Background
Stress is a multifaceted phenomenon referring to the physical and psychological reactions to imposing demands [
1]. While there are many definitions of stress in the literature, the concept of stress generally refers to: a) contextual stressors, such as the frequency and impact of stressful life events, b) neuroendocrinological effects, for example the release of adrenal glucocorticoids, and c) cognitive processes, reflecting the subjective perception of stress and subsequent emotional and behavioral responses [
2]. In researching stress, measures have relied heavily on the two former perspectives, that is, focusing on either the topography of stressful life events or biological markers. However, given an increased interest in the idiosyncratic appraisal of stressors and how it seems to affect individuals differently, Cohen, Kamarck [
3] developed the Perceived Stress Scale (PSS), which intends to assess “the degree to which respondents found their lives unpredictable, uncontrollable, and overloading” (p. 387). This includes queries into more general aspects of stress as experienced during the last month, for example, “…how often have you been upset because of something that happened unexpectedly” (item 1). The original version of the PSS includes 14 items that are scored on a five-point Likert-scale: from 0 (“Never”) to 4 (“Very often”). Items 4–7, 9–10, and 13 are reversed, meaning that they are framed in a positive manner, e.g., “how often have you felt that things were going your way” (item 7). In the original study by Cohen, Kamarck [
3] the PSS was administered to two samples of college students (
N = 332 and 114), and as part of one trial of smoking cessation (
N = 64). Moreover, additional self-report measures were also used to explore the association between the PSS and other constructs. The results showed that the PSS had good internal consistency, Cronbach’s α = 0.84-0.86 depending on the sample, and moderate to strong positive correlations with symptoms of social anxiety, depression, and physical issues (
rs = 0.52-0.76). However, the relationship with life events and their impact were lower,
rs = 0.17-0.49, particularly for the samples of college students. No attempt at studying the dimensionality of the PSS was made.
The PSS has since then been translated to at least 25 languages and been the subject of a large number of psychometric evaluations. Moreover, two shorter versions of the PSS with either 10 (PSS-10) or four (PSS-4) items have also been put forward. The latest systematic review of the field was performed by Lee [
4], who found 19 eligible studies, including translations into languages such as Spanish, Turkish, and Arabic, with samples derived from a general population, college students, and patient groups. Overall, the PSS exhibited good internal consistencies, i.e., Cronbach’s α > 0.70, in 11 out of 12 cases for the 14-item version, in all 12 cases for the 10-item version, but only in one out of three cases for the 4-item version. Most investigations of its dimensionality suggested a one-factor solution for the 14-item version, but a two-factor solution for the 10-item version, based on exploratory factor analyses. Confirmatory factor analyses were seldom used, but suggested two-factor solutions for all versions. Lee [
4] also argued that the test–retest reliability was satisfactory if the time period between two measurement points was shorter than four weeks, but less satisfactory for longer intervals.
Further research on the PSS has been made during the last decade. This includes several studies of additional samples and translations [
5,
6], establishing norms for the PSS-4 [
7], using confirmatory factor analysis [
8], bifactor analysis [
9], and Rasch analysis [
10‐
13]. In line with previous studies, higher scores are repeatedly shown to be associated with negative outcomes, such as increased anxiety, depression, and fatigue [
5]. Regarding its different versions and their respective dimensionalities, most research point toward two dimensions. Taylor [
8] identified two factors for the PSS-10 in English, referred to as “perceived helplessness” and “perceived self-efficacy” (often called “perceived stress” and “perceived [lack of] control” in the literature), highlighting that the multidimensionality of the PSS can make scores difficult to interpret if stress is in fact conceived as unidimensional and global construct. Likewise, for the Spanish translation, Juárez-García, Merino-Soto [
9] argued that the two factors for the PSS-10 and PSS-14 probably are a result of the wording of the items, recommending these to be rephrased into either only negatives or positives, or treated separately. Similar findings have also been put forward by Nielsen, Ornbol [
10] regarding the Danish translation, identifying two dimensions on the PSS-10 (i.e., one negative and one positive), albeit still exhibiting problems of model fit. Another study by Nielsen and Dammeyer [
11], on a different Danish translation of the PSS-10, confirmed these two dimensions, with improved fit after removing item 6 (“…how often have you felt confident about your ability to handle your personal problems?”). Comparable results have been obtained in other studies of the PSS in English and Spanish [
12,
13], suggesting that the 14- and 10-item versions likely consist of two dimensions, with most problems of model fit emanating from the positively worded items (i.e., “perceived [lack of] control”).
The Swedish translation
Stress is a topic of much debate and research interest, particularly because of its association with many health-related outcomes. Stress has for instance been linked to many somatic conditions and mental distress, such as cardiovascular diseases and depression [
2]. Moreover, in many countries, work-related stress has also been gaining more attention. This is particularly true of Sweden where a surge in the incidence of long-term sick-leave caused by exhaustion disorder started two decades ago (a non-traumatic stress-related condition which is included in the Swedish version of the International Classification of Disorders, tenth version) [
14]. Measuring stress by means of self-reports has therefore become increasingly important, both in clinical and research settings, with the PSS being widely used for screening purposes and determining the outcome of treatment. For example, in a systematic review and meta-analysis of Internet-based cognitive behavior therapy [
15], nine out of 13 identified trials (69.2%) had administered the PSS (three with the 14-item version, five with the 10-item version, and one with the 4-item version).
The first translation and psychometric evaluation of the PSS in Swedish consists of an unpublished report from 1996 by Eskin and Parr [
16]. The study does not provide any information about how it was translated from English, but the instrument (14 items) was administered to 87 university students, together with 13 questions about stressful life events that might have occurred during the last six months (e.g., “moving”), and self-report measures of depression and perceived social support. The results indicated that the PSS had good internal consistency, Cronbach’s α = 0.82, no association with stressful life events (
r = 0.09), a moderate positive correlation with depression (
r = 0.66), and weak negative correlations with perceived social support that ranged from
rs = -0.29 for friends, and -0.33 for family. However, no attempt was made to investigate its dimensionality.
Since then, two other psychometric evaluations of the PSS in Swedish have been made. Nordin and Nordin [
17] used the same translation as Eskin and Parr [
16] but with 10 items, distributing the instrument to 3406 individuals as part of a larger survey study on environmental health issues. The PSS had good internal consistency, Cronbach’s α = 0.80-0.86 (depending on the age range) and revealed moderate to strong positive correlations for depression (
r = 0.57), anxiety (
r = 0.68), and exhaustion (
r = 0.71). Furthermore, an exploratory factor analysis identified two factors, with factor 1 consisting of the negatively worded items (explained variance of 33.8%), and factor 2 being comprised of the positively worded items (explained variance 24.1%). However, Nordin and Nordin [
17] argued that due to the lack of a theoretical explanation for a two-factor solution, the PSS should be conceived as a unidimensional and global construct, thus disregarding the obtained factor solution. Meanwhile, Eklund, Bäckström [
18] administered the PSS-14 to a small sample recruited via the Internet (
N = 171) and women with stress-related disorders (
N = 84), revealing good internal consistencies, Cronbach’s α = 0.90 (Internet-sample) and 0.84 (stress-sample), and moderate negative correlations with mastery (
r = -0.66), and coping ability (
r = -0.51). As for its dimensionality, Eklund, Bäckström [
18] found a two-factor solution after removing item 12, as determined using a confirmatory factor analysis, but argued that the two factors can be collapsed because of the high correlation between them.
The current study
In Sweden, the PSS is recommended by many regional authorities responsible for the public healthcare sector in the country, making it popular among clinicians. Yet, there is no consensus on which version to distribute to patients, which may affect how scores are interpreted and compared. Moreover, different response patterns on the PSS might also complicate its use, particularly in relation to gender [
11]. For example, administering the Turkish translation of the PSS-10 in a sample of 508 university students revealed that women scored higher than men (Cohen’s
d = 0.30) [
19]. Gitchel, Roessler [
20] demonstrated a similar finding, with women scoring higher than men,
d = 0.24, using an English version of the PSS with 11 items administered among a sample of 1079 adults with multiple sclerosis. More specifically, women scored higher overall and higher on the negatively worded items, but women and men were similar in terms of the positively worded items. In addition, Martinez-Garcia, Nielsen [
13], distributing the Spanish PSS-10, found that gender interacted with educational type and year in a sample of 399 university students, such as female students studying for a professional degree on the second year in fact experienced less stress than their male counterparts, regardless of educational type or year. Meanwhile, in a study of the English PSS-14, Ribeiro Santiago, Nielsen [
12], Differential Item Functioning (DIF) was explored, i.e., to test the unequal probability of providing a certain response depending on a specific attribute. In relation to gender, this was found for items 1, 3, 6, and 10 (see Table
1 for an overview of the items), which was attributed to gender roles typical of many western countries, such as men being less likely to recognize negative emotions but more likely to demonstrate self-confidence. Similar findings has been put forward by Nielsen and Dammeyer [
11] for the Danish PSS-10, illustrating DIF related to gender for item 1 and 3. In other words, gender effects may have to be acknowledged and accounted for when administering the PSS, although it should be noted that other studies have not found such an influence [
9,
21‐
23]. Regarding the Swedish translation of the PSS, no published investigation of different response patterns or DIF exists, which would be helpful given its widespread use among clinicians and researchers with a wide array of patient demographics. It is important to note that differences in response patterns as examined by group differences in sum score levels are not to be taken as evidence for issues with invariance or DIF.
Table 1
Items of the perceived stress scale
q1 | Hur ofta har du under den senaste månaden känt dig upprörd på grund av att något oväntat har inträffat? | In the last month, how often have you been upset because of something that happened unexpectedly? |
q2 | Hur ofta har du under den senaste månaden känt att du inte kunnat kontrollera viktiga saker i ditt liv? | In the last month, how often have you felt that you were unable to control the important things in your life? |
q3 | Hur ofta har du under den senaste månaden känt dig nervös och stressad? | In the last month, how often have you felt nervous and “stressed”? |
q4 | Hur ofta har du under den senaste månaden framgångsrikt hanterat vardagsproblem och irritationsmoment? | In the last month, how often have you dealt successfully with day to day problems and annoyances? |
q5 | Hur ofta har du under den senaste månaden känt att du effektivt kunnat hantera viktiga förändringar som inträffat i ditt liv? | In the last month, how often have you felt that you were effectively coping with important changes that were occurring in your life? |
q6 | Hur ofta har du under den senaste månaden känt tilltro till din egen förmåga att hantera personliga problem? | In the last month, how often have you felt confident about your ability to handle your personal problems? |
q7 | Hur ofta har du under den senaste månaden känt att saker och ting gått din väg? | In the last month, how often have you felt that things were going your way? |
q8 | Hur ofta har du under den senaste månaden tyckt att du inte kunnat klara av allt du skulle ha gjort? | In the last month, how often have you found that you could not cope with all the things that you had to do? |
q9 | Hur ofta har du under den senaste månaden kunnat kontrollera irritationsmoment i ditt liv? | In the last month, how often have you been able to control irritations in your life? |
q10 | Hur ofta har du under den senaste månaden känt att du har haft kontroll på saker och ting? | In the last month, how often have you felt that you were on top of things? |
q11 | Hur ofta har du under den senaste månaden blivit arg på saker som har hänt och som du inte kunnat kontrollera? | In the last month, how often have you been angered because of things that happened that were outside of your control? |
q12 | Hur ofta har du under den senaste månaden kommit på dig själv med att tänka på saker som du måste göra? | In the last month, how often have you found yourself thinking about things that you have to accomplish? |
q13 | Hur ofta har du känt under den senaste månaden att du haft kontroll över hur du använder din tid? | In the last month, how often have you been able to control the way you spend your time? |
q14 | Hur ofta har du under den senaste månaden tyckt att svårigheter har tornat upp sig så mycket att du inte kunnat hantera dem? | In the last month, how often have you felt difficulties were piling up so high that you could not overcome them? |
Although previous psychometric evaluations of the PSS in Swedish have contributed to its widespread use, several issues remain to be explored. This relates to its dimensionality and whether a one- or two-factor solution is most appropriate, the reliability of the different versions of the measure, and potential response patterns that may affect its scores. The current study aims to address these issues by applying a modern test theory approach, Rasch analysis, which produces estimates that help to understand measurement validity and reliability from the perspective of both items and persons [
24,
25]. For a measure like the PSS this means that the response to a specific item reflects both the individual’s level of stress and the level of stress underlying the item, which is not possible to determine using classical test theory approaches like exploratory factor analysis. Rasch analysis also has the advantages of being more robust against missing data, having the ability to test possible item bias and identify items that do not contribute to the measure, and to investigate potential DIF, such as the unequal probability of providing a certain response depending on a specific attribute like gender [
26]. This might further the understanding of the Swedish translation of the PSS and how it can be used, thereby increasing its utility in different settings.
Discussion
The current study explored the psychometric properties of the PSS in Swedish by using Rasch analysis. Similar to Nordin and Nordin [
17], who employed an exploratory factor analysis for the PSS-10, and Eklund, Bäckström [
18] applying a confirmatory factor analysis for the PSS-14, two factors were identified, which are comprised of either negatively or positively worded items. In both of these prior studies the reasoning was nevertheless to regard the instrument as unidimensional, either from a theoretical standpoint or because of the high correlation between the two factors. However, a different perspective is that they actually address different, albeit related, constructs. The negative subscale seems to reflect the concept of negative stress (i.e., distress), e.g., “…felt difficulties were piling up so high that you could not overcome them?” (item 14), in line with the idea of measuring more general aspects of stress [
3]. In contrast, the positive subscale seems to convey an ability to effectively deal with stressors, e.g., “…felt that you were on top of things?” (item 10), which instead appears to resemble a sense of mastery. This is in line with prior research [
8‐
13], demonstrating a negative subscale referred to perceived stress and a positive subscale of perceived [lack of] control, which resembles the idea of stress being comprised of appraisal and coping [
44]. In line with the recommendations by Juárez-García, Merino-Soto [
9], all items should be phrased as either negatives or positives, depending on what concept one intends to study. Researching the negative impact of stress, it seems more reasonable to include only negatively worded items, while the capability to manage stressors is treated as a distinct construct or captured using another self-report measure, e.g., the General Self-Efficacy Scale [
45]. Another option would be to treat the two factors separately when administering the PSS, scoring the stress-dimension and the [lack of] control-dimension individually to gain an understanding of both the negative impact stressors might have and the capacity to deal with stressful life events. However, combining the two into a sum score on general aspects of stress is not supported.
As for the two shorter versions, only the reliability for the negative subscale of the PSS-10 was acceptable, analogous to the PSS-14. Hence, should the PSS-10 be used, a similar issue concerning the wording of the items facing the full instrument is important to consider. This is different from the recommendations by Nordin and Nordin [
17] who proposed that the PSS-10 could be used in the same manner as the PSS-14 and that it in fact captures a unidimensional and global construct. As for the PSS-4, no previous attempt has been made to determine its reliability in Swedish, but the results from the current study imply that it is unsuitable as a unidimensional scale measuring more general aspects of stress.
In addition to its dimensionality, all items of the PSS adhered to its predefined scale-steps, with the possible exception of item 14 (“…felt difficulties were piling up so high that you could not overcome them?”), having minor issues with the second highest category. In other words, it seems reasonable to retain the present five-point Likert-scale (0–4). However, it should be noted that vague quantifiers, e.g., “Never” and “Often”, are often interpreted differently depending on subjective experiences [
46]. This should be particularly true for such a multifaceted phenomenon as stress. Research on how these scale-steps are conceived is therefore recommended, for example by using cognitive interviews [
47]. As for response patterns, no difference was observed with regard to gender in the current study. Although similar findings have been obtained in the research [
9,
21‐
23], there are numerous cases were DIF is in fact evident for gender. Ribeiro Santiago, Nielsen [
12] found differences in responding to items 1, 3, 6, and 10 in English, and Nielsen and Dammeyer [
11] to items 1 and 3 in Danish. This might in turn be attributed to gender roles and their influence on how stress is perceived, experienced, and responded to, which should be accounted for when administering the PSS. The reason as to why similar findings were not demonstrated in the current study is unclear. Data in the current study was collected from university students who may differ from population-wide samples, warranting further research on the subject of gender differences and stress in Sweden. The current study did on the other hand find an influence of age. This was true for item 12 (“…found yourself thinking about things that you have to accomplish?”), comparing those under or above the age of 20, although this needs to be interpreted cautiously given the small sample size. More crucial is perhaps the age effect found for all positively worded items (with the exception of item 6), and particularly item 4 and 9. It might be that individuals over the age of 31 share experiences that are different from those who are younger based on their life situation, such as being more likely to have children and struggling with work-life-balance. Those over the age of 43 may in turn experience their circumstances differently from those being 32–43 years. Further research on response patterns and age should be conducted as it might reflect differences in perceived difficulties of the items. This could also account for such factors as occupation, marital status, and social support. Juárez-García, Merino-Soto [
9] for example found a relationship between both age and occupation and negatively worded items, suggesting that younger people as well as workers experience more stress than older people and university students.
Finally, the issue of content validity is relevant to consider. As demonstrated by the item hierarchy for the negative items (see Fig.
6), the two items at the top featured aspects of anger and being upset. Irritation, frustration, and anger are often considered signs of a non-traumatic stress-related condition, including exhaustion disorder [
14]. However, these are not symptomatic for everyone experiencing stress, may be more evident in the initial phase of exhaustion disorder, and seem only weakly associated with the underlying construct of exhaustion disorder in the Karolinska Exhaustion Disorder Scale [
48]. In a recent study on the topic, responses to open-ended questions about symptoms of exhaustion disorder were completed by 670 participants and analyzed using qualitative content analysis also suggests that frustration/irritability may not be the most typical signs of the condition, as experienced by individuals themselves. Instead, other mood and emotional symptoms were more common, such as depression, emotional regulation, anxiety, and worry [
49]. Hence, the PSS might benefit from replacing the two items on anger and being upset, should it be used to measure more general aspects of stress.
There are some limitations important to considering when reviewing the results of the current study. First, while a sample size of 793 is deemed sufficient to conduct Rasch analysis, the constitution of the participants is restricted in that it was based on data from university students. Although being relatively heterogenous in gender and age, those studying in higher education may differ from a working population or specific patient groups when it comes to experiencing and dealing with stressful life events. Previous research on the PSS have used similar samples (c.f., [
9,
19]), but the results from the current study should nevertheless be replicated in other settings. Second, given that the participants were self-recruited, this might have created self-selection bias. Because recruitment was made via the Internet and being completely anonymous, investigating the motifs to participate is not possible, including factors that may have affected their responses to the PSS, e.g., motivation, stress levels, or other present living conditions. However, it is plausible that the purpose of the cross-sectional research project, i.e., stress and wellbeing of university students in Sweden, might have attracted individuals that are more interested in the topic, or alternatively suffer from difficulties related to stress and wellbeing.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.