Introduction

Stress, defined as the relationship between the person and the environment that is regarded as personally significant and as taxing or exceeding one’s resources for coping,¹ affects people worldwide and represents a considerable health burden.^2,3 Stress has been associated with the development of unhealthy behaviors (e.g., alcohol consumption and cigarette smoking),^4,5 mental and physical health conditions,^6,7 and higher mortality risk.⁸

It is proposed that a sustained body response to stress can trigger inflammation processes that increase the risk of developing different health conditions including some types of cancer.⁹ This stress-induced response leads to increased norepinephrine levels which results in tumor inflammation and aberrant arachidonic acid metabolism. In addition, stress responses occurring via the autonomic nervous system affect the number and function of immune cells.¹⁰ Consequently, a number of studies have emerged in recent years evaluating the relationship between perceived stress and cancer.^11,12,13 Depending on the population, the type of cancer and the instrument used to measure it, the prevalence of perceived stress varies between 22 and 69%.^14,15

The Perceived Stress Scale (PSS) is one of the most widely used measures of self-reported stress.¹⁶ The scale originally included 14 items, but 10- and 4- item versions have also been developed.¹⁷ To date, PSS has been translated into several languages, including Spanish. The PSS offers a method to assess stress among cancer patients because it can provide clinical information about the extent to which patients consider their lives as stressful.^18,19 When utilized in stress models, the scale can possibly distinguish the role of perceived stress in important cancer outcomes; for instance, patients’ quality of life and adherence to treatment.^14,20,21 Due to the increased burden of cancer in Mexico and similar countries,²² understanding its association with perceived stress is of paramount importance. However, almost no studies have been performed to validate the use of the PSS among cancer patients in Mexico or Central American countries. As a first step towards that goal, the factor structure of the PSS requires assessment in this population.

To the best of our knowledge, only one study has assessed the factor structure of the PSS-4 in a Mexican population. This study was conducted among a large sample of men and women from Northern Mexico and supported a two-dimensional structure of the scale. Studies among non-Spanish speakers have reported mixed results.^23,24 Studies exploring the factor structure of Spanish versions of the PSS-10 have favored one-²⁵ and two-dimensional structures.²⁶ These studies have been limited to evidence from small samples,^27,28 and conducted predominantly among US-based Latino populations.^26,27 Further validation of this scale is warranted in different populations, such as women living with cancer, as individuals’ experiences with stress may vary across different social and cultural contexts.²⁹

Therefore, we aimed to examine the factor structure of the Spanish version of the PSS-4 and 10 in a large sample of Mexican women participating in a prospective cancer cohort study.

Materials and methods

Study population

Data for this analysis comes from female teachers participating in the Mexican Teacher’s Cohort (MTC).³⁰ The MTC is an ongoing prospective cohort study aimed at assessing the association of lifestyle factors with chronic non-communicable diseases, including cancer. At baseline, 2% of participants reported a cancer diagnosis and 5% had family cancer history. Between September 2012 and November 2013, a subsample of 2 230 women aged ≥40 years were invited to participate in a clinical assessment as part of an ancillary study on cardiovascular risk. The participants lived within 50 kilometers of five clinical sites in two states in Southern Mexico (Chiapas and Yucatán). A total of 1 625 (73%) attended a clinical appointment. Among them, 1 310 (81%) completed the PSS-10 and were included in the current analysis (figure 1). Among the study participants, 902 completed the Patient Health Questionnaire 9 (PHQ-9) in a follow-up between 2012 and 2013. All participants provided written consent. The study was approved by the Institutional Review Board of the National Institute of Public Health (INSP; project number 1221).

MTC: Mexican Teacher’s Cohort

PSS: Perceived Stress Scale

Figure 1 Study flowchart 

Results

Table I shows the descriptive characteristics of the participants. The mean (SD) age was 48.2 (4.3) years. Most women were non-indigenous, lived in urban areas, had at least a university-level education, were of medium to high socioeconomic level, and were married or living with a partner. The median (IQR) scores of PSS-10, PSS-4 and PHQ-9 were 14 (9-18), 4 (2-6) and 3 (0-6), respectively. The characteristics of PHQ-9 respondents vs. non-respondent are presented in Supplementary Table I.⁴⁰ Briefly, non-respondents were more likely to be indigenous, had a lower education level, were less likely to be married/cohabiting and, and were more likely to be in the highest tertiles of perceived stress.

Overall, there were no differences between the sub-samples used for the EFA and CFA (Supplementary Table II).⁴⁰ Results from the EFA are presented in table II and suggested a one-factor solution for PSS-4 and a two-factor solution for PSS-10. The unidimensional model for PSS-4 explained the variance by 49%, and all factor loadings were higher than 0.4. This model displayed a high model fit according to the TLI (0.72) as well as a moderate internal consistency (α= 0.72). For PSS-10. The first factor was defined by six items corresponding to negatively-worded items, and the second factor was defined by four items corresponding to the positively stated items. All items -except for item six- had factor loadings above 0.40, ranging from 0.49 (item 7) to 0.73 (item 3). The first factor accounted for 68.6%, and the second, for 18.7% of the variance. This model displayed a high model fit according to TLI (0.89) and a moderate internal consistency (α= 0.71 and 0.79). Table II also shows the EFA results of a one-factor solution for PSS-10.

Table II shows the fit statistics and standardized factor loadings for the competing CFA models tested in the split sample of 655 women. Results favored the two-factor structure of both PSS-4 and PSS-10. All but one of the factor loadings were above 0.4, and the two-factor solution for PSS-4 had the best fit to the data (RMSEA= 0.00; CFI= 1.00; NFI= 0.99). For PSS-10, the two-factor model showed lower yet adequate fit indices (RMSEA= 0.08; CFI= 0.92; NFI= 0.90). Fit statistics for one-factor models are also shown in table III.

Internal consistency was high for the full scale in both 4- and 10-item versions (α= 0.72 and 0.83, respectively) and slightly lower for each factor of the two-dimensional solutions (ranging from 0.69 to 0.79).

Also, we compared PSS-10 and -4 scores and found that Spearman’s correlation was positive and strong (r=0.91, p<0.001), and the agreement between score tertiles was substantial (Kappa=0.66, p<0.001).

We found a positive correlation between PSS and PHQ-9 scores (r=0.41, p<0.001 for PSS-4, and r=0.46, p<0.001 for PSS10). Correlations for each of the factors in the two-factor model and the PHQ-9 score were also positive and significant (PSS-4: r=0.41, p<0.001 for factor 1, and r=0.29, p<0.001 for factor 2; PSS-10: r=0.47, p<0.001 for factor 1, and r=0.32, p<0.001 for factor 2)

Discussion

This study adds to the sparse literature assessing the factor structure of the Spanish version of the PSS-4 and PSS-10 among Mexican women participating in a prospective cancer cohort study. Overall, our results support two-factor solutions for both PSS-4 and PSS-10, with factor loadings similar to those previously reported in the literature.^23,41,42,43 Interestingly, the two-factor model of PSS-4 not only showed acceptable internal consistency but the highest fit indices and a positive correlation with depressive symptomatology. Furthermore, the correlation between the PSS-4 and 10 scores was high.

The PSS is a tool used to assess perceived stress, and in recent studies has been administered to cancer patients. In the present study, the Spanish versions of the PSS-4 and PSS-10 showed adequate factor properties to evaluate psychological stress and related depressive symptomatology. This suggests that both PSS-4 and PSS-10 might be used to assess stress-related symptoms in Mexican female teachers and could be useful as screening tests in longitudinal studies.

Studies assessing the factor structure of the Spanish version of the PSS-10 are in partial agreement with our study. González-Ramírez and colleagues²⁵ found a two-factor structure of the PSS-10 in a sample from Northern Mexico with factor loadings and item distributions similar to those reported in our study. Perera and colleagues,²⁶ on the other hand, reported a bifactor model among a Hispanic/Latino sample residing in the United States. In this bifactor model, all positive items loaded onto a “General perceived stress” factor, and the four-reverse-worded items load onto a “Reverse-worded nuisance” factor. However, Baik and colleagues²⁷ found a better fit for the bifactor model only for English-speaking participants, but not for Spanish-speaking participants, among 436 Hispanic Americans who answered the PSS-10, in which the data were best explained by the two-factor model. In our study, the two-factor model showed the best fit to the data. Despite the analytical support behind the two-factor solution of the PSS-10, most of the studies conclude that a unidimensional conceptualization of the scale relies on better theoretical support for measuring perceived stress.²³ As suggested by the PSS developers, this is because any distinction between multiple scale factors is irrelevant¹⁶ and only reflects the sentence structure of the scale.

The psychometric properties of PSS-4 have been less explored, and results have also been inconsistent in regard to its factor structure.²³ However, our findings are in line with those of the only other study that assessed the factor structure of the Spanish version of this scale²⁵ and with those reported in a large sample of cardiac patients in China where a two-factor solution was found to better fit the data.²⁴ This scale may also be appropriately interpreted as a unidimensional construct, given that the superficial nature of the reversed wording remains.

In Mexico and other Latin American countries, the scarcity of studies focusing on the association between stress and cancer contrasts with the increasing morbidity and mortality rates that have occurred in this region in recent years.⁴⁴ Evidence from other studies shows differences in regard to the association between stress and the development of cancer.^13,45 Stress may favor inflammatory processes related to excessive secretion of cortisol, generating alterations of the endocrine and immune systems, which in turn have been associated with the development of certain types of cancer.⁴⁶ The use of PSS-10 and PSS-4 in Latin America may contribute to the understanding of the role of stress in cancer incidence and survival.

We found a positive correlation between perceived stress and depressive symptomatology in the expected direction. This result is consistent with previous studies²³ and with the report by Cohen and colleagues, who state that “there is some overlap between what is measured by depressive symptomatology scales and by the PSS, since the perception of stress may be a symptom of depression”.¹⁶ The current literature suggests that stress exposure increases the risk for poor clinical outcomes across a variety of major health conditions, including depression.⁴⁷ A strong association between perceived stress and depression was found in a large sample of 229 293 community-dwelling adults from 44 low and middle income countries.⁴⁸ Likewise, a high correlation between PSS-10 score and depressive symptoms has been reported among older Vietnamese women.⁴¹ In a case-control study carried out in Poland, women with stress in their daily activities and with depression had a 3.7 times higher risk of developing breast cancer, compared to those who were not exposed to such stress.⁴⁹ In a recent study in China, perceived stress measured with the PSS was a predictor for depressive symptoms among oral cancer patients.⁵⁰ According to national estimates, in Mexico the prevalence of depressive symptoms is 15%, while detection coverage is 9.9%.⁵¹

The results of our study should be interpreted in light of its limitations. First, due to the nature of the study, men were not included in our sample. However, previous studies assessing the factor structure of the Spanish version of the PSS-10 have not found differences by gender.²⁵ Likewise, although this is a population-based study, the generalizability of its findings is limited since the sample only included teachers living in certain areas of Mexico. Finally, we only performed exploratory analyses evaluating the correlation between stress and depression symptoms. Future studies should consider examining how PSS constructs relate with other risk factors or domains associated with stress such as positive affect or life satisfaction and provide information on the scale’s concurrent and discriminant validity.

Despite these limitations, our results present important strengths and implications. First, analyses were performed with robust methods in a large sample with a population-based design. Additionally, the good fit to data and high factor loadings reported here suggest that the Spanish versions of the PSS-4 and PSS-10 are useful measures of perceived stress. We conclude that the Spanish versions of the PSS-4 and PSS-10 maintain psychometric properties suitable to assess perceived stress. For theoretical and practical reasons, their use as unidimensional scales (global score) is recommended. Both the PSS-4 and -10 are valuable research tools for assessing symptoms related to stress and may assist in the assessment of psychological stress in women. The PSS-4 may be more useful for large samples, including multicountry studies on stress and cancer, due to its simpler structure, low cost, and short application time.