^{1}

^{*}

^{1}

^{1}

^{2}

Workplace stress is a common problem with broad effects in professional life. This study aimed to understand how workplace stressors affect job satisfaction among biologics development professionals. A cross-sectional survey was conducted at a biologics development organization. Multiple linear regression analysis was performed using years of experience, ambiguity, job conflict, perceived control, social support, job demands, self-esteem, and self-rated workplace stress as independent variables and job satisfaction as dependent variable (response). The regression model indicated that the workplace stressors and their two-level interactions significantly predicted employees’ job satisfaction, which explained 89% of the variance in level of job satisfaction (R<sup>2</sup> = 0.89, <i>F</i>(17, 16) = 7.251, <i>p</i> = 0.0001). The interaction between perceived control and job demand and interaction between self-rated stress and job conflict had the biggest effect size on job satisfaction. This model was further used in Monte Carlo simulation to predict the outcome of job satisfaction under different work conditions. The findings will help the management to develop strategies to improve employee job satisfaction.

Workplace stress has long been recognized as a common problem with broad effects. Its high prevalence is associated with costs to individuals and organizations [

lack of control, emotional needs, lack of support, irregular working schedules, undefined roles, and lack of reward [

The purpose of this study was to measure workplace stress, and to assess how workplace stressors affect employee job satisfaction among professionals working at a biologics development organization. A research model (

Schematic illustration of the proposed research model of workplace stress and job satisfaction

Evidence from previous studies revealed negative association between workplace stressors and job satisfaction in professionals such as nurses [

A cross-sectional survey was conducted to measure workplace stressors and job satisfaction. Participants were recruited from a biologics development organization with 110 employees. The questionnaire was built online at the eSurv website (http://esurv.org), and distributed electronically to the participants. All employees working part-time or full-time at this organization at the time of survey were qualified study participants. The exclusion criteria included temporary service contractors, such as janitors, consultants, etc. The study proposal was reviewed and approved by the A.T. Still University Institutional Review Board prior to study initiation.

The survey response was collected from April 15 to May 15, 2013. At the closing date of the survey, 42 responses were obtained from 110 employees. This represented a response rate of 38%. Sample demographics are summarized in

. Sample demographics

Frequency | Percent | Valid Percent | Cumulative Percent | |||
---|---|---|---|---|---|---|

Gender | Valid | Female | 15 | 36 | 37 | 37 |

Male | 25 | 59 | 62 | 100 | ||

Total | 40 | 95 | 100 | |||

Missing | 2 | 5 | ||||

Total | 42 | 100 | ||||

Age Group | Valid | 21 - 30 | 4 | 9 | 10 | 10 |

31 - 40 | 18 | 43 | 44 | 54 | ||

41 - 50 | 15 | 36 | 37 | 90 | ||

51 - 60 | 4 | 9 | 10 | 100 | ||

Total | 41 | 98 | 100 | |||

Missing | 1 | 2 | ||||

Total | 42 | 100 | ||||

Role | Valid | Contributor | 31 | 74 | 74 | 74 |

Manager | 11 | 26 | 26 | 100 | ||

Total | 42 | 100 | 100 | |||

Years of Experience | Valid | 0 - 10 | 21 | 50 | 52 | 52 |

11 - 20 | 17 | 40 | 42 | 95 | ||

21 - 30 | 2 | 5 | 5 | 100 | ||

Total | 40 | 95 | 100 | |||

Missing | 2 | 5 | ||||

Total | 42 | 100 | ||||

Function | Valid | Analytic | 15 | 36 | 37 | 37 |

Document | 3 | 7 | 7 | 45 | ||

Process | 20 | 48 | 50 | 95 | ||

Project/ | 2 | 5 | 5 | 100 | ||

Total | 40 | 95 | 100 | |||

Missing | 2 | 5 | ||||

Total | 42 | 100 |

“Percent” refers to the percentage of total responses that includes both valid and missing responses; “Valid Percent” refers to the percentage of valid responses; “Cumulative Percent” refers to the cumulative percentage of valid responses.

The survey instrument was an abbreviated version of the National Institute of Occupational Safety and Health (NIOSH) General Job Stress Questionnaire [

Internal reliability of the items was verified by computing the Cronbach’s α, which is summarized in

. Descriptive and reliability statistics of measured results

Measurement | Number of Items | Mean | SD | Cronbach’s α | |
---|---|---|---|---|---|

This Study | Reference Data^{a} | ||||

Role Ambiguity | 6 | 2.94 | 0.81 | 0.85 | 0.74 |

Role Conflict | 8 | 3.34 | 0.71 | 0.79 | 0.82 |

Intragroup Conflict | 8 | 2.78 | 0.68 | 0.74 | 0.86 |

Intergroup Conflict | 8 | 3.08 | 0.78 | 0.86 | 0.85 |

Job Future Ambiguity | 4 | 3.27 | 0.99 | 0.76 | 0.65 |

Resource Control | 2 | 3.25 | 0.54 | −0.37 | 0.82 |

Task Control | 8 | 3.19 | 0.51 | 0.57 | 0.85 |

Physical Environment Control | 2 | 3.03 | 0.90 | 0.64 | 0.79 |

Decision Control | 4 | 2.67 | 0.90 | 0.82 | 0.74 |

Support from Supervisor | 4 | 4.02 | 0.80 | 0.83 | 0.88 |

Support from Peer | 4 | 3.84 | 0.71 | 0.64 | 0.84 |

Support from Family/Friends | 4 | 4.44 | 0.56 | 0.67 | 0.85 |

Quantitative Workload | 11 | 3.79 | 0.67 | 0.87 | 0.85 |

Variance in Workload | 3 | 3.59 | 0.92 | 0.92 | 0.86 |

Skill Underutilization | 3 | 2.74 | 1.04 | 0.82 | 0.73 |

Responsibility for People | 4 | 2.70 | 1.09 | 0.86 | 0.62 |

Mental Demands | 5 | 4.16 | 0.74 | 0.72 | 0.75 |

Self-esteem | 4 | 3.97 | 0.61 | 0.82 | 0.85 |

Workplace Stress | 10 | 2.37 | 1.00 | 0.91 | 0.93^{b} |

Job Satisfaction | 7 | 3.73 | 0.98 | 0.77 | 0.83 |

All measurement results were converted to 1 - 5 scale, based on a method reported in “Transforming different Likert scales to a common scale,” by IBM [^{a}700 Newfoundland nurses [^{b}1794 general adult UK population [

All measurement results were converted to a 1 - 5 scale, based on a method reported by IBM [

The data were analyzed using descriptive statistics such as mean, standard deviation, standard error of mean, 95% confidence interval, percentage, and frequency. Data were fitted using two-tailed Normal fitting first. The Normality was checked using Shapiro-Wilk W test. The null hypothesis (Ho) posited that the data was from the Normal distribution; a p-value smaller than or equal to the significance level (0.05) rejects Ho. If the goodness- of-fit test failed (p < 0.05), other continuous fitting was performed. The fitting with the smallest AICc (Akaike information criterion with correction of sample size) was selected. Inter-correlations of the variables were assessed using pairwise correlation analysis, and the resulting Pearson (product-moment) correlation coefficients were summarized in a correlation table (

The research hypothesis was tested with multiple linear regression analysis, which not only tested the statistical significance (p-value) of the effect, but also quantified the effect size of the independent variables and their interactions on the responses. Multiple linear regression analysis was used to construct statistical models, using the independent variables as factors and dependent variables as responses. Model fit was assessed by both goodness of fit (R^{2} and adjusted R^{2}) and lack of fit indices, with significance (p-value) set at 0.05. Model significance was tested by multivariate ANOVA. Each factor and interaction pair was correlated with an estimated regression coefficient to form a term in the regression model. The significance of each term as a factor of the model was evaluated by t-test comparing the estimated coefficient to the standard deviation of that coefficient. Model terms with significant effects on the measured response were identified by a p-value of less than 0.05 of getting a greater t-ratio by chance. The model was refined to keep only the terms that have statistically significant effects (p-value less than 0.05). The collinearity was assessed by the variable inflation factor (VIF), and the terms with VIF values of 8 and above were removed to avoid over fitting. Based on the regression model, prediction profiles were plotted to discern factor importance (indicated by the steepness of a trace), interactions (indicated by changes in trace slope or curvature in response to shifts in other parameter settings), and conditions that achieved a desired response.

The established regression model was used in Monte Carlo simulation to predict the level of job satisfaction with independent variables set under different scenarios. Two scenarios were simulated to illustrate the application of model. The first scenario was the current work condition, where the mean values and standard deviations of independent variables from the survey data were used. To compensate for the uncontrolled variations, the

. Pearson correlation of the variables (n = 42)

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | |
---|---|---|---|---|---|---|---|---|---|

1. Years of Experience | - | ||||||||

2. Ambiguity | −0.271 | - | |||||||

3. Job Conflict | 0.076 | 0.225 | - | ||||||

4. Perceived Control | 0.203 | −0.164 | −0.341^{*} | - | |||||

5. Job Demand | 0.066 | 0.150 | −0.267 | 0.038 | - | ||||

6. Social Support | −0.088 | 0.274 | 0.190 | −0.111 | −0.368^{*} | - | |||

7. Self-esteem | 0.209 | −0.206 | −0.183 | 0.050 | 0.306 | −0.108 | - | ||

8. Workplace Stress | −0.070 | −0.003 | −0.140 | −0.196 | 0.208 | 0.016 | −0.254 | - | |

9. Job Satisfaction | −0.207 | 0.148 | 0.149 | −0.046 | 0.092 | −0.014 | 0.153 | −0.336^{*} | - |

^{*}Statistically significant correlation (p < 0.05, two tailed).

standard deviation of the model residuals was added in the simulation as a random noise. The mean value of response (job satisfaction) of the observations obtained from the survey was used as the low limit in the simulation. The number of simulation runs was set at 10,000 to improve the prediction accuracy. The simulation results gave the distribution of predicted response, desirability to achieve the expected outcome, and failure rate to meet the limit, either in percentage or ppm. The second scenario analyzed using the regression model was the work condition to achieve the highest level of job satisfaction. This was done by select “Maximize Desirability” function and the condition was automatically picked by JMP software.

Scores from each item of the measurements (exogenous observed variables) were transformed into a common scale of 1 - 5, and were averaged to give the results of each measurement.

The descriptive statistics of the variables are summarized in

Multiple regression analysis was used to test if workplace stressors significantly predicted employees’ job satisfaction. A regression model was constructed using the independent variables (or factors, which include years of experience, ambiguity, job conflict, perceived control, social support, job demands, self-esteem, and self-rated workplace stress) and their two-level interactions. The regression model was refined to contain only statistically significant terms (independent variables and their two-level interactions) using a backward stepwise approach. The regression is illustrated in the Actual by Predicted Plot in

. Distribution of variables

n | Mean | SD | Sample Range | 95% CI | |
---|---|---|---|---|---|

Years of Experience | 40 | 10.25 | 5.99 | 5 - 25 | 8.34 - 12.16 |

Ambiguity | 41 | 3.11 | 0.75 | 1.33 - 4.61 | 2.87 - 3.34 |

Job Conflict | 42 | 3.07 | 0.58 | 2.00 - 4.49 | 2.89 - 3.25 |

Perceived Control | 40 | 3.05 | 0.53 | 1.94 - 4.16 | 2.88 - 3.22 |

Social Support | 40 | 4.10 | 0.50 | 3.08 - 5.00 | 3.94 - 4.26 |

Job Demands | 41 | 3.40 | 0.44 | 1.97 - 4.24 | 3.26 - 3.54 |

Job Satisfaction | 42 | 3.73 | 0.98 | 1.50 - 5.00 | 3.43 - 4.04 |

Self-Esteem | 42 | 3.97 | 0.61 | 2.70 - 5.00 | 3.78 - 4.16 |

Workplace Stress | 41 | 2.37 | 1.00 | 1.00 - 5.00 | 2.05 - 2.68 |

Demands | 40 | 3.19 | 0.38 | 2.32 - 4.20 | 3.07 - 3.31 |

Resources | 38 | 3.73 | 0.30 | 3.16 - 4.38 | 3.63 - 3.82 |

n = sample size; SD = standard deviation; CI = confidence interval.

Summary of Fit | ||||
---|---|---|---|---|

R^{2} | Adjusted R^{2} | RMSE | Response Mean | Observations |

0.89 | 0.76 | 0.48 | 3.80 | 34 |

Summary of Fit

R^{2}

Adjusted R^{2}

RMSE

Response Mean

Observations

0.89

0.76

0.48

3.80

34

ANOVA | |||||
---|---|---|---|---|---|

Model | Sum of Squares | df | Mean Square | F | p |

Regression | 28.499 | 17 | 1.676 | 7.251 | 0.0001^{*} |

Residual | 3.699 | 16 | 0.231 | ||

Total | 32.198 | 33 |

^{*}regression model is statistically significant (p < 0.05); blue dashed line = mean of the response (job satisfaction); red dashed curves = 95% confidence interval; red solid line = regression line.

that the independent variables and their two-level interactions significantly predicted job satisfaction, which explained 89% of the variance (R^{2} = 0.89, F(17, 16) = 7.251, p = 0.0001) (^{2} value depicted the goodness of fit of the model, and the adjusted R^{2} was a modification of R^{2} that adjusted for the number of explanatory terms (factors and two-level interactions used in the model, as listed in ^{2} (0.89) and the adjusted R^{2} (0.76) are high, indicating good model fitting. The multi-collinearity was assessed by the variable inflation factor (VIF); the terms with a VIF value of 8 and above were removed to avoid over fitting. In the constructed regression model, all the terms have VIF values below 8 (

The characteristics of the regression model are summarized in

. Regression coefficients and collinearity of independent variables and interactions

Independent Variable and Interaction | Unstandardized Coefficient | Standardized Coefficient | Effect Size^{a} | t Ratio | p | VIF | |
---|---|---|---|---|---|---|---|

B | SE | β | |||||

Intercept | −1.161 | 1.682 | 0 | 6.855 | −0.69 | 0.4999 | |

Years of Experience | −0.123 | 0.021 | −0.705 | −2.469 | −5.77 | <0.0001^{*} | 2.080 |

Self-esteem | −0.265 | 0.201 | −0.172 | −0.608 | −1.32 | 0.2070 | 2.386 |

Workplace Stress | −0.514 | 0.107 | −0.556 | −2.054 | −4.81 | 0.0002^{*} | 1.865 |

Ambiguity | −0.419 | 0.148 | −0.335 | −1.373 | −2.84 | 0.0119^{*} | 1.948 |

Job Conflict | 0.482 | 0.196 | 0.269 | 1.105 | 2.46 | 0.0256^{*} | 1.660 |

Perceived Control | −0.067 | 0.191 | −0.037 | −0.148 | −0.35 | 0.7320 | 1.566 |

Job Demand | 2.413 | 0.377 | 1.116 | 5.476 | 6.41 | <0.0001^{*} | 4.221 |

(Years of Experience-10) *(Self-esteem-3.90458) | 0.284 | 0.051 | 1.141 | 6.536 | 5.55 | <0.0001^{*} | 5.890 |

(Years of Experience-10) *(Workplace Stress-2.37255) | 0.067 | 0.023 | 0.426 | 2.666 | 2.88 | 0.0109^{*} | 3.053 |

(Years of Experience-10) * (Job Conflict-3.17419) | 0.273 | 0.063 | 0.769 | 6.265 | 4.34 | 0.0005^{*} | 4.358 |

(Self-esteem-3.90458) *(Workplace Stress-2.37255) | −1.003 | 0.231 | −0.667 | −4.613 | −4.34 | 0.0005^{*} | 3.285 |

(Self-esteem-3.90458) *(Perceived Control-3.01374) | −2.250 | 0.426 | −0.782 | −5.742 | −5.28 | <0.0001^{*} | 3.053 |

(Workplace Stress-2.37255) *(Ambiguity-3.17271) | 0.980 | 0.183 | 0.759 | 6.416 | 5.35 | <0.0001^{*} | 2.801 |

(Workplace Stress-2.37255) *(Job Conflict-3.17419) | −1.525 | 0.317 | −0.776 | −6.990 | −4.81 | 0.0002^{*} | 3.617 |

(Workplace Stress-2.37255) *(Perceived Control-3.01374) | −1.274 | 0.306 | −0.744 | −5.652 | −4.16 | 0.0007^{*} | 4.459 |

(Job Conflict-3.17419) *(Perceived Control-3.01374) | 2.448 | 0.612 | 0.697 | 6.224 | 4 | 0.0010^{*} | 4.239 |

(Perceived Control-3.01374) *(Job Demand-3.34626) | 3.245 | 0.662 | 0.866 | 8.169 | 4.9 | 0.0002^{*} | 4.345 |

Dependent variable: Job satisfaction; ^{a}effect size on job satisfaction; ^{*}statistically significant effect (p < 0.05); VIF = variable inflation factor.

The constructed regression model was further used in Monte Carlo simulation to predict the outcome of response (job satisfaction) under different work condition scenarios.

Another simulation was performed to find out the work conditions that potentially lead to the highest level of job satisfaction, which is illustrated in

satisfaction. The predicted mean value of job satisfaction was 27.0, with a standard deviation of 7.91. The defect rate was predicted to be 0.00, which indicated that none of the employees in the organization would have a job satisfaction level below the mean value of 3.80, under this scenario. This was predicted based only on the regression model from available data, which needs to be verified in real situation.

Workplace stress is a common problem with broad effects in professional life. It is important to gain a better understanding of work-related stress and its effects on employee job satisfaction. Although a number of studies suggested negative association between workplace stress and job satisfaction, the studies on workplace stress are “hindered by lack of understanding of how sources of stress vary between different practice areas, lack of predictive power of assessment tools, and a lack of understanding of how personal and workplace factors interact”(p. 633) [

Like other empirical studies, this study has a number of limitations. First, this study was limited by the small population size in a specific setting. The findings of this study may not apply to the general population of biologics development professionals. Second, the survey was delivered online with e-mail notification, which has the drawback of low response rate and self-selection bias [

Nonetheless, this study was the first to measure workplace stressors and quantify the effects of such stressor on job satisfaction among biologics development professionals within this organization. The results suggested that the workplace stressors significantly predicted the level of job satisfaction. The interaction between perceived control and job demand and interaction between self-rated stress and job conflict had the biggest effect size on the level of job satisfaction. The simulation results in

Although regression analysis results revealed the correlation relationship between the independent variables and dependent variable, the causal relationships among the variables cannot be determined. Thus, it is recommended to verify the findings of this study in real situations, and continue to monitor in longitudinal studies. In addition, further studies with increased sample size at different biologics development organizations are needed to further elucidate the predictors for job satisfaction for general working population in this industry.

The results of this study revealed that the workplace stressors have significant effect on employees’ level of job satisfaction within the organization studied. Under the current work condition, 59% of the employees in the organization had a job satisfaction level below the mean value of 3.80. The conditions to achieve the highest level of job satisfaction were predicted in Monte Carlo simulation, which include employees with long working experience, under low levels of ambiguity, self-rated stress, and high levels of self-esteem, perceived control, job conflict, and job demand. This was an early stage study due to the limitations; however, it serves to establish a baseline to measure the working conditions and corresponding job satisfaction for further longitudinal studies at the particular setting under study. The findings may guide the management to develop an environment that supports improved job satisfaction. Further studies with a larger population at different settings are needed to elucidate the predictors for job satisfaction for general working populations in this industry.