There is interest in how work organizations can contribute to improving employee well-being [1
]. Models of employer–employee relations predict that good social environments in the workplace are associated with worker well-being. These models relate to perceived organizational support (POS) [4
], organizational climate (OC) [5
], social identity (SI) [6
], and organizational justice (OJ) [7
]. Theoretical perspectives suggest that POS, OC, SI, and OJ may be closely inter-related [4
]. There is evidence that employers can accrue benefits from good social relations between employer and employees [9
], and good workplace social environments enable easier implementation of complex organizational change [10
]. Therefore, it may be the case that interventions to improve POS, OC, SI, and OJ improve employee well-being, and are an important tool for management practice.
However, there is no current synthesis of interventions aimed at improving workplace social environments, representing a significant gap in the literature. Moreover, interventions to improve POS, OC, SI, and OJ may prove complex, and some interventions may produce adverse effects [12
]. Therefore, in the present study, we use a systematic review methodology to examine whether interventions that seek to improve various aspects of social environments in organizations, including POS, OC, SI, and OJ, also improve well-being. Because employers may benefit from improved social environments, we also examine effects on performance.
In the present review, we focus on the interventions to improve psychological well-being. Psychological well-being is a core component of well-being [13
], and has the following major components [14
]: (a) subjective assessments of satisfaction, which in the work context can include job satisfaction; (b) hedonic experience, such as a positive affect (e.g., joy, enthusiasm) and the relative absence of a negative affect (e.g., lack of anxiety, feeling calm); and, (c) eudemonic well-being, which includes feelings of autonomy, mastery, personal growth, positive relations with others, purpose in life, and self-acceptance [16
]. Our definition of performance is broad, so that we can encompass multiple levels of analysis (e.g., work group or individual), and multiple indicators of performance and performance-relevant outcomes (e.g., productivity, absence, intent to quit).
Before proceeding with our systematic review, we begin by summarizing existing evidence on the four theoretical perspectives that provided the basis for our expectations that interventions to improve workplace social environments will also improve psychological well-being.
1.1. Perceived Organizational Support
According to Rhoades and Eisenburger [4
], POS refers to workers’ global beliefs that the organization values workers’ contributions, and cares for their well-being. Rhoades and Eisenburger hypothesize that POS relates to well-being and work performance because: POS influences workers’ rewards expectations; POS fulfils workers’ socio-emotional needs related to social identity (see below), and a perception that help is available from the organization if required; and, through social exchange processes, POS encourages worker reciprocity through enhanced performance and more favorable job attitudes.
Three meta-analyses have examined the associations between POS and well-being [4
], and two papers are based on the same review [18
]. The meta-analyses found moderate to large meta-correlations between POS and: indicators of well-being (0.28 to 0.61, indicators included affective experience, strain, and job satisfaction); small but statistically reliable to moderate meta-correlations, with indicators of in-role and extra-role performance (0.16 to 0.42); and, small but statistically reliable to moderate meta-correlations with indicators of withdrawal behaviors, including turnover intentions, turnover, and absence (0.11 to 0.49).
1.2. Organizational Climate
One meta-analysis [5
] and two systematic reviews focused on healthcare contexts [20
] have examined OC and wellbeing. Gershon et al. [21
] defined OC in relation to group and management support. They found only four studies in their review that pertained to well-being (specifically burnout), and these studies were inconclusive. Bronkhorst et al. [20
] defined OC as perceptions of the social and interpersonal aspects of work, and included communication, participation, group and leader relations in the global concept of OC. They concluded that OC was related to better mental health, especially for those aspects of OC linked to leader and group relations. Benzer and Horner [5
] distinguished between task and relational climates. Task climate refers to goal setting, organizational innovation, and responsiveness. Relational climate refers to work group relations and social recognition and acknowledgement, and therefore relates most closely to social environments in the workplace. Benzer and Horner found that relational climate had statistically reliable and moderate to large meta-correlations with job satisfaction (0.47), perceived stress (−0.25), managers’ ratings of worker performance (0.17), and workers’ ratings of their performance (0.42).
1.3. Social Identity
In the workplace, SI refers to the extent to which individuals internalize their membership of a work group or organization into their self-concept [22
]. Therefore, SI can reflect a sense of belonging to an organizational community, and embeddedness in the workplace [22
]. As a component of SI, organizational identity reflects the extent to which a worker ascribes organizational attributes to their self-concept [6
]. High levels of SI may improve well-being through promoting a sense of belonging, providing a sense of purpose, providing a sense of control, providing a sense of self-affirmation, and/or providing a sense that help is available if needed [22
Two meta-analyses have examined SI at work and well-being [6
]. Lee et al. [6
] found that organizational identification had moderate to strong associations with: job satisfaction (0.45), self-rated in-role performance (0.27), self-rated extra-role performance (0.42), others’ ratings of worker performance (0.19), and others’ ratings of worker extra-role performance (0.29). Steffens et al. [22
] found that organizational and work group identification had moderate associations with health (both 0.21), with a stronger association for psychological health (0.23) than physical health (0.16), although the association with physical health is still statistically reliable. Steffens et al. also found that positive indicators of well-being had a stronger association with identity (0.27) than negative indicators such as stress (−0.18).
1.4. Organizational Justice
In general terms, OJ refers to the fairness in how workers are treated. It has four major components: distributive justice, which refers to the perceived fairness of outcomes such as pay; procedural justice, which refers to the perceived fairness of how rewards are allocated; interpersonal justice, which refers to the perceptions of politeness and respect in how reward decisions are communicated; and informational justice, which refers to the accuracy and truthfulness of how reward decisions are communicated [7
]. The four aspects of OJ are strongly inter-related [7
The associations between OJ and well-being have been subject to six meta-analyses [7
], and a systematic review of prospective studies [28
]. Across the meta-analyses, there tended to be a small but usually statistically reliable to strong meta-correlations with indicators of well-being (0.10 to 0.49), including indicators of affective experience, job satisfaction, burnout, and health. There were also small, but usually statistically reliable to strong meta-correlations between indicators of job performance (0.10 to 0.69). Ndjaboué et al.’s systematic review [28
] of prospective studies found that OJ is consistently related to indicators of mental health and well-being.
1.5. Review Questions
Previous reviews have indicated that POS, OC, SI, and OJ are reliably associated with well-being. However, most studies included in the reviews have been cross-sectional and there is little or no evidence from intervention studies in the reviews cited. Therefore, in the present review, we focus on reviewing evidence on interventions that seek to improve the various aspects of social environments in organizations, including POS, OC, SI, and OJ. Our specific questions are:
Do interventions that seek to improve social environments in organizations promote well-being?
Do interventions that seek to improve social environments in organizations improve performance?
In answering the first question, we will also seek to determine whether the interventions work as hypothesized, namely we will examine whether the indicators of social environments in organizations change as a result of the intervention, as well as indicators of well-being. The second question pertains to exploring whether there are gains for employers for implementing interventions to improve the social environments in organizations. Although we did not specifically search for studies with performance as an outcome, where studies reported on performance as well as well-being outcomes, we extracted performance data too. Consistent with recommended practice for systematic reviews of complex well-being interventions, we also examined the interventions for contextual factors that may have affected the implementation of the interventions, or acted to accentuate or attenuate the success of interventions [29
Prior to the review, we developed a review protocol outlining the process for the review. The review team sought input from experienced researchers working in related fields to the review for their advice on relevant search terms to include in the protocol. The protocol was designed according to the best practice PRISMA-P reporting guidelines [30
], and registered on PROSPERO, The International Prospective Register for Systematic Reviews.
To determine inclusion/exclusion criteria for studies, we were guided by the PICOS approach (population, intervention, comparators, outcomes, and study design) (see Appendix A
Population. We included studies that focused on well-being in the working population in advanced industrial democracies (e.g., EU-15 countries, USA, Australia, Japan). Studies in countries where economic conditions (and therefore work conditions and organizational context) differ markedly from the advanced industrial democracies were excluded. The decision to focus on the advanced industrial democracies was based on institutional factors that may influence well-being in work, including but not limited to: Greater levels of employment protection through legislation; employees’ expectations of their work environment; expectations regarding corporate social responsibility; health and safety legislation; and, widespread and professionalized expertise in occupational health, work psychology, human resource management and other related disciplines in universities and consultancies. Although this bounds the scope of the present review, it does allow for synthesis and practical application of evidence from more homogenous institutional contexts than would be the case if research from other contexts had been included in the review.
Intervention. We focused on interventions to change the social environment in work organizations, such as interventions to change POS, OC, SI, or OJ.
Comparison. Ideally, we wanted to compare a group or groups who had been subject to a change or intervention in the workplace with a control group who had not. We also included studies where the only comparator was the level of well-being before the intervention.
Outcomes. Change in well-being. Studies that investigated performance alongside well-being were also included.
Study Designs. Qualitative or quantitative studies that included a longitudinal element were used.
Other. Peer reviewed empirical research published in an English/non-English language peer reviewed journal that met the other inclusion criteria as specified above. Articles not containing empirical research were excluded.
The search terms were developed on the basis of the research questions, consultation with subject matter experts, and the inclusion/exclusion criteria detailed above. The search terms are shown in the Appendix
. We did not apply restrictions on date, language, or publication type in the searches, and all citation data were downloaded using reference management software (EndNote X7.4) (Clarivate Analytics, 1500 Spring Garden, Philadelphia, PA, USA). The electronic searches were performed up to 27 September 2016, and targeted the following databases: EconLit, PsycINFO, PubMed Central (PMC), Web of Science, Business Source Complete, and Academic Search Complete.
2.2. Study Selection
At all stages in the selection of studies, we erred on the side of caution and included studies for the next stage of sifting if there was any doubt whether the study did or did not meet the inclusion criteria. At each stage of sifting, the included papers were sifted independently by two review authors. Any disagreements were recorded, and these were resolved by discussion between the members of the review team. Figure 1
summarizes the process of study selection.
Our first search identified 1396 titles as ‘hits’ to be sifted according to the inclusion criteria. This was performed independently by the three review authors who then met to check reliability. Cohen’s Kappa rating indicated a moderate level of agreement between the reviewers (range 0.59 to 0.65). We then sifted the abstracts according to the inclusion criteria. This was preceded by a pilot sift of 50 abstracts (chosen at random), conducted by all members of the review team to ensure consistency. The abstracts were then sifted independently by the review authors. All disagreements were recorded and these were resolved by discussion. Cohen’s Kappa scores indicated good levels of agreement between the reviewers (range 0.73 to 0.84). All studies that made it through the abstract sift were then assessed as full papers to ascertain whether they did meet the inclusion criteria. Cohen’s Kappa scores indicated moderately good levels of agreement between reviewers at this stage (Kappa was 0.66 in all instances). Out of the original search results, 12 papers made it through to the data extraction phase of the review, which was carried out by all of the review authors. Following the data extraction, four studies were removed from the review because they did not meet inclusion criteria. The decision to exclude these papers was made collectively by the review team following discussion, and after two reviewers had read each discarded paper.
2.3. Data Extraction
Data extraction sheets captured detailed information about the nature of the intervention and outcomes. Prior to the data extraction and to ensure consistency, all authors extracted data from one paper as a test case. The papers were then divided between the reviewers for coding. The first author coded all of the papers. To ensure the consistency of coding, all papers were double coded by one of the other authors. Second coders’ comments on papers were incorporated into the first coders’ comments on the data extraction sheets.
Once the data were extracted, the first author synthesized the data extraction sheets into an evidence summary table, which categorized studies into type of intervention, and provided the basis for a narrative summary of evidence around each type of intervention, as well as a summary evidence statement for each type of intervention. Consistent with recommended practice for systematic reviews of complex well-being interventions [29
], we then assigned quality gradings for the evidence base underpinning each evidence statement, rather than grading the quality of the individual studies. The final quality grading for evidence was based on recommendations made for reviews of complex interventions targeted at well-being [29
]. Snape et al. [29
] provide four categories of evidence: “Strong evidence”, in which there is confidence that an intervention has an impact in stated group and context; “Promising evidence”, which suggests an impact may occur but requires further investigation; “Initial evidence”, which requires further investigation and although an effect may occur, there is less confidence than for “promising evidence”; “Evidence not yet strong enough for conclusions”, where there is insufficient evidence to make conclusions. The four categories of evidence are developed from the GRADE approach specified in the Cochrane Centres handbook for quantitative studies [32
], and the CERQual approach for qualitative studies [33
]. Both approaches have been developed to assess the overall quality of evidence underpinning the evidence findings of the reviews, which is informed by the methodological limitations of individual studies. All three authors met to review the draft evidence summary table and the evidence statements to discuss and reach a consensus on the evidence, how it should be interpreted, and the accuracy of the evidence statements.
Where there was sufficient statistical information presented in papers on a homogenous range of outcomes, we calculated a sample size adjusted meta-analytic Cohen’s d to provide a formal statistical test of the effects of interventions. We first calculated Cohen’s d for each study, which is the ratio of change in the mean scores from before to after the intervention, divided by the pooled standard deviation of scores from before and after the intervention. To estimate a standard error for meta-analytic Cohen’s d, we used the standard error of the raw mean d across the five studies.
The evidence that we have reviewed indicates that activities based on increasing the frequency of shared activities between workers can improve worker well-being and performance via improved social environments at work. The evidence suggests that such activities need to be sustained, have some external facilitation, and have different components for different types of shared activities (for example, externally facilitated training workshops alongside group social activities, internal mentoring programs, and action planning groups around specific organizational issues). It also seems that interventions are more likely to be successful if workers have favorable attitudes towards such interventions. For these types of interventions, there is more extensive and more consistent evidence, albeit with weaker study designs than for fairness interventions. Interventions based on shared activities may be relatively cost-effective when compared to interventions that require changes in organizational processes, such as job redesign or purchase of external services such as employee assistance programs. Although some external facilitation may be required, this might be relatively brief and limited to a few workshops [35
]. Other components of the intervention could include social activities [36
], internal mentoring programs [39
], and communities or practice or action planning groups around specific organizational issues [35
]. Fairness interventions have strong theoretical and epidemiological support, but there is no strong support from multiple intervention studies.
One surprising finding of the study is that the literature has so few intervention studies related to social environments in the workplace, yet there is a wealth of other research, summarized in several reviews and meta-analyses and well-articulated theories of POS, OC, SI, and OJ. Most of the intervention studies that do exist have been focused primarily on one indicator of well-being, namely job satisfaction. Therefore, there is a need to increase the number of intervention studies, and use a wider range of outcomes, especially given the potential cost effectiveness of interventions based on shared activities. Given the dominance of interventions for healthcare workers in studies of interventions based on shared activities, the absence of studies in manufacturing in this review, and the focus on research in advanced economies, there is a need for studies in a wider range of contexts. However, interventions based on social activities appear to have benefits for well-being across different national contexts, including individualist contexts (UK, US).
As well as the size of the evidence base on interventions, one limitation of the evidence reviewed here is the heterogeneity of the interventions in the different studies. Given both the current size of the evidence base, and the heterogeneity of interventions, we grouped interventions according to high level generic features, rather than specific grouping together of interventions that were very similar. For interventions based on shared social activities, we were able to conclude that such interventions require several components. However, the heterogeneity of the current evidence base means that it is not possible to make any conclusions concerning the best combination of components, or the order in which those components are introduced. One possible exception is that the evidence suggests that workers require favorable attitudes to the intervention prior to the intervention, and therefore actions to gauge the extent of, and/or to inculcate favorable worker attitudes to the shared activities may be required before any other actions.
Similarly, there was a potential heterogeneity for fairness interventions. It is reasonable to propose that fairness around performance appraisal, which has potential implications for wages, may have more salience than fairness around internet monitoring. Even so, there were null results for fairness interventions, which could reflect at least four things. First, the interventions themselves may have been weak influences on justice perceptions. Second, the studies may have had weak statistical power. Third, fairness interventions may have stronger effects if they are targeted at specific and disadvantaged groups (e.g., disabled, ethnic minorities). Fourth, there may have been a contextual effect. The earliest fairness study was published in 1995, and both fairness studies were based in the US. By 1995, most developed economies would have had clear equality and discrimination laws. It may be the case that fairness interventions are best enacted through legislative frameworks that apply across all organizations in a given nation. In this respect, comparative and longitudinal analysis of policy changes in different countries may be informative on the effects of fairness interventions.
Another limitation on the evidence concerns the quality of the studies that comprise the evidence base on shared social activities. Generally, these studies used weak designs, with four out of six studies using a relatively weak pre-post-test only, with no control group designs. Three out of six studies did not report formal statistical tests for some of the effects of the interventions, in spite of using data collected using quantitative ratings scales. Moreover, the studies of shared social activities were not based on explicit theories of well-being, and competing theories of how the interventions may have worked were not examined. That is, no study examined formally whether the effects of the interventions on well-being were mediated by indicators of POS, OC, SI, OJ, or other possible theories. This places a limitation on what can be concluded, in that it is not possible to make explicit the mechanisms that underpin successful interventions. Notwithstanding, it was possible to make some statistical inference across the studies as a whole, with a meta-Cohen’s d suggesting a statistically reliable small to medium sized effect for interventions based on shared activities. Consistent results across the studies do indicate that interventions based on shared social activities are promising and require further investigation.
Given the limitations on the size and nature of the evidence base for interventions targeted at improved social relations, it is important for future intervention studies to use stronger designs (e.g., suitably powered randomized controlled designs, non-equivalent control group designs) with long-term follow-ups, strong statistical tests, and a more extensive range of well-being outcomes. Intervention studies could also make explicit use of the realist evaluation approach [42
], which focuses researchers on the mechanisms of how interventions work, and whether those mechanisms vary in strength in different contexts, or for different sub-groups exposed to the interventions. Uncovering such mechanisms is theoretically important because intervention studies can provide relatively strong causal inference in ecologically valid settings [43
]. Intervention studies require researchers to operationalize specific actions and changes that should alter a theoretical construct, and realist evaluations of interventions can help denote dependencies or redundancies between models [8
]. For example, in relation to interventions based on shared activities, it is not clear whether the effects on well-being and better social environments derive from enhanced SI, POS, OC, or some combination thereof.