FormalPara Key Points for Decision Makers

Despite clear evidence that poor mental health is associated with lost productivity at work, more evidence is required to understand the extent to which mental illness decreases productivity and the mechanisms through which this occurs in order to provide appropriate policy responses.

A better understanding of the relationship between mental illness and worker productivity is needed to understand the trade-offs between presenteeism and absenteeism.

Workplace policies that limit and help workers manage job stress can help improve workers’ productivity.

1 Introduction

Mental health disorders in the workplace, such as depression and anxiety, have increasingly been recognised as a problem in most countries. Using a human capital approach, the global economic burden of mental illness was estimated to be US$$2.5 trillion in 2010 increasing to US$$6.1 trillion in 2030; most of this burden was due to lost productivity, defined as absenteeism and presenteeism [1]. Workplaces that promote good mental health and support individuals with mental illnesses are more likely to reduce absenteeism (i.e., decreased number of days away from work) and presenteeism (i.e., diminished productivity while at work), and thus increase worker productivity [2]. Burton et al. provided a review of the association between mental health and worker productivity [3]. The authors found that depressive disorders were the most common mental health disorder among most workforces and that most studies examined found a positive association between the presence of mental health disorders and absenteeism (particularly short-term disability absences). They also found that workplace policies that provide employees with access to evidence-based care result in reduced absenteeism, disability and lost productivity [3].

However, this review is now outdated. Prevalence rates for common mental disorders have increased [4], while workplaces have also responded with attempts to reduce stigma and the potential economic impact [5], necessitating the need for an updated assessment of the evidence. Furthermore, given that most of the global economic burden of mental illness is due to lost productivity [1], it is important to have a good understanding of the existing literature on this outcome. While the previous review focused on the prevalence of certain mental health conditions and the available interventions and workplace policies, this review focused on the measures of lost productivity and the instruments used, as well as the data and methods employed, which the previous review did not examine in depth. Thus, the objectives of this paper were to update the Burton et al. review [3] on the association between mental health and lost productivity, and undertake a critical review of the literature that has been published since then, specifically how researchers have studied this relationship, the type of data and databases they have employed, the methods they have used, their findings, and the existing gaps in the literature.

2 Methods

We undertook a critical review, i.e., a review that presents, analyses and synthesises evidence from diverse sources by extensively searching the literature and critically evaluating its quality [6], ultimately identifying the most significant papers in the field.Footnote 1 We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement [7] to guide our analysis. Our review focused on all studies published since 2008, which examined the relationship between mental health and workplace-related productivity among working-age adults. We used the Population, Intervention, Control, Outcomes, and Study design (known as PICOS) criteria to guide the development of the search strategy.

2.1 Eligibility Criteria

The populations of interest comprised working-age adults (18–65 years old). Studies focusing solely on volunteers and/or caregivers (i.e., unpaid workers) were excluded. The intervention(s), or rather more appropriately the exposure(s), had to be a diagnosis of any mental disorder/illness or self-reported mental health problem(s). Any studies that examined substance use and/or physical health in addition to mental health were included if results were reported separately for mental health-related outcomes. The control or comparator group, where applicable, included working age individuals without a mental disorder/illness or mental health problem(s). The outcome(s) included lost workplace productivity measured by absenteeism, presenteeism, sick leave, short- and/or long-term disability, or job loss. Studies that examined productivity of home-related activities (e.g., housework) were excluded. Studies with an observational study design and/or regression analysis were included; randomised control trials, cost-of-illness studies and economic evaluations were excluded (the first two were only included if they examined the relationship between mental health and lost productivity). Only original studies were considered; however, relevant reviews were retained for reference checking to find relevant studies, which may not have been captured by the search strategy.

2.2 Search Strategy

We searched literature published in English from 1 January 2008 to 31 May 2020. Structured searches were done in MEDLINE and EconLit to capture the most relevant literature published in the medical and economics fields, respectively. We also undertook relevant searches in Google and on specific websites of interest (e.g., UK Parliament Hansard, the National Institute for Health and Care Excellence, the Centre for Mental Health, the Health Foundation, the Institute for Fiscal Studies and the King’s Fund) and a hand search of the references of key papers [8]. Search terms or strings were developed on the basis of four concepts: population or workplace, intervention/exposure (i.e., presence of mental disorder/illness), work-related outcomes, and study design (see Table 1).

Table 1 Concepts and search terms used to identify relevant studies
Full size table

2.3 Study Selection

After duplicate records were removed, one reviewer (LB) screened all titles and abstracts while additional reviewers (CdO and RJ) were brought in for discussion, if/where necessary. Articles were excluded either because they did not examine the relationship between mental health and lost productivity (e.g., some cost-of-illness studies) or were mainly focused on physical health. Subsequently, all relevant full-text articles were retrieved and screened by one reviewer (LB) to confirm eligibility; additional reviewers (MS, RJ or CdO) were brought in, if/where necessary.

2.4 Data Extraction

Two reviewers (LB and MS) undertook the data extraction, and an additional reviewer (RJ or CdO) was assigned to resolve any disagreements. The research team developed a data extraction form, based on the Cochrane good practice data extraction form, which included study information (author(s), year of publication), country (where the study was published or conducted), aims of study, study design (cross-sectional, longitudinal), data source(s) (i.e., database(s), surveys/questionnaires), study population (sample size, age range), mental disorder(s) examined, workplace outcome examined (absenteeism, presenteeism, short-term disability, long-term disability, job loss, other), methods employed (statistical analysis, regression model employed), and results/key findings.

2.5 Quality Assessment

We reviewed the methods employed in the studies to assess their quality and robustness, drawing loosely on the Newcastle–Ottawa Scale, a risk-of-bias assessment tool for observational studies [9]. We paid particular attention to whether studies were able to move beyond simple associations and attempted to address causal inference, where necessary, and whether they took account of endogeneity (i.e., cases where the explained variable and the explanatory variable are determined simultaneously) and/or unobserved heterogeneity (i.e., cases where the presence of unexplained (observed) differences between individuals are associated with the (observed) variables of interest), which are common issues when examining the relationship between mental health and lost productivity. All studies that recognised and/or accounted for these issues were considered high quality. We also examined the type of data/databases employed (i.e., cross-sectional or longitudinal data and representative, population-based samples), findings, and limitations (and the extent to which these impacted the findings), which were also considered when determining the quality of a study.

2.6 Data Synthesis

Given the heterogeneity of studies examined, undertaking a meta-analysis was not possible. Therefore, we undertook a narrative synthesis of the relevant literature, where we synthesised the existing evidence by mental disorder/illness and workplace outcome (absenteeism, presenteeism, sick leave, short- and long-term disability, or job loss), if/where appropriate.

3 Results

3.1 Study Selection

After all citations were merged and duplicates removed, our search produced 648 unique records, of which 89 full texts were assessed; four studies were obtained from other sources (e.g., Google searches). Ultimately, 38 studies were included in the final review [10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47] (see Fig. 1) and relevant data were extracted (see Table 2 and Table A1 in the Appendix for more details).

Fig. 1
figure 1

PRISMA flow diagram

Full size image
Table 2 Details of included studies
Full size table

3.2 Overview of Studies

All studies focused on individuals typically between the ages of 18 and 64/65 years. Some studies (n = 5) examined individuals 20 or 25 years and older [11, 12, 16, 28, 38] to account for younger individuals who might still be in school and thus not working, while other studies had different lower and upper age limits (e.g., age 15 [25, 47] and age 60 [11, 38] years, respectively). Most studies were from the USA (n = 10; 26%) and the Netherlands (n = 6; 16%); this result is line with the findings from a review of economic evaluations of workplace mental health interventions [48]. The remaining studies were from Australia (n = 4), Japan (n = 4), South Korea (n = 3), multiple countries (n = 4), Brazil (n = 1), Colombia (n = 1), Denmark (n = 1), Finland (n = 1), Norway (n = 1), Singapore (n = 1), and the United Kingdom (n = 1). Many studies did not specify the setting or industry or state the size of the firm where the study was undertaken (also found elsewhere [48]); consequently, this information was not included in the data extraction form.

3.3 Measures and Instruments/Tools Used

3.3.1 Mental Health

Most studies (n = 16) examined depression/depressive symptoms, major depressive disorder or other mood disorders (see Fig. 2). Two studies examined anxiety and five studied both anxiety and depression. A smaller number of studies examined other disorders—three studies examined attention-deficit hyperactivity disorder (ADHD), two studies focused on bipolar disorder, one examined panic disorder, one studied binge-eating disorder, and one looked at other disorders including mental disorders (depressive symptoms and cognitive function). Three studies looked at mental health broadly speaking (two studies examined poor mental health and another studied common mental disorders). Finally, four studies examined multiple mental disorders (e.g., depression, bipolar disorder, anxiety disorders, emotional disorders, substance use disorders, ADHD). Some studies used a binary indicator for the presence/absence of a mental disorder/poor mental health, while other analyses used different aggregate measures of mental illness or psychological distress, based on the number of recorded symptoms.

Fig. 2
figure 2

Studies by mental disorder. ADHD attention-deficit hyperactivity disorder

Full size image

A variety of instruments/tools were used to measure mental health, depending on the disorder. Depression was measured using the Kessler Psychological Distress Scale (K6 scale) [49], Patient Health Questionnaire (PHQ-9) depression scale [50], Center for Epidemiologic Studies Depression Scale (CES-D) [51], Short General Health Questionnaire (GHQ-12) [52], Major Depression Inventory (MDI) [53], Hamilton Rating Scale for Depression (HAM-D) [54], and Mental Health Inventory (MHI-5) [55]. In studies that examined both anxiety and depression (n = 2), the authors used either the Hospital Anxiety and Depression Scale (HADS) [56] or the Composite International Diagnostic Interview (CIDI) [57]. In one study [42], severity of anxiety and depressive symptoms was assessed using the Beck Anxiety Inventory [58] and the Inventory for Depressive Symptomatology questionnaire [59], respectively. In another study [33], mood disorder was measured using the Mood Disorder Questionnaire (MDQ) [60]. In one study [41], ADHD was assessed using the WHO World Mental Health (WMH) survey [61]; in another [35], it was assessed using the WHO Adult ADHD Self-Report Scale [62]. Panic disorders were measured using the Panic Disorder Severity Scale [63] in one study [16].

3.3.2 Lost Productivity

Nineteen studies examined both absenteeism and presenteeism, eight studies examined absenteeism only, two studies examined presenteeism only, and nine examined other or several workplace outcomes, such as employment, absenteeism, presenteeism, workplace accidents/injuries, short- and/or long-term disability, activity impairment and/or job loss (see Fig. 3).

Fig. 3
figure 3

Studies by workplace outcome

Full size image

Five studies used the Work Productivity and Activity Impairment (WPAI) questionnaire [64] (Beck et al. [27], Jain et al. [36], Able et al. [30], Asami et al. [31], Ling et al. [44]); three used the WHO’s Health and Work Performance Questionnaire (HWP) [65] (Hjarsbech et al. [18], Woo et al. [38], Park et al. [16]) to determine absenteeism and presenteeism. A recent systematic review also found that that the WPAI was most frequently applied in economic evaluations and validation studies to measure lost productivity [66]. Two studies [12, 20] used the Work Limitations Questionnaire [67]. Other studies used a variety of different instruments to measure lost productivity, such as the Trimbos/iMTA questionnaire for Costs Associated with Psychiatric illness (TiC-P) [68] (Bokma et al. [26]), the Short-Form Health and Labour Questionnaire [69] (Bouwmans et al. [45]), the WHO Disability Assessment Schedule (WHO-DAS) [70] (de Graaf et al. [23]) and the Endicott Work Productivity Scale [71] (McMorris et al. [33]). One study [43] made use of four work performance measures to examine lost productivity: WPAI [64], Work Limitations Questionnaire (WLQ) [66], Endicott Work Productivity Scale (EWPS) [71] and Functional Status Questionnaire Work Performance Scale (WPS) [72].

3.4 Data Sources and Methods

3.4.1 Data

Most studies (n = 20) employed data collected through surveys/questionnaires, though some used publicly available datasets, such as the Medical Expenditure Panel Survey [29], the National Comorbidity Survey Replication [28] and the National Latino and Asian American Study [28], the US National Health and Wellness Survey [44], the Household, Income and Labour Dynamics in Australia survey [25, 47], and the Singapore Mental Health Study [24]. One study used administrative claims data [32]. Three studies made use of linked data, such as Hjarsbech et al. [18], which linked questionnaires to the Danish National Register of Social Transfer Payments; Erickson et al. [43], which utilised questionnaires linked to medical records, and Mauramo et al. [34], which used survey data from the Helsinki Health Study linked to employer's register data on sickness absence. Only one study employed trial data [45]. Most studies (n = 29; 76%) employed cross-sectional data; few used longitudinal data (n = 9; 24%).

3.4.2 Methods

Several studies (n = 8) used regression analysis to examine the relationship between mental health and lost productivity, namely linear regression [11, 17] and logistic regression models [25, 29, 42, 45]. Two studies employed two-part models, where the first part examined the probability/odds of workers experiencing absenteeism, while the second part modeled the number of hours of absenteeism [10] or the number of work days missed [29]. One paper employed Poisson regressions to model the rate of work-lost days (absenteeism) and work-cut days (presenteeism) [34]. Another study computed Kaplan–Meier survival curves to estimate the mean and median duration of sickness absence due to depressive symptoms [40], and one estimated a Cox's proportional hazards model to analyse whether and to what extent depressive symptoms at baseline predicted time to onset of first long-term sickness absence during the 1-year follow-up period [18]. Only one study employed instrumental variables to address the potential endogeneity of the mental illness variable employed [28] and four employed longitudinal data models [13, 20, 25, 47].

3.5 Evidence Synthesis

Almost all studies (n = 36) found a positive (and, many times, a strong) association between the presence of mental illness/disorders or poor mental health and productivity loss measured by absenteeism and/or presenteeism. Nevertheless, there were a few exceptions—one study found that mood disorders were associated with decreased presenteeism (i.e., work performance) but found no significant relationship between mood disorders and absenteeism [11]. Another study found that individuals with binge-eating disorders reported greater levels of presenteeism and lost productivity than those without but found no effect for absenteeism [44].

Many studies (n = 6) on depression examined both absenteeism and presenteeism where the presence of the former was positively associated with the latter (as was the case for studies, which examined only absenteeism and only presenteeism), and the latter was higher among those with higher severity of depression. These findings held in studies examining major depressive disorder and bipolar disorder (though one study found that symptoms of mania or hypomania were not significantly associated with absenteeism) [14]. Studies examining depression and anxiety (and anxiety alone, including panic disorder) generally examined both absenteeism and presenteeism and found that these disorders were significantly associated with lost productivity. One study found that workers with binge-eating disorder reported greater levels of presenteeism than those without but no differences in absenteeism. All studies on ADHD (n = 3) examined both absenteeism and presenteeism and found ADHD was associated with more days of missed work and poor work performance. Studies looking at mental health (broadly defined) typically examined absenteeism only, finding a positive relationship between both, though the magnitude of the effect was found to be modest in one study [47]. Studies examining multiple disorders (n = 4) also examined both absenteeism and presenteeism. Overall, having a mental disorder was positively associated with lost productivity; however, one study found no significant relationship between mood disorders and alcohol use/dependence and absenteeism [11].

Many studies (n = 6) found that higher severity of the disorder or co-occurring mental health conditions was associated with greater productivity loss. For example, Knudsen et al. found that while comorbid anxiety and depression and anxiety alone were significant risk factors for absenteeism, depression alone was not [37].

Some studies examined outcomes separately for men and women (n = 5) or examined specific groups (n = 1). For example, Ammerman et al. examined high-risk, low-income mothers with major depression and found that depression significantly increased the likelihood of absenteeism (i.e., missing workdays) among this group [29]. However, beyond gender, studies did not report on differences by ethnicity/race and/or age.

Overall, we found that the literature on this topic continues to examine the most common mental disorders (e.g., depression and anxiety) using similar data sources and analysis techniques as the Burton et al. review [3] (see Table 3). However, more recent literature shows that the positive relationship between the presence of mental disorders and lost productivity may not hold in all instances.

Table 3 Comparison between the Burton et al. [3] and the de Oliveira et al. [current paper] reviews
Full size table

4 Discussion

The goal of this review was to provide a comprehensive overview and critical assessment of the most recent literature examining the relationship between mental health and workplace productivity, with a particular focus on data and methods employed. It provides clear evidence that poor mental health is associated with lost productivity, defined as increased absenteeism (i.e., more missed days from work) and increased presenteeism (i.e., decreased productivity at work). However, overall, only three studies were of high quality [25, 28, 47]. Studies with greater rigour and more robust methods, which accounted for unobserved heterogeneity for example, found a similar positive relationship but a smaller effect size [25, 47].

Other reviews have also found large significant associations between measures of mental health and lost productivity, such as absenteeism [3, 73,74,75]. For example, Burton et al. [3] found that depressive disorders were the most common mental health disorder among most workers, with many studies showing a positive association between the presence of mental health conditions and absenteeism, particularly short-term disability absences [3]. However, we found that studies employing superior methodological study design have shown the strength of the observed association may be smaller than previously thought.

Overall, our findings are in line with those from other reviews [73,74,75] and the Burton et al. study [3]. We too found that the most common disorder examined was depression, followed by depression and anxiety, the most studied workplace outcomes were both absenteeism and presenteeism, and that there was an association between mental disorders and both absenteeism and presenteeism. We found that studies employed a variety of data sources, from data collected from surveys/questionnaires to existing surveys and administrative data. Regression analysis was commonly used to examine the relationship between mental health and lost productivity, though there were some studies where the most appropriate regression model was not used given the outcome examined (e.g., linear regression models were used regardless of the type of outcome examined).

Some studies employed small sample sizes [20, 43], which are not representative of the broader population and can thus impact the generalizability of findings, and other studies that did use nationally representative population samples employed cross-sectional designs [11, 42, 46], which can limit causal inference. Therefore, the vast majority did not examine the causal effect of mental health on lost productivity, but rather only the association between the two. A notable exception was Banerjee et al. [28], who examined the potential endogeneity of the mental illness variable used. Moreover, few studies employed longitudinal data, which can help account for unobserved heterogeneity (that may be correlated with both mental health and lost productivity) and minimise the potential for reverse causality and omitted variable bias; Wooden et al. [47] and Bubonya et al. [25] were notable exceptions. Wooden et al. found that the association between poor mental health and the number of annual paid sickness absence days was much smaller once they accounted for unobserved heterogeneity and focused on within-person differences [47]. For example, the incidence rate ratios for the number of sickness absence days for employed women and men experiencing severe depressive symptoms were 1.31 and 1.38, respectively, in the negative binomial regression models but dropped to 1.10 and 1.13, respectively, once the authors controlled for unobserved heterogeneity through the inclusion of correlated random effects. Thus, it may be that previous research has overstated the magnitude of the association between poor mental health and lost productivity. More studies with rigorous causal inference are required to help strengthen the ability to make informed policy recommendations.

Few studies explored the factors that might explain absenteeism and/or presenteeism due to mental health. Again, the study by Bubonya et al. was a notable exception [25], providing several important insights on the relationship between mental health and lost productivity. According to the authors, initiatives that limit and help workers manage job stress seem to be the most promising avenue for improving workers’ productivity. Furthermore, the authors found that presenteeism rates among workers with poor mental health were relatively insensitive to work environments, in line with other research from the UK [76]; consequently, they suggested that developing institutional arrangements that specifically target the productivity of those experiencing mental ill health may prove challenging. These findings are particularly important in the context of the COVID-19 pandemic due to changes in work arrangements and workplaces (e.g., working from home while trying to balance work with home and care responsibilities, hybrid working arrangements, and ensuring workplaces have COVID-19-secure measures in place). This work will be of particular interest to employers and decision makers looking to improve worker productivity.

Most literature examined either depression or anxiety or both, the most common mental disorders. Few studies examined mental disorders such as ADHD, bipolar disorder and eating disorders, and no studies examined schizophrenia and other psychotic disorders, personality disorder or suicidal/self-harm behaviour. More work is needed on these mental disorders, which, although less prevalent and thus less studied, are potentially more work disabling (despite already low employment rates for individuals with these conditions) [77, 78]. Other research suggests there are important gender differences [25, 28]. For example, Bubonya et al. found that increased job control can help reduce absenteeism for women with good mental health, though not for women in poor mental health [25]. Banerjee et al. found that the impact of poor mental health on the likelihood of being employed and in the labour force is higher for men [28]. Future research should ensure that gender differences, as well as other differences (e.g., age, industry, job conditions), are examined to ensure tailored polices are developed and implemented.

There is also a need to better understand the extent to which mental illness decreases productivity at work and the mechanisms through which this occurs, as this could help inform the role of employment policy and practices to minimise presenteeism [25]. Some research suggests that conducive working conditions, such as part-time employment and having autonomy over work tasks, can help mitigate the negative impact of mental health on presenteeism [76]. Alongside this, it is important to learn more about the dynamics of the relationship between mental illness and worker productivity to understand the trade-offs between presenteeism and absenteeism [25]. For example, it would be helpful to understand whether policies that incentivise workers with mental ill health to take time off improve overall productivity by reducing presenteeism. None of the studies in this review explored this trade-off. Finally, more rigorous research on this topic would help achieve a better understanding of the overall economic impact of mental disorders.

This review is not without limitations. It only included studies obtained from a few select databases and did not include grey literature, and only one reviewer screened the titles and abstracts (though the purpose was not to undertake a systematic review); however, it examined papers and reports from select websites of interest. Furthermore, this review only focused on the relationship between mental health and lost productivity. Although lost productivity is an important labour market outcome, there are other outcomes that mental health can impact such as labour force participation, wages/earnings, and part-time versus full time employment. Finally, this review only included studies published in English and therefore may have missed other relevant studies. Nonetheless, this review has several strengths. It provides an updated review on this topic, thus addressing a critical gap in the literature, and examined the type of data and databases employed, the methods used, and the existing gaps in the literature, thus providing a more comprehensive overview of the research done to date.

5 Conclusion

This review found clear evidence that poor mental health, typically measured as depression and/or anxiety, was associated with lost productivity, i.e., increased absenteeism and presenteeism. Most studies used survey and administrative data and regression analysis. Few studies employed longitudinal data, and most studies that used cross-sectional data did not account for endogeneity. Despite consistent findings across studies, more high-quality studies are needed on this topic, namely those that account for endogeneity and unobserved heterogeneity. Furthermore, more work is needed to understand the extent to which mental illness decreases productivity at work and the mechanisms through which this occurs, as well as a better understanding of the dynamics of the relationship between mental illness and worker productivity to understand the trade-offs between presenteeism and absenteeism. For example, future research should seek to understand how working conditions and work arrangements as well as workplace policies (e.g., vacation time and leaves of absence) impact presenteeism.