Bastianon et al. BMC Psychiatry          (2020) 20:450 
https://doi.org/10.1186/s12888-020-02851-2
RESEARCH ARTICLE Open Access
Perceived Stress Scale (PSS-10)
psychometric properties in migrants and
native Germans
Christina Diane Bastianon1* , Eva M. Klein2, Ana Nanette Tibubos2, Elmar Brähler2, Manfred E. Beutel2† and
Katja Petrowski1†
Abstract
Background: With the increasing diversity of the German population, it is important to test the psychometric
validity and reliability of the German version Perceived Stress Scale (PSS-10) specifically between German natives
and residents with a migration background.
Methods: Using nationally representative data (N = 2527), this study conducted an Exploratory Factor Analysis (EFA) to
determine the most appropriate factor structure, a Multi-Group Confirmatory Factor Analysis (MGCFA) to compare the
validity of the two-factor structure and tested the PSS-10 measurement invariance between the German native and
migrant sub-samples. Lastly, reliability of the PSS-10 was examined via Cronbach’s alpha, omega and individual item
analyses across the two sub-samples.
Results: The EFA results support a two-factor structure in the migrant sample. The MGCFA showed adequate model fit
for both sub-samples and the PSS-10 is strict invariant between German natives and migrants. Cronbach’s alpha and
omega for Perceived Helplessness (PHS: factor 1) and Perceived Self-Efficacy (PSES: factor 2) demonstrate good internal
consistency in both German and migrant sub-samples.
Conclusions: The key conclusions are: (1) the German version PSS-10 is suitable for German residents with a migration
background. (2) Despite good internal consistency for the total scale, the PSS-10 measures two aspects: (a) perceived
helplessness and (b) perceived self-efficacy. Future research would profit from analyzing the two subscales separately,
not only using the total score.
Keywords: PSS, Psychometric properties, Stress, Migration, Measurement invariance
Background opportunities for mental health well-being, specifically rec-
Mental health and specifically stress have gained attention ognizing vulnerable or at risk groups. The Stress Report
as public health concerns by the European Regional Office Germany 2012 [2] echoes the importance of bringing atten-
of the World Health Organization (WHO) and by the Ger- tion to and researching the impacts of stress and mental
man national government. Objective 1 of The European health, especially in the ever-changing work environment.
Mental Health Action Plan 2013–2020 [1] focuses on equal The Stress Report Germany 2012 [2] found that perceived
stress and the number of health complaints increased from
the 2005/2006 report. Most importantly, both the WHO
* Correspondence: cbastian@uni-mainz.de European Regional Office and Stress Report 2012 emphasize
Manfred E. Beutel and Katja Petrowski shared Senior Authorship
1Medical Psychology and Medical Sociology, University Medical Center of the the inequality of how stress and mental health impact
Johannes Gutenberg, University Mainz, Mainz, Germany various groups, particularly vulnerable and disadvantaged
Full list of author information is available at the end of the article
© The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if
changes were made. The images or other third party material in this article are included in the article's Creative Commons
licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons
licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain
permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the
data made available in this article, unless otherwise stated in a credit line to the data.
Bastianon et al. BMC Psychiatry          (2020) 20:450 Page 2 of 9
groups which may be characterized by ethnicity, sex, age, The Perceived Stress Scale developed by Cohen,
religion, sexuality, refugee or immigrant status, socioeco- Kamarck and Mermelstein [13] is a widely used self-report
nomic status and physical and/or mental disability [1, 2]. measure assessing “the degree to which situations in one’s
Germany has a rich immigration history that has led to life are appraised as stressful”(p.387). The scale measures,
the diversity of its population. Work-visa programs from over the past month, the degree to which life has been ex-
the mid-1950s until 1973 explain, in large part, the earlier perienced as unpredictable, uncontrollable and overloaded.
settlement of Italians and Spaniards, but also the Turkish, The original scale included 14 items but was later reduced
Polish and Romanian populations within Germany [3, 4]. to 10 items due to low factor loadings on 4 items; this
In more recent years, immigration flows consisted not change marginally improved the scale reliability shown via
only of economic migrants, but increasingly people fleeing Cronbach’s alpha [14]. An additional 4 item version was
persecution and war-torn countries. The total foreign developed for telephone interviews or situations with time
population in Germany in 2014, including EU28, is re- restrictions [14], however, this short form has not fared as
ported at just over 8 million [5]. Citizens of Turkey well as the full 14- and 10-item versions [15]. In a review of
(11.2%), Poland (8.6%), Syria (7.1%), Romania (6.9), and the PSS psychometric properties, Lee [15] showed across
Italy (5.1%) represent the highest percentages of the the 19 studies included, the PSS-10 was found to be super-
foreign-born population of Germany in 2018 [6]. ior to the 14-item version. Cronbach’s alpha constantly sur-
Migrants (first- and second-generation) have long been a passed the standard .70 threshold ranging between .74–.91
focus of stress research. Regardless of whether migration [15]. The review also found consistent results supporting a
occurred voluntarily or as a result of natural disasters or two factor structure, which is in contrast to the original
displaced by conflict, many aspects of the migration process one-factor structure presented by Cohen et al. [13]. This
are seen as stressors [7]. Even after settling in a new coun- debate was sparked after Hewitt, Flett and Mosher [16]
try, those who have migrated may still experience stress as challenged the one-factor structure, recognizing that both
they adapt to their new home through acculturation and factors explained unique variances of depression and that
integration stressors [8]. In Russian and Iranian migrants in factor 1 comprised of ‘adaptational symptoms’ while factor
Germany, Haasen, Demiralay, and Reimer [9] found a sig- 2 reflected ‘coping ability’. Roberti, Harrington and Storch
nificant correlation between acculturative stress and mental [17] further supported the two-factor structure naming fac-
distress; the length of residency in Germany did not have a tor 1 perceived helplessness (PHS) and factor 2 perceived
significant effect. Therefore, it is important to take a closer self-efficacy (PSES). The 10-item version has proven to be a
look at stress within the population of residents with a valuable tool for stress research as it maintains consistent
migration background [10]. However, it is not known, if test-retest reliability across various timespans, acceptable
generalized mental health or stress are also responsive to Cronbach’s alpha, and strong factorial validity in measuring
migrant specific stressors such as integration, discrimin- perceived stress across various populations and languages
ation, acculturation or PTSD. Therefore, it is important in [18–22]; see review: [15]. Although the PSS-10 has been
advancing stress research to test whether widely used scales used in specialized populations, in particular minority
maintain psychometric strength and measurement equiva- groups, many psychometric studies on the PSS-10 call for
lence in migrant specific populations. continual testing in more diverse and representative popu-
Within the German population, only a few studies exist lations [15, 18].
testing the measurement equivalence of mental stress mea- In a representative German population, Klein et al. [23]
sures between populations with and without a migration tested the translated German version PSS-10 showing a
background. Tibubos et al. [11] demonstrated strict meas- strong Cronbach’s alpha = .84, further emphasizing the
urement invariance of the Patient Health Questionnaire-9 strength of the scale’s internal consistency. Reis and col-
between migrant status (nonimmigrants, 1st generation leagues [24] tested the factor structure of the German ver-
migrants and 2nd generation migrants) and country of ori- sion PSS-10 using bifactor modeling bringing a deeper
gin in a German cohort study, the Gutenberg Health Study understanding to the multidimensionality of the scale.
(GHS), with 13,973 participants. In another study with N > Most recently, Schneider et al. [25] tested the PSS-10 two-
26,000 participants from the German Socio-economic factor structure between clinical and nonclinical samples,
Panel (SOEP) including nonimmigrant, migrant and refu- showing strict measurement invariance. Although the
gee populations in Germany from 16 different countries, PSS-10 has been translated and tested in numerous lan-
the Patient Health Questionnaire-4 and the Short Form guages, populations and contexts, the psychometrics of the
Health Survey (SF-12) showed scalar invariance (1) be- German version PSS-10 have not been tested for migrant
tween men and women, (2) between groups stratified by populations within Germany. Considering the popularity
migration status, (3) between survey languages, (4) between of the PSS-10 as a generalized measure of stress and its
country of origin, (5) between sex and country of origin, various applications, this study aims to fill this gap by test-
and (6) between age groups [12]. ing the validity, reliability, and measurement equivalence
Bastianon et al. BMC Psychiatry          (2020) 20:450 Page 3 of 9
of the German version PSS-10 between native Germans Statistical procedure
and migrants. Analyses were computed using R v3.5.0 [26] and associ-
ated packages of Lavaan [27], psych [28], GPArotation
Methods [29] and MBESS [30]. Descriptive statistics were tabu-
Data collection lated by migrant status (German native and migrant) to
This study, including the consent procedure, was ap- provide a breakdown of the nationally representative
proved by the institutional ethics review board of the sample. T-tests assuming equal variance on key sociode-
University of Leipzig (Az 063–14-10,032,014). Further- mographic variables were calculated to determine the
more, the study adhered to ICH-GCP-guidelines along comparability of the two samples.
with the ICC/ESOMAR International Code of Marketing There were few missing data on the PSS, four persons
and Social Research Practice. All participants were in- did not answer any item and were dropped completely. In
formed of the study procedures, data collection and the remaining group, two persons left three items un-
anonymization of all personal data. According to Ger- answered, three persons two items, and 55 persons one
man law, all participants provided verbal informed con- item. Also, across the sub-samples there were few missing
sent, which was noted by the interviewer before starting item values, Item 7 had the highest with 12 missing values
with the survey. in the native sample (0.05%), all together the missing PSS
This study uses data from a representative survey of item values do not exceed 2.4% in the native sample and
the German population and the same data as in Klein 4.2% in the migrant sample. Hardt et al. [31] suggests as a
et al. [23]. Data were collected by USUMA (Unabhängi- rule of thumb with less than approximately 5% missing
ger Service für Umfragen, Methoden und Analysen; data, no measures such as multiple imputation should be
Berlin) between February and April 2014. The sample performed. In response to missing data and non-
consisted of a total of N = 2527 participants between the normality, maximum likelihood (ML) estimation with ro-
ages of 14 and 95 years old (detailed sample description bust standard errors and Satorra-Bentler scaled test-scores
is found below in Table 1). Face-to-face interviews were was deemed the most appropriate method [32, 33]. Cron-
conducted by trained interviewers via a stratified, bach’s alpha was calculated on pair wise covariances, for
random-route procedure in line with the ADM (Arbeit- correlations and EFA no adjustments were made.
skreis Deutscher Markt- und Soziolforschungsinstitute A single EFA using maximum likelihood estimation
e.V.) sampling guidelines. Questionnaires were inde- with oblimin rotation on the migrant sample was com-
pendently completed by the participant in the presence pleted to determine the appropriate PSS-10 factor struc-
of the interviewer. ture [34]. Separate CFAs were calculated on the native
and migrant samples to test the established two-factor
Measures structure [18, 23], followed by a multigroup confirmatory
Migration factor analysis (MGCFA). The negatively worded items
In line with Beutel et al. [10] a variable distinguishing Ger- form factor one, while positively worded items form factor
man natives from migrants was generated in accordance two. Goodness of fit was evaluated based on Satorra-
with the German micro census definition. The variable Bentler [32] adjusted chi-square (χ2), standardized root
combines information on participants’ citizenship and birth- mean square residual (SRMR), comparative fit-index
place of both mother and father. Therefore, in this study, (CFI), Tucker-Lewis Index (TLI), as well as the root mean
the term ‘migrant’ is used to identify first- and second- square error of approximation (RMSEA) and its 90% con-
generation migrants. First generation migrants include non- fidence interval (90% CI). Higher values of CFI and TLI
German citizens who migrated to the Federal Republic of (> 0.95) indicate a better model fit, while SRMR and
Germany after 1949, while second-generation migrants are RMSEA values below 0.08 are recommended [35].
all non-German citizens born in Germany and all citizens In line with Milfont and Fischer [36], measurement in-
born in Germany with at least one migrated parent. variance of the PSS-10 between the native and migrant
samples was tested using four hierarchical models: (1)
Perceived stress scale (PSS-10) configural, (2) metric, (3) scalar, and (4) strict. Model (1)
The German version of the PSS-10 was translated and tests that the PSS-10 two-factor structure is invariant in
standardized by Klein et al. [23]. Respondents report the both groups. Model (2) holds factor loadings equal across
degree to which situations in one’s life have been unpre- groups, followed by model (3) that additionally constrains
dictable, uncontrollable and overloaded in the past item-intercepts. Lastly, model (4) constrains factor load-
month on a 5-point Likert scale (0 = never, 1 = almost ings, intercepts and error variances between native and
never, 2 = sometimes, 3 = fairly often, 4 = very often). migrant samples. Measurement invariance was evaluated
Scores for the four positively stated items (Items 4, 5, 7, by changes (Δ) in goodness of fit indices including: Δχ2,
8) are reversed. ΔCFI, ΔTLI, ΔSRMR and ΔRSMEA. When sample sizes
Bastianon et al. BMC Psychiatry          (2020) 20:450 Page 4 of 9
Table 1 Descriptive Characteristics
Natives Migrant t- p
N = 2195 N = 328 test
87% 13%
Gender Male 46.33 48.48 0.72 0.46
Female 53.67 51.52
Age Mean 49.86 46.46 3.23 0.001***
SD 17.74 18.19
Range 14–95 14–86
Marital Status Single 27.47 25.91 2.48 0.01**
Married 45.15 52.44
Separateda 2.28 0.91
Divorced 13.90 13.72
Widowed 11.12 7.01
Education Less than 10 36.99 47.56 1.30 0.19
In years 10–13 50.48 37.50
Greater than 13 10.11 10.6
Other 2.41 4.88
Employment Full-time 39.91 35.06 −1.88 0.06
Part-time 11.75 12.80
Hourly 2.51 3.35
Volunteerb 0.78 0.92
Unemployed 5.60 8.54
Retired 29.25 20.73
Householdc 3.51 7.93
Apprenticeship 1.18 3.05
Student 5.10 7.32
Household Incomed < 750 € 3.98 3.76 −0.21 0.82
750 to < 1250 € 15.21 12.23
1250 to < 2000 € 26.82 32.29
≥ 2000 € 54.01 51.72
Religione Protestant 38.77 17.99 −2.50 0.01**
Catholic 31.25 35.06
Muslim 0.18 17.68
Other 0.46 9.76
No religion 29.09 18.77
Note a separated but still legally married; b Volunteer includes those on parental leave; c Household refers to those not working but not unemployed; d Household
income per month; e 12 participants did not respond, i.e. 9 German natives and 3 migrants
are unequal, as is the case in this study, Chen [37] recom- [38–40]. Means of individual items were examined between
mends the following cutoff criteria for testing levels of invari- German natives and migrants via t-tests and Cohen’s d ef-
ance: ΔCFI ≥− .005, ΔRSMEA ≤ .010 or a ΔSRMR ≤ .025 (≤ fect size. Cohen’s d effect size tests the standardized mean
.005 for intercept and residual invariance) indicate invariance. difference for each item between German natives and mi-
Reliability of the German version PSS-10 was evaluated grants. Cohen’s d values are evaluated as irrelevant
via Cronbach’s alpha (α), omega (ω), and associated 95% (d < .20), small (.20 ≤ d < .50), medium (.50 ≤ d < .80), and
confidence intervals. Alpha and omega were calculated large (d ≥ .80) [41]. Lastly, PHS and PSES scores were com-
using a 1000-repetition bootstrap method to obtain per- pared using t-tests and associated Cohen’s d effect size be-
centile confidence intervals. Higher values of omega and tween migrants and natives as well as between first- and
alpha (.70 and greater) reflect stronger internal consistency second-generation migrants.
Bastianon et al. BMC Psychiatry          (2020) 20:450 Page 5 of 9
Results Table 2 PSS item descriptives
Sample descriptive statistics Item Natives Migrants Cohen’s d t-test p-value
The nationally representative sample included N = 2523 Factor 1 PHS
participants, of which n = 328 participants had a migra- 1 M 1.19 1.23 −0.051 −0.87 0.381
tion background and n = 2195 were German natives. Mi-
SD 0.92 0.98
grants were significantly younger than German natives
by approximately three years (t2525 = 3.23, p < .001). The
rIT 0.55 0.59
two sub-samples did not statistically differ based on sex 2 M 0.88 1.01 −0.141 −2.38 0.017**
(χ 21 = .52, p = .46) or household income (χ
2
3 = 4.96, p = SD 0.94 0.98
.17). First-generation migrants account for n = 76 partici- rIT 0.69 0.65
pants, while n = 252 are considered second-generation 3 M 1.38 1.65 −0.266 −4.50 < 0.001**
migrants. The two generational groups did not statisti-
SD 0.98 1.07
cally differ based on sex, age, or household income.
rIT 0.59 0.52
6 M 1.00 1.10 −0.112 −1.89 0.057
Descriptive item analysis
SD 0.93 0.94
Table 2 displays the mean (M), standard deviation (SD),
and the corrected item-rest correlation (rit) for each PSS
rIT 0.67 0.69
item separated by sub-samples. T-tests for each item com- 9 M 1.59 1.71 −0.122 −2.06 0.039*
pared the means of German natives to migrants showing SD 1.02 0.98
that six out of the ten items were significantly different with rIT 0.54 0.46
a p-value of .05 or lower. However, Cohen’s d effect size re- 10 M 0.99 1.10 −0.105 − 1.77 0.076
ports that only item 3 has a small effect size where migrants
SD 0.98 1.05
differ from natives by more than .20 standard deviations,
while all other items have an irrelevant effect size. rIT 0.73 0.70
Factor 2 PSES
4 M 1.37 1.52 −0.143 −2.42 0.015**
Factor analyses
An EFA tested the German version PSS-10 factor solu- SD 1.10 1.10
tion with the migrant sample. Results showed a two- rIT 0.69 0.62
factor solution where 56% of the variance was explained 5 M 1.51 1.63 −0.119 −2.01 0.044*
by Factor 1 (PHS) with loadings from .55–.78; while Fac- SD 1.02 1.00
tor 2 (PSES) explained 44% of the variance with loadings rIT 0.69 0.65
ranging between .68–.83.
7 M 1.40 1.52 −0.110 −1.85 0.063
Model fit indices of the individual CFAs for German na-
tives and migrants are reported in Table 3. The Sattora- SD 1.07 1.05
Bentler scaled χ2 test is significant at the 1% level for both rIT 0.67 0.71
natives and migrants, indicating poor model fit. As χ2 is 8 M 1.14 1.28 −0.144 −2.43 0.015**
highly sensitive to sample size and non-normality of data, SD 0.96 0.97
additional scaled indicators were evaluated for model fit. rIT 0.72 0.65
In both samples, SRMR and RMSEA < .08 and CFL > .95
all indicate good model fit and while TLI is not greater PHS M 7.01 7.76 −0.172 −2.89 0.003***
than .95, it is close with .93. Overall, the two-factor PSS
SD 4.34 4.42
fits the data in both native and migrant samples.
The hierarchical measurement invariance models were PSES M 5.41 5.95 −0.157 −2.62 0.008***
computed with robust maximum likelihood method.
SD 3.46 3.36
Measurement invariance was evaluated by the change
(Δ) in goodness of fit indices. Table 4 shows the changes Note: Sample Size (N), mean (M), standard deviation (SD), item-rest correlation(rIT). Sample size for items range between 2183 and 2195 for natives and 324–
in robust model fit between the invariance models. Al- 328 for migrants. Sig. difference shows the p-value of a two-tailed t-test and
though χ2 values violate invariance assumptions, in test- significance value * p < 0.1, **p < 0.05, *** p < 0.01. Effect size reports the
Cohen’s d. Degrees of freedom for the t-tests varied between 2506 and 2518
ing the loading, intercept and residual invariance, ΔCFI due to missing data
exceeds the −.005 threshold, ΔSRMR and ΔRSMEA are
lower than the .025 and .010 cutoff, indicating invari- Reliability
ance. Thus, the two-factor PSS-10 is a strict invariant As both the EFA and CFA results show and confirm a
measure between German natives and migrants. two-factor structure, Cronbach’s alpha and omega for
Bastianon et al. BMC Psychiatry          (2020) 20:450 Page 6 of 9
Table 3 Individual & Multigroup CFA model fit
N χ2 (df) CFI TLI SRMR RMSEA 90% CI
Native 2146 446.21(34)*** 0.947 0.929 0.053 0.082 0.076–0.089
Migrant 317 81.44(34)*** 0.945 0.928 0.060 0.077 0.056–0.099
Configural 505.17(68)*** 0.947 0.929 0.049 0.082 0.075–0.088
Metric 521.65(76)*** 0.946 0.937 0.051 0.077 0.071–0.084
Scalar 547.53(84)*** 0.946 0.942 0.051 0.074 0.068–0.080
Strict 490.40(94)*** 0.945 0.948 0.052 0.070 0.064–0.077
Note: robust estimates, ***sig. p < 0.001
each factor was calculated, higher values of omega and migrants, it additionally shows good internal consistency
alpha indicate good internal consistency. Perceived helpless- as found in other diverse populations, see review [15].
ness (PHS-factor 1) showed a good internal consistency in Omega and Cronbach’s alpha for the PSS-10 sub-scales
both the native sample (α = .85 SD = .25 95% CI = .83–.85; ranged from .83 to .85 thus supporting the reliability of
ω = .85 SD = .01 95% CI = .84–.86) and in the migrant sam- the German version PSS-10. Migrants reported higher
ple (α = .83 SD= .32 95% CI = .80–.86; ω = .83 SD= .02 95% scores on both PSS sub-scales, indicating higher levels of
CI = .80–.86). Similarly, PHS Perceived self-efficacy (PSES- perceived helplessness as well as lower levels of perceived
factor 2) showed good internal consistency for the native self-efficacy since the latter scale is reversed. However, by
(α = .85 SD = .34 95% CI = .84–.87; ω = .85 SD = .01 95% taking into account Cohen’s d effect size, there is no rele-
CI = .84–.87) and migrant (α= .83 SD= .37 95% CI = vant effect between the two samples on PHS or PSES.
.79–.86; ω = .83 SD= .02 95% CI = .79–.87) samples. When comparing first- and second-generation migrants,
perceived self-efficacy showed no significant difference,
Perceived stress of migrants however, first-generation migrants report significantly
In testing the PSS sub-scale scores between natives and mi- lower perceived helplessness scores with a small effect size.
grants, natives scored significantly lower on both the PHS Although this finding is modest, it is at odds with previous
(t2494 = − 2.89, p < .01) and the PSES (t2485 = − 2.63, p < .01). literature, which generally show that first-generation mi-
Cohen’s d effect size for PHS (d = −.17) and PSES (d = −.16) grants report higher rates of mental distress compared to
show irrelevant effects although the t-tests are significantly natives and second-generation migrants [10, 11]. Factors of
different. In comparing first- and second-generation mi- language skills, unemployment, perceived discrimination,
grants, first-generation migrants score significantly lower integration or acculturation, and cultural barriers often re-
on the PHS scale than second-generation migrants (t320 = veal differences between first- and second- generation mi-
1.67, p < .05), albeit a Cohen’s d of only .22 reveals a small grants [10, 11, 42]. It is important to note the sample size
effect size. No significant difference of the PSES was found. of the current study, as first-generation migrants make up
a very small proportion of total migrant sample (n = 76).
Discussion While the second-generation sample may still emulate
Lee [15] called for validating the PSS in representative representative characteristics (n = 252), it is unlikely that 76
populations and Klein et al. [23] responded by testing migrants represent the linguistic, religious, and cultural
the German version PSS-10 in a representative German diversity of the first-generation migrants residing in
population. The current study took the analysis in the Germany. For many first- and second- generation migrants,
same sample one step further by testing whether the the cultural background plays an important role. For in-
German version PSS-10 is also a reliable and valid meas- stance, migrants of Turkish origin report higher rates of de-
ure of perceived stress in a migrant sample. pression, anxiety, and suicidal ideation compared to Polish
Overall the PSS-10 maintains a two-factor structure origin migrants and German natives [10]. Similarly among
that is strictly invariant between German natives and first-generation migrants, Tibubos et al. [11] illustrate that
those with a Turkish origin report higher rates of mental
Table 4 Change in goodness of fit MGCFA distress symptoms compared to other geographical origins.
Δχ2 (df) ΔCFI ΔTLI ΔSRMR ΔRMSEA Self- and group- attribution as migrants are additional fac-
Metric 8.8(8) 0 0.007 .001 −0.004 tors that may influence not only differences between mi-
Scalar 16.7(8) −0.001 0.005 .001 −0.003 grants and native Germans, but also differences between
Strict 11.5(10) 0 0.006 0 −0.004 first- and second-generation migrants. Nesterko et al. [43]
Note: showing the difference of the model from the model explored attribution as a migrant as a predicting factor of
before (metric = metric-config) PTSD, depression and anxiety, finding that many second-
Bastianon et al. BMC Psychiatry          (2020) 20:450 Page 7 of 9
generation migrants do not self-attribute as a migrant. on the research focus, furthermore one must take caution
These types of factors are likely to influence perceptions of when interpreting means of the total score as well as co-
discrimination and ultimately levels of perceived stress. variances or correlations with other concepts.
With respect to the PSS-10 factor structure, the on- Although the PSS-10 was not intended to operate as a
going one- or two-factor debate led this study to test the diagnostic measure [44], the scale may signal early signs
scale factor structure using an EFA specifically in the of mental distress for a broad range of stressful living
migrant sample. Results showed a two-factor solution conditions. Perceptions of helplessness and self-efficacy
with good factor loadings, thus reflecting the established are key factors in stress processes and stress regulation
two-factor structure presented by Hewitt et al. [16] and [25]. In clinical settings, it could be used to adjust stress
Roberti et al. [17], where negatively worded items repre- management strategies, understand worsening symp-
sent factor 1 (perceived helplessness) and positively toms due to stress, and facilitate the development of
worded items make up factor 2 (perceived self-efficacy). coping skills [13, 17, 18]. Schneider et al. [25] support
Item-rest correlation values are generally consistent the use of the German PSS-10 in clinical populations as
across the native and migrant samples, emphasizing that a first-screening instrument. Potential clinical applica-
items fit to the total scale similarly across the samples. tions of the German PSS-10 are relevant for residents
Separate group-specific CFA and combined MGCFA with a migration background as it may identify individ-
model fit indices were in line with the recommended uals with heightened levels of stress, indicating potential
values of Schermelleh-Engel, Moosbrugger, & Müller risk factors for other mental health concerns.
[35]. Therefore, the current two-factor structure shows Overall, the German version PSS-10 is a valid, strictly in-
adequate model fit in the native and migrant samples. variant and reliable measure of perceived stress among Ger-
As for measurement invariance, the two-factor model is man natives and residents with a migrant background.
strictly invariant between German natives and migrants. These results are valuable as implications of the PSS-10 can
Given that χ2 is sensitive to large sample sizes and viola- aid German health officials as well as stress researchers to
tions of normality [33, 36, 37], with a total sample of N = understand stress in migrant populations in Germany.
2523, it was expected to be significant showing poor
model fit. Chen [37] discusses similar biases caused by Strengths & Limitations
large or unequal sample sizes in evaluating measurement To the authors’ knowledge this is the first study to evalu-
fit indices; as with increasing sample size, the standard de- ate and compare the psychometric properties of the Ger-
viations of fit indices will decrease leading to higher man version PSS-10 between native Germans and a
chances of rejecting the model. Additionally, unequal sam- migrant sample. One major strength of this study stems
ples may conceal non-invariance as the larger group will from the total sample size and quality of the nationally
dominate calculations of many fit indices, leading to small representative data. Of the many studies that translated
changes between invariance models [37]. This is likely the and standardized the PSS-10, most studies used university
case with the current samples, ultimately showing very student samples [15] or very specific samples [20–22].
low changes in all fit indices across models. Even so, con- Although results of the MGCFA show acceptable fit and
sidering that the data violates normality and has a large, strict measurement invariance, the stark difference in mi-
particularly unequal, sample size, ΔCFI, ΔSRMR and grant (n = 328) and native (n = 2195) sample size should
ΔRSMEA still meet strict invariance. raise caution. As mentioned above, there are associated
Despite this study’s support of the two-factor structure biases that come with large and unequal samples [37]. In
and results of others showing independent explanatory ideal circumstances, factor analyses should further separate
power of the individual factors [24], the general PSS total first- and second-generation migrants in comparison to
scale also consistently demonstrated good Cronbach alpha German natives. Previous studies have presented stark dif-
values offering a simplified interpretation of perceived ferences between first- and second-generation migrants [10,
stress [15]. Klein et al. [23] report good internal 11] where second-generation migrants are more comparable
consistency for the German version PSS-10 total scale to German natives and seen as “adjusted” rather than first-
(Cronbach alpha = .84). Cohen and Williamson [14] chal- generation migrants [10]. Therefore, by merging both
lenge the content relevance of the second factor and Reis groups together there is a potential weakened effect of mi-
et al. [24] rather argue for a bifactor solution where each gration status in comparison to German natives. However,
factor is content relevant. Given these contradicting re- due to small sample size, factor analysis on migrant groups
sults it is difficult to recommend using one single solution was deemed inappropriate [33, 34]. This small sample size,
for all analyses – rather, while many studies find two fac- specifically of first-generation migrants, is unlikely to repre-
tors, the total PSS score may still be appropriate. There- sent the diversity of first-generation migrants in Germany.
fore, as Reis et al. [24] elaborate, the decision of In a similar vein, language skills required to complete
implementing the total score or sub-scales depends highly the interview process may have led to a sampling bias
Bastianon et al. BMC Psychiatry          (2020) 20:450 Page 8 of 9
towards first-generation migrants with adequate German Abbreviations
language. Additional cultural aspects need to be consid- PSS: Perceived Stress Scale; PSES: Perceived Self-Efficacy; PHS: Perceived
Helplessness; EFA: Exploratory Factor Analysis; MGCFA: Multi-Group
ered such as religious backgrounds or country of origin, Confirmatory Factor Analysis; Baua: Bundesanstalt für Arbeitsschutz und
as they are expected to have an influence on not only Arbeitsmedizin [Federal Institute for Occupational Safety and Health];
the outcome of scales but also the understanding of WHO: World Health Organization; MI: Measurement Invariance; ML: Maximum
likelihood; CFA: Confirmatory Factor Analysis; Χ2: Chi-square;
scales [8, 9, 11]. Lastly, the migrant sample was signifi- SRMR: Standardized root mean square residual; RMSEA: Root mean square
cantly younger (by 3 years) compared to the German na- error of approximation; CFI: Comparative Fit Index; TLI: Tucker-Lewis Index
tives, which may also impact the validity of the scale, as
age is continuously shown as an influencing factor of AcknowledgementsNot applicable.
perceived stress [19, 21, 23].
Authors’ contributions
CB contributed to the analysis and drafting the manuscript. KP contributed
to critical revisions of the manuscript, and study supervision. MB, EK, AT and
Future research EB each contributed to the study concept and design, acquisition of the
Future research must critically evaluate the PSS-10 two- data and revision of the manuscript. All authors read and approved the final
factor structure. With continued support for two-factors, manuscript.
the research community must then adapt to reporting Funding
the two factors separately rather than the total score The authors have no support or funding to report. Open Access funding
value. Reis et al. [24] recommends that researchers provided by Projekt DEAL.
understand their use of the scale in determining whether Availability of data and materials
to use it with a one- or two-factor structure. The data used in this manuscript is not publicly available but may be
In efforts to advance stress research in migrant popu- available from Prof. Brähler upon reasonable request.
lations, the PSS-10 should be further tested for its psy-
Ethics approval and consent to participate
chometric properties among first- and second- This study, including the consent procedure, was approved by the
generation migrant samples, as the two migrant groups institutional ethics review board of the University of Leipzig (Az 063–14-
may experience stress differently. Additionally, taking 10032014). Furthermore, the study adhered to ICH-GCP guidelines as well as
the guidelines of the ICC/ESOMAR International Code of Marketing and So-
into account one’s migration status (e.g. asylum seeker, cial Research Practice. All participants were informed of the study proce-
refugee, economic migrant, international student etc.) dures, data collection and anonymization of all personal data. Parents
would expand the current understanding of how various provided verbal and written informed consent for all minors. Moreover, a de-
tailed data privacy statement was delivered by the study assistant. According
migrant classifications influence not just one’s perceived to German law, all participants provided verbal informed consent, which was
stress but also the validation of the PSS-10. Representa- noted by the trained interviewer before starting with the survey.
tive studies would also benefit from more detailed ac-
Consent for publication
counts of migrant group representation and should Not applicable.
therefore include all migrant groups within a country in
data collection strategies. Lastly, future research needs Competing interests
The authors declare that they have no competing interests.
to not only focus on acute perceived stress but also the
more dangerous chronic forms of stress. Thus, testing Author details
1
the psychometric properties of chronic stress scales such Medical Psychology and Medical Sociology, University Medical Center of the
Johannes Gutenberg, University Mainz, Mainz, Germany. 2Department of
as the Trier Inventory of Chronic Stress (TICS) in refu- Psychosomatic Medicine and Psychotherapy, University Medical Center of
gee and migrant populations would further advance the Johannes Gutenberg University Mainz, Mainz, Germany.
stress research.
Received: 30 April 2020 Accepted: 3 September 2020
Conclusion References
This study tested the validity and reliability of the two- 1. WHO. The European Mental Health Action Plan 2013-2020. Copenhagen;
2015. http://www.euro.who.int/pubrequest. Accessed 5 Sep 2019.
factor structure of the German version PSS-10 and 2. Lohmann-Haislah A. Stressreport Deutschland 2012 - Psychische
measurement invariance in native and migrant subsam- Anforderungen, Ressourcen und Befinden. Berlin; 2012. www.baua.de.
ples. The key conclusions are: (1) the German version Accessed 11 Sep 2019.
3. Castles S, De Haas H, Miller MJ. The age of migration: international
PSS-10 is suitable for comparing German residents with population movements in the modern world. 5th ed. Houndmills, New
and without a migration background. (2) Despite good York: Palgrave Macmillan; 2014.
internal consistency for the total scale, the PSS-10 mea- 4. Massey DS, Arango J, Hugo G, Kouaouci A, Pellegrino A, Taylor JE. Worlds in
motion: understanding international migration at the end of the
sures two aspects: (a) perceived helplessness and (b) per- millennium. Oxford: Clarendon Press; 2009.
ceived self-efficacy. Future research would profit from 5. Statistisches Bundesamt. Foreign population by selected citizenships and
analyzing the two subscales separately, not only using years. 2019. https://www.destatis.de/EN/Themes/Society-Environment/
Population/Migration-Integration/Tables/foreigner-citizenship-time-serie.
the total score. html. Accessed 3 Apr 2020.
Bastianon et al. BMC Psychiatry          (2020) 20:450 Page 9 of 9
6. Statistisches Bundesamt. Foreign population by place of birth and selected 29. Bernaards CA, Jennrich RI. Gradient projection algorithms and software for
citizenships. 2019. https://www.destatis.de/EN/Themes/Society-Environment/ arbitrary rotation criteria in factor analysis. Educ Psychol Meas. 2005;65:676–
Population/Migration-Integration/Tables/foreigner-place-of-birth.html. 96 https://doi.org/10.1177/0013164404272507.
Accessed 3 Apr 2020. 30. Kelley K. MBESS: the MBESS R package. 2019. https://cran.r-project.org/
7. IOM. Manual on Community-based mental health and psychological package=MBESS.
support in emergencies and displacement. Geneva: IOM; 2019. 31. Hardt J, Herke M, Leonhart R. Auxiliary variables in multiple imputation in
8. Berry JW. Acculturation: living successfully in two cultures. Int J Intercult regression with missing X: a warning against including too many in small
Relations. 2005;29:697–712. https://doi.org/10.1016/J.IJINTREL.2005.07.013. sample research. BMC Med Res Methodol. 2012;12:184. https://doi.org/10.
9. Haasen C, Demiralay C, Reimer J. Acculturation and mental distress among 1186/1471-2288-12-184.
Russian and Iranian migrants in Germany. Eur Psychiatry. 2008;23:10–3. 32. Satorra A, Bentler PM. Corrections to test statistics and standard errors in
https://doi.org/10.1016/S0924-9338(08)70056-7. covariance structure analysis. In: von Eye A, Clogg CC, editors. Latent
10. Beutel ME, Jünger C, Klein EM, Wild P, Lackner KJ, Blettner M, et al. Depression, variable analysis: applications for developmental research. Thousand Oaks,
anxiety and suicidal ideation among 1st and 2nd generation migrants - results CA: SAGE; 1994. p. 399–419.
from the Gutenberg health study. BMC Psychiatry. 2016;16:1–10. 33. Iacobucci D. Structural equations modeling: fit indices, sample size, and
11. Tibubos AN, Beutel ME, Schulz A, Klein EM, Brähler E, Michal M, et al. Is advanced topics. J Consum Psychol. 2010;20:90–8. https://doi.org/10.1016/j.
assessment of depression equivalent for migrants of different cultural jcps.2009.09.003.
backgrounds? Results from the German population-based Gutenberg health 34. Gorsuch R. Exploratory factor analysis: its role in item analysis. J Pers Assess.
study (GHS). Depress Anxiety. 2018;35:1178–89. https://doi.org/10.1002/da.22831. 1999;68:532–60.
12. Tibubos AN, Kröger H. A cross-cultural comparison of the ultrabrief mental 35. Schermelleh-Engel K, Moosbrugger H, Müller H. Evaluating the fit of
health screeners PHQ-4 and SF-12 in Germany. Psychol Assess. 2020;32(7): structural equation models: tests of significance and descriptive goodness-
690–7. https://doi.org/10.1037/pas0000814. of-fit measures. Methods Psychol Res Online 2003;8:23–74. https://www.
13. Cohen S, Kamarck T, Mermelstein R. A global measure of perceived stress. J dgps.de/fachgruppen/methoden/mpr-online/issue20/art2/mpr130_13.pdf.
Health Soc Behav. 1983;24:385–96. https://doi.org/10.2307/2136404. Accessed 8 Feb 2019.
14. Cohen S, Williamson G. Perceived stress in a probability sample of the 36. Milfont TL, Fischer R. Testing measurement invariance across groups:
United States. In: Spacapan S, Oskamp S, editors. The social psychology of applications in cross-cultural research. Int J Psychol Res. 2010;3:111.
health: Claremont symposium on applied social psychology. Newbury Park, 37. Chen FF. Sensitivity of goodness of fit indexes to lack of measurement
CA: Sage; 1988. p. 31–67. invariance. Struct Equ Model. 2007;14:464–504.
15. Lee E-H. Review of the Psychometric Evidence of the Perceived Stress Scale. 38. Nunnally JC, Bernstein IH. Psychometric theory. 3rd edition. New York, NY:
Asian Nurs Res (Korean Soc Nurs Sci). 2012;6:121–7. https://doi.org/10.1016/j. McGraw-Hill, inc.; 1994.
anr.2012.08.004. 39. McDonald RP. Test theory: a unified approach. Mahwah, New Jersey:
16. Hewitt PL, Flett GL, Mosher SW. The perceived stress scale: factor structure Lawrence Erlbaum Associates, Inc.; 1999.
and relation to depression symptoms in a psychiatric sample. J 40. Padilla MA, Divers J. Bootstrap Interval Estimation of Reliability via
Psychopathol Behav Assess. 1992;14:247–57. https://doi.org/10.1007/ Coefficient Omega. J Mod Appl Stat Methods. 2013;12:13. https://doi.org/10.
BF00962631. 22237/jmasm/1367381520.
17. Roberti JW, Harrington LN, Storch EA. Further psychometric support for the 41. Cohen J. A power primer. Psychol Bull. 1992;112:155–9.
10-item version of the perceived stress scale. 2006. doi:https://doi.org/10. 42. Aichberger MC, Bromand Z, Rapp MA, Yesil R, Montesinos AH, Temur-Erman
1002/j.2161-1882.2006.tb00100.x. S, et al. Perceived ethnic discrimination, acculturation, and psychological
18. Taylor JM. Psychometric Analysis of the Ten-Item Perceived Stress Scale distress in women of Turkish origin in Germany. Soc Psychiatry Psychiatr
2014. doi:https://doi.org/10.1037/a0038100. Epidemiol. 2015;50:1691–700. https://doi.org/10.1007/s00127-015-1105-3.
19. Lesage F-X, Berjot S, Deschamps F. Psychometric properties of the French 43. Nesterko Y, Friedrich M, Brähler E, Hinz A, Glaesmer H. Mental health among
versions of the perceived stress scale. Int J Occup Med Environ Health. 2012; immigrants in Germany - the impact of self-attribution and attribution by
25:178–84. https://doi.org/10.2478/s13382-012-0024-8. others as an immigrant. BMC Public Health. 2019;19:1697. https://doi.org/10.
1186/s12889-019-8060-y.
20. Khalili R, Sirati Nir M, Ebadi A, Tavallai A, Habibi M. Validity and reliability of
44. Helios M, Rosario D. Package “pwr” title basic functions for power analysis.
the Cohen 10-item Perceived Stress Scale in patients with chronic
2020.
headache: Persian version. Asian J Psychiatr. 2017;26:136–40. https://doi.org/
10.1016/J.AJP.2017.01.010.
21. Remor E. Psychometric properties of a European Spanish version of the Publisher’s Note
perceived stress scale (PSS). J Psychol. 2006;9:86–93. Springer Nature remains neutral with regard to jurisdictional claims in
22. Chaaya M, Osman H, Naassan G, Mahfoud Z. Validation of the Arabic published maps and institutional affiliations.
version of the Cohen perceived stress scale (PSS-10) among pregnant and
postpartum women. BMC Psychiatry. 2010;10:111. https://doi.org/10.1186/
1471-244X-10-111.
23. Klein EM, Brähler E, Dreier M, Reinecke L, Müller KW, Schmutzer G, et al. The
German version of the perceived stress scale-psychometric characteristics in
a representative German community sample. BMC Psychiatry. 2016;16.
https://doi.org/10.1186/s12888-016-0875-9.
24. Reis D, Lehr D, Heber E, Erbert DD. The German version of the perceived
stress scale (PSS-10): evaluation of dimensionality, validity, and
measurement invariance with exploratory and confirmatory Bifactor
modeling. Assessment. 2019;26:1246–59. https://doi.org/10.1177/
1073191117715731.
25. Schneider EE, Schönfelder S, Domke-Wolf M, Wessa M. Measuring stress in
clinical and nonclinical subjects using a German adaptation of the
perceived stress scale. Int J Clin Heal Psychol. 2020.
26. R Core Team. R: A language and eviornment for statistical computing. 2018.
https://www.r-project.org/.
27. Rosseel Y. Lavaan: an R package for structural equation modeling. J Stat
Softw. 2012;48:1–36 http://www.jstatsoft.org/v48/i02/.
28. Revelle W. psych: Procedures for Psychological, Psychometric, and Personality
Research. Evanston, IL; 2019. https://cran.r-project.org/package=psych.