BEU, Alm.del - 2024-25 - Bilag 310: Orientering om videnskabelig artikel fra Nationale Forskningscenter for Arbejdsmiljø, fra beskæftigelsesministeren

Beskæftigelsesudvalget 2024-25
BEU Alm.del Bilag 310
Offentligt

Downloaded from www.sjweh.fi on August 18, 2025

Scand J Work Environ Health

Online-first -article

https://doi.org/10.5271/sjweh.4244

Published online: 18 Aug 2025

Economic gains from hypothetical improvements in the

psychosocial work environment: A cohort study of 71 207

workers in Denmark

Graversen BK, Hansen KS, Rugulies R, Sørensen JK, Larsen AD

This study estimated the economic effects of hypothetical

interventions improving different aspects of Danish workers’

psychosocial work environment. We found substantial economic gains,

mostly driven by savings related to sickness absence, from

simultaneous improvements of all aspects. Economic effects from

improvements in specific aspects varied a lot. The results may be

useful when considering implementing future real-life interventions.

Affiliation:

National Research Centre for the Working Environment,

Lersø Parkallé 105, DK-2100 Copenhagen, Denmark. [email protected]

Refers to the following texts of the Journal:

2010;36(4):313-318

2018;44(5):458-474 2021;47(7):489-508 2021;47(6):456-465

2021;47(6):466-474 2022;48(4):249-252 2023;49(5):315-329

2024;50(2):61-72 2024;50(2):49-52

Key terms:

cohort study; cost; Denmark; economic gain; healthcare

use; occupational health; parametric g-formula; psychosocial work

environment; sickness absence; simulation study

Additional material

Please note that there is additional material available belonging to

this article on the

Scandinavian Journal of Work, Environment & Health

-website.

This work is licensed under a

Creative Commons Attribution 4.0 International License.

Print ISSN: 0355-3140 Electronic ISSN: 1795-990X

BEU, Alm.del - 2024-25 - Bilag 310: Orientering om videnskabelig artikel fra Nationale Forskningscenter for Arbejdsmiljø, fra beskæftigelsesministeren

riginal article

This work is licensed under a Creative Commons Attribution

4.0 International License.

Scand J Work Environ Health – online first. doi:10.5271/sjweh.4244

Economic gains from hypothetical improvements in the psychosocial work environment:

A cohort study of 71 207 workers in Denmark

by Brian Krogh Graversen, PhD,

Kristian Schultz Hansen, PhD,

Reiner Rugulies, PhD,

1,2

Jeppe Karl Sørensen, PhD,

Ann

Dyreborg Larsen, PhD

Graversen BK, Hansen KS, Rugulies R, Sørensen JK, Larsen AD. Economic gains from hypothetical improvements in the

psychosocial work environment: A cohort study of 71 207 workers in Denmark.

Scand J Work Environ Health

– online first.

Objectives

There is increasing interest in the economic effects of improving working conditions, however,

evidence is sparse. This study aims to estimate the economic effects of hypothetical improvements in the psy-

chosocial work environment (PSWE) experienced by Danish workers.

Methods

We included 71 207 workers, reporting information on their psychosocial working conditions in the

“Work Environment and Health in Denmark” survey and linked these workers to population-based register data.

We used the parametric g-formula method to estimate the economic effects of hypothetical improvements of

the general PSWE, in terms of costs related to sickness absence and healthcare use. We further examined which

PSWE factors contributed most to the economic effects.

Results

A hypothetical improvement of the PSWE – from the least to the most desirable situation – resulted in

an annual gain of €1685 [95% confidence interval (CI) €1234–2135] per worker. When analyzing an improve-

ment from the observed to the most desirable situation, the gain became weaker (€305, 95% CI €134–476).

Gains were largely driven by reductions in sickness absence and were larger for women than men and for public

sector workers than private sector workers. The PSWE factors with the largest contribution were eliminations

of threats of violence and improvements in quality of leadership and social support from colleagues (least to

most desirable) and improvements in social support from colleagues, influence at work and quality of leadership

(observed to most desirable), respectively.

Conclusions

Hypothetical improvements in the PSWE resulted in substantial economic gains, mostly driven by

savings related to sickness absence.

Key terms

cost; healthcare use; occupational health; parametric g-formula; sickness absence; simulation study.

Exposure to adverse psychosocial working conditions

can be hazardous for workers’ physical and mental

health (1, 2). Despite the potential risks, however, many

workers face such exposure (3). Encouragingly, a recent

overview review suggests that organizational-level

interventions aimed at improving psychosocial work-

ing conditions can be effective for improving workers’

health and well-being (4).

In occupational and public health research, there

is increasing interest not only in whether psychosocial

workplace interventions protect and improve workers’

health but also whether such interventions generate

economic gains, both at the company level (eg, by

reducing absenteeism and increasing productivity) and

the societal level (eg, by reducing public healthcare use

and disability pensioning) (5–9). However, as pointed

out in recent editorials, research on the economic effects

of workplace interventions is still in its infancy, with few

studies of acceptable quality (5, 6).

In the present article, we first estimate how the

societal costs of sickness absence (SA) and healthcare

use of workers are associated with exposure to various

psychosocial working conditions. Subsequently, we

estimate the reduction in costs of SA and healthcare

use from simulated hypothetical improvements of these

working conditions.

1 National Research Centre for the Working Environment, Copenhagen, Denmark.

2 Section of Epidemiology, Department of Public Health, University of Copenhagen, Denmark.

Correspondence to: Brian Krogh Graversen, PhD, National Research Centre for the Working Environment, Lersø Parkallé 105, DK-2100

Copenhagen, Denmark. [E-mail: [email protected]]

Scand J Work Environ Health – online first

BEU, Alm.del - 2024-25 - Bilag 310: Orientering om videnskabelig artikel fra Nationale Forskningscenter for Arbejdsmiljø, fra beskæftigelsesministeren

Economic effects of improving the psychosocial work environment

The estimation of the effects of hypothetical work

environment improvements on specific outcomes is a

relatively new approach in occupational and public health

research. The approach originates from causal inference

debates and the use of the parametric g-formula that

allows simulating hypothetical interventions using obser-

vational data (10, 11). Our article is particularly inspired

by two recent studies of Mathisen et al who used the

parametric g-formula to estimate the effects of hypotheti-

cal improvements in the psychosocial work environment

(PSWE) on job turnover (12) and SA rates (13) among

Danish hospital workers. Our study differs, though, from

Mathisen et al's studies in that it estimated the economic

effects of the hypothetical interventions and was not

restricted to hospital workers but rather used data from a

nationally representative survey of Danish workers.

More specifically, the present study aimed to esti-

mate: (i) the economic effects of hypothetical improve-

ments of the general PSWE experienced by Danish

workers, in terms of changes in costs of SA and health-

care use, and (ii) the economic effects of hypothetical

improvements of specific PSWE factors, ie, to identify

those PSWE factors whose improvement offers the

greatest potential economic gains.

Methods

Study population

The study population consisted of 18–64-year-old work-

ers invited to participate in one or more of the four

waves of the national survey “Work Environment and

Health in Denmark” (WEHD) conducted in 2012, 2014,

2016 and 2018 (14–16). The workers were selected by

a mix of random sampling from the general population

of workers and stratified sampling from selected work-

places. Respondents from the general worker population

in the 2012 and 2016 waves were invited to participate

in subsequent waves (ie, in 2014, 2016 and 2018 for

respondents from 2012 and in 2018 for respondents

from 2016). By design, workers can be respondents in

one to four waves.

The WEHD data provides information about respon-

dents’ work (including work environment characteris-

tics), health and health behaviors. We linked this infor-

mation with detailed register information on employ-

ment periods, wage payments, job characteristics, work-

place characteristics, sociodemographic characteristics,

SA and healthcare use. There were 86 787 workers

responding in one or more survey waves (figure 1). On

average, each of these workers responded in 1.5 waves,

yielding 128 514 observations. Across the four waves,

the response rate was 55.9%.

For each survey wave, Statistics Denmark have

calculated sample weights for the respondents to make

the sample of respondents representative of all workers

in Denmark meeting the standard requirements to be

included in the WEHD survey: (i) liable to pay taxes

in Denmark, (ii) registered with an address in Den-

mark, (iii) ≥35 monthly working hours and (iv) earning

≥€402 (3000 Danish kroner) per month. In some cases,

respondents from 2012 and 2016, who were invited in

subsequent waves, did not meet these requirements. Sta-

tistics Denmark disregarded observations not meeting

the requirements when calculating the sample weights

and thus did not calculate sample weights for such

observations.

We excluded responses with missing sample weights

and responses from individuals who were not wage

earners at questionnaire completion. Furthermore, we

excluded responses from workers employed in pri-

vate companies with <10 full-time workers, workers

employed in agriculture, forestry or fishing, and work-

ers employed in a subsidized job for disabled workers.

Such workers are not included in the SA register and are

therefore uninformative about the association between

PSWE factors and SA.

Finally, we excluded responses from individuals not

living in Denmark at the start of the year of question-

naire completion. We ended up with a study sample

containing 71 207 workers and 102 379 observations.

For part of our analyses (when estimating regression

models), we used a smaller sample excluding responses

from workers with missing information on SA from

their main job (defined as the job with most working

hours at questionnaire completion). This sample, which

we will refer to as the “regression sample”, included

48 434 individuals and 67 780 observations. The main

reason for missing SA information is that only a selected

sample of private companies with 10–249 full-time

workers have to report information on their workers’

SA to the SA register. The probability for a company of

such size to be included in the sample having to report

SA increases with the number of full-time workers in the

company. All public workplaces and private companies

with ≥250 full-time workers report SA information on

their workers.

Psychosocial work environment factors

We measured 17 PSWE factors, including 7 multiple

item scales and 10 single item scales. The 17 PSWE fac-

tors cover key concepts from the Copenhagen Psycho-

social Questionnaire II (COPSOQ II) (17), and belong

to 6 domains: (i) Demands at work (measured by the

factors quantitative demands, work pace and emotional

demands); (ii) Work organization and job contents (mea-

sured by influence at work); (iii) Interpersonal relations

Scand J Work Environ Health – online first

BEU, Alm.del - 2024-25 - Bilag 310: Orientering om videnskabelig artikel fra Nationale Forskningscenter for Arbejdsmiljø, fra beskæftigelsesministeren

Graversen et al

18-to 64-year-old workers invited to

participate in WEHD survey

N=169 262 (229 819 observations)

Figure 1.

Flow-chart.

Non-respondents

N=82 475 (101 305 observations)

Respondents

N=86 787 (128 514 observations)

Sampling weight exists

N=84 858 (123 978 observations)

No sampling weight

N=1929 (4536 observations)

Was a worker at questionnaire

completion

N=78 961 (114 416 observations)

Was not a worker at questionnaire completion according to

interview or register information

N=5897 (9562 observations)

Job type is not included in sickness absence register

(employed in a private company with less than 10 full-time workers;

employed in agriculture, forestry or fishing;

employed in a subsidized job for disabled workers,

i.e. a flexjob or a subsidized job for recipients of early retirement pension)

N=7749 (12 030 observations)

Job type is included in

sickness absence register

N=71 212 (102 386 observations)

Was not living in Denmark at the start of the year of

questionnaire completion

N=5 (7 observations)

Was living in Denmark at the start of

the year of questionnaire completion

N=71 207 (102 379 observations)

Study sample

Information on sickness absence from

the main job exists

N=48 434 (67 780 observations)

Regression sample

No information in the sickness absence register on sickness absence

from the main job (main job at questionnaire completion)

N=22 773 (34 599 observations)

and leadership (role clarity, quality of leadership, social

support from colleagues, recognition from colleagues);

(iv) Work-individual interface (job insecurity); (v) Val-

ues at the workplace (justice, social inclusiveness); and

(vi) Offensive behaviors (conflicts and quarrels, bully-

ing, witnessing bullying, physical violence, threats of

violence, sexual harassment). The six domains, the 17

PSWE factors, and the 35 items measuring these fac-

tors are listed in table S1 of the supplementary material

(www.sjweh.fi/article/4244). In this table, we also list

the Cronbach’s alphas for the 7 multiple item scales.

The 35 items were partly derived from the COPSOQ

II and partly developed specifically for the WEHD

questionnaire, in collaboration with researchers and

stakeholders, including employer organizations, labor

unions, and the Danish Working Environment Authority.

We collapsed items into scales by converting the

individual item 5-point Likert values to scores of 0–100

(ie, 0, 25, 50, 75 or 100), with higher scores indicating

a more desirable PSWE, and then averaging the scores

of items included in a given scale. If some of the item

scores were missing, averaging was made over the non-

missing items only, and if more than half of the scores

of the included items in a scale were missing, the scale

value was set to missing.

For the first 11 scales belonging to domains (i)–(v),

we categorized non-missing scale values into three

groups by the 25

and 75

sample-weighted percentiles

(or as close as possible) in the study sample, indicat-

ing “least desirable”, “medium” and “most desirable”

working conditions, respectively. We categorized obser-

vations with missing scale values into a fourth group

named “unknown”. As a special case, for the quality

of leadership scale, we split observations with missing

scale values into two groups: “no superior” (miss-

ing scale value because worker had no superior) and

“unknown” (missing scale value because too few ques-

tions included in the scale were answered).

For the last 6 scales belonging to domain (vi) and

describing various types of offensive behavior, we cat-

egorized scale values into three groups: “least desirable”

(had experienced the given offensive behavior; scale

value 0–75), “most desirable” (had not experienced

the given offensive behavior; scale value 100) and

“unknown” (missing scale value).

Scand J Work Environ Health – online first

BEU, Alm.del - 2024-25 - Bilag 310: Orientering om videnskabelig artikel fra Nationale Forskningscenter for Arbejdsmiljø, fra beskæftigelsesministeren

Economic effects of improving the psychosocial work environment

Outcome variables

When information on SA from a worker’s main job was

available in the SA register, we measured the worker’s

hours of SA in the main job within a one-year window

starting at the date of questionnaire completion. If the

job ended (or information on SA in the job was no longer

available) before the end of the one-year window, SA

was only measured as long as the job (or SA informa-

tion) lasted. Applying the human capital approach, we

calculated the societal costs (ie, the reduction in soci-

ety’s production value) of a worker’s SA as hours of

SA times the worker’s productivity per working hour,

measured by the hourly wage rate (18).

Based on information from healthcare registers, we

measured the costs of workers’ healthcare use over the

same period as we measured hours of SA, ie, a period

of up to one year. We computed four healthcare cost

measures for each worker: costs of prescription drug

use, costs of primary healthcare use (includes, among

other things, services delivered by general practitioners

and specialist doctors), costs of non-psychiatric hospital

treatment and total healthcare costs (defined as the sum

of the three preceding healthcare cost measures).

Costs of prescription drug use included costs of pre-

scription drugs collected during the outcome measure-

ment period. Costs of primary healthcare use included

costs of primary healthcare services invoiced by health-

care providers during the outcome measurement period

(visit dates were not available in the data). Costs of

outpatient hospital treatment included costs of outpatient

treatment received during the outcome measurement

period. Costs of inpatient hospital treatment included

costs of inpatient treatment received during the outcome

measurement period. If an inpatient hospital treatment

episode started before or ended after the outcome mea-

surement period, we included a proportion of the treat-

ment episode costs corresponding to the proportion of

the treatment episode that overlapped with the outcome

measurement period.

In Denmark, healthcare is predominantly publicly

provided and financed and our data included information

on publicly financed or subsidized healthcare services

only (19, 20). Costs of prescription drug use included

public subsidies and individuals’ out-of-pocket co-

payments. Costs of primary healthcare use and costs of

hospital treatment only included publicly financed costs.

All costs are in 2023 euros.

Covariates

(work type, sector, industry, occupation, seniority, actual

weekly working hours, time of day working, commuting

time, number of full-time workers at the local workplace

and number of full-time workers in the overall com-

pany), (iii) physical work environment characteristics

(exposure to various physical working conditions and

physical strenuousness of work), (iv) health status char-

acteristics (body mass index and variables describing

current or previous treatment for diseases) and (v) health

behaviors (smoking behavior, alcohol consumption and

exercise behavior). All these covariates were measured

at or before questionnaire completion.

We also included a survey wave indicator and a

variable measuring length of the period over which we

measured hours of SA and costs of healthcare use (since

our outcome measures increased with measurement

period length).

Supplementary Appendix 2 provides more details

on the covariates.

Analytical framework

As covariates, we included (i) sociodemographic char-

acteristics (sex, age, education, immigrant status, fam-

ily type and interaction of age of the youngest child in

the family with worker’s sex), (ii) job characteristics

All analyses were performed using sample weights to

make the results representative for the source population

from which our study sample was drawn (the WEHD

source population excluding workers in private compa-

nies with <10 full-time workers, workers in agriculture,

forestry or fishing, and workers in subsidized jobs for

disabled workers).

Like Mathisen et al (12, 13), we applied a simplified

version of the parametric g-formula without time-vary-

ing variables (21, 22) to estimate changes in costs of

SA and healthcare use from hypothetical improvements

of the PSWE. The parametric g-formula method uses

regression models to estimate potential population-level

outcome effects from a hypothetical intervention by

comparing simulated outcomes for a given population in

two scenarios: without and with the intervention.

We analyzed two specific types of hypothetical inter-

ventions. The first intervention type changed the PSWE

from a situation where all workers experienced the least

desirable PSWE to a situation where all workers experi-

enced the most desirable PSWE (measured by either one

specific PSWE factor or all PSWE factors). This interven-

tion resembles a randomized controlled trial comparing

an all-exposed group with an all-unexposed group (23).

The second intervention type changed the PSWE

from a situation where the workers experienced their

observed PSWE to a situation where all workers expe-

rienced the most desirable PSWE (again measured by

either one or all PSWE factors). This intervention is

equivalent to an intervention removing a harmful expo-

sure from a real-world setting (23).

We used hurdle models to generate outcome predic-

tions for the scenarios that we compared. Hurdle models

Scand J Work Environ Health – online first

BEU, Alm.del - 2024-25 - Bilag 310: Orientering om videnskabelig artikel fra Nationale Forskningscenter for Arbejdsmiljø, fra beskæftigelsesministeren

Graversen et al

are appropriate because our outcome variables (costs of

SA and healthcare use) have many zero observations

and a right-skewed distribution for the observations

with positive value. Supplementary Appendix 3 provides

details on the computational algorithm.

As previous studies have reported gender differences

in the association between PSWE factors and SA, all

analyses were performed for men and women combined

and separately (24, 25). Furthermore, we made separate

analyses for workers in the public sector, private compa-

nies, and five different industries (the five industries with

most observations in our regression sample).

For all calculations, we used SAS 9.4 and Stata 18.

Sensitivity analyses

Table 1.

Means of outcome variables in the source population

Annual hours of sickness absence

Hourly wage rate (euros)

Annual costs of sickness absence (euros)

Annual costs of prescription drug use (euros)

Annual costs of primary health care use (euros)

Annual costs of hospital treatment (euros)

Total annual health care costs (euros)

Excludes workers in private companies with <10 full-time workers, workers

in agriculture, forestry or fishing, and workers in subsidized jobs for disabled

workers. Hourly wage rates and all costs are in 2023 euros.

Calculated from predictions of a hurdle model including the full set of ex-

planatory variables listed in supplementary appendix 2.

The sum of the means of annual costs of prescription drug use, primary

healthcare use and hospital treatment does not exactly add up to the mean of

total annual health care costs because the different mean values were calcu-

lated from non-linear models.

All

Men Women

52.0

38.6

66.1

35.4

38.6

32.1

1706

1367

2053

161

157

177

279

216

341

1056

849

1245

1539

1308

1780

SA history strongly predicts future SA and may be

associated with workers’ perception of their working

conditions (26–28). In sensitivity analyses, we therefore

repeated the main analyses while adding self-reported

days of SA within the last year as a covariate. The

sensitivity analyses were performed using observations

from the 2014 to 2018 waves only, since they were the

waves for which self-reported information on days of SA

within the last year was available. It was not possible

to use information on workers’ SA history from the SA

all workplaces report information on their workers’ SA,

and the workplaces reporting change from year to year).

Among the observations in the regression sample,

2740 (4.0%) had missing values for the PSWE factors,

and 8335 (12.3%) had missing values for the PSWE

factors or the covariates. To avoid losing statistical

power, we chose not to exclude observations with miss-

ing values on the explanatory variables. Instead, for

each explanatory variable with missing data, that were

all categorical, we created an extra category to hold

the cases with missing data. Such a missing-group

method can produce biased regression estimates (29).

As a sensitivity check, we performed analyses excluding

observations with missing values on PSWE factors, and

complete-case analyses excluding all observations with

missing values on explanatory variables.

for women than men. The average annual SA costs

per worker was €2053 for women and €1367 for men,

and the average annual costs of overall healthcare use

per worker was €1780 for women and €1308 for men.

Low-wage workers experienced more SA than high-

wage workers. Therefore, the average annual SA costs

per worker was lower than the product of the average

annual hours of SA per worker and the average hourly

wage rate.

Economic gains from hypothetical improvements of the

general psychosocial work environment

Results

Supplementary tables S2–S7 provide descriptive statis-

tics on the explanatory variables, ie, the PSWE factors

and the covariates.

Table 1 shows that in the source population the aver-

age annual SA per worker was 52 hours and the aver-

age annual costs of SA and healthcare use per worker

were €1706 and €1539, respectively. Costs were higher

Figure 2 presents the estimated annual economic gains

per worker from hypothetical interventions changing all

PSWE factors from least to most desirable (upper part

of figure) and from observed to most desirable (lower

part of figure).

When we estimated the effects of a hypothetical

improvement that changed the general PSWE from the

least to the most desirable, we found an annual economic

gain of €1685 (95% CI €1234–2135) per worker. When

we estimated the effects of a hypothetical improvement

that changed the general PSWE from the observed to

the most desirable, the economic gain was €305 (95%

CI €134–476).

In both analyses, the economic gain was overwhelm-

ingly driven by savings in SA. Savings in healthcare use

contributed only marginally (or not at all) to economic

gain (figure 2).

Supplementary figure S1 shows the estimates strati-

fied by sex. Hypothetical PSWE improvements were

associated with economic gain among both women and

men. However, the annual economic gain per worker

was considerably higher among women than among

men, both for improvements from the least to the most

desirable [€2321 (women) and €943 (men), respectively]

and from the observed to the most desirable (€505

(women) and €128 (men), respectively).

Scand J Work Environ Health – online first

BEU, Alm.del - 2024-25 - Bilag 310: Orientering om videnskabelig artikel fra Nationale Forskningscenter for Arbejdsmiljø, fra beskæftigelsesministeren

Economic effects of improving the psychosocial work environment

Supplementary table S8 provides separate estimates

for different sectors and industries. The estimated eco-

nomic gains from hypothetical PSWE improvements

(from least to most desirable and from observed to most

desirable) were larger for workers in the public sec-

tor than workers in private companies. Partly due to a

smaller number of observations in the industry-specific

analyses, differences between the estimated gains in

the five selected industries were generally statistically

insignificant.

Economic gains from hypothetical improvements of

specific psychosocial work environment factors

Figure 3 shows the economic effects of hypothetical

improvements of specific PSWE factors from the least to

the most desirable. We observed the strongest economic

gains for elimination of threats of violence and for

improvements of quality of leadership and social sup-

port from colleagues. When we analyzed the economic

effects of improvements from the observed to the most

desirable situation, we observed the strongest economic

gains for improvements in social support, influence at

work and quality of leadership (figure 4).

Improvements in recognition from colleagues and

in social inclusiveness were associated with economic

losses both for improvements from least to most desir-

able and improvements from observed to most desir-

able. In addition, elimination of sexual harassment was

associated with an economic loss in the analyses of

improvements from least to most desirable.

Supplementary Appendix 4 provides separate esti-

mates for SA and healthcare use and for women and

men (figures S2–S17), and supplementary Appendix 5

Figure 2.

Estimates with 95% confidence

intervals for annual economic gains per

worker from two hypothetical improve-

ments of the general psychosocial work

environment. Parametric g-formula analy-

ses with adjustment for sociodemographic

characteristics, job characteristics,

physical work environment characteristics,

health status and health behaviors.

Figure 3.

Estimates with 95% confidence

intervals for annual economic gains per

worker from hypothetical improvements

(least to most desirable) of specific psy-

chosocial work environment factors. Para-

metric g-formula analyses with adjustment

for sociodemographic characteristics, job

characteristics, physical work environment

characteristics, health status and health

behaviors.

Scand J Work Environ Health – online first

BEU, Alm.del - 2024-25 - Bilag 310: Orientering om videnskabelig artikel fra Nationale Forskningscenter for Arbejdsmiljø, fra beskæftigelsesministeren

Graversen et al

Figure 4.

Estimates with 95% confidence

intervals for annual economic gains per

worker from hypothetical improvements

(observed to most desirable) of specific

psychosocial work environment factors.

Parametric g-formula analyses with

adjustment for sociodemographic char-

acteristics, job characteristics, physical

work environment characteristics, health

status and health behaviors.

shows separate estimates for the public sector, private

companies, and selected industries (figures S18–S61).

Sensitivity analyses

When including self-reported days of SA within the

last year in the set of covariates, the economic gains

from hypothetically improving the general PSWE

became weaker. The estimated annual economic gains

per worker from improving the general PSWE from

the least to the most desirable and from the observed

to the most desirable were reduced to €1193 and €201,

respectively (supplementary figure S62). Furthermore,

with the inclusion of days of SA within the last year, we

no longer found economic losses from improvements

in recognition from colleagues and from improvements

from least to most desirable in social inclusiveness

(supplementary figures S63–S64).

Excluding observations with missing values on

explanatory variables from the analyses did not change

the results substantially (supplementary table S9 and

figures S76–S87).

general PSWE from the least to the most desirable and

from the observed to the most desirable were €1685

and €305, respectively. The gains were largely driven

by savings in costs related to SA, and they were higher

for women than men and for public sector compared to

private sector workers.

When analyzing the economic effects of hypothetical

improvements of specific PSWE factors, we found the

largest economic gains for improvements from least to

most desirable for elimination of threats of violence and

improvements in quality of leadership and social sup-

port from colleagues. For improvements from observed

to most desirable, we found the largest economic gains

for improvements in social support from colleagues,

influence at work and quality of leadership.

In contrast, elimination of sexual harassment was

associated with an economic loss in the analyses of

improvements from least to most desirable. Improve-

ments in recognition from colleagues and social inclu-

siveness were also associated with economic losses in

the main analyses, but these associations largely disap-

peared in our sensitivity analyses.

Comparison with previous studies

Discussion

In this large prospective cohort study, using data from a

nationally representative survey of Danish workers, we

estimated economic effects of hypothetical improve-

ments of the general PSWE and of specific PSWE

factors.

We found substantial economic gains from hypo-

thetically improving the general PSWE. The estimated

annual economic gains per worker from improving the

To our knowledge, this is the first study applying the

parametric g-formula to estimate economic effects of

hypothetical improvements in the general PSWE and

in several different specific PSWE factors, making

direct comparisons of our results with previous find-

ings difficult.

A limited number of cost-of-illness studies have esti-

mated the costs associated with different PSWE risks,

such as job strain, bullying and work-related violence

(30–32). The estimated costs per worker of specific psy-

Scand J Work Environ Health – online first

BEU, Alm.del - 2024-25 - Bilag 310: Orientering om videnskabelig artikel fra Nationale Forskningscenter for Arbejdsmiljø, fra beskæftigelsesministeren

Economic effects of improving the psychosocial work environment

chosocial risks vary a lot across studies due to, among

other things, differences in methodologies and statistical

techniques, differences in included types of costs and

differences between different countries’ labor market

structures and welfare systems.

Our estimates of the potential costs savings from

hypothetical improvements of the observed PSWE are

on the low end when compared to previous Danish stud-

ies on the effects of PSWE risks (16, 33–35).

For example, a study investigating the association

between work-related psychosocial and physical risks

and SA reported that such risks explained 28% of Danish

workers’ SA (33). Assuming the average annual costs of

SA per worker were €1706 (table 1), a simple calculation

implies that the combined costs of psychosocial and phys-

ical risks were around €478 (0.28×€1706) per worker. The

study did not provide estimates of the separate effects of

the psychosocial and physical risks, but both types of

risks contributed to SA. Therefore, the contribution of

psychosocial risks to annual SA costs per worker were

substantially smaller than €478 and closer to our estimate

of the economic gain from improving the general PSWE

from the observed to the most desirable (€305).

Another Danish study, which included costs of SA,

disability pensioning, early death and healthcare use,

estimated the annual costs of job strain (high quantita-

tive demands and low influence at work) to be around

€180–1080 per person in the labor force (with costs

converted to 2023 euros) (34).

Furthermore, two Danish studies have estimated the

costs of work-related stress, which to a large extent may

be attributed to poor psychosocial working conditions.

Pedersen et al (16) reported that the annual costs of

work absenteeism (SA and unemployment) associated

with work-related stress was €950 per worker, of which

€620 related to SA. The Economic Council of the Labor

Movement estimated the annual costs of work-related

stress (coming from reductions in working hours and

increases in SA) to be €3200 per worker (35).

The larger potential economic gains from hypotheti-

cal improvements of the general PSWE for women than

men are in line with previous research showing that

women have more SA (36, 37).

That improvements in social inclusiveness and rec-

ognition from colleagues and the elimination of sexual

harassment were associated with economic losses was

surprising. With regard to inclusiveness and recognition,

one could speculate that SA is more socially acceptable

in workplaces with a high degree of social inclusiveness

and collegial recognition and that this association may

explain the unexpected result.

It seems, however, implausible that elimination

of sexual harassment would be associated with eco-

nomic loss. Previous studies on the relationship between

various forms of sexual harassment and long-term SA

reported that sexual harassment was associated with an

increased risk of long-term SA (38, 39). One explanation

for our results regarding sexual harassment could be that

the empirical model that we used was not sufficiently

rich to capture the true relationship between the PSWE

factors and the outcome variables. The estimated effect

of sexual harassment could be biased in the direction

of finding that sexual harassment decreases the risk of

SA and healthcare use if the effect from simultaneously

improving several PSWE factors is smaller than the

sum of the effects from separate improvements in each

of these factors. This is because, in our sample, workers

exposed to sexual harassment were significantly more

likely to experience several other unfavorable PSWE

aspects than those not exposed to sexual harassment.

In our study, the ordering of PSWE factors by the

potential reductions in costs of SA achievable from

hypothetical improvements of these factors were differ-

ent from the ordering in Mathisen et al (13) of similar

PSWE factors by the potential reductions in SA. For

example, we found improvements in quality of leader-

ship to be one of the PSWE factors with the largest

potential to reduce costs of SA (supplementary figures

S2–S3), whereas Mathisen et al found improvements in

this factor to be among the PSWE factors with the low-

est potential to reduce SA. In contrast, improvements in

influence at work were among the PSWE improvements

with the most favorable effects in both our study and

in Mathisen et al. A detailed comparison of our results

with Mathisen et al is difficult. Among other things, we

estimated the economic effects of hypothetical PSWE

improvements whereas Mathisen et al estimated the

effects on hours of SA. Our study population is a sample

of all workers, whereas Mathisen et al restricted their

sample to public hospital workers. Finally, we examined

working conditions that only partly overlapped with the

working conditions examined by Mathisen et al.

Strengths and limitations

The study has some major advantages: we analyzed

a large nationwide cohort of more than 70 000 Dan-

ish workers from different industries, which increases

statistical power and external validity. We had access to

detailed self-reported data on a broad range of psycho-

social working conditions.

We applied the novel parametric g-formula, allowing

us to estimate the potential economic gains associated

with hypothetical improvements in the general PSWE

and across 17 PSWE factors, while controlling for a

wide array of potential confounders. Moreover, by pair-

wise comparison of hypothetical PSWE scenarios — the

least to the most desirable, and the observed to the most

desirable — we provided a nuanced understanding of the

potential economic effects.

Scand J Work Environ Health – online first

BEU, Alm.del - 2024-25 - Bilag 310: Orientering om videnskabelig artikel fra Nationale Forskningscenter for Arbejdsmiljø, fra beskæftigelsesministeren

Graversen et al

The study also has limitations. When estimating the

economic gains from hypothetical improvements of the

PSWE, we restricted attention to costs of workers’ SA

and healthcare use, which we could measure in registers.

However, we expect a poor PSWE to have other costs.

The PSWE can affect productivity at the workplace

not only through SA, but also through workers’ work

motivation, worker turnover and presenteeism (working

while sick) (12, 18). Previous research indicates that

productivity losses due to presenteeism might likely

be larger than those due to SA (30). Additionally, an

adverse PSWE can impose costs on workers in the form

of reduced quality of life and reduced ability to fulfil

domestic and leisure roles (18). Furthermore, exposure

to a poor PSWE can be associated with higher risk of

unemployment and early labor market exit (40–42).

By excluding relevant costs of a poor PSWE, we are

likely to underestimate the cost reductions from PSWE

improvements.

The response rate across the four WEHD waves

was 55.9%, which can be considered satisfactory com-

pared to other large-scale surveys conducted in random

samples of a national workforce in the Nordic countries

(43). We have previously analyzed in detail the demo-

graphic characteristics of the non-responders in the 2012

WEHD wave. Non-responders were more often men,

of younger age, and with lower educational attainment

(14). This selective non-response may have introduced

selection bias into our analyses. To address this poten-

tial bias, we adjusted our analyses for a wide range of

covariates – including sex, age, and education – and

used appropriate sample weights.

Residual confounding is an inherent concern in

observational studies, and we therefore adjusted for a

wide range of possible confounders. However, adjust-

ing for many confounders might also introduce over-

adjustment, potentially accounting for mediators and

leading to more conservative estimates of the economic

gains from improving the PSWE.

Several prior studies have linked various adverse

psychosocial working conditions to SA. Given the inter-

related nature of these conditions, interventions targeting

specific aspects of the PSWE may indirectly affect other

PSWE aspects, potentially broadening the impact on SA

and economic effects. Moreover, it should be noted that

the estimated potential economic gains presented are

from hypothetical scenarios where PSWE factors would

be improved to the most favorable level. These estimates

represent upper bounds on the potential economic gains

achievable through PSWE improvements, in contrast

to what may be realistic in real-life settings where

often smaller PSWE improvements are observed (4).

Finally, some PSWE improvements might inadvertently

reduce productivity. For example, reducing work pace or

quantitative demands might lower worker output. Thus,

PSWE improvements do not always result in economic

gains, especially when considering the direct costs of

implementing these changes.

Implications for real interventions

The estimated economic gains from hypothetical inter-

ventions presented in this study are economic gains

from a societal perspective that does not consider the

direct interventions costs. These economic gains can

be used as an input to obtain well-informed estimates

of the net economic societal gains of actual workplace

interventions (for given intervention costs) or estimates

of the maximum intervention costs that can be spent if

interventions should not result in net economic societal

losses.

For example, assume that a nationwide intervention

aiming to improve quality of leadership is being consid-

ered for implementation. We estimated an annual eco-

nomic gain of €93 per worker from improving quality of

leadership from observed to most desirable. Therefore,

if the intervention has maximum efficiency, actually

improving quality of leadership to most desirable for all

workers, the annual net economic gain per worker from

the intervention would be €93 minus the intervention

costs per worker. If the intervention only brings 10%

of workers with “least desirable” or “medium” quality

of leadership to “most desirable” quality of leadership,

the annual net economic gain per worker would be €9.3

minus the intervention costs per worker.

If estimates of other economic gains from the inter-

vention than reductions in costs of SA and healthcare

use exist (eg, reduced turnover or presenteeism costs),

these economic gains could of course be included when

estimating the net economic gains.

In cases, where the aim is to estimate the net eco-

nomic gains of an intervention for a specific group of

workers, our estimates should be adjusted to account for

the degree to which the workers are exposed to unfavor-

able PSWE aspects and the costs of SA and healthcare

use before the intervention (measuring the potential

for cost reductions). The larger the exposure and initial

costs of SA and healthcare use, the larger the expected

net economic gains from the intervention.

Finally, it is worth mentioning that the economic

gains from a workplace intervention is smaller from an

employer’s perspective than from the societal perspec-

tive. This is so because the reductions in the costs of

healthcare use and part of the reductions in societal

costs of SA (equal to the gains in production value) does

not fall into the hands of the employer. Employers will

generally have to pay higher salaries when SA decreases,

because they pay reduced salaries while workers are sick

or because they receive partial reimbursement for sala-

ries paid to workers with >30 days of SA. Consequently,

Scand J Work Environ Health – online first

BEU, Alm.del - 2024-25 - Bilag 310: Orientering om videnskabelig artikel fra Nationale Forskningscenter for Arbejdsmiljø, fra beskæftigelsesministeren

Economic effects of improving the psychosocial work environment

an employer’s economic gain from reductions in SA is

the increase in production value minus the increase in

salaries net of reimbursements. A policy implication

is that public subsidies may sometimes be relevant to

increase the incentives to implement interventions pro-

viding net economic societal gains.

Concluding remarks

Steel J, Godderis L, Luyten J. Productivity estimation in

economic evaluations of occupational health and safety

interventions: a systematic review. Scand J Work Environ

Health 2018 Sep;44(5):458–74.

https://doi.org/10.5271/

sjweh.3715.

Tompa E, Verbeek J, van Tulder M, de Boer A. Developing

guidelines for good practice in the economic evaluation

of occupational safety and health interventions. Scand J

Work Environ Health 2010 Jun;36(4):313–8.

https://doi.

org/10.5271/sjweh.3009.

In summary, we found that interventions improving the

PSWE could potentially lead to economic gains, in par-

ticular improvements for female workers and workers in

public sector workplaces. When considering individual

PSWE factors, interventions towards improving the

work environment with regard to threats of violence,

quality of leadership, social support from colleagues and

influence at work held the largest economic potential.

10. Robins J. A new approach to causal inference in mortality

studies with a sustained exposure period - application to

control of the healthy worker survivor effect. Math Model

1986;7(9-12):1393–512.

https://doi.org/10.1016/0270-

0255(86)90088-6.

11. Taubman SL, Robins JM, Mittleman MA, Hernán MA.

Intervening on risk factors for coronary heart disease: an

application of the parametric g-formula. Int J Epidemiol

2009 Dec;38(6):1599–611.

https://doi.org/10.1093/ije/

dyp192.

12. Mathisen J, Nguyen TL, Jense JH, Rugulies R, Rod NH.

Reducing employee turnover in hospitals: estimating the

effects of hypothetical improvements in the psychosocial

work environment. Scand J Work Environ Health 2021

Sep;47(6):456–65.

https://doi.org/10.5271/sjweh.3969.

13. Mathisen J, Nguyen TL, Jensen JH, Mehta AJ, Rugulies

R, Rod NH. Impact of hypothetical improvements in the

psychosocial work environment on sickness absence rates:

a simulation study. Eur J Public Health 2022 Oct;32(5):716–

22.

https://doi.org/10.1093/eurpub/ckac109.

14. Johnsen NF, Thomsen BL, Hansen JV, Christensen BS,

Rugulies R, Schlünssen V. Job type and other socio-

demographic factors associated with participation in a

national, cross-sectional study of Danish employees. BMJ

Open 2019 Aug;9(8):e027056.

https://doi.org/10.1136/

bmjopen-2018-027056.

15. Nordentoft M, Rod NH, Bonde JP, Bjorner JB, Madsen IE,

Pedersen LR et al. Effort-reward imbalance at work and risk

of type 2 diabetes in a national sample of 50,552 workers in

Denmark: A prospective study linking survey and register

data. J Psychosom Res 2020 Jan;128:109867.

https://doi.

org/10.1016/j.jpsychores.2019.109867.

16. Pedersen J, Graversen BK, Hansen KS, Madsen IE. The labor

market costs of work-related stress: A longitudinal study of

52 763 Danish employees using multi-state modeling. Scand

J Work Environ Health 2024 Mar;50(2):61–72.

https://doi.

org/10.5271/sjweh.4131.

17. Pejtersen JH, Kristensen TS, Borg V, Bjorner JB. The second

version of the Copenhagen Psychosocial Questionnaire.

Scand J Public Health 2010 Feb;38(3 Suppl):8–24.

https://

doi.org/10.1177/1403494809349858.

18. Tompa E, Culyer AJ, Dolinschi R. Economic evaluation of

interventions for occupational health and safety: Developing

good practice. New York: Oxford University Press; 2008.

19. Christensen BJ, Gortz M, Kallestrup-Lamb M. Medical

References

Niedhammer I, Bertrais S, Witt K. Psychosocial work

exposures and health outcomes: a meta-review of 72

literature reviews with meta-analysis. Scand J Work Environ

Health 2021 Oct;47(7):489–508.

https://doi.org/10.5271/

sjweh.3968.

Rugulies R, Aust B, Greiner BA, Arensman E, Kawakami

N, LaMontagne AD et al. Work-related causes of mental

health conditions and interventions for their improvement in

workplaces. Lancet 2023 Oct;402(10410):1368–81.

https://

doi.org/10.1016/S0140-6736(23)00869-3.

Eurofound. Psychosocial risks to workers’ well-being:

Lessons from the COVID-19 pandemic. Luxembourg:

Publications Office of the European Union; 2023.

Aust B, Møller JL, Nordentoft M, Frydendall KB, Bengtsen

E, Jensen AB et al. How effective are organizational-

level interventions in improving the psychosocial work

environment, health, and retention of workers? A systematic

overview of systematic reviews. Scand J Work Environ

Health 2023 Jul;49(5):315–29.

https://doi.org/10.5271/

sjweh.4097.

van Dongen JM, van der Beek AJ. Economic evaluations in

occupational health: what brings the best bang for the buck?

Scand J Work Environ Health 2022 May;48(4):249–52.

https://doi.org/10.5271/sjweh.4026.

Rugulies R, Burdorf A. What are the economic costs of a

poor work environment? Scand J Work Environ Health 2024

Mar;50(2):49–52.

https://doi.org/10.5271/sjweh.4143.

van Dongen JM, van Wier MF, Tompa E, Bongers PM,

van der Beek AJ, van Tulder MW et al. Trial-based

economic evaluations in occupational health: principles,

methods, and recommendations. J Occup Environ

Med 2014 Jun;56(6):563–72.

https://doi.org/10.1097/

JOM.0000000000000165.

Scand J Work Environ Health – online first

BEU, Alm.del - 2024-25 - Bilag 310: Orientering om videnskabelig artikel fra Nationale Forskningscenter for Arbejdsmiljø, fra beskæftigelsesministeren

Graversen et al

Spending in Denmark. Fisc Stud 2016;37(3-4):461–97.

https://doi.org/10.1111/j.1475-5890.2016.12119.

20. Vrangbaek K. The Danish healthcare system. In: Tikkanen

R, Osborn R, Mossialos E, Djordjevic A, Wharton G,

editors. International profiles of healthcare systems. New

York: The Commonwealth Fund; 2020.

21. Ahern J, Hubbard A, Galea S. Estimating the effects of

potential public health interventions on population disease

burden: a step-by-step illustration of causal inference

methods. Am J Epidemiol 2009 May;169(9):1140–7.

https://

doi.org/10.1093/aje/kwp015.

22. Hernán MA, Robins JM. Causal Inference: What If. Boca

Raton: Chapman & Hall/CRC; 2024.

23. Westreich D. From Patients to Policy: Population Intervention

Effects in Epidemiology. Epidemiology 2017 Jul;28(4):525–

https://doi.org/10.1097/EDE.0000000000000648.

24. Casini A, Godin I, Clays E, Kittel F. Gender difference

in sickness absence from work: a multiple mediation

analysis of psychosocial factors. Eur J Public Health 2013

Aug;23(4):635–42.

https://doi.org/10.1093/eurpub/cks183.

25. Nielsen ML, Rugulies R, Christensen KB, Smith-Hansen

L, Kristensen TS. Psychosocial work environment

predictors of short and long spells of registered sickness

absence during a 2-year follow up. J Occup Environ

Med 2006 Jun;48(6):591–8.

https://doi.org/10.1097/01.

jom.0000201567.70084.3a.

26. Hultin H, Lindholm C, Malfert M, Möller J. Short-term sick

leave and future risk of sickness absence and unemployment

- the impact of health status. BMC Public Health 2012

Oct;12(1):861.

https://doi.org/10.1186/1471-2458-12-861.

27. Sumanen H, Pietiläinen O, Lahelma E, Rahkonen O. Short

sickness absence and subsequent sickness absence due

to mental disorders - a follow-up study among municipal

employees. BMC Public Health 2017 Jan;17(1):15.

https://

doi.org/10.1186/s12889-016-3951-7.

28. Stapelfeldt CM, Nielsen CV, Andersen NT, Krane L,

Borg V, Fleten N et al. Sick leave patterns as predictors

of disability pension or long-term sick leave: a 6.75-year

follow-up study in municipal eldercare workers. BMJ

Open 2014 Feb;4(2):e003941.

https://doi.org/10.1136/

bmjopen-2013-003941.

29. Jones MP. Indicator and Stratification Methods for Missing

Explanatory Variables in Multiple Linear Regression. J Am

Stat Assoc 1996;91(433):222–30.

https://doi.org/10.1080/0

1621459.1996.10476680.

30. Hassard J, Teoh KR, Visockaite G, Dewe P, Cox T. The

cost of work-related stress to society: A systematic review.

J Occup Health Psychol 2018 Jan;23(1):1–17.

https://doi.

org/10.1037/ocp0000069.

31. Hassard J, Teoh KR, Visockaite G, Dewe P, Cox T. The

financial burden of psychosocial workplace aggression: A

systematic review of cost-of-illness studies. Work Stress

2018;32(1):6–32.

https://doi.org/10.1080/02678373.2017.1

380726.

32. Hassard J, Teoh KR, Cox T. Estimating the economic burden

posed by work-related violence to society: A systematic

review of cost-of-illness studies. Saf Sci 2019;116:208–21.

https://doi.org/10.1016/j.ssci.2019.03.013.

33. Analytics Squared. Arbejdsmiljø og sygefravær [Working

environment and sick leave]. Copenhagen, Denmark; 2021.

34. Juel K, Sørensen J, Brønnum-Hansen H. Risikofaktorer og

folkesundhed i Danmark [Risk factors and public health

in Denmark]. Copenhagen, Denmark: Statens Institut for

Folkesundhed; 2006.

35. Economic Council of the Labour Movement. Stress koster

55 mia. kr. om året [Stress costs 55 billion Danish kroner per

year]. Copenhagen, Denmark; 2024.

36. Timp S, van Foreest N, Roelen C. Gender differences in

long term sickness absence. BMC Public Health 2024

Jan;24(1):178.

https://doi.org/10.1186/s12889-024-17679-

37. Arcas MM, Delclos GL, Torá-Rocamora I, Martínez

JM, Benavides FG. Gender differences in the duration

of non-work-related sickness absence episodes due to

musculoskeletal disorders. J Epidemiol Community Health

2016 Nov;70(11):1065–73.

https://doi.org/10.1136/jech-

2014-204331.

38. Hogh A, Conway PM, Clausen T, Madsen IE, Burr H.

Unwanted sexual attention at work and long-term sickness

absence: a follow-up register-based study. BMC Public

Health 2016 Jul;16:678.

https://doi.org/10.1186/s12889-

016-3336-y.

39. Blindow K, Bondestam F, Johansson G, Bodin T, Westerlund

H, Nyberg A. Sexual and gender harassment in Swedish

workplaces: A prospective cohort study on implications

for long-term sickness absence. Scand J Work Environ

Health 2021 Sep;47(6):466–74.

https://doi.org/10.5271/

sjweh.3971.

40. Knardahl S, Johannessen HA, Sterud T, Härmä M, Rugulies

R, Seitsamo J et al. The contribution from psychological,

social, and organizational work factors to risk of disability

retirement: a systematic review with meta-analyses. BMC

Public Health 2017 Feb;17(1):176.

https://doi.org/10.1186/

s12889-017-4059-4.

41. Stansfeld SA, Carr E, Smuk M, Clark C, Murray E,

Shelton N et al. Mid-life psychosocial work environment

as a predictor of work exit by age 50. PLoS One 2018

Apr;13(4):e0195495.

https://doi.org/10.1371/journal.

pone.0195495.

42. Sundstrup E, Thorsen SV, Rugulies R, Larsen M, Thomassen

K, Andersen LL. Importance of the Working Environment

for Early Retirement: Prospective Cohort Study with

Sep;18(18):9817.

https://doi.org/10.3390/ijerph18189817.

43. Magnusson Hanson LL, Leineweber C, Persson V, Hyde

M, Theorell T, Westerlund H. Cohort Profile: The Swedish

Longitudinal Occupational Survey of Health (SLOSH).

Int J Epidemiol 2018 Jun;47(3):691–692i.

https://doi.

org/10.1093/ije/dyx260.

Received for publication: 30 October 2024

Scand J Work Environ Health – online first