Psychometric and structural properties of the Karolinska Exhaustion Disorder Scale: a 1,072-patient study

Chronic psychosocial stress negatively affects individuals’ health and has high societal costs through reduced work capacity and sickness absence [1, 2]. Based on egalitarian values of social democracy, Sweden has an extensive public health care system and social insurance that aim to be accessible to all citizens on equal terms [3]. In Sweden, stress-related disorders currently account for over 50% of all sickness absences due to mental disorders [4]. In 2005, Exhaustion Disorder (ED) was introduced as a new medical diagnosis into the Swedish version of the International Statistical Classification of Diseases and Related Health Problems (ICD-10; see Table 1 for diagnostic criteria). ED is characterized by persistent mental and physical fatigue believed to develop in the wake of prolonged exposure to intractable stressors [5]. Diagnostic criteria were developed by a task force of researchers commissioned to investigate the rapid increase in sick leave rates in Sweden after a period of economic recession in the late 1990s. Interviews and clinical observations (unpublished to this day) indicated that many individuals on sick leave due to depression presented a clinical picture dominated by fatigue and cognitive complaints and attributed their mental health problems to work-related and psychosocial stressors.

A recent review of all published empirical ED studies found that research on the validity of this new diagnosis remains limited [6]. The clinical picture of ED is similar to that of burnout [7, 8] and chronic fatigue [9], and the overlap with anxiety and depressive disorders is substantial [6]. Amidst an international debate regarding whether burnout should be conceptualized as a medical disorder [10], ED has not been included in international versions of the ICD or the Diagnostic and Statistical Manual of Mental Disorders (DSM). In Sweden, however, the number of individuals diagnosed with ED has increased rapidly over the years with prevalence estimates approaching those of major depression [11, 12]. Furthermore, ED alone accounts for more long-term sickness absences than any other psychiatric or somatic disorder in the country [4]. Given the functional disability and suffering that can be associated with ED, continued investigation into the new diagnostic construct is merited.

To date, most research on ED has relied on self-rating scales developed to assess burnout, such as the Shirom-Melamed Burnout Questionnaire (SMBQ; [8]). However, several Swedish self-rating scales have been developed to pinpoint ED diagnostic criteria. One of these measures – The Karolinska Exhaustion Disorder Scale (KEDS) – has been widely implemented in clinical practice and workplace settings across Sweden. Despite the popularity of the KEDS, the instrument’s psychometric properties remain largely uninvestigated. In the only published study that has investigated the KEDS’s psychometric properties in a sample of ED patients, principal component analysis reached equivocal results regarding the scale's dimensionality, with the emergence of one- and two-component solutions [13]. A recent Danish study that investigated symptoms of exhaustion in patients with stress-related disorders and other psychiatric and somatic disorders using the KEDS found no difference in total scores between patients diagnosed with major depression and patients diagnosed with a stress-related diagnosis [14]. In addition, the KEDS has been found to correlate only moderately (below 0.50) with other inventories deemed to assess ED symptoms [15]. To further our understanding of the clinical utility of the KEDS, the present study investigated the dimensionality, homogeneity, reliability, and measurement invariance of the KEDS based on data from a large ED sample.

The present study examined the psychometric and structural properties of the KEDS in a sample of 1,072 ED patients. While the KEDS was designed to capture a unitary construct, results from our CFA and ESEM bifactor analysis provided little support for the scale’s essential unidimensionality. Consistent with our factor analytic findings, a homogeneity analysis indicated that the use of the KEDS’s total score is likely ill-advised. One-third of the items did not align well on the latent continuum assumed to underlie the measure. Furthermore, the total-score reliability of the instrument was modest, with values in the 0.70 s. Such values, which reveal considerable measurement error, are considered insufficient for scales meant to be used in basic research, and clearly problematic for instruments intended to be employed in applied settings [27, 28].

The limited internal consistency of the KEDS found in our study aligns with findings from the only previous psychometric evaluation of the KEDS among ED patients, conducted by Besèr et al. [13]. Such a finding is a cause for concern if the KEDS is to be used in applied settings involving clinical decision-making (e.g., for granting sick leave or prescribing treatment). Examining the latent structure of the KEDS with principal component analysis, Besér et al. obtained equivocal results. However, the authors concluded that a one-component structure best described the scale. In our study, we did not find evidence that the KEDS can be considered unidimensional. The divergent conclusions from the two studies might be due to variations in the statistical methods used, and the small sample size (N = 200) in the study of Besér et al. as compared with our large sample size (N = 1,072). It may also be that the samples used in the two studies represent somewhat different patient populations. ED patients in the study by Besèr et al. were recruited at a time when the ED diagnosis was recent in Sweden (2005–2010), whereas patients in our study were recruited between 2017 and 2022. Between the years of 2005 and 2022, the use of the ED diagnosis has increased markedly [4]. Even though the sociodemographic data indicate that the topographic qualities of the two samples are similar (middle-aged, highly educated women on sick leave), it is plausible that there, over the years, has been a shift in diagnostic trends such that some patients that would previously have been diagnosed with, for example, major depression or anxiety disorders where mental fatigue is a common symptom [33, 34], have instead been diagnosed with ED. In other words, it is conceivable that the patient group diagnosed with ED has become more heterogeneous with time. In Sweden, the recommendations for sick leave provided by the Swedish Social Insurance Agency stipulate that ED patients may be granted sick leave reimbursement for up to one year, whereas patients with other common mental disorders (e.g., depression) are usually eligible for part-time sick leave periods of 3 to 6 months. Such differentiated regulations might trigger healthcare professionals to use the ED diagnosis for a range of fatigued patients suffering from functional impairment.

There are at least two potential explanations of our findings. First, it might be that ED indeed constitutes a unitary diagnostic construct and that the KEDS simply fails to assess ED properly. The advanced statistical techniques employed in our study allow investigators to ascertain whether a scale can be used based on its total score despite the presence of a degree of multidimensionality. Measures such as the PHQ-9, a depression scale that covers as many as nine symptoms, have been found to meet such requirements in both clinical and nonclinical samples [35, 36].

A second way of interpreting our results is that, rather than dealing with an operationalization problem (i.e., a problem at the level of the instrument), we are dealing with a conceptualization problem. A conceptualization problem would imply that the diagnostic construct of ED is poorly devised, rendering any measures of the construct psychometrically unsound. Conceptualization issues would bear on the very definition and clinical validity of ED. As mentioned previously, ED is a relatively new medical diagnosis that has only been accepted into the Swedish version of the ICD-10. Very few studies have investigated the clinical validity and the specificity of ED in relation to other diagnostic constructs [6]. It is, however, well-established that the core symptom of ED (exhaustion/fatigue) is common to a range of psychiatric and somatic disorders [33, 37], and is a widespread experience in the general population [38]. The large symptom overlap of ED with anxiety and depressive disorders found in previous studies [6], together with the wide range of somatic symptoms expressed by ED patients [39, 40], suggest that any measure attempting to capture ED diagnostic criteria risks masking considerable multimorbidity.

Given the limited knowledge regarding ED as a diagnostic construct, and the finding that the KEDS may lack psychometric and structural robustness when employed with ED patients, a future avenue to further the understanding of ED might be to focus on the core symptom, namely, fatigue. Several internationally acknowledged fatigue self-rating scales have been found to be unidimensional with good reliability and validity [41]. The use of such scales could increase the comparability and integration of results across (clinical) samples, trials, and countries. This, in turn, could promote knowledge accumulation regarding the Swedish ED diagnosis and its relation to other fatigue-dominated conditions.

This study investigated the psychometric and structural properties of the KEDS in a large sample of patients diagnosed with ED. Results indicate that the scale does not meet the requirements for essential unidimensionality and lacks reliability, suggesting that using the KEDS’s total score for the assessment and screening of ED may be ill-advised. Because the KEDS closely adheres to the current diagnostic criteria for ED, our findings may have implications for ED as a nosological and diagnostic category. The sturdiness of ED as a pathological entity requires further investigation.