Stress is a commonly reported trigger of smoking and relapse after quitting. Previous studies have shown that stress increases the amount of smoking, reduces the likelihood of successful quitting, and predicts relapse [1, 2]. Higher stress is associated with less self-efficacy in quitting and less in not smoking among working adults [3], and less predicted success in quitting prospectively [4].

Heart disease is a leading cause of death in both China and Hong Kong [5, 6]. Smoking is a modifiable health risk-factor for cardiac patients [7], but many of them continue to smoke regardless of the gain in health benefits from quitting including a reduction in both morbidity and mortality. There are calls for smoking cessation interventions which target smokers with high perceived stress levels [8]; this is particularly applicable to patients with chronic diseases who usually have high levels of perceived stress.

Cohen and colleagues developed the original 14-item English version of the Perceived Stress Scale (PSS-14) as a global measure of stress by asking respondents to report whether their lives seem to be unpredictable, uncontrollable or overloaded [9]. The PSS is one of the most frequently used tools to measure stress in chronic conditions and situations often not listed on other life-event scales, such as the Daily Hassles Scale [10] and Daily Stress Inventory [11]. The PSS is also available in two shortened versions of 10 items (PSS-10) and 4 items (PSS-4), with the items selected from the pool of 14 in the original version of PSS. The PSS is considered to be a brief scale measuring perceived stress that can be administered in a few minutes [12].

The scale has been translated into several languages, including Spanish, Swedish, Japanese and Chinese [13–16]. A few studies have examined the psychometric properties of the PSS, with five using principal component analysis (PCA) and three confirmatory factor analysis (CFA). The PCA studies provide important yet limited psychometric information on the PSS, as the statistical method has been criticised by methodologists [17]. Two CFA studies were conducted in clinical populations [18, 19] and one focused on college students [13], but mixed results on the structure of the PSS were reported. All the three CFA validation studies found a two-factor structure of the PSS, while others reported a one-factor structure, in particular for the 4-item version [12, 20]. Moreover, it is unclear whether all the three versions of the PSS have a similar two-factor structure as well as relationships with certain variables.

The PSS has been applied extensively as a health-related outcome in studies examining the effect of smoking cessation interventions [21, 22] and the Chinese version of the PSS has also been used in research areas such as mental health and physical activities [23, 24]. Although the Chinese version of PSS was found to have high internal consistency and test-retest reliability, its construct and concurrent validity have not yet been established among smokers in Chinese societies, precluding the confident use of the instrument in this large population of varying cultures. The goals of the present study were to assess the construct validity, reliability and concurrent validity of the three versions of the Perceived Stress Scale (PSS), and to compare their appropriateness when used among Hong Kong Chinese cardiac patients who smoke.

This study has explored and extended the factor structure of the Chinese version of the Perceived Stress Scale and its shorter variants (PSS-10 and PSS-4) from clinic populations in the West to a Chinese cardiac population with a smoking habit. The results show that construct validity, internal consistency and concurrent validity of the Chinese versions of PSS-14, PSS-10 and PSS-4 and their corresponding subscales were generally supported in a community-based sample of Chinese smoking adults with cardiac disease.

Our study provides evidence to support the manifestation of the PSS as two-dimensional, and that the two-factor models give better approximations of the data of PSS-14, PSS-10 and PSS-4 with items loaded on their designated factor. The results obtained are consistent with previous CFA studies using PSS-14 and PSS-12 [13, 18, 19], but contrary to those using PSS-4, which was hypothesised as one latent factor [12, 20]. The factor loading of item 11 of the negative factor was found in the present study to be far below 0.4 in both PSS-14 (0.27) and PSS-10 (0.26). The low factor loadings of this item compared with the rest of the negative factor could be due to the potential interpretation of 'anger' as expressing the perceived distress externally by the subjects. Alternatively, it could simply have been produced by the translation process but we could not find any study utilizing the Chinese PSS reporting the associated factor loadings for comparison and hence more studies are needed to verify this possibility. Indeed, a low factor loading of 0.28 for item 7 has also been reported among a sample of ambulatory care patients with asthma in the US [19]. Future studies are warranted to address equivalence in items of the PSS across sub-populations such as sex, ethnicity and education.

Cronbach's alpha values demonstrate that each of the two subscales of PSS-14 and PSS-10 are internally reliable and form a cohort set of the construct. Similar to previous studies conducted in the Asian region, including Japan, Singapore and Taiwan [24, 35, 36], high internal consistency reliability of the three overall scales of the PSS was also observed in the current study. Although the alpha value of this subscale of PSS-4 was below the cut-off point of 0.7, it has been argued that a reliability coefficient as low as 0.5 should not seriously attenuate validity [37] and the alpha coefficient increases with the instrument's length [38]; this subscale is therefore still considered reliable.

The subscales of the PSS correlate with smoking- and health-related variables in the anticipated directions except daily cigarette consumption, providing some preliminary evidence on the concurrent validity of the PSS scale in the three versions. Our results on the low correlations of the PSS with smoking-related variables were consistent with previous studies that low and insignificant correlations of the PSS scores with other smoking-related variables were also reported in Cohen and Lichtenstein's study among smokers [1]. In the large community-based study in the US, high stress was significantly associated with reports of increased tobacco consumption and less confidence in not smoking again but the magnitude of the association was also small after controlled for demographic variables [3]. Results from our study and the two previous studies show that PSS is associated with smoking-related variables but it might not act as a proxy for them.

The opposite correlations of the positive and negative factors of PSS with these variables provide further evidence of the manifestation of two components in PSS. The greater correlations with 'perceived distress' than with 'perceived coping' support the assertion that individuals react to a stressor or threatening event with a primary appraisal, and judge their ability to cope with the stressor or event with a secondary appraisal in the stress and coping model [39]. Further research will need to test the differential predictive validity of the two PSS subscales.

Consistent with earlier findings [40, 41], we have also found that women reported a significant higher total score for perceived stress. In addition, we have found women had significantly higher scores in the negative and lower scores in the positive subscales than men, regardless of the PSS version. This result is also consistent with previous findings, that women smoke more for tension reduction and relaxation [42, 43].

Our results show that the psychometric properties of all three versions are satisfactory and similar. However, it is important to consider which version of the PSS should be used in particular applications. As in a previous study [40], the PSS-10 not only provides an adequate measure of perceived stress and similar correlations with smoking- and health-related measures as the complete version, but has also shown a higher reliability among our Chinese patients. In addition, the PSS-4 also has satisfactory construct validity, but a somewhat lower reliability in the negative subscale. From a practical point of view, because a shorter questionnaire would be more appropriate for studies with multiple measures, we recommend the use of PSS-10 in future research to focus on the two components of perceived stress; and the use of PSS-4 if such separation is not essential and space for multiple measures is limited [44].

This study had several limitations. First, the present validation study of the PPS was part of a randomised controlled trial carried out among a group of cardiac outpatients who smoke. Given the large sample size, our results can be generalised to smoking cardiac patients, but not to those who do not smoke. Moreover, the large proportion of male participants in the study may limit its generalisability to female smokers. However, the potential of the current findings to be generalised is enormous, given that smoking in Hong Kong and elsewhere in Asia is strongly sex-linked with much higher smoking prevalence in men (men >40%, women <10%) [45] and that cardio-vascular disease is the number one killer in China [6]. Nevertheless, further research using population-based study and on different subgroups among the Chinese and other Asian populations is warranted. Second, the use of single-item measures of smoking and health-related variables in the study (except FTND) should include a substantial amount of measurement error in each of these variables. The imperfect measures of these variables however would lead to underestimation of the correlations with other variables, which could explain the low observed correlations with the PSS scores [46]. Third, we were unable to assess test-retest reliability of the PSS in the current study as it was a secondary data analysis of an RCT and such information were not collected at follow-up surveys. Further studies on the psychometric properties of the Chinese versions of the PSS should be carefully designed to include both precise measures of potential related variables to assess concurrent validity and a test-retest component to examine stability of the instrument.

The study corroborates the observations of others in demonstrating that the Chinese-translated PSS has adequate construct validity, reliability and concurrent validity, and hence is acceptable in a clinical smoking population with cardiac disease. Researchers may choose the shorter versions of PSS (PSS-10 and PSS-4) depending on the practical needs of their work.