The effectiveness of workplace nutrition and physical activity interventions in improving productivity, work performance and workability: a systematic review

A two-step search strategy has been followed. Firstly, a comprehensive literature search in Medline, Cochrane Library and PROPSERO was conducted, using Medical Subject Headings (MeSH) terms and relevant keywords, in order to identify previous systematic reviews with resembling objective as in this search regarding the impact of workplace nutrition and physical activity interventions on productivity, work performance and workability [7, 9, 18]. The MeSH is a controlled vocabulary for describing various biomedical topics which has been shown to greatly facilitate document retrieval [22]. Studies were identified from the reference lists of systematic reviews which met the inclusion criteria and were included in the review. In addition, Google Scholar search engine was searched.

Since the most recent systematic reviews with resembling objective had captured relevant studies until the year 2014 [7, 9, 18], an update search strategy following PICOS was also developed. The PICOS tool focuses on the Population, Intervention, Comparison, Outcome and Study design of an article. It is a framework designed to make the process of the literature searching more rapid and accurate, limiting the number of irrelevant articles [23]. Two experienced university librarians (CG, SG) developed the search strategy (Additional file 1) based on the provided PICOS. The search was conducted in four electronic databases: Medline, EMBASE.com, Cochrane Library and Scopus and covered year 2015 to September 2016.

The search strategies included studies conducted in High Income Countries (HICs), due to the different Occupational Health and Safety (OHS) context between the developed and the developing countries [14]. Searches were restricted to include studies with healthy adults over 18 years of age working in a full- or part-time capacity, and written in English language. Furthermore, studies which constitute “grey literature”, such as editorials, letters, working papers, reports and reviews were excluded.

Any workplace nutrition and physical activity intervention examining the organizational-, environmental- or multi-level effect on work-related outcomes such as productivity, absenteeism, presenteeism, work performance and workability was included. There was no restriction on the length of the intervention in order to study the short-term as well as the long-term effects of interventions. The research studies were either randomized controlled trials (RCTs) or non-randomized controlled study designs (NRSs) with a predefined control group or comparator group. Schelvis, Hengel [24] strongly suggest researchers who conduct systematic reviews to include studies applying alternative research designs such as NRSs. There were no restrictions on the type of comparator used in the study. The search process was done in accordance with the recommendations of the Cochrane Collaboration [25].

The screening process of abstracts and titles of systematic reviews was conducted by one reviewer. The systematic reviews were included if they met the inclusion criteria. Subsequently a reference list was conducted containing the studies of the selected systematic reviews.

Three reviewers (AG, EA, LK) independently selected studies by screening abstracts and titles of intervention studies, based on the inclusion criteria. Subsequently, three groups of two reviewers (group 1: AG, EA; group 2: AG, LK; group 3: EA, LK) independently determined the eligibility of studies on the basis of a review of the full texts, using a predesigned criteria form (Additional file 1). Any disagreement about selection in included studies were solved through discussion or by involving a third reviewer who did not participate to the group (AG or EA or LK).

In order to evaluate the methodological quality of included studies, the Cochrane Collaboration Risk of Bias Tool (CCRBT) for RCTs [25] and the Risk of Bias in NRSs of interventions tool (ROBINS-I) were used [26, 27]. ROBINS-I is mainly distinct from CCRBT because of the randomization. For the first three domains, randomization, if properly implemented, protects against biases that arise before the start of intervention but not after. Therefore, there is substantial overlap for the last four domains between the two assessment tools [25, 26]. The Cochrane Collaboration is strongly encouraging all reviewers to use these tools to establish consistency and avoid discrepancies in assessment of methodological quality among all review groups.

Methodological quality of each study was independently evaluated by two review authors (AG, EA), using the assessment tools. Disagreements were resolved by discussion until consensus was reached. A senior expert (GB) contributed to the assessment procedure whenever considered to be necessary.

To identify the key elements from existing guidelines and texts, as well as from relevant systematic reviews, we followed the same data extraction procedure as in a previous systematic review [28]. A data extraction form was developed, reviewed and refined by the authors to better capture the key aspects that are essential for evaluation, synthesis and presentation, ensuring the adequacy of the tool. The data extraction form included information on publication (title, authors, year), location, occupation / industrial sector of target population, number of participants, company size and type, type of study design, measurement time period, type and description of the intervention, studies’ outcomes, our primary and secondary outcomes. Three reviewers extracted the data (AG, EA, LK) from equal number of studies. It was not possible to conduct a meta-analysis due to the heterogeneity of study designs, populations, interventions, and outcomes. We provided a narrative synthesis of the findings from the effective included studies, structured around the level of intervention (i.e. organizational-, environmental-, multi-level) and the type of the intervention (nutrition, physical activity, both).

Seventy three systematic reviews were assessed, resulting in a list of 699 RCTs and NRSs. After duplicates were removed (n = 284), a total of 415 citations remained for screening. After the update search, a total of 2209 citations were screened. One hundred and fifty three articles retrieved in full text, and 39 of them fulfilled the inclusion criteria (see Fig. 1).

Nineteen included studies were RCT, while 20 included studies used NRS design, such as quasi-experimental controlled study design and pretest/ posttest controlled study design. Eighteen studies were carried out in the United States, seventeen studies in Europe and four studies in Australia. The studies included working populations from the following sectors: health care provider or insurance, services and administration, manufacturing, communication, education and multiple sectors. Sample sizes ranged from 25 to 155,543 employees (249,175 in total; mean sample size: 6557; median sample size: 407.5). Seven studies were conducted in the 1990s, nine studies were conducted during the following decade, while twenty three of the included studies were published after 2010. The follow-up period ranged from 3 months to 8 years, with upwards of 12 months as the most common duration for 22 studies. The studies reported on physical activity (n = 22), nutrition (n = 1) or both physical activity and nutrition (n = 16) interventions. The studies addressed either environmental-level (n = 4) or multi-level interventions (n = 35). The multi-level interventions were organizational- and individual-level interventions (n = 14), environmental- and individual-level interventions (n = 10) or environmental-, organizational- and individual-level interventions (n = 11). The included intervention studies targeted work-related outcomes such as absenteeism, presenteeism, work performance, workability and productivity and diverse types of health-related outcomes, such as sedentary behavior, physical activity, musculoskeletal comfort, weight loss and health risks. Two environmental-level [29, 30] and three multilevel intervention studies [31‐33] aimed either at physical activity or nutrition measured work performance using subjective and objective measurements. An environmental-level [34] and two multilevel intervention studies [35, 36] aimed at physical activity measured workability using Work Ability Index (WAI). Six multilevel intervention studies [37‐42] aimed either at physical activity or physical activity and nutrition measured productivity using objective and subjective measurements, such as HPQ (Health & Work Performance Questionnaire), WPAIQ (Work Productivity and Activity Impairment Questionnaire), Work Limitations Questionnaire (WLQ). Fourteen multilevel intervention studies aimed either at physical activity or physical activity and nutrition measured absenteeism. Nine of them used objective measurements (company records) [43‐51], while five studies used subjective measurements [52‐56]. An environmental-level [57] and ten multilevel intervention studies [58‐67] aimed either at physical activity or physical activity and nutrition measured multiple work-related outcomes using objective and subjective measurements.

Table 1 presents the characteristics of the included studies (N = 39) and the interventions they studied, while an overall description of the interventions and details of the study designs are presented in Additional file 2. The studies were stratified according to the level of intervention and the type of the intervention. Eight categories were compiled: environmental-level interventions aimed at physical activity (n = 4); organizational- and individual-level interventions aimed at physical activity (n = 7), at nutrition (n = 1) and at both physical activity and nutrition (n = 6); environmental- and individual-level interventions aimed at physical activity (n = 7) and at both physical activity and nutrition (n = 3); environmental-, organizational- and individual-level interventions aimed at physical activity (n = 4) and at both physical activity and nutrition (n = 7).

The overview of risk of bias assessment of the 19 RCTs is summarized in Additional file 3. The majority of the studies were assessed with a low or unclear risk of bias (see Cochrane Collaboration’s recommendations). Studies conducted before 2011 were assessed with an unclear risk of bias [31, 32, 36, 41, 55, 56], while more recent studies were assessed either with low [35, 37, 39, 57, 61, 63, 64] or unclear [30, 42, 58, 62] or high risk of bias [51, 66].

The overview of risk of bias assessment of the 20 NRSs is summarized in Additional file 3. None of the studies discussed deviations from intended interventions. According to Cochrane Collaboration’s recommendations, if there is a lack of information in one or more key domains of bias it is difficult to extract safe conclusion on which to base a judgment about risk of bias, as a result of which these studies are classified as no information. Thus, the studies were judged taking into consideration all the pre-intervention and at-intervention domains except bias due to deviations from intended interventions. The intervention studies were judged as having moderate risk of bias, as the majority of them provide sound evidence for a NRS but cannot be considered comparable to a well-performed randomized trial. Older studies (before 2011) were rated as having either moderate risk of bias [33, 43, 46, 47, 50, 52, 60] or no information [44, 48], while more recent studies were rated as having either low [29, 53, 54, 59] or moderate risk of bias [34, 38, 40, 45] or no information [65, 67].

Fourteen studies (4 RCTs and 10 NRSs) were evaluated as being effective regarding the work-related outcomes, while two more NRSs showed effects only between subgroups. However, only two studies were rated as having low risk of bias [35, 57]. The effective studies were stratified according to the level of intervention and the type of the intervention. Three categories were compiled and presented below: (i) physical activity interventions aimed at changes in the workplace physical environment (n = 3); (ii) nutrition and physical activity interventions aimed at changes at multi-levels of the workplace (organizational- and individual-level: n = 4; environmental-, organizational- and individual-level: n = 6); (iii) physical activity interventions aimed at changes at multi-levels of the workplace (organizational- and individual-level: n = 2; environmental- and individual-level: n = 1). The overview of effective interventions is summarized in Table 2.

This systematic review sought to evaluate the effectiveness of workplace health promotion interventions, aimed at physical activity and/or nutrition, on productivity, absenteeism, work performance and workability. Only multi-level interventions, with components aimed at the workplace organization and/or physical work environment were included. The most frequently evaluated interventions were multi-level initiatives that included components focused on individuals, such as counseling, combined with components focusing on workplace environmental and/or organizational levels. A total of 39 studies, 19 RCTs and 20 NRSs, were included. The majority of studies were conducted in USA and Europe. There was a substantial increase in relevant published studies from 2010 onwards. Nearly 28% of the included studies were high quality, while 56% were of medium quality. The medium quality of the included studies was due to missing information considering the domains of bias.

Most of the studies measured absenteeism rather than productivity. This was perhaps to be expected, given that absenteeism data are easily and objectively assessed using workplace personnel records. Productivity on the other hand is arguably a more complex construct to measure [9]. Presenteeism, going to work while sick, is a significant problem, which accounts for a larger proportion of productivity losses compared to absenteeism and needs to be considered. However, the presenteeism literature is young and heterogeneous, thus, there was only limited evidence of a relationship between physical activity and presenteeism. Nevertheless, existing studies suggest that encouraging employees to be more physically active and reducing sitting can reduce presenteeism [10, 40].

Due to the lack of studies that have evaluated the impact of workplace interventions on other work-related outcomes, such as workability and work performance, knowledge appears to be limited. It is therefore difficult to draw general conclusions about the effects of interventions on a particular outcome as only a few studies have been conducted. In addition, there is an inherent problem in how the various work-related outcomes are measured, as there is no so-called gold standard for these types of outcomes. Few of the included studies have used the same questionnaires, and in some cases the same questionnaires are used to measure different outcomes. For instance, the World Health Organization Health and Work Performance Questionnaire (WHO-HPQ) was used to measure productivity, presenteeism and work performance. This makes it difficult to compare the results of the studies and to draw general conclusions about the impact of the initiatives.

Finally, the knowledge about the impact of the effects is limited as many of the included studies have had too short follow-up times, such as 4 weeks. Changes in work-related outcomes may have occurred after the final follow-up measurements have been completed. It takes time for a health promotion effect to lead to an improvement in physical activity and/or eating habits, which in turn is expected to lead to an improvement in work-related outcomes.

Overall, fourteen interventions were evaluated as being effective. Seven of these showed significant reductions in absenteeism (6 studies with moderate risk of bias and 1 study with no information), two studies showed effects on work performance (1 study with unclear risk of bias and 1 study with low risk of bias), three studies showed effects on workability (1 study with low risk of bias, 1 study with high risk of bias and 1 study with moderate risk of bias), one study showed effects on productivity (moderate risk of bias) and one study showed effects on both workability and productivity (no information). Two more studies showed effects on absenteeism only between subgroups (1 study with moderate risk of bias and 1 study with no information).

One of the strengths of this systematic review is the comprehensive search strategy used across major electronic databases, that facilitated a more evidence-based approach to literature searching. Moreover, inclusion of study designs other than RCTs is important in evaluating complex interventions, such as workplace health promotion interventions, as the implementation of an RCT may be difficult and/or ethically inappropriate in practice. Each included study in this review was comprehensively selected, assessed, data extracted and quality assessed by two review authors independently to minimize potential biases in the review process. Another important strength is the use of two methodological quality assessment tools, both recommended by Cochrane Collaboration, to assess the risk of bias of the included RCTs and NRSs. A limitation of this review is that studies in languages other than English, as well as unpublished studies (“grey” literature) were excluded. As a consequence, some useful and relevant studies might have been missed. In addition, all included studies were from high-income countries, restricting the generalizability of the results. Furthermore, the heterogeneous study components rendered a meta-analysis impossible.