A Systematic Review of Occupational Health and Safety Business Cases

Abstract

Business cases are arguments developed to secure management commitment and approval for investment in an intervention. This systematic review evaluated 12 experimental and quasi-experimental studies on occupational health and safety interventions (OHSI) in various settings. The search engines used in this systematic review include PubMed, CINAHL, and Scopus. A cost and benefit analysis of OHSI was completed at the organizational level in these studies. The focus of this analysis included sample, design, theoretical framework, interventional strategies, and threats to validity and outcomes. Positive returns on investment of OHSI outcomes were shown in 10 of the studies. The other two studies concluded that their chosen OHSI were not cost-effective.

According to the U.S. Bureau of Labor Statistics (BLS), nearly 3 million nonfatal workplace injuries and illnesses were reported by private industry employers in 2014. This number of injuries and illnesses reflects a rate of 3.2 cases per 100 full-time workers. Also, the U.S. BLS (2015) has preliminarily reported a total of 4,679 fatal work injuries in 2014. Liberty Mutual’s Workplace Safety Index of 2016 concluded that disabling and nonfatal workplace injuries resulted in nearly US$62 billion of workers’ compensation costs. These costs translate into more than US$1 billion a week spent by businesses on workplace injuries (Liberty Mutual Workplace Safety Index, 2016). On the contrary, a study by Goldman Sachs JBWere Investment Research (2007) revealed that companies adequately managing workplace safety and health performed better financially than those that did not.

A previous systematic review showed a direct positive correlation between investments in safety, health, and environmental performance and subsequent returns on investments (ROIs) from multiple industries and countries, including metal in Finland, aviation in the Netherlands, wood processing in the United States, pet food production in the United Kingdom, automotive industries in Canada, and electronics in Malaysia (Verbeek, Pulliainen, & Kankaanpaa, 2009).

Promoting an effective occupational health and safety intervention (OHSI) system can significantly reduce injuries and illnesses and lower associated costs, including workers’ compensation payments, health care expenses, and lost productivity (Goldman Sachs JBWere Investment Research, 2007; Noben et al., 2015). Existing systematic reviews have examined the effectiveness of OHSI (Verbeek et al., 2009); however, this review critiques the study designs and examines threats to validity of each study. The purpose of this systematic review was to (a) analyze and critique the experimental and quasi-experimental studies using OHSI and (b) examine the efficacy of OHSI. Findings from this review may provide business cases which could be used to secure management commitment and approval of investments in OHSI.

Method

For this integrated review, search engines included PubMed, CINAHL, and Scopus. Search terms include economics, occupational health and safety, business case, cost–benefit analysis (CBA), ROI, and productivity. Inclusion criteria included studies of employed and actively working populations, intervention outcomes based on measurement of costs and benefits in monetary terms, intervention study design (experimental or quasi-experimental studies), and published in English between 2011 and 2016. Exclusion criteria included populations below 18 years of age and OHSI studies that only demonstrated intervention outcomes, not economic outcomes. The search yielded 269 references in PubMed, 99 in CINAHL, and 321 in Scopus. After removal of duplicate references, 328 articles remained. Studies were then screened by title and abstract, and another 282 references were excluded. Finally, after applying inclusion criteria, 12 articles were retained (Figure 1).

Figure 1.

Search flow diagram.

Results

Samples from all studies included employed and actively working individuals. Study samples were selected from various business settings, including a bank, research institutes, government organizations, medical centers, shipbuilding, forestry, universities, information technology firms, and shoe manufacturers (Table 1). Samples were included from six countries (i.e., the United States, the United Kingdom, the Netherlands, Brazil, Finland, and Bangladesh; Table 1). Sample sizes varied among the studies, ranging from 55 (Allen, Lewis, & Tagliaferro, 2012) to 7,936 (Nyman, Abraham, Jeffery, & Barleen, 2012) and are described in Table 3. Only one study (Rantonen et al., 2016) reported power calculations: 85% power using two-tailed tests with the alpha of 0.05 for a sample of 73 employees in each group. Characteristics of the samples were well described among the studies, but only two studies (Barbosa et al., 2015; Rantonen et al., 2016) reported clear diagrams of sample formation and inclusion and exclusion criteria. Noben et al.’s (2014) study provided simple narrative inclusion and exclusion criteria. Van Dongen et al. (2016) only described exclusion criteria, and Van Dongen et al. (2013) study only provided inclusion criteria.

Table 1.

Summary of Reference, Country, Industry, Intervention, Comparison, and Intervention Outcome

Reference	Country/industry	Intervention	Comparison	Intervention outcome
Allen, Lewis, and Tagliaferro (2012)	United States/university	Health risk screening plus lifestyle education	Health risk screening plus minimal information	Improved low-density lipoproteins and coronary heart disease risk
Barbosa et al. (2015)	United States/information technology firm	Network intervention named STAR (support, transform, achieve, results)	Usual practice	Reduction in self-reported presenteeism, reductions in employee health care costs, and lower turnover
Dallat, Hunter, Tully, Cairns, and Kee (2013)	United Kingdom/government office	Financial incentive intervention	No incentive	Employees in the intervention group improved their level of productivity over 6 months more than the employees in the control group
Geraedts et al. (2015)	Netherlands/two banks, two research institutes, one security company, and one university	Happy@Work intervention. Brief web-based intervention delivered with minimal guidance	Care as usual; only received an email with the randomization outcome	Self-help intervention for employees with depressive symptoms was effective in reducing depressive symptoms compared with care-as-usual group
Guimarães, Ribeiro, and Renner (2012)	Brazil/shoe manufacturer	Sociotechnical intervention (macro-ergonomic changes)	Traditional model	Reduction in human resource costs (80% reduction in industrial accidents, 100% reduction in work-related musculoskeletal disorders. 45.65% reduction in absenteeism)
Noben et al. (2015)	Netherlands/health care	Referral to the occupational physician for screen-positive nurses in need of intervention	Screening without feedback and without referral to the occupational physician	Increased productivity due to less absenteeism and presenteeism
Noben et al. (2014)	Netherlands/health care	Screening and feedback followed by referral to occupational physician	Screening without feedback and without referral	Significant improvement in work functioning occurred
Abraham, Jeffery, and Barleen (2012)	United States/university	Health promotion program (disease management and lifestyle management)	NA	Disease management program was effective in reducing health care expenditures, resulting in a larger average annual savings
Rantonen et al. (2016)	Finland/forestry company	Personal copy of booklet that contained information related to low back pain. Booklet reviewed face to face by occupational health nurse	Not invited to study visits	Booklet intervention with additional face to face information resulted in lower health care costs, but reduction in sickness absence was minimal. Physical impairment decreased with both interventions
Thiede and Thiede (2015)	Bangladesh/shipbuilding industry	Comprehensive personal protective equipment, occupational health and safety training	NA	Worker injury rate dropped from over 500 injuries per month. Injury figures remained low. Improvement in productivity, precision, and efficiency of the workforce
Van Dongen et al. (2013)	Netherlands/health care	Worksite vitality intervention Vital@Work: yoga, coaching, aerobic exercise, and fruit	Usual care	Intervention group workers increased their general vitality by 2.5 (range, 0 to 100) points and their work-related vitality by 0.12 (range, 0 to 6) points
Van Dongen et al. (2016)	Netherlands/academic research institutes	Mindfulness-based worksite intervention	Usual care	Statistically significant difference in work engagement was found for the control group

Designs

The OHSI studies for this systematic review used either an experimental or quasi-experimental design (Table 2). Seven studies used a randomized control trial design. Barbosa et al. (2015) used adaptive randomization to ensure balance on job function, leadership, and size of the study group. The next three studies clearly described randomization and allocation methods. Van Dongen et al. (2013) used random allocation software, and the operator for this randomization had no information about the samples to ensure concealment of treatment allocation. Noben et al. (2014) employed cluster randomization at the hospital ward level to prevent contamination between participants working in the same ward. Noben et al.’s study also used a prerandomization procedure with an “incomplete-double-consent” design, in which participants were only informed about their own group to further minimize the risk of contamination. Rantonen et al.’s (2016) study chose a randomization scheme for a trial that was prepared by an independent biostatistician before the study began. Rantonen et al.’s study used a computer-generated randomization table with a block size of eight.

Table 2.

Summary of Study Population, Study Design, Economic Evaluation Method, and Economic Outcome

Reference	Study population	Study design	Economic evaluation	Economic outcome
Allen, Lewis, and Tagliaferro (2012)	Employees of the University of New Hampshire. Analytic sample consisted of 55 individuals Intervention: 26 Control: 29	Quasi-experimental design Two-group pre–post intervention study	CER of the intervention	Cost-effectiveness of the intervention was US$10.17 per percentage-point reduction of low-density lipoprotein cholesterol and US$454.23 per point reduction in coronary heart disease risk
Barbosa et al. (2015)	Employees of an information technology firm. Analytic sample consists of 946 individuals Intervention: 473 Control: 473	Group-randomized multisite controlled experimental study	ROI calculated	For each participant, model-adjusted costs of the intervention were US$690 and company savings were US$1,850. The ROI was 1.68.
Dallat, Hunter, Tully, Cairns, and Kee (2013)	Employees working at main government offices. Analytic sample consists of 406 workers. Intervention: 199 Control: 207	Researcher-blinded quasi-experimental trial	ICER of the incentive group compared with the no-incentive group calculated	The ICERs were EU£2,900/QALY and EU£2,700 per percentage increase in overall employee productivity.
Geraedts et al. (2015)	Employees from six companies including two banks, two research institutes, one security company, and one university. Analytic sample consists of 231 workers Intervention: 116 Control: 115	Randomized control trial	CEA, CUA, and ROI were calculated	Societal perspective: CEA and CUA For depressive symptoms, an ICER of 314 was found.
Guimarães, Ribeiro, and Renner (2012)	Employees from shoe manufacturer in Brazil Analytic sample consists of 1,800 workers Intervention: 100 Control: 1,700	Longitudinal study	CEA and CBA were calculated	The total intervention cost of US$70,132, over a 42-month period, was compensated by annual savings calculated as US$503,479. The net value of the ergonomics intervention was US$433,347. The benefit-to-cost ratio being 7.2.
Noben et al. (2015)	Employees from an academic medical center in the Netherlands. Analytic sample consists of 413 workers Intervention: 207 Control: 206	A pragmatic-cluster randomized controlled trial	CBA conducted from the employer’s perspective	Net-savings of EU£244 per nurse when only absenteeism was considered, and EU£651 when presenteeism was also taken into account. This corresponds to a return-on-investment of EU£5 up to EU£11 for every euro invested.
Noben et al. (2014)	Employees working in one Dutch academic medical center Analytic sample consists of 617 workers Intervention 1: 207 Intervention 2: 204 Control: 206	Pragmatic-cluster randomized controlled trial	CBA was conducted from the societal perspective	The median ICER for the occupational physician condition versus the control condition was dominant, suggesting cost savings of EU£5049 per treatment responder. The ICER for the e-mental health condition versus the control condition was estimated at EU£4054 per treatment responder.
Abraham, Jeffery, and Barleen (2012)	Employees from the University of Minnesota Analytic sample consists of 7,936 workers The year 2006: 1,825 The year 2007: 2,649 The year 2008: 3,462	A difference-in-difference equations approach	ROI calculated	The study found that a 1.76 ROI occurred in the third year of operation that was generated solely by the effect of disease management program participation in reducing health care expenditures.
Rantonen et al. (2016)	Employees from forestry company Analytic sample consists of 232 workers Intervention 1: 92 Intervention 2: 89 Control: 51	Randomized controlled trial	CEA was completed	Booklet reduced health care costs by EU£196 per employee and SA by 3.5 days per year. In the combined intervention group, the figures were EU£107, 0.4 days.
Thiede and Thiede (2015)	Employees from Bangladeshi shipbuilding industry Total number of analytic sample size not provided	Quantifying past costs and benefits and discounting future ones	CBA conducted and ROI calculated	The complete ROI for the OHS measures and certification was negative over a 5-year period. Total costs were US$407,000 for the period including the original certification and OHS measure implementation.
Van Dongen et al. (2013)	Employees from two Dutch academic hospitals Analytic sample consists of 730 workers Intervention: 367 Control: 363	Randomized controlled trial	CBA was conducted and ROI calculated	The Vital@Work intervention was neither cost-effective from the societal perspective nor cost saving from the employer perspective.
Van Dongen et al. (2016)	Employees of two Dutch governmental research institutes Analytic sample consists of 257 workers Intervention: 129 Control: 128	Randomized controlled trial	CBA and ROI calculated	The intervention was neither cost-effective from the societal perspective nor from the employer perspective. Also, the intervention did not produce a positive financial return to the employer.

Note. CER = cost-effective ratio; ROI = return on investment; ICER = incremental cost-effectiveness ratio; QALY = quality-adjusted life years; CEA = cost-effectiveness analysis; CUA = cost-utility analyses; CBA = cost–benefit analysis; SA = sickness of absence; OHS = occupational health and safety.

Five studies were quasi-experimental designs. Dallat, Hunter, Tully, Cairns, and Kee’s (2013) study was a researcher-blinded quasi-experimental trial. Both the intervention group and control group could use electrical cards to self-monitor their physical activity levels at work. However, the intervention group received points that could be redeemed for rewards.

Theoretical Frameworks

The theoretical frameworks used for each study were not described in any of the articles included in this systematic review. Three main keywords were used throughout the included articles reviewed, “cost-effectiveness,” “costs and benefits,” and “return on investment.” Because the purpose of the business case is an economic evaluation of OHSI, a positive monetary return is a key research driver. The word “cost-effectiveness” is used in the title of eight included studies.

Economic Evaluation

Four different methods were used for calculating the economic effectiveness of OHSI. A cost-effective ratio (CER) was used in two studies. The CER was calculated by dividing the cost of the intervention by unit of effectiveness. Cost–benefit analysis was used in six studies. The CBA is a systematic process for calculating and comparing benefits and cost of a decision (Table 2). Cost-effectiveness analysis (CEA) was calculated in three studies. The CEA compares the relative costs and outcomes (effects) of different courses of action. Return on investment was calculated in four studies. Return on investment is used to evaluate the efficiency of an investment. A higher ROI means the investment gains compare favorably to investment costs. Seven studies used only one economic evaluation method; the other studies used more than two of the previously listed methods.

Economic Outcome

Monetary economic outcomes are described in Table 2. Van Dongen et al.’s (2013; Van Dongen et al., 2016) two studies concluded their interventions were neither cost-effective nor cost saving. The other studies showed their interventions resulted in positive monetary savings or return. Thiede and Thiede (2015) study’s unique conclusion showed that complete ROI for the OHSI was negative over a 5-year period.

Interventions

Comparison of interventions is described in Table 1. Occupational health and safety interventions in this review study are either health promotion or injury prevention strategies. Interventions for eight studies were health promotion or injury prevention education and training, including lifestyle change training, online-supported cognitive therapy, a disease management program, yoga, exercise, or an educational booklet. Thiede and Thiede (2015) described the scale of an organizational intervention. The study was implemented in the Bangladeshi shipbuilding industry, and the study intervention included comprehensive personal protective equipment, occupational health and safety training, and the establishment of a primary health care center. The level of intervention and size of settings varied in this systematic study.

Intervention Outcome

Although Van Dongen et al.’s (2013; Van Dongen et al., 2016) two studies concluded no monetary savings or ROI resulted from the studied intervention, a positive intervention outcome was identified from both of their studies and the other studies that were evaluated. Measuring effectiveness of their intervention methods varied depending on the intervention. Intervention outcomes (i.e., lowering heart disease risk, depressive symptoms, turnover rate, weight, injury rates, health care costs, and self-reported presenteeism or improving productivity and vitality) are described in Table 1.

Discussion

Threats to Validity

Table 3 describes threats to validity. Mortality was the most frequent threat to internal validity. Nine studies reported “dropout” as a study limitation. Because the purpose of the studies was to examine the efficacy of the OHSI, 6 months to 12 months of monitoring was required to examine economic outcomes. Retaining all participants until the end of the study period was extremely challenging. Researchers could have contacted the participants more frequently than was reported to encourage participants to stay in the study until the end. Asking open-ended questions of participants regarding factors that would motivate them to participate in the study for longer periods of time may have been an effective approach. Selection bias was the second most frequently found internal validity threat. Eight study authors reported selection bias as a limitation of their studies. The convenience sampling method could have resulted in overselecting participants who volunteered, or high-achieving employees. Random sampling could decrease this threat.

Table 3.

Threats to Validity

Reference	Statistical conclusion validity	Construct validity	Internal validity	External validity
Allen, Lewis, and Tagliaferro (2012)	Low statistical power	Compensatory rivalry	Self-selection bias Design contamination History Mortality: 12-month monitoring	Interaction effect of testing
Barbosa et al. (2015)	Unreliability of measures: outcomes from self-reported data Regression to mean	Novelty and disruption effects	Selection bias Homogeneous group	Interaction effect of testing Interaction of the causal relationship with settings: one information technology firm
Dallat, Hunter, Tully, Cairns, and Kee (2013)	Inaccurate effect size estimation	Resentful demoralization Novelty and disruption effects	Selection bias Design contamination: Participants contact each other	Interaction of the causal relationship with settings: one governmental organization
Geraedts et al. (2015)	Low statistical power Unreliable measures: outcomes from self-reported data	NA	Mortality Testing Selection bias Maturation	Interaction effect of testing
Guimarães, Ribeiro, and Renner (2012)	Unreliable measures: unable to access significant data	Reactivity to the experimental situation	Maturation Mortality History	Interaction of the causal relationship with settings: one shoe company in Brazil
Noben et al. (2015)	Unreliable measures: Outcomes were based on self-report	Treatment diffusion: control group may talk to occupational physician (intervention)	Mortality: 3- and 6-month monitoring	Interaction of the causal relationship with settings: one hospital
Noben et al. (2014)	Unreliable treatment implementation Unreliable measures: Outcomes were based on self-report	Treatment diffusion: may share information among participants	Mortality: 6-month monitoring	Interaction of the causal relationship with settings: one hospital
Abraham, Jeffery, and Barleen (2012)	“Fishing” and error rate problem: A large number of regressions were run	Treatment diffusion	Selection bias Mortality	Interaction of the causal relationship with settings: one university
Rantonen et al. (2016)	NA	Treatment diffusion	Selection bias Mortality	Interaction of the causal relationship with settings: one company
Thiede and Thiede (2015)	NA	Treatment diffusion	History Maturation	Interaction of the causal relationship with settings: one company
Van Dongen et al. (2013)	Unreliable measures: Outcomes were based on self-reported retrospective questionnaires. Health insurance data inaccessible	NA	Selection bias Mortality: 6- and 12-month monitoring	Interaction of the causal relationship with settings: two Dutch academic hospitals
Van Dongen et al. (2016)	NA	NA	Selection bias Mortality: 12-month randomized controlled trial History	Interaction of the causal relationship with settings: two Dutch academic hospitals

Interaction of the causal relationship with settings was noted in 10 studies as a threat to external validity. Most of the studies collected data from one or two settings, the result of both financial and situational factors. Results from the studies may be transferable to similar settings but not every business setting. Selecting multiple settings would increase the generalizability of study findings.

Treatment diffusion was found to be a construct validity threat in five studies. When data are collected in only one or two settings, participants may share information about interventions or treatments. Compensatory rivalry and demoralization might have also been noted in Allen et al.’s (2012) and Dallat et al.’s (2013) studies. Selecting study populations from a variety of units and settings and using blind or double-blind study designs may reduce this threat.

Limited measurement reliability was noted in six studies, raising questions concerning statistical conclusions and validity. Certain outcomes, and participant’s perspectives on the efficacy of the interventions, were measured based on self-report. Based on the results, it is difficult to conclude that participant reports were accurate and outcome measures were reliable. Use of tools that measure objective outcomes may reduce this threat.

Strengths

The strength of this literature review is that business cases were searched systematically using Internet search engines. This literature review emphasized CBA to secure decision makers’ commitment and approval to implement OHSI. As shown above, all but two studies resulted in improved employee health and monetary benefits to employers. The cases provide meaningful information on how OHSI can create a positive balance between cost and benefit.

Limitations

The outcome of the interventions was either total dollars saved annually or ROI. Most studies were longitudinal and the assumption that the benefits would continue at the same level as the first year could be false. It is possible that this review missed case studies reported in the gray literature (Google or Yahoo general search engines). Business cases that demonstrated unfavorable results are less likely to be published. Consequently, published studies included in this review may bias interpretation of the evidence regarding OHSI results.

Implications for Practice and Research

The following are areas for improvement. Intangible benefits such as reputation damage and risk of disobeying regulations should be assessed. Reliable and valid guidelines for reporting business cases are needed. The value of the cases can be underestimated when structures and consensus among business administrators are missing. None of the studies from this review were conducted based on a theoretical framework or model. Theory-based studies provide stronger evidence of a positive correlation between OHSI and economic effectiveness. Selecting study populations from a variety of geographic regions and business settings, with worker racial and socioeconomic diversity, could result in a more representative sample.

Conclusion

This integrative review suggests the economic benefits of OHSI. Administrators will use OHSI based on data, evaluation of health and economic impact, and their assessment of OHSI importance to the company. The foundation of policies and regulations related to occupational safety and health must be economically and ethically sound. However, because economic value is a critical factor that drives administrator decision making, conducting well-designed business case studies should be expanded.