In Australia, the Australian Psychological Society (APS) conducted a ‘State-of-the-Nation’ Stress & Well-Being Survey (SWBS) from 2011 to 2015 to investigate stress at a national level [
1,
2]. The results showed that almost two in three Australians (64%) reported that stress was impacting their mental health, while approximately one in five (17%) reported that stress was strongly impacting their physical health [
3]. The findings from the last survey, which had 1731 respondents, indicated that compared to 2011 the levels of stress increased, and the levels of well-being decreased in the Australian population. One of the concerning findings was that, among those with severe levels of distress, 61% drank alcohol, 41% gambled, 40% smoked and 31% used recreational drugs as a coping mechanism [
2]. The surveys also revealed gender differences. Women were consistently more affected by stress than men and reported financial and health issues as their main sources of concern [
3].
One of the many psychological instruments used in the SWBS was the Perceived Stress Scale (PSS) [
4]. The PSS is the world’s most widely used instrument to measure perceived stress [
5] and since its development has been continuously applied in empirical research [
6,
7]. The PSS was developed based on the theoretical perspective of Lazarus [
8], which rather than focusing on external environmental stressors, postulated that the stress response is determined by the
perception of these environmental stressors. According to Lazarus [
8], life events, such as divorce or losing a job, only cause stress when they are
appraised as threatening (e.g. “I don’t have another job”
) and there is a
perception of insufficient coping resources (e.g. “I don’t know anyone who could employ me”). The measurement of stress has then been operationalized in two ways: the
environmental perspective (e.g. using life-event scales) and the
psychological perspective (e.g. using perceived stress scales) [
9,
10]. The PSS was developed to measure stress from the
psychological perspective, diverging from the life-event scales regularly used at that time [
11]. The initial validations conducted by Cohen [
4,
12] led to the creation of two shortened scales derived from the original 14 item-version (PSS-14): the PSS-10 and the PSS-4.
The results of the SWBW surveys were nationally reported by the Australian media (see “Australian women feel more stressed than men, mental health survey finds” [
13]). However, the reports did not specify which PSS version was used and indicated only that the “level of stress was derived by summing the scores of the 11 scale items” [
2]. Additionally, evidence of validity was not provided. Considering the high levels of stress reported in the Australian population, it is necessary to ensure that psychological measures applied to measure stress in Australians are valid and reliable, so it is possible to have confidence in the interpretation of test results. In the present study, we aim to investigate the psychometric properties of the PSS-14 in the general Australian population and examine whether this instrument can provide a valid measure of perceived stress for future research. To evaluate the PSS-14 validity and reliability we used data collected for the Australia’s National Survey of Adult Oral Health (NSAOH) 2004–2006, a broad project originally aimed to determine the psychosocial determinants of oral health in Australia. Despite being conducted prior to the SWBW, the NSAOH 2004–2006 has a large national sample (
n = 3857) that can provide evidence of the PSS-14 validity in the Australian general population.
The present research
The psychometric properties of the PSS have been evaluated in multiple countries [
14]. There are, however, two main limitations regarding the generalizability of its psychometric properties to an Australian population. Firstly, the majority of studies evaluated the PSS-14 in small and/or non-representative samples [
14]. For example, in China, the PSS-14 was evaluated in a sample of 1860 cardiac patients who smoked [
15], while the PSS-10 was evaluated in a sample of policewomen [
16]. Secondly, other studies were conducted in countries culturally and economically diverse from Australia, such as the application of the PSS-10 to 479 adults in Thailand [
17], a country known for its “collectivist Eastern culture” [
18]; or the application of the PSS-14 to 941 adults in Greece [
19], which recently experienced financial crisis [
20]. Among all countries studied, Canada is the western developed nation most similar to Australia due to its “large geography, low population density and similar health care challenges” [
21]. However, the PSS-14 was initially applied in Canada to 96 psychiatric patients [
22] and the PSS-4 was later evaluated in 217 pregnant women [
23]. The peculiarity of the samples from Canada (i.e. psychiatric patients) and most countries in general makes it difficult to generalize the results to typical members of the Australian general population. For the most part, the PSS has been validated in samples experiencing stressful environments (i.e. patients, students, policemen) rather than in general populations [
14].
The most relevant study in a population similar to Australia continues to be the validation conducted by Cohen and Williamson [
12] in a representative sample of 2387 Americans. Both countries, Australia and United States (US), are large high income countries [
24], with a history of English colonization [
25] and populations with similar demographic characteristics [
26] and morbidity patterns [
27,
28]. Nevertheless, there are important dissimilarities in terms of social-political context between these countries. For example, in the US, the national health system is a private employer-based and individual insurance program that provides coverage to 90% of the population, while Australia has a universal public insurance program covering 100% of the individuals [
26]. Although finances are the main source of stress both in Australia [
2] and the US [
29], these are structural differences regarding how these environmental stressors are experienced by each population (i.e. concerns with
health costs are more prominent in the US).
One important characteristic of the Australian population is the cultural background of its Indigenous groups, namely Aboriginal Australians and Torres Strait Islanders (ABTSI). The Aboriginal Australians experiences of well-being are rather distinct from western individuals [
30] and “Western psychological concepts are inappropriate and potentially damaging to Indigenous people” [
31]. One example is the PSS-14, which was recently validated for an Aboriginal population and the findings showed a weak latent correlation between the “Perceived Stress” and “Perceived Coping” subscales (
r = 0.14), a result distinct from the moderate (
r = 0.50) to strong (0.70) correlations found in western societies [
32]. For these reasons, we followed the recent recommendations by Kowal, Gunthorpe [
31] and Santiago, Roberts [
32] that ABTSI are a culturally distinct group in which psychological instruments should be evaluated separately from the general Australian population.
Hence, the present study aims to (1) investigate the psychometric properties of the PSS-14 in the general Australian population. We hypothesize that the functioning of the PSS-14 in the Australian population is similar but not equal to its functioning in other high-income countries. In addition, we aim to (2) updated the evidence about the PSS-14 functioning in developed countries using a large national sample and (3) further advance the knowledge regarding the PSS psychometric properties using item-response theory to investigate issues of differential item functioning (DIF) and local dependence (LD). The previous research about stress in Australia showed that “Australian women feel more stressed than men” [
13]. Although this result is common in many western countries, a long-established questioning is whether those differences are due to measurement bias [
14,
33]. Therefore, we aim to (4) investigate gender difference in PSS scores, and whether differences were due to measurement bias.
Finally, we aim to evaluate criterion validity by inspecting convergence and divergent validity with two psychological constructs (social support and stress at work) of the perceived stress’ nomological network [
34]. Social support has been shown by a large body of research as a protective (or
buffering) factor against stress [
35]. Social support refers to the functions performed by family, friends, and significant others when an individual encounters an external environmental stressor [
36]. In this case, family, friends or significant others can help to change the situation (e.g. helping with a task at work) or change
the meaning of the situation (e.g. help interpreting the event from a less distressing or extreme perspective) [
37]. In both cases, the individual has additional resources to deal with the
environmental stressor and this decreases his
perception of how stressful the situation is [
38].
On the other hand, psychological stress can be experience at work due to a demanding environment. One theoretical model that explains how the work environment generates stressful experiences is the
effort-reward imbalance [
39]. The model indicates that when the rewards received at work did not correspond to the efforts employed (‘high cost/low gain’), the imbalance can lead adverse stress responses [
40]. Therefore, it is expected that participants with high perceived stress will have low social support from friends, family and significant others and experience more efforts with less rewards at work.
To achieve these aims, we analysed data from Australia’s National Survey of Adult Oral Health (NSAOH) 2004–2006, a broad project originally designed to determine the psychosocial determinants of oral health in the Australian population. The NSAOH was chosen since it provides the best available data for the evaluation of the PSS-14 validity in the Australian population. Firstly, the NSAOH sample comprises the largest national Australian sample (
n = 3857) in which the PSS-14 has been applied. Secondly, the NSAOH achieved high standards of response quality for surveys [
41], including high response rates (77.4%) [
42] and low missingness of individual items (0.0 to 1.3%). Survey response rates have declined over the decades, with average rates below 50% been consistently reported since the 1990s [
43]. In summary, the large sample recruited at a national level and the high-quality PSS-14 item responses qualified the NSAOH as the preferred data for our research question.