Effects of organisational-level interventions at work on employees’ health: a systematic review

Based on the occupational health principle of “hierarchy of controls” [1] it was proposed that interventions addressing the level of work organisation or work environment may produce more sustainable effects on the health of employees than interventions focusing mainly on individual-level characteristics [2]. This argument is also implicit in influential policy statements and initiatives such as WHO’s Global Strategy for Occupational Health for All (1994) [3], the UK Management Standards [4], or the European Directive 89/391 – OSH, among others [5]. Thus, workplace interventions improving the working conditions are expected to result in substantial reductions of work-related ill health. Such interventions modify specific circumstances under which work is performed [6]; for instance, exposure to physical and chemical agents, working time and intensity, type of employment contract, psychosocial factors at work, work-life balance, and health and safety policies within the organization. Nevertheless, in contrast to rather consistent findings of several meta-analyses of individual-level interventions reporting statistically significant effects on selected health outcomes (e.g. [7–10]), the results of organisational-level interventions have yielded inconsistent findings so far [11–14]. Several systematic reviews have examined the effects of organisational-level interventions on specific health-related outcomes such as work-life balance [15], mental health [16], general health and well-being [17], job stress [2], injury prevention [18], and psychosocial and health effects [19]. In these reviews, a lack of consistency of intervention effects was demonstrated and was discussed in methodological and practical terms. For instance, difficulties of engaging employers, but also employees in intervention activities, interference of the intervention with organisational changes and personnel turnover [20], or inability to adjust for a variety of confounding factors were mentioned [21, 22]. Moreover, as employees and (line) managers play an active role in determining how and why interventions work [23] there are potential process-related influences which can hardly be controlled [24]. Understanding how employees appraise the intervention itself, how they are involved in the planning, implementation, and process of interventions, and how line managers’ behaviours may drive or hinder the intervention process are important factors to be taken into account when designing successful worksite intervention studies [23, 25, 26].

In methodological terms, the paucity and difficulty of implementing organisational-level interventions with highest degree of evidence, i.e. randomized controlled trials (RCT), is one such limiting factor. Even though some authors claim that RCTs are impractical and inadequate in organisational interventions - arguing that they do not take into account psychological and social processes having a causal relationship with the intervention outcomes [24] -, it should be mentioned that interventions as such can be regarded as the distal causes of organisational change and, thus, should be subject to systematic inquiry in terms of RCTs. Another methodological concern of available intervention studies relates to the high degree of heterogeneity of intervention designs, study protocols, intervention measures and study populations. This heterogeneity prevents rapid progress of cumulative knowledge, and it creates obstacles against systematically reviewing available results.

In this contribution we set out to tackle this latter difficulty by evaluating all organisational-level intervention studies identified by our search strategy on the basis of a systematic classification. This classification aims to synthesise the major modifications of the working conditions implemented by the interventions. We distinguish work organisation-directed changes from work time-directed changes and from changes of the material substrate of work (see Methods). By improving the comparability of the reported study findings we aim at identifying the conditions that may explain the effectiveness of interventions at organisational level. To this end, we apply a logistic regression model to estimate significant intervention effects.

The search in the databases and the manual search resulted in 18 145 initial records after checking for duplicates. Based on titles and abstracts 699 studies were submitted to further appraisal. Applying the selection criteria mentioned in the Methods Section 77 studies were selected for detailed analysis. Of these, 39 intervention studies meeting fully the criteria were included in the systematic review. A detailed description of main characteristics of the studies and their implementation is reproduced in Additional file 2: Tables S1 and S2.

An analysis of the frequencies of specific outcomes showed that interventions focused mostly on burnout (six studies), absenteeism (four studies), musculoskeletal disorders of the upper body (six studies), and depressive symptoms (three studies). The studies were conducted in the following industrialized countries: Australia, Canada, Italy, Germany, the Scandinavian countries, USA, Japan, UK and The Netherlands.

At the same time, interventions modifying material and organisational conditions of work also report more frequently significant effects (six out of ten studies). These interventions involved, among others, the introduction of mechanical lifting equipment and employee training [48], improvements of machine performance and communication between workers and supervisors [49], reduction of lifting loads and rotation schedules [50], increased compliance with the use of protective substances and establishment of improved control procedures of occupational risks [51], improvement of technical equipment and health surveillance [52], as well as substituting chemicals and increased safety guidelines in the organisation [53]. The health-related outcomes of these interventions included back pain, injuries, sick leave, blood pressure, and eczema.

If the results of each intervention study are compared according to their level of evidence, three out of four high-evidence studies did not report significant intervention effects [37, 54, 55], whereas 14 out of 25 medium evidence level studies reported significant improvements of health-related outcomes, such as ischemic heart disease risk [56], burnout [45, 46, 57, 58], lost time injury [47], perceived health [59], blood pressure [60], decreased mental distress and better sleep [61], reduction of sick-leave length [49], back-pain related lost working days [50], eczema incidence [53], and mental health [62]. Finally, four out of ten low evidence studies reported statistically significant changes in outcomes such as self-rated health [63] and injury rates [48, 52, 64].

In this systematic review 39 organisational-level workplace interventions were analysed with regard to their effects on employee health. We tackled the large heterogeneity of study designs, intervention targets and health outcomes by classifying the studies according to three general categories of intervention targets (material, organisational, work time-related). About half of the studies yielded statistically significant intervention effects. Favourable health outcomes were reported for self-rated mental and general health, and for reduction of injury rates. Taking the potential impact of study quality and occupational class into account, our main finding revealed that more comprehensive interventions addressing several organisational-level targets simultaneously had a higher chance of reporting significant health improvements than those restricted to one intervention target. Before discussing limitations and strengths of this study the following questions are addressed. First, how can we explain the relatively low success rate of these interventions? And second, how can we interpret the finding that more comprehensive interventions are more likely to result in positive health outcomes among employees?

The success rate of the 39 interventions reviewed in this study is substantially lower than the one reported in an important earlier systematic review by LaMontagne et al. [2] where about two thirds of the 18 intervention studies demonstrated a significant effect. It seems particularly worrying that three of the four studies with highest level of evidence in our review failed to achieve a positive outcome [37, 54, 55]). Therefore, it is instructive to learn how the authors themselves explained the limited success of their interventions. In a synthesis of their arguments the following reasons can be distinguished. First, in some cases there was not sufficient participation of employees in preparing the intervention, or there was a lack of communication and motivation to support the intervention and to comply with organisational changes [65–68]. Second, the implementation of some interventions was difficult because some aspects did not work as anticipated or could not be realised as originally designed [39, 55].Third, there were difficulties in developing or maintaining the intervention due to lack of support from employers and managers [62, 66, 69]. Fourth, the intervention was threatened by external events or conditions that were beyond the control of the responsible team (e.g. job turnover, organisational restructuring or merging) [52, 53, 70]. Finally, the lack of significant effects was eventually due to a short follow-up time, to a weak treatment effect, or to a failure of controlling for potential confounders of intervention effects [21, 58, 65, 68, 69, 71]. These reasons clearly point to the need of developing interventions in a participatory approach, where “organizational interventions (are) to be understood as a collective of initiatives and change activities, competing and intertwining with a multitude of concurrent events” [72]. Moreover, careful process evaluation should be a mandatory part of any intervention trial.

How can we interpret the finding that more comprehensive interventions are more likely to result in positive health outcomes among employees? It is of interest to note that individual-level stress management interventions demonstrated exactly the opposite trend. Effect sizes were strongest if the treatment focused on cognitive behavioural therapy or on an intervention that strengthened employees’ resources of coping with stressful work, whereas the effects of multimodal interventions were much weaker [9]. In case of organisational-level interventions, the situation is different. Given the many difficulties of implementation, including the interference of a structural change with established organisational procedures, it may be crucial to develop a comprehensive set of modifications that can exert sufficient impact on employees who are exposed to a complex work environment. This is, for instance, the case if material improvements of work tasks are linked with changes in the division of work or other features of work organisation, and/or with increased flexibility of work schedules. Thus, the combination of a hierarchy of controls approach with specific improvements of technical, chemical or physical elements of the work environment may contribute to a reduction of risk exposure and related adverse health effects.

Despite the heterogeneity of the 39 organisational-level workplace interventions underlying this review, we were able to compare their effects by applying broad classification categories. About half of the studies reveal significant effects on employees’ health. Success rates were higher among more comprehensive interventions tackling material, organisational and work-time related conditions simultaneously. To increase the number of successful organisational-level interventions in the future commonly reported obstacles against the implementation process should be addressed in developing these studies.