Background
Stress is the reaction when human perceives a discrepancy in his resources and/or the ability to respond to an event or stimulus or stressor [
1]. Eventually, stress has been conceptualized into three perspectives: (i) biological, physiology of the stress stimulus and response; (ii) environmental, related to life events; and (iii) psychological, assessment of subjective stress and dealing methods [
1,
2]. Psychological stress is associated with asthma, upper respiratory tract infections, smoking, depression, diabetes, epilepsy, HIV/AIDS, herpes viral infections, autoimmune diseases, wound healing, self-reported measures of health behavior and help-seeking [
3‐
5].
Perceived stress is a risk factor for poor sleep quality in Ethiopian university students [
6]. Stress is common in various sections of the Ethiopian population such as university students [
6], epilepsy patients [
7], HIV-infected patients [
8], nurses [
9], and females students with childhood sexual abuse [
10]. Moreover, many risk factors, i.e., substance use [
11‐
13], sleep problems [
6,
11,
12], HIV [
14], food insecurity [
15], poverty [
15], and risky sexual behavior [
14] for stress and related mental problems are commonly prevalent in Ethiopian populations. It can therefore arguably be considered that there is a prospect of the undiagnosed and under-recognized magnitude of stress in the Ethiopian population. Indeed, similar to most places, depression seems to be more researched in Ethiopia than stress and other related psychological problems. The preponderance of depression in Ethiopian psychometric research is evidenced by the availability of validated tools to measure depression in Ethiopians [
16‐
19]. However, no tool has been validated in Ethiopians to assess stress.
Therefore, in this study, we examined the psychometric validation of one of the most widely used questionnaire tools to evaluate psychological stress, i.e., the Perceived Stress Scale [
2,
20]. This tool measures the extent and/or severity of self-reported appraisal of the stressors effect on respondent’s life [
20]. There are three versions; 14-item scale called PSS-14, 10-item scale in short PSS-10, and 4-item scale called PSS-4 [
3,
20]. The psychometric properties of the PSS have been assessed in various cultures across the globe, but have never been validated in Ethiopians. PSS has been translated into many languages and has been found to have adequate validity and reliability in various demographics of the population [
2,
20]. The convergent validity of the PSS has been evaluated by assessing correlation of the PSS scores with the measures of anxiety including the Generalized anxiety disorder-7 (GAD-7) scale and the Hospital Anxiety and Depression Scale (HADS) [
2,
21]. However, research is required to ascertain some aspects of the psychometric properties of the PSS. The items of the PSS are ordinal in nature; therefore, it would be better to investigate internal consistency using ordinal alpha or McDonald’s Omega and factorial validity employing polychoric correlation matrix with estimation method suitable for ordinal data like robust diagonally weighted least squares (robust DWLS). However, most of the studies investigating the psychometric properties of the PSS employed the Cronbach’s alpha [
2]. Similarly, statistical discrepancies are evident in previous works like use of maximum likelihood estimation for factor analysis, which is more suitable for normally distributed continuous data [
22,
23]. Furthermore, the psychometric properties of the PSS-4 are not well studied in the student population. Therefore, this study assessed the psychometric properties, i.e., ceiling/floor effect, factorial validity, internal consistency, item discrimination, and criterion validity like convergent validity, of PSS-10 and PSS-4 in Ethiopian university students. PSS-10 and PSS-4 are both brief measures but PSS-4 is still shorter. If both PSS-10 and PSS-4 show similar psychometric properties then the use of PSS-4 may be favorable.
Discussion
This is the first study to evaluate the psychometric validation measures of the PSS-10 and the PSS-4 in Ethiopian Africans in general and university students in particular. The investigation found sufficient level of the ceiling/floor effect, item discrimination, internal consistency, convergent validity, and factorial validity for PSS-10 in the study population.
Preliminary item analysis
The absence of both the floor and ceiling effects entails that even at the lowest or the highest scores of the PSS total and factor scores for both the PSS-10 and the PSS-4, the variance of the measurement is not unaccounted [
28]. This favors the structural validity of the PSS-10 and the PSS-4 in Ethiopian university students as a self-reported measure of stress. Of the few previous studies that investigated this aspect of the validity of the PSS, Wu and Amtmann reported that there was no major floor and/or ceiling effects in Americans Multiple Sclerosis patients [
31].
Factorial validity
CFA (Fig.
2 and Table
4) favored the 2-Factor model for the PSS-10 in the Ethiopian university students. Most of the previous studies have also favored the 2-Factor model of the PSS-10 and PSS-4 [
2,
21,
32,
33]. However, some of the studies reported a bi-factor model [
34] while some reported a 1-Factor model [
2]. Similarly, previous works have validated a 2-Factor model for the PSS-10 among Americans, Thai and Turkish university students [
2,
35]. Incidentally, the 2-Factor structures of both the PSS-10 and the PSS-4 are theoretically favored over a unidimensional model because some of the items measure stress, while others assesses the coping strategy to the stress [
20].
Internal consistency and item discrimination
The internal consistency as assessed by the McDonald’s Omega value for one of the factors of the PSS-10 was slightly lower than the minimum acceptable value of 0.70 (Additional file
3). However, internal consistency was very poor for the PSS-4: Factor-2. Most of the previous studies have reported the Cronbach’s alpha (0.74 to 0.91) for assessing internal consistency of the PSS, [
2]. The item discrimination index, i.e., item-Factor correlations were all above 0.5 for both the PSS-10 as well as the PSS-4. This implies that item scores of the PSS had the ability to distinguish between high and low scoring individuals in the study population [
23].
Stress conditions are closely associated with anxiety but they represent different psychological constructs [
36], therefore previous studies have investigated the relationship between the PSS and measures of anxiety including the GAD-7 to establish its criterion validity [
2]. However, noticeably this difference in the construct is perhaps accountable for the moderate level of correlation between the PSS and the GAD-7 scores in this study population of Ethiopian university students. Nevertheless, it can reasonably be concluded that the correlation between the PSS scores, i.e., PSS-10 total, PSS-4 total, PSS-10: Factor-1 and PSS-4: Factor-1 scores with the GAD-7 (Additional file
4, Table
4) favors the convergent validity of the PSS-10 and the PSS-4 in this population of Ethiopian university students. Previous studies have also supported the convergent validity of the PSS in different populations by assessing its correlation with measures of anxiety [
21,
32‐
34]. Maroufizadeh et al. 2014 found a moderate association between the PSS and anxiety subscale of the Depression Anxiety Stress-21 scale in Persian asthmatic adults [
33]. The convergent validity of the PSS had been supported by assessing correlations employing measures of anxiety-such as Spielberger Trait Anxiety Inventory in Spanish Americans [
34], Hospital Anxiety and Depression Scale among American patients with systemic lupus erythematosus [
21], and the GAD-7 among community-dwelling Hispanic Americans [
32]. PSS Factor-2 did not correlate with these measures (Table
4) because it does not assess stress but coping strategy to stress conditions [
20].
In summary, PSS-10 was found to have no major issues of the ceiling/floor effect, favorable factorial validity for 2-Factor model, internal consistency, item discrimination, and convergent validity among Ethiopian university students. However, psychometric properties were not adequate for the PSS-4 in the study population.
Limitations of the current study include a small number of female student’s participants, non-assessment of diagnostic validity, and test-re-test reliability. The questionnaire was not administered in the first language of the respondents which may have limitations, but it is worth mentioning that the medium of instruction is English in the Ethiopian universities. The lower response rate among female participants was one of the important reasons that led to bias in gender representation in the final sample. Future works employing diagnostic clinical interview to explore the concurrent validity of the PSS in Ethiopians are needed. Nevertheless, there are notable merits of this study. We found adequate psychometric validation for the PSS-10 in a population which has limited access to expert medicine professionals and facilities. Many risk factors for behavioral problems are prevalent in Ethiopia. Therefore, the availability of a validated measure of stress is very important. The psychometric properties assessed in this study do support application of the PSS-10 to screen stress among Ethiopians.