course of illness. The 8-item Somatic Symptom Scale (SSS-8) was recently developed as a brief, patient-reported outcome measure of somatic symptom burden, but its reliability, validity, and usefulness have not yet been tested.
To investigate the reliability, validity, and severity categories as well as the reference scores of the SSS-8.
Design, Setting, and Participants
A national, representative general-population survey was performed between June 15, 2012, and July 15, 2012, in Germany, including 2510 individuals older than 13 years.
Main Outcomes and Measures
The SSS-8 mean (SD), item-total correlations, Cronbach α, factor structure, associations with measures of construct validity (Patient Health Questionnaire–2 depression scale, Generalized Anxiety Disorder–2 scale, visual analog scale for general health status, 12-month health care use), severity categories, and percentile rank reference scores.
The SSS-8 had excellent item characteristics and good reliability (Cronbach α = 0.81). The factor structure reflects gastrointestinal, pain, fatigue, and cardiopulmonary aspects of the general somatic symptom burden. Somatic symptom burden as measured by the SSS-8 was significantly associated with depression (r = 0.57 [95% CI, 0.54 to 0.60]), anxiety (r = 0.55 [95% CI, 0.52 to 0.58]), general health status (r = −0.24 [95% CI, −0.28 to −0.20]), and health care use (incidence rate ratio, 1.12 [95% CI, 1.10 to 1.14]). The SSS-8 severity categories were calculated in accordance with percentile ranks: no to minimal (0-3 points), low (4-7 points), medium (8-11 points), high (12-15 points), and very high (16-32 points) somatic symptom burden. For every SSS-8 severity category increase, there was a 53% (95% CI, 44% to 63%) increase in health care visits.
Conclusions and Relevance
The SSS-8 is a reliable and valid self-report measure of somatic symptom burden. Cutoff scores identify individuals with low, medium, high, and very high somatic symptom burden.
Somatic symptoms are the common ground of many medical conditions.1 Presenting symptoms are often the impetus for further diagnostic evaluation, and symptom patterns may point to specific medical diseases. In clinical consultations, somatic symptoms are used as indicators for the severity and course of medical diseases as well as for monitoring treatment progress. They are associated with decreased quality of life, increased psychological distress, and increased use of health care services.2,3 Somatic symptoms are characteristic of various medical conditions such as cancer or coronary heart disease,2,3 and they are also present in psychiatric conditions such as depressive disorders, anxiety disorders, or somatoform disorders.4- 6 Previous studies7- 9 suggest that individual somatic symptoms frequently cluster into 4 groups: cardiopulmonary, gastrointestinal, pain, and general.
Although brief self-report instruments are frequently used for screening and severity assessment of depression and anxiety, they are less commonly used for the measurement of somatic symptom burden.10,11 However, somatic symptom scales might identify symptoms that are not the direct focus of the medical consultation. In addition, they could serve as screening instruments for elevated somatic symptom burden, as indexes for illness severity, and as measures to gauge response to treatment. Moreover, the use of self-report scales to complement medical interviews for symptom-based conditions can increase patient confidence that physicians are correct in their diagnosis and are making systematic efforts to assess and manage their problems.12
A recent systematic review8 indicated that many self-report questionnaires for common somatic symptoms exist but vary considerably with respect to their length, scaling, dimensionality, reliability, validity, and relevance to populations and regions. Among the 40 self-report somatic symptom scales investigated in this review, the Patient Health Questionnaire–15 (PHQ-15)13 and the 12-item Symptom Checklist–90 somatization scale14 were identified as the most appropriate measures for large-scale studies because of their well-established psychometric properties, relevance of symptoms, brevity, and availability in multiple languages.8 Derived from the PHQ-15, the 8-item Somatic Symptom Scale (SSS-8) (Figure 1) was recently developed as a brief patient-reported measure of somatic symptom burden. The SSS-8 has been used as a reference measure in the Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) (DSM-5)15 field trials as an instrument to facilitate the diagnosis of somatic symptom disorder. However, the psychometric properties of the instrument have not been investigated, and clinical reference scores are unavailable.
Therefore, the aims of this general-population study were 2-fold: (1) to investigate the reliability, factorial validity, and construct validity of the SSS-8 and (2) to provide reference scores for low, medium, high, and very high overall somatic symptom severity. The ultimate goal was to find a useful scale to evaluate and monitor somatic symptom burden in research and clinical practice. The study was approved by the medical ethics board at Leipzig University, Leipzig, Germany. Oral informed consent was obtained from the participants.
The study was part of a national, representative general-population survey that approached people older than 13 years in Germany. The data were collected in 2 waves between June and July 2012 by professional demographic consultants (Unabhängige Serviceeinrichtung für Umfragen, Methoden und Analysen, Berlin). A random-route sampling procedure with 320 sample points revealed that 4480 households should be contacted as part of the study. Of these, 4436 households were eligible to participate (ie, were not vacant or without individuals who met the inclusion criterion). The households were visited by trained face-to-face interviewers who recorded participants’ demographic information; the other information was collected through a paper/pencil self-report. Interviewers made a maximum of 4 contact attempts per household.
The SSS-8 is an abbreviated 8-item version of the PHQ-15 questionnaire.13 The PHQ-15 scale assesses the presence and severity of common somatic symptoms and is among the most widely used and best-validated self-report measures of somatic symptom burden.8,13,16 The SSS-8 was first developed for the DSM-5 field trials that investigated the newly established somatic symptom disorder.15 Three PHQ-15 items (menstrual problems, sexual problems, and fainting) were not included in the SSS-8 because of low communalities with other items, low symptom prevalence, and low associations with measures of functioning, quality of life, and health care use. In addition, 5 PHQ-15 items regarding gastrointestinal and cardiovascular symptoms were collapsed into 2 new items (SSS-8 items 1 and 5). A 5-point response option (0-4) for each SSS-8 item (instead of the 3-point response option of the PHQ-15) and a 7-day time frame (instead of the 4-week time frame of the PHQ-15) were used to mirror the response options and time frame of the Patient Reported Outcomes Measurement Information System measures used in the DSM-5 field trials. The SSS-8 score can, therefore, range from 0 to 32. Initially labeled as the PHQ-SSS in the DSM-5 field trials,15 the scale was renamed to shorten the name and reflect the number of items. The German version of the SSS-8 was developed using several steps of translation and blind back-translation according to state-of-the-art methods of test translation.17
To examine the construct validity of the SSS-8, participants also completed the PHQ-2 depression scale,10,18 the Generalized Anxiety Disorder–2 scale,10,19 and a visual analog scale to rate their current general health status (0, worst; to 100, best). Health care use was measured as the self-reported numbers of contacts with health care providers in the past 12 months (eg, days in the hospital, general practitioner visits, and emergency treatments). We chose these measures because somatic symptoms often are associated with depression and anxiety,16 decreased general health status,13 and increased health care use.13
We computed mean (SD) and corrected item-total correlations as descriptive statistics. Reliability was measured using Cronbach α. Factorial validity was analyzed using confirmatory factor analyses that tested 2 models theoretically derived from previous studies7- 9: a simple general factor model with all items loading on 1 factor and a higher-order general factor model composed of an overarching somatic symptom factor and 4 lower-order symptom clusters (gastrointestinal, pain, cardiopulmonary, and fatigue). Factorial invariance for sex and age was tested using multigroup confirmatory factor analyses with equality constraints.20 For these analyses, 4 groups based on sex and age were generated by applying a median split: males younger than the median age, males at the median age or older, females younger than the median age, and females at the median age or older. Construct validity was examined using (1) bivariate correlations between SSS-8, PHQ-2, Generalized Anxiety Disorder–2 (GAD-2), and visual analog scale general health and (2) negative binomial regression on the count of health care visits, with SSS-8 score as the predictor and anxiety and depression scores as covariates. The SSS-8 reference scores were presented as percentile ranks. Severity categories were determined in accordance with these percentile ranks. Severity categories were validated using negative binomial regression to predict health care visits at every severity level while adjusting for anxiety and depression scores. Missing values were imputed with a hot deck technique.21
Within the 4436 eligible households, interviewers had contact with 3736 persons. Of those, 2538 agreed to participate (67.9% cooperation rate22). However, 23 individuals were excluded because of severe illness that made participation impossible and 5 individuals were excluded for other reasons. As a result, data from 2510 individuals were included in the analyses (56.6% minimum response rate22). Less than 1% of the values in the data set were missing. Missing values showed no systematic pattern and were imputed.21Table 1 reports the demographic characteristics of the study sample.
The SSS-8 had good reliability (Cronbach α = 0.81), and all corrected item-total correlations exceeded 0.40. The item means varied between 0.16 and 0.70, reflecting an acceptable range of item difficulty. In addition, the SDs were comparable among the items (0.50-0.96) and the overall SSS-8 sum score mean was 3.23 (3.96), with an observed range of 0 to 26. Females had slightly higher mean scores compared with males (3.49 vs 2.94; t2508 = 3.50; P < .001; Cohen d = 0.14), and age was positively correlated with SSS-8 scores (r = 0.32 [95% CI, 0.28-0.35]). Therefore, we calculated age- and sex-specific reference scores (Table 2).
Confirmatory factor analysis revealed good fit indices for the higher-order general factor model (Tucker-Lewis index, 0.95; comparative fit index, 0.97; root mean square error of approximation, 0.08 [90% CI, 0.08-0.09]).23 This model, displayed in Figure 2, consisted of an overarching somatic symptom factor and 4 lower-order symptom clusters (gastrointestinal, pain, cardiopulmonary, and fatigue). Multigroup confirmatory factor analyses also demonstrated strict factorial invariance for age and sex for this model (Supplement [eTable]). In contrast, the simple general-factor model achieved lower fit indices and was, therefore, rejected (Tucker-Lewis index, 0.91; comparative fit index, 0.94; root mean square error of approximation, 0.11 [90% CI, 0.10-0.12]). Altogether, the results of the confirmatory factor analyses support the organization of individual somatic symptoms in 4 clusters according to a higher-order general-factor model.
In terms of construct validity, SSS-8 scores were significantly correlated with PHQ-2 depression scores (r = 0.57 [95% CI, 0.54 to 0.60]), GAD-2 anxiety scores (r = 0.55 [95% CI, 0.52 to 0.58]), and the ratings of general health status (r = −0.24 [95% CI, −0.28 to −0.20]). Even after controlling for depression and anxiety, SSS-8 scores significantly predicted health care visits in the past 12 months: each 1-point increase in the SSS-8 score was associated with a 12% increase in health care visits (incidence rate ratio, 1.12 [95% CI, 1.10 to 1.14]).
We determined severity categories in accordance with the SSS-8 percentile ranks (Table 2). Somatic symptom burden was classified as high if the sum score was between 12 and 15 points (ie, between the 95th and 98th percentile). From 12 points, we went downward in equal 4-point steps and classified 11 to 8 points as medium, 7 to 4 as low, and 3 to 0 as no to minimal burden. Finally, we added a very high category ranging from 16 to 32 points (ie, ≥98th percentile). We chose these cutoffs for 4 reasons. First, in primary care settings, approximately 10% of patients present with high or very high somatic symptom burden.13 In the present study, however, we surveyed the general population, so the rates of high and very high symptom burden should be lower than those in a typical clinical sample. Second, measures of symptom burden usually follow a left-skewed distribution in a nonclinical sample, ie, most individuals should fall into the categories no to minimal and low. This is also the case for our chosen severity categories. Third, pragmatic cutoffs are needed for clinical applications, and our equal 4-point increments are both valid and memorable (ie, 4, 8, 12, and 16 points representing low, medium, high, and very high levels of somatic symptoms, respectively). Finally, a negative binomial regression analysis revealed that an increase in severity category is significantly associated with an increase in health care use (Figure 3): for every SSS-8 severity category increase, there was a 53% increase in health care visits (incidence rate ratio, 1.53 [95% CI, 1.44-1.63]), which provides further evidence of the construct validity of our severity categories.
The present study evaluated the SSS-8, a new 8-item self-report measure for the assessment of somatic symptom burden. The results revealed good reliability of the scale and strong evidence of its validity. Cutoff points allow for pragmatic categorization of no to minimal (0-3 points), low (4-7 points), medium (8-11 points), high (12-15 points), and very high (16-32 points) somatic symptom burden. Simply stated, SSS-8 scores of 4, 8, 12, and 16 represent valid and easy-to-remember cutoff points for low, medium, high, and very high levels of somatic symptoms, respectively. Completion of the scale by the patient takes approximately 1 minute, and the scoring is easily done within another minute. Within this short period of time, health care professionals can obtain a valid score to gauge the total severity of the patients’ somatic symptoms and initial information regarding the pattern of the somatic symptoms themselves to inform diagnosis and guide further evaluations.
The SSS-8 assesses somatic symptom burden as a subjective, patient-reported phenomenon, without making any assumptions concerning its cause. The SSS-8 is potentially useful in many medical settings. First, SSS-8 scores could serve as quantitative markers of somatic symptom burden in patients with chronic diseases, such as cancer2 or coronary heart disease,3 who often experience various somatic symptoms that decrease their quality of life and psychological and social functioning. The general-population reference scores allow for comparison with people of the same age and sex. Second, the SSS-8 can facilitate the diagnosis of somatic symptom disorder, which was recently introduced in DSM-5.24 The A criterion of this disorder concerns the presence of burdensome somatic symptoms. Such symptoms can be easily assessed by using the SSS-8, the measure used in the DSM-5 field trials.15 Third, serial administration of the SSS-8 can be used to monitor somatic symptom burden and optimize symptom management (eg, reevaluation and modification of treatment if necessary). Fourth, the SSS-8 may detect burdensome and untreated somatic symptoms that are not cardinal symptoms of a particular disease, for example, musculoskeletal pain in patients with coronary heart disease.3 Fifth, because somatic symptom burden is associated with high health care use and costs,25,26 completion of the SSS-8 by frequent attenders of outpatient clinics and emergency departments could be helpful in determining whether high somatic symptom burden is a contributing factor to more frequent use of health care services. Furthermore, the scale could be used in managing the patient panels of primary care physicians or in reimbursement for such patients. Sixth, the SSS-8 could be used in risk adjustment of patient populations given that somatic symptom burden is an independent predictor of hospitalization and death, even when controlling for clinical characteristics, chronic medical conditions, self-rated health, and depression.27
The SSS-8 was designed to be an abbreviated version of the established PHQ-15 scale.13,16 Although the SSS-8 is approximately half as long as the PHQ-15, it features comparable construct validity: SSS-8 scores were significantly associated with health care use, anxiety, depression, and general health status.13,16 Interestingly, the associations between somatic symptom burden, depression, and anxiety were even stronger than the association between somatic symptom burden and perceived general health status. This finding is plausible considering the high comorbidity and partial overlap of somatic, anxiety, and depressive symptoms sometimes referred to as the somatization-anxiety-depression triad.16,28 This triad suggests that somatic, anxiety, and depressive symptoms seldom occur as “pure forms,”16(p352) meaning that most patients experience combinations of these symptoms. Nevertheless, studies show that somatic, anxiety, and depressive symptoms “have independent, additive and differential effects on multiple domains of health-related quality of life, functional status, disability and health care use,” which, therefore, gives support to the differentiation of these 3 symptom domains.16(p352)28 Given the high comorbidity of somatic symptom burden, anxiety, and depression, we recommend complementing the SSS-8 with either the brief 4-item PHQ-4 anxiety and depression screening tool10 or the slightly longer GAD-7 anxiety and PHQ-9 depression scales.16,29,30
The higher-order general-factor structure revealed by our analyses has been shown in previous studies9 investigating the latent dimensions of somatic symptoms that have revealed an overarching somatic symptom factor. Similarly, the symptom clusters identified as subcomponents of this general factor in the present study closely mirror symptom clusters reported in earlier studies.8,9 From a clinical viewpoint, these clusters correspond to 4 common medical syndromes (ie, pain, gastrointestinal symptoms, cardiopulmonary symptoms, and fatigue). Moreover, this factor structure allows for aggregation of the individual item scores into a simple, easily interpretable sum score ranging from 0 to 32 points. The analyses showed that the factor structure is strictly invariant for age and sex. This is important for the use of the scale in diverse populations because it permits the comparison of SSS-8 scores from respondents of different ages and sexes.9,20
The 5 severity categories derived from our data were based on the empirical distribution of the SSS-8 sum score. The high and very high symptom burden categories lay in the upper 5% of the severity distribution (ie, ≥95th percentile), which is common in clinical practice. By determining 5 severity categories (no to minimal, low, medium, high, and very high), we follow the taxonomy of frequently used self-report measures, for example, the PHQ-9 depression scale.16,30 Each increase in severity category is associated with a large increase in health care visits, which emphasizes the validity of the categories. In addition to an interpretation based on the severity category, there is a second way to interpret the SSS-8 score: the sex- and age-specific percentile reference scores can be used to locate an SSS-8 score’s relative position within the general population. The reference scores can also be used to aid patient communication because percentage ranks are easy to understand and some patients may feel relieved to know that they are at a certain level of somatic symptom burden (eg, “About X of 100 people experience lower somatic symptom burden than you do.”)
To facilitate the assessment of changes in the severity of somatic symptoms, the SSS-8 refers to recent symptoms (ie, in the past 7 days), whereas the PHQ-15 referred to more chronic symptoms (ie, in the past 4 weeks). Therefore, we believe that the SSS-8 is particularly suitable to assist the close monitoring of somatic symptom burden and to assess response to treatment. Similar to the PHQ-9 depression scale, for which responsiveness to treatment has been demonstrated in longitudinal studies,31,32 we believe that future studies will demonstrate that the SSS-8 is useful in assessing changes in the severity of somatic symptoms over time.
This study has several strengths, including its large sample size and the representativeness of the study sample. However, there are also several limitations. First, the response rate of 56.6% may be seen as low and could affect the generalizability. However, response rates in population studies are generally lower than those in clinical studies, and our response rate is substantially higher than the average response rates of questionnaire surveys (ranging between 45% and 50%) as reported in 2 meta-analyses.33,34 Second, because our study was cross-sectional, longitudinal studies are needed to determine the scale’s test-retest reliability (ie, the stability of SSS-8 scores over a defined period during which no change in somatic symptom burden occurred), its responsiveness to treatment, and its minimum clinically important difference.35 The minimum clinically important difference is of special interest for treatment evaluation because it identifies which intraindividual SSS-8 score difference is clinically meaningful. Third, our sample was drawn from the general population; the SSS-8 requires further study in patients receiving clinical care. However, the SSS-8 has been applied successfully in patient samples as part of the DSM-5 field trials.15 The DSM-5 field trials combined with our study findings suggest that the SSS-8 may be applicable to both clinical and general-population samples.
An important question, however, concerns the representativeness of the German study sample and the ability to generalize the study results to other countries. There is good reason to assume that the results regarding validity and normative data of the SSS-8 may be applicable to other Western countries. Studies using the PHQ-9 depression measure (which showed substantial correlations with the SSS-8) revealed similar distributions and cutoff scores in large North American30 and German36 samples. Therefore, we do not expect substantial differences in the distribution of SSS-8 scores among Western countries.
Because of the high prevalence and the functional impairment associated with somatic symptoms, their careful consideration remains especially important in patient care. Evidence-based treatments are available for patients with high somatic symptom burden,37- 39 and the SSS-8 may serve as a patient-reported outcome measure to evaluate these treatments. Given the time constraints and competing demands for busy clinicians,40 practical instruments for symptom assessment are needed to ease clinicians’ workload and facilitate research. The SSS-8 is efficient in that it is brief, reliable, and valid and can be completed entirely by the patient. As with the PHQ-9 depression measure30 and the GAD-7 anxiety measure,29 which have had substantial uptake,16 the SSS-8 somatic symptom measure should have considerable usefulness in busy medical settings and clinical research.
Accepted for Publication: September 3, 2013.
Corresponding Author: Benjamin Gierk, MSc, Department of Psychosomatic Medicine and Psychotherapy, University Medical Center Hamburg-Eppendorf, Martinistrasse 52, 20246 Hamburg, Germany (firstname.lastname@example.org).
Published Online: November 25, 2013. doi:10.1001/jamainternmed.2013.12179.
Author Contributions: Mr Gierk and Dr Löwe had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Gierk, Kohlmann, Zenger, Löwe.
Acquisition of data: Spangenberg, Zenger, Brähler.
Analysis and interpretation of data: Gierk, Kohlmann, Kroenke, Zenger, Löwe.
Drafting of the manuscript: Gierk, Kohlmann, Löwe.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Gierk, Kohlmann, Zenger, Brähler.
Administrative, technical, or material support: Gierk, Spangenberg, Zenger, Löwe.
Study supervision: Kohlmann, Zenger, Löwe.
Conflict of Interest Disclosures: None reported.
Additional Contributions: Alexandra M. Murray, DPhil, Department of Psychosomatic Medicine and Psychotherapy, University Medical Center Hamburg-Eppendorf, and Schön Clinic Hamburg Eilbek, Germany, provided comments on the manuscript. No financial compensation was given for this service.
Correction: This article was corrected on December 11, 2013, to restore an omitted line of text in the Discussion section.