Workplace Stressors and Negative Health Outcomes

What is the headline saying?

Why Your Workplace Might Be Killing You – Stanford scholars identify 10 work stressors that are destroying your health.

What is the news article saying?

The workplace environment that is created by an organization impacts the mental and overall health of employees which, in turn, contributes to rising health care costs. Stressors such as job insecurity, working long hours, lack of health insurance and increased demands at work contribute to poor health and even mortality that is comparable to the 4th or 5th largest causes of mortality in the U.S. Workplace wellness programs that are implemented in work environments with management practices that create stress are not effective because overall, “It costs more to remediate the effects of toxic workplaces than it does to prevent their ill effects in the first place.” The article makes sure to point out the main limitations provided by the researchers: they can only claim association, not causation, because the studies are observational. The researchers also only looked at 10 stressors that could be attributed to management practices.

Does the headline ultimately support claims made by the news article? Does it truly summarize the key points of the news article?

Yes and no. The article seems to be mostly focused on management practices that affect employee wellbeing. Also, the headline makes it seem as if the 10 stressors definitively lead to poor health and increased mortality- These are claims that the researchers say their study can’t make. However, in the video at the end of the article, one of the researchers does say that the workplace is killing employees as a result of long hours and stress.

What are the implications of this headline?

Workplaces may be contributing to ill health in America as a result of specific stressors.

What are the implications of this news article?

Physical AND psychological effects of the workplace harm employees and the nation as a whole. Workplace wellness programs cannot overlook the negative psychological aspects of the workplace that influence the decisions employees make regarding their health (such as eating healthy and exercising), especially in the face of various workplace stressors.

What evidence currently exists to counter or support these implications?
Countering views:

Supporting views:

Work stress and risk of death in men and women with and without cardiometabolic disease: a multicohort study

State of the American Workplace

Burn-out an “occupational phenomenon”: International Classification of Diseases

Employee Burnout, Part 1: The 5 Main Causes

Workplace Mental Health: Data, Statistics, and Solutions

Are there similar and/or opposing headlines from other news outlets? Do the news outlets only link back to other news outlets?

There are similar headlines from other news outlets. The majority of these news articles link to different data sources (see above).

What are the data sources (i.e. memo, official statement, official document, research study, validated surveillance system, official report, etc.) supporting the article?

The Relationship Between Workplace Stressors and Mortality and Health Costs in the United States

Agency for Healthcare Research and Quality (2011a) Medical
expenditure panel survey. Accessed November 7, 2011.

Agency for Healthcare Research and Quality (2011b) Total health services—Mean and median expenses per person with expense and distribution of expenses by source of payment: United States, 2008. Medical Expenditure Panel Survey, Household Component data. Generated interactively. Accessed November 7, 2011.

National Opinion Research Center (2011) General Social Survey. Accessed November 7, 2011.

Are these data sources credible when applied to the news story? Why or why not?

Yes. The researchers apply relevant methodologies and incorporate national-level datasets in their analyses. Thorough justification is provided for each analysis performed as well.


228 articles were included in a meta-analysis that was conducted (115 studies
used longitudinal data (these included panel studies), 115 studies used cross-sectional data, and 2 studies used both types of data). Both a multiplicative model and conservative model were used to derive change in cost and change in mortality:

The model focuses on the U.S. civilian labor force in 2010 and divides the analysis according to four subpopulations: (men, women) _ (employed, unemployed). The unemployed are assumed to be exposed to only two stressors: unemployment and no insurance. The employed are exposed to all the stressors except unemployment. The model estimates the increased prevalence of four categories of poor health (henceforth termed outcomes) associated with the 10 workplace stressors (henceforth termed exposures) and then combines them with separate estimates of the increase in health spending associated with each of the categories of poor health. The four outcomes that we consider are those that are commonly measured in the medical literature: poor self-rated physical health, poor self-rated mental health, presence of physician-diagnosed health conditions, and mortality. Self-rated health measures are included because (a) they have been shown to be excellent proxies for actual health and mortality (e.g., Marmot et al 1995, Idler and Benyamini 1997); (b) they are easy to assess in surveys, including surveys of healthcare costs; and (c) they are commonly used in epidemiological studies.

First, we assume each of the four subpopulations considered to be statistically homogeneous. This allows us to focus our analysis on a characteristic individual within each subpopulation and estimate for that individual the annual healthcare spending and the probability of mortality associated with workplace stressors. The corresponding population-level estimates are obtained by scaling the individual-level estimates (by the subpopulation’s size) and summing across the four subpopulations. Second, we assume that exposures to the 10 stressors and outcomes are binary; that is, we do not account for a more nuanced interaction between stressors and outcomes that takes into account the duration of the exposure to a stressor. This is because most of the studies used to obtain the parameters in our model also employ a binary model of exposures and outcomes.

Input Parameters

-Joint probability distribution of exposures – Average prevalence of (and correlation between) stressors faced by workers in the United States (based on data from the General Social Survey (GSS) and from the health insurance component of the Current Population Survey (CPS) were used to calculate the joint probability distribution of exposures in order to capture)

-Relative risk for each exposure-outcome pair – Incremental probability/relative risk of individuals having a certain outcome that were exposed to a certain stressor compared to those who were not exposed to the stressor (based on data from a meta-analysis of relevant epidemiological studies

-The probability of observed prevalence of each category of poor health in the United States – Status quo prevalence of each outcome (based on data from the Medical Expenditure Panel Survey (MEPS) and mortality data from the Centers for Disease Control and Prevention)

-Excess healthcare spending per year associated with each category of poor health in the United States or the average increase in healthcare spending for those with a certain outcome, compared to those without the outcome – incremental cost of each outcome per year (based on data from MEPS, also controls for overlapping healthcare cost contributions from multiple health outcomes

*Estimate confidence intervals for cost and mortality using Monte Carlo simulations and/or mathematical characterizations along with standard distributional assumptions


The model was designed to overcome limitations with using inputs from multiple data sources. Specifically, the model separately derives optimistic and conservative estimates of the effect of multiple workplace exposures on health, and uses optimization to calculate upper and lower bounds around each estimate, which accounts for the correlation between exposures.

Sensitivity Analyses to address some limitations

1. The meta-analysis computed pooled relative risks by combining study populations from various countries. The assumption is that these estimates are relevant to our U.S.-based target population. To test this assumption, we performed sensitivity analyses in which we restricted the studies for the metaanalysis calculations to populations drawn from G8 countries and high-income Organisation for Economic Co-operation and Development (OECD) countries (Sensitivity Analysis 2).

2. To generate relative risk estimates for the mortality outcome in our meta-analysis, we pooled studies that estimated the risks of all-cause mortality and cause-specific mortality. To test the effect of this assumption, we repeated our analysis but excluded studies with cause-specific mortality (Sensitivity Analysis 3).

3. We pooled studies using longitudinal and cross-sectional data to estimate the relative risks in the base model. Because cross-sectional data have limitations as outlined before, we conducted Sensitivity Analysis to study how only our final estimates change if only studies that use longitudinal data are included in the meta-analytic sample.

4. In our base model, the meta-analytic sample contains studies that use either logistic regressions or Cox regressions. We test this assumption in Sensitivity Analysis 5, where we excluded studies that use Cox regressions.

5. To derive the relative risk estimates for NOINSURE, we included studies that group respondents with public insurance (Medicaid) together with the limitations uninsured. In Sensitivity Analysis 6, we excluded studies that performed this pooling.

6. Un-insurance was assumed to be independent of the remaining exposures for the employed subgroup. In Sensitivity Analysis 7, we extended the robustness analysis to allow correlation between these exposures and no insurance.

7. Our definition of physician-diagnosed medical condition included any respondent in the MEPS data who had one or more health conditions within a list of conditions. To test sensitivity to that assumption, we repeated our estimation, but varied the threshold of conditions present needed to determine whether someone had a physician-diagnosed medical condition (Sensitivity Analysis 8).

8. We pooled exposure prevalence data from 2002, 2006, and 2010 in our base model, which assumes that the exposures were similar in those years. We tested this assumption in Sensitivity Analysis 9, where we repeated the analysis separately for each year 2002, 2006, and 2010 by using exposure data that were specific to that year.

What are the data sources saying? Are they being interpreted correctly in the article and are limitations provided? Are there multiple ways to interpret the data or various conclusions that may be drawn from the data?

We find that more than 120,000 deaths per year and approximately 5%–8% of annual healthcare costs are associated with and may be attributable to how U.S. companies manage their work forces. Our results suggest that more attention should be paid to management practices as important contributors to health outcomes and costs in the United States.

The data sources conclude that stressors in the work environment can be associated with health outcomes and costs in the U.S., based off of various estimates that were generated from the models. The researchers report that, in all instances (using all models/estimates), there were more than 120,000 excess deaths each year associated with key workplace stressors. Incremental costs related to the workplace comprised 5-8% of the total national healthcare expenditure in 2008.

Observations made by researchers from the results:

1. Estimates generated by our model are consistent with estimates reported previously in the literature. In particular, our results show that not having insurance is associated with about 50,000 excess deaths per year, a number quite close to the 45,000 reported by Wilper et al. (2009). This provides some confidence that our other estimates, derived and presented here for the first time, are likely to be reliable.

2. Absence of insurance contributes the most toward excess mortality, followed closely by unemployment. Low job control is, however, also an important factor contributing an estimated 31,000 excess deaths annually.

3. Not having health insurance, being in jobs with high demands, and work–family conflict are the major exposures that contribute to healthcare expenditures.

4. The exposures that contribute the most to healthcare expenditures differ from the highest contributors to mortality. This is because incremental costs stop when someone dies, so exposures with higher deaths are not necessarily associated with higher costs.

5. Although each of the exposures contributes to healthcare expenditure, not all of them contribute, at least from our estimates, to incremental deaths. This is partly due to data limitations: our analysis excluded relative risk estimates that were generated only by two or fewer studies. From Table 3, we observe that several exposures for mortality fall into this category.

6. Because of the nonlinear manner in which each workplace exposure contributes to the final estimate of either expenditure or mortality, the sum of the marginal contributions from each exposure does not add up to the totals reported in Table 6.

Our model estimates significantly lower workplace-associated expenditures and mortality for 2006, which is when the U.S. economy was doing well, relative to estimates in the base model and for 2010, when the U.S. economy was bruising from the global financial crisis. The estimates for 2002 were moderately lower, which was around the time of an economic recession in many developed countries. Overall, these results corroborate the intuition that people experience greater workplace stressors during times of economic turbulence, and that these can have significant impact on health costs and outcomes. This suggests that workplace exposures could be used to better understand how the economic climate affects health, which is a subject that is an interesting direction for future research.


…The estimated effect of these workplace stressors is substantially large, with the number of deaths associated with such stressors exceeding the number of
deaths from diabetes, for instance, and with a reasonable estimate of the total costs incurred in excess of $180 billion. Our analysis suggests that these stressors could potentially be fruitful avenues for policy attention to improve health outcomes and costs.

The results reported in this paper suggest that the association between employer actions and healthcare outcomes and costs is strong. Although we stop short of claiming that employer decisions have a definite effect on these outcomes and costs, denying the possibility of an effect is not prudent either. Analyzing how employers affect health outcomes and costs through the workplace decisions they make is incredibly important if we are to more fully understand the landscape of health and well-being.

What does this mean for the general public?

Poor management practices can create stressful workplace environments which contribute to the negative health outcomes of employees in the United States. Although workplaces cannot be free of all stressors, improvements can be made to management practices. Organizations should take into consideration the workplace environment created by management practices and not just individual-level interventions such as exercising, smoking cessation, and healthy eating- Activities usually promoted through workplace wellness programs. Employees should also consider how their workplace environment might be contributing to poor health decisions they may be making as a result of stress.

Scrutinizer Product
work sc


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s