How Different Stressors Affect Quality of Life: An Application of Multilevel Latent Class Analysis on a Large Sample of Industrial Employees

Introduction

Quality of life (QoL) is one of the most important outcomes in health-care programs.^1,2 QoL is an individual, multidimensional and latent concept that involves both objective and subjective factors, reflecting the physical, mental, and social health of a person.^3–5 The World Health Organization also defines QoL as: “individuals’ perceptions of their position in life in the context of the culture and value systems in which they live and in relation to their goals, expectations, standards and concerns“⁶

QoL influences the quality of work life and work characteristics such as productivity.^7–11 QoL has been found to be negatively associated with stress.^12,13 Increased stress levels are related to decreased job performance, higher medical costs, and impaired QoL.^14,15 A number of studies have reported significant association between psychological stressors and poor physical and mental health.^16,17 Furthermore, stressful life events are inversely related to QoL and well-being.^18–20 Stressful life events were defined by the occurrence of both positive and negative situations in an individual’s life that lead to significant changes in usual activities.¹⁸

The body’s response to stress is attributed to the activity of several axes, such as the hypothalamic-pituitary-adrenal (HPA) axis and sympathetic adrenomedullary circuits. Stressful life events through the HPA axis can affect the function of neuroendocrine and immune systems, physical and psychological well-being.^21–23 Long-term stress by changing many physiological systems, such as metabolic and inflammatory pathways and immunological defense systems in the body leads to function and cell senescence impairment.²⁴

Stressful life events have been identified as a risk factor for a variety of physical and mental illnesses such as cardiovascular disease, cancer, metabolic syndrome, and depression.^25–28

Stressful life events can also lead to changes in job-related behaviors such as reduced job involvement, impaired performance effectiveness, and reduced job satisfaction and eventually absenteeism, tardiness and turnover.²⁹

Numerous studies are available on the relation between QoL and stressful life events across different populations such as diseases-specific populations^30–32 and older adults^20,33 and less attention has been paid to evaluate it among the workforce or general population.^5,34

Furthermore, in many studies among the workforce population, the association between QoL and job stress has been widely studied but other stressors have been neglected.^35–38 However, individuals are exposed to multiple stressors which may affect their QoL. Understanding the relationships between QoL and stressors as well as job stress may help in planning interventions to reduce negative effects of stressors and consequently improve QOL among the workforce.

Moreover, most of the previous studies on the determinants of QoL have applied simple statistical analysis and the subscale scores were combined to obtain an overall, or total QoL score as a single measure. However, QoL is a multidimensional concept with distinct domains. To make statistical analysis about QoL as a multidimensional and latent variable, an advanced statistical procedure is required that can perform a comprehensive assessment of QoL in the study participants according to all five dimensions of QoL.

Latent class analysis (LCA)³⁹ as an advanced statistical method that provides such a comprehensive evaluation of QoL by classifying the subjects into mutually exclusive and exhaustive and homogenous groups. In the current study, an extended version of LCA was used to assess QoL while simultaneously addressing dependencies due to multilevel structure of study population (ie, employees as lower level units nested in job categories as a higher level).

To our knowledge, no large-scale study has focused on the relationship between stressors and QoL among industrial employees in Iran.

Therefore, the objectives of the present study were: (1) to classify the participants as well as the job categories based on QoL; and (2) to evaluate the association of stressful life events with QoL using multilevel latent class regression (MLLCR) analysis. This method provides a comprehensive feature about QoL through classifying employees objectively based on all domains of QoL and simultaneously enable us to evaluate the association of stressful life events with QoL classes in the presence of various confounders including demographic, lifestyle, and job-related variables.

Materials and Methods

Study Design and Participants

This cross-sectional study was conducted among full-time employees and contractual workers of the Esfahan Steel Company (workforce of 16,000). The sample size was determined based on an epidemiological survey of chronic diseases in manufacturing employees by Roohafza et al (ESCOME)⁴⁰ carried out to estimate the prevalence of psychological disorders (depression and anxiety) among the workforce. The sample size was estimated to be 3500 by considering 0.1, 0.05, and 0.01 as prevalence of psychological disorders, type one error rate, and sampling error rate, respectively.

Three thousand and sixty-three volunteers returned complete questionnaires (response rate: 0.87) and were included in the statistical analysis. The inclusion criterion was work experience for at least one year and being willing and agreeing to participate in the study. Volunteers who did not answer more than 10% of the questionnaire were excluded from the analysis.

We applied multistage cluster sampling, in which clusters were the main seven departments and their sections, stratified by job categories. Sample sizes in the clusters and strata were proportional to the size of respective departments. Due to a low number of women workers (n=800), we relied on convenience sampling to recruit women volunteers (n=260).

Demographic data were gathered through self-administered questionnaires at the company premises with the help of study coordinators, who also monitored questionnaire administration over the six months of data gathering. The data were quality checked for inclusion and exclusion criteria as they were entered in the computer for statistical analysis. The study design and its implementation has been presented elsewhere in more detail.^40,41

Employees are mainly divided into two categories including white-collar and blue-collar employees. Employees in managerial and supervisory positions were considered as high-level white-collar employees (4.9%) and office employees were considered as low-level white-collar employees (15.2%). Blue-collar employees were also classified as high-level (skilled workers) (16.4%) and low-level (semi-skilled and nonprofessional workers) (63.5%).⁴⁰

Employees based on their job groups, employee rank, and work conditions have been divided into 71 job categories, such as manager, technical supervisor, nontechnical supervisor, first class electrician repairman, second class electric repairman, first class mechanical repairman, second class mechanical repairman, first-class engineer, second class-engineer, senior engineer, firefighter, welder, driver, etc.

All participants were informed about the study protocol and signed informed consent. The Medical Research Ethics Committee of the Isfahan University of Medical Sciences approved the study protocol (project numbers #87115 and #395482). This study was conducted in accordance with the Declaration of Helsinki.⁴²

Study Instruments and Assessment of Variables

Quality of Life

A standardized and generic form of the Euro QoL-five dimensions (EQ-5D-3L)^43,44 was used for assessing QoL. This self-report instrument comprises two parts of the self-classifier (descriptive system) and the visual analog scale (VAS).

The EQ-VAS was considered as a single value for assessing of overall health status. This is a VAS, ranging from 0 (worst imaginable health) to 100 (best imaginable health). Respondents by marking the scale with a single vertical mark, rate their current health status. EQ-5D self-classifier describes health state of subjects in five domains: mobility, self-care care, normal activities, pain/discomfort, anxiety and depression. Each domain consists of three levels: no problem (1), some problems (2), severe problems (3). We combined the last two categories into a single category in LCA due to poor response rate in last category. Higher EQ-5D scores represent worse health status. The Iranian validated version of EQ-5D has been used in the current study and it has shown good reliability and validity.^45,46 We evaluated the internal reliability of the questionnaire in a pilot sample of 300 participants in current study and Cronbach alpha was 0.51. The EQ-5D score was able to differentiate patients according to the severity of the disease based on the EQ-5D VAS (P <0.0001). The ICC for EQ-5D and EQ-5D VAS were 0.753 and 0.896, respectively. The correlation of total score of the EQ-5D questionnaire with the total scores of SF-36 was (r=0.445, P<0.05) and dimensions of EQ-5D questionnaire with counterpart dimensions from SF-36 questionnaire were (r=0.37–0.45) indicating a moderate correlation.⁴⁷

Stressful Life Events

The number and intensity of experienced life stressors were measured by the Stressful Life Events (SLE) questionnaire.⁴⁸ Participants were asked about the occurrence of stressors within the past six months. The SLE is a 46-item self-administered scale which consists of eleven dimensions, including home life (measured with seven items), financial problems (five items), social relations (four items), personal conflicts (five items), job conflicts (four items), educational concerns (four items), job security (five items), loss and separation (four items), sexual life (four items), daily life (two items), and health concerns (two items). The items are rated on a six-point Likert scale (0: never, 1: very mild, 2: mild, 3: moderate, 4: severe, 5: very severe) and the higher score indicates higher stress level. Internal consistency of the SLE questionnaire was 0.92.⁴⁸

Assessment of Other Variables

Variables that were considered as potential confounders included demographics, age (years), gender (male/female), marital status (married/single), education (0–5 years/6–12 years/over 12 years), lifestyle variables, sleep duration (hours) and physical activity (hours per week), BMI (weight kg/height m²), job-related variables, job stress (effort-reward imbalance), and second job (yes/no). Physical activity was evaluated with the International Physical Activity Questionnaire (IPAQ), which included 11 questions. The internal reliability of this questionnaire was reported as good by Moghaddam et al⁴⁹ based on Cronbach’s alpha of 0.7 and Spearman–Brown correlation coefficient of 0.9.

Statistical Analysis

Two-level LCA^50–52 (employees as level one and job categories level two units) was employed to identify homogeneous latent classes of employee according to the their responses to five indicators of QoL. Multilevel LCA is an extension of traditional LCA that handles situations where there is a multilevel construct. In our study, employees (level one) are nested within job categories (level two).

The two-level latent class model not only classifies the employees but also the job categories based on the distribution of QoL of employees nested in job categories. The modeling process consists of the following steps: in the first step, with ignoring the nesting structure of data, LCA was used to classify individuals based on their response to the five items of the EQ-5D. To determine the appropriate number of classes at the employee-level, we started by fitting a one-class model and sequentially increased the number of classes until to yield the best fit. Bayesian Information Criterion⁵³ was used to select the best-fitting model. In addition, interpretability of the identified classes was also considered.

In the next step, two-level latent class model was estimated to take (the clustered nature of the data) into account the multilevel structure of our data. A two-level model with two classes at level one and two classes at the second level was selected as the best fit.

Finally, we used the MLLCR for evaluating the predictors of QoL. The proposed predictors of QoL in current study were stressors. For evaluating their association with QoL, at first, we used exploratory factor analysis on eleven domains of life stressors. Two factors were extracted and labeled as “personal stressors” and “socioeconomic stressors” and these domains were used as latent predictors of QoL of employees. The association of two extracted stressor domains with quality of life in the structure of MLLCR was quantified as OR and 95%CI for OR.

We adjusted the association of those potential confounding variables that were statistically significant at P<0.1 in univariate analyses.

Continuous and categorical variables were represented as mean (SD) and number (percentage) respectively. Independent samples t-test/Mann–Whitney U-test and chi-squared test were used for comparing continuous and categorical variables between studied groups, respectively. MLLCR was fitted in Mplus statistical software (Muthén and Muthén, Los Angeles, CA, USA; version 8).⁵⁴

Results

The current study consists of 3063 participants aged 21–64 mean (SD): 36.73 (7.30) years. Most of the participants were male (91.5%), 90% married and 29.4% of the employees had a university education.

Table 1 presents the two classes at the employees’ level based on five dimensions of QoL identified by LCA. The first class is characterized by higher probabilities of expressing “no problem” for all items of QoL. Accordingly, this class was labeled as high QoL (consisted of 80% participants). The second class is characterized by higher probabilities of expressing “some/severe problems” for items of mobility, pain/discomfort and anxiety/depression. Accordingly, we named this class as low QoL (consisted of 20% participants).

Table 1 Latent Classes of QoL Identified by Latent Class Analysis

Using exploratory factor analysis, two factors were identified from eleven life stressors (Table 2). Financial problems, social relations, personal conflicts, job conflicts, educational concerns, job security and daily life were loaded more on first factor. Accordingly, this factor was labeled as “socioeconomic stressors”. Home life, loss and separation, sexual life and health concerns stressors were highly loaded on the second factor. Therefore, this factor was labeled as ”personal stressors”.^41,55

Table 2 Factor Loadings of Two Extracted Factors from Stressful Life Events

Differences in demographic, life style and job-related variables between two extracted classes of QoL are shown in Table 3. These results indicated that the subjects who were classified as high QoL, were more likely to be younger, male, and have lower BMI and higher sleep duration than those in the low QoL class.

Table 3 Demographic Characteristics, Lifestyle, Job-related Variables of Participants in Two Classes of QoL

Differences between the two classes in terms of the scores of stressful life events domains and two extracted socioeconomic and personal stressor dimensions are reported in Table 4. Compared with those people in the high QoL class, individuals in the low QoL class had higher mean stress scores in all eleven stressor domains as well as socioeconomic and personal stressors (Table 4).

Table 4 Mean Scores of Stressful Life Events Across Two Classes of QoL

Figure 1 shows the two-level latent class model with two classes at level one (employees) and two classes in the second level (job categories) and the distribution of employee’s classes in terms of QoL at the classes of job categories. “Job categories with low QoL” consists of a higher percentage of employees with low QoL (27.2%) whereas a larger percentage of “job categories with high QoL” contains employees with high QoL (84.5%).

Figure 1 Distribution of level-one latent classes (n=3063 employees) within level-two latent classes (n=71 job categories).

The crude and adjusted ORs (95%CI for OR) for the association of socioeconomic and personal stressors and all domains of stressful life events with QoL in the presence of demographic, lifestyle and job-related confounding variables are presented in Table 5. Higher scores of socioeconomic (OR: 1.15, 95%CI: 1.13–1.17) and personal stressors (OR: 2.56, 95%CI: 2.02–3.24) are associated with increased risk of being in the poor QoL class in crude model. In this model an increase of one unit in socioeconomic or personal stressors’ score, increases the risk of being in the low QoL class by 1.15 and 2.56 times, respectively.

Table 5 Crude and Multivariable Adjusted Odds Ratio (OR) and 95% Confidence Interval (95%CI for OR) of the Association of Stressful Life Events Scores and QoL

The association of socioeconomic (OR: 1.12, 95%CI: 1.12–1.16) and personal stressors scores (OR: 2.48, 95%CI: 1.96–3.14) with low QoL remained statistically significant after adjustment for demographic variables including age, sex, and household size. A one-unit increase in socioeconomic stressors’ score was associated with 12% increase in odds of being in low QoL class and a one-unit increase in personal stressors’ score was associated with a 2.48 times increase in risk of being in the low QoL class.

These associations, remained significant even after adjustment for the association of other confounders including BMI, sleep duration and shift work. In the fully adjusted model, it was observed that one unit increase in socioeconomic stressors’ score was significantly associated with 14% increased risk for being in the low QoL class (OR:1.14, 95%CI: 1.12–1.16) and one-unit increase in personal stressors score, enhances the risk of being in the low QoL class by 2.36 times (OR: 2.36, 95%CI: 1.87–2.98).

All dimensions of stressful life events were found to be significantly associated with increased risk of belonging to the low QoL class in both crude and adjusted models. Health concerns, daily life, and home life had the strongest association with low QoL which resulted in ORs of 1.40, 1.23 and 1.17, respectively.

Discussion

This cross-sectional study examined the association between stressful life events and QoL using MLLCR analysis among 3063 Iranian industrial employees. Two classes of employees, as lower-level units, and two classes of job categories, as higher-level units, were identified according to QoL indicators. These classes were labeled as high, low or poor QoL. Poor QoL, was significantly associated with higher scores of social and personal stressor profiles extracted from 11 dimensions of stressful life events. However, the strength of the associations was markedly higher for personal stressors than socioeconomic stressors. Higher scores of all dimensions of stressful life events were significantly associated with increased risk of low QoL. Among them health concerns, daily life and home life had the greatest associations with low QoL.

To the best of our knowledge, the current study is the first one that was conducted on employees and classified individuals by considering all aspects of QoL and constructed two parsimonious and comprehensive measures from all dimensions of stressful life events and evaluated their association with QoL using MLLCR analysis.

We found that the social and personal stressor profiles were associated with low QOL. These findings are in line with some previous studies that have focused on cumulative stressful life events in relation to QoL and well-being.^{5,18,30,31,33} In a population-based study of individuals with chronic obstructive pulmonary disease (COPD) in Singapore, higher stressful life event scores were associated with lower physical and mental health scores.³⁰ The study of Staniute et al on 560 coronary artery disease patients in Lithuania indicated negative relationship between stressful life events and QoL.³¹ In a study among adults aged 60 years and older in Bhutan, stressful life events during childhood and adulthood were negatively associated with overall QoL.³³

Our findings are consistent with some parts of previously reported studies, which have supported a negative association between stressful life events and QoL.^34,56 Damush et al, in their study on college students, have reported a statistically significant relationship between autonomous life events, distressful life events, family/parental life events, sexual life events and poor QoL.⁵⁶ Similarly, in a study among Indian undergraduate medical and dental students, stress related to academics, self-expectations, relationships and living conditions were associated with decreased QoL including physical, psychological, social, and environmental dimensions and stress related to health and value conflict leads to decrease in physical and social components of QoL.⁵⁷

Our findings are inconsistent with some previous results.^34,56,58 Villalonga-Olives et al, in a study among Spanish adolescents and youths, did not find significant association between family and extra family events and desirable events and QoL.⁵⁸ Damush et al, in their study on college students, did not find significant relationship between accident/illness, deviance and relocation events with QoL.⁵⁶ Cleland et al, in a longitudinal study, found that new job/promotion and parenthood events positively impact on physical QoL.³⁴

There are several possible reasons for this difference. (1) One possibility may be due to population difference. (2) Impact of stressful life events on wellbeing and QoL may vary depending on the appraisal of stressors³⁴ and compassionate skills⁵⁹ by different individuals. (3) Differences in cultural characteristics can also be a factor that may change the impact of events on QoL among individuals.⁶⁰

The possible mechanism that underlies the relationship between stressful life events and physical QoL includes neuroendocrine responses. Stressful life events can make people susceptible and vulnerable to chronic and acute diseases through changes in immune competence that are mediated by the neuroendocrine system. Stressful life events may be associated with QoL indirectly through impairing their psychological health.⁶¹ Some psychological responses to stress include changes in cognitive functions such as limitation of perceptions, reduced ability to concentrate, change in emotional responses such as anxiety and depression.⁶² Potential mediators, including elevated levels of cortisol and neurotoxicity might underlie the impact of stressful life events on physical and mental health.²⁰ Long-term stress can lead to dysregulation of the allostatic system by altering physiological processes within the body, and thus contribute to health morbidities^24,63 and consequently, QoL. Another possible pathway for explaining the stress effects on health and finally QoL, is directed through behavioral changes such as increased smoking, decreased exercise and sleep duration and quality.^64,65

In the present study, it was found also that personal stressors were more strongly related to poor QoL than socioeconomic stressors. This result is in agreement with Holmes and Rahe Stress Scale in which death of a spouse, divorce, marital separation, death of close family member and personal injury or illness ranked as the highest stressors.⁶⁶

This finding emphasizes that adequate supports are important for people with exposure to life events such as health concerns and bereavement.³⁴

Although in current study, personal stressors had a strong association with QoL but sexual life dimension had the lowest strength. This result may be attributed to the fact that some people probably have not correctly answered to the questions in the area of this type of stressor because of cultural constraints and limitations about declaration of sexual issues in public settings.

Study Strengths and Limitations

The major strength of the present research is the use of an advanced statistical model that provides a comprehensive evaluation of both life stressors and QoL and their association. Other strengths of the current study include the large sample of industrial employees and that the estimated association between life stressors and QoL have been adjusted for many potential confounding variables. The present study has several limitations that it is necessary to mention. Because of the cross-sectional design of the study, causality between life stressors and QoL cannot be inferred. All information in the present study was self-reported by participants which may lead to misclassification of study participants. As study participants were employees of the Esfahan Steel Company, caution should be taken in generalizing the findings of present study to the Iranian population. Although we used an Iranian validated version of EQ-5D and its validity has been approved previously, we obtained a low internal consistency based on Cronbach’s alpha in our study sample, accordingly it is suggested to evaluate the validity and reliability of this instrument in a workforce population.

Conclusion

We found significant negative association between life stressors and QoL of employees. Personal stressors were more strongly associated with poor QoL than socioeconomic stressors. Proper management of stressors is required to reduce their potential adverse effects on QoL. Managing the stressors and educating stress-coping styles can play an effective role in improving QoL of employees and it could be finally effective on promoting organizational performance and job productivity.