(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Effectiveness of tailored digital health interventions for mental health at the workplace: A systematic review of randomised controlled trials [1]

['Thirimon Moe-Byrne', 'Department Of Health Sciences', 'University Of York', 'York', 'United Kingdom', 'Jessie Shepherd', 'Dorota Merecz-Kot', 'Institute Of Psychology', 'Lodz', 'Marjo Sinokki']

Date: 2022-11 

Mental health problems in the workplace are common and have a considerable impact on employee wellbeing and productivity. Mental ill-health costs employers between £33 billion and £42 billion a year. According to a 2020 HSE report, roughly 2,440 per 100,000 workers in the UK were affected by work-related stress, depression, or anxiety, resulting in an estimated 17.9 million working days lost. We performed a systematic review of randomised controlled trials (RCTs) to assess the effect of tailored digital health interventions provided in the workplace aiming to improve mental health, presenteeism and absenteeism of employees. We searched several databases for RCTs published from 2000 onwards. Data were extracted into a standardised data extraction form. The quality of the included studies was assessed using the Cochrane Risk of Bias tool. Due to the heterogeneity of outcome measures, narrative synthesis was used to summarise the findings. Seven RCTs (eight publications) were included that evaluated tailored digital interventions versus waiting list control or usual care to improve physical and mental health outcomes and work productivity. The results are promising to the advantage of tailored digital interventions regarding presenteeism, sleep, stress levels, and physical symptoms related to somatisation; but less for addressing depression, anxiety, and absenteeism. Even though tailored digital interventions did not reduce anxiety and depression in the general working population, they significantly reduced depression and anxiety in employees with higher levels of psychological distress. Tailored digital interventions seem more effective in employees with higher levels of distress, presenteeism or absenteeism than in the general working population. There was high heterogeneity in outcome measures, especially for work productivity; this should be a focus of attention in future studies.

Work-related stress has become a major issue at the work place. It is associated with increased physical and mental health risk. Work stress is also associated with long term economic impact due to absenteeism and presenteeism. There have been several workplace interventions to tackle this issues, with varying results. In recent years, digital health interventions have become increasingly popular to promote employees’ physical and mental wellbeing. Some have shown moderate effects improving employees’ psychological wellbeing. However, most interventions were not tailored to the individual employee. So far, no review has explored the effect of tailored digital interventions addressing stress in the workplace compared to usual care. Also, no reviews have taken blended e-health models into account explicitly. This systematic review seeks to address this gap. We found that tailored digital interventions did not reduce anxiety or depression in the general working population but they significantly reduced anxiety in employees with higher levels of psychological distress. However, as the outcomes measured were so diverse, it is difficult to draw conclusions. There is a need for uniformity in the use of assessment tools and outcomes reporting in future research, especially in terms of work productivity.

Competing interests: The authors have declared that no competing interests exist except for CvdFC who received honoraries for lectures from the Loyds Register Foundation and Janssen UK.

Funding: The study is funded by the European Union’s Horizon 2020 Research and Innovation Programme under Grant Agreement No. 848180. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Copyright: © 2022 Moe-Byrne et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

This systematic review seeks to evaluate the impact of tailored digital health interventions aiming to improve mental health, stress-related physical symptoms, presenteeism and absenteeism in the workplace in employees in case of mental health issues or work-stress.

In the past decade, the advancement of digital technology has led to the creation of digital health interventions in physical and mental health. Nowadays, web based and smartphone based digital health interventions have become increasingly popular within the health industry due to their easy implementation [ 30 ]. In addition, digital interventions have the potential to be scaled up to provide care to large populations and could offer greater anonymity and stigma reduction compared to traditional face-to-face interventions [ 16 , 31 ]. Several reviews published in 2017 and 2018 found that digital interventions could effectively improve employees’ psychological wellbeing and reduce stress at the workplace. They explored the impact of digital interventions without further support from health professionals on health-related outcomes, and most of the interventions examined were not tailored to the individual employee [ 30 , 32 , 33 ]. Tailoring in this context differs from mode tailoring as described elsewhere [ 34 ]. It indicates an automated process that adapts the intervention based upon the employee’s input, for example, by applying an algorithm to the individual responses from the baseline questionnaire to generate specific feedback and provide modules relevant to the user. Also, blended eHealth interventions delivered by healthcare professionals combined with a digital intervention tailoring to the treatment is now available. So far, their application and effect in the workplace setting has not been explored.

Non-digital workplace interventions have been developed to promote employees’ wellbeing and potentially prevent or treat mental disorders [ 24 ] and a few studies showed some positive results with workplace interventions such as cognitive-behavioural therapy, mindfulness and relaxation [ 24 – 28 ]. Also, one review which explored the effect on absenteeism and productivity found a small number of studies that suggested some effect for short interventions [ 29 ].

Understanding how employee health impacts productivity has advanced substantially in the past two decades [ 19 – 22 ]. According to a survey conducted in 2019, 69% of UK line managers were aware that supporting employee wellbeing is a core management requirement; however, only 13% have received any form of wellbeing or mental health training [ 23 ].

Mental ill-health costs employers between £33 billion and £42 billion a year, with an annual cost to the UK economy of between £74 billion and £99 billion [ 13 ]. According to a Health and Safety Executive report in 2020 there were an estimated 828,000 workers affected by work-related stress, depression, or anxiety; these three conditions are responsible for 51% of all cases of work-related ill health and 55% of all days lost due to work-related ill health in the UK [ 14 , 15 ]. Work-related stress is a significant risk factor for a range of physical and mental health conditions such as cardiovascular disease, depression and related mortality [ 16 , 17 ]. In addition, a 2020 survey undertaken on 3,614 UK workers reported that 41% of employees surveyed have experienced mental health symptoms that were caused or worsened by work and 51% said their symptoms were due to pressure they experienced at work [ 18 ].

Globally, mental health problems as a cause of the burden of illness are on the rise and account for a fifth of all years of living disabled [ 10 ]. The Global Burden of Disease study reported that in 2017, an estimated 792 million people are affected by mental ill-health worldwide [ 11 ] and a 2018 Lancet Commission report on mental health estimates mental health disorders will cost the global economy 16 trillion US dollars by 2030 [ 12 ].

Mental ill-health conditions such as depression, anxiety, and stress are common in the workplace. They impact employee wellbeing, productivity, absenteeism (sickness absence), and presenteeism [ 1 , 2 ]. Presenteeism is working with difficulty to do the tasks at hand [ 3 ]. Originally coined as “showing up at work while being sick” [ 4 ] because of chronic medical conditions [ 5 ] or because of work or personal characteristics, the emphasis in interpretation has shifted toward worker slowdowns in general and the economic costs associated with that [ 6 ]. The prevalence of presenteeism is high, amounting to an average of 40% in a survey conducted amongst workers in 34 countries [ 7 – 9 ].

Due to the lack of homogeneity across outcome measures, we did not consider formal meta-analysis a feasible option. Therefore, a narrative approach was used to summarise the findings as presented in tables in the Results section. We have also presented the percentage range for adherence and attrition in each group. We considered less than 25% as low, less than 50% as moderate and 50% and above as good.

We produced the overall risk of bias assessment for each study (low, high, uncertain) based on reviewing all sources of bias identified in the individual trial. Two reviewers (TMB and JS) independently assessed the quality of the studies. Disagreements were resolved through consensus, and if necessary, a third reviewer (CFC) was consulted. The funnel plot to assess the publication bias was not conducted as there were less than ten trials in the review [ 41 ].

The quality of the RCTs was assessed according to the Cochrane Risk of Bias tool (low, high, uncertain) [ 39 ] supported by a standardised data extraction form. The tool assessed several sources of bias, including selection, performance, detection, attrition, and reporting bias. The clustered RCTs were assessed using the same tool but amended by adding the following sources of bias (recruitment bias, baseline imbalance, loss of clusters, incorrect analysis, and comparability with individual randomised trials) which are particular to the Clustered RCTs [ 40 ].

Data were extracted in duplicate by two reviewers independently (TMB and JS). Disagreements were resolved through consensus, and if necessary, a third reviewer (CFC) was consulted. Data from studies with multiple publications were extracted and reported as a single study. The study authors were contacted for further information when appropriate.

Data extraction was carried out using a standardised data extraction form. This extraction form was piloted (by filling in the first two studies) and refined. Data were extracted as follows: study design (e.g., RCT, Cluster randomisation or Stepped wedge), study participants (age, sex) and interventions (population, sample size, intervention/comparator details), outcome measures (outcome description and measures), results, effect size and findings. We also extracted the types of personalisation and tailoring used for participants and the theoretical background upon which the digital intervention materials were based, and created a table showing triage, personalisation, and intervention tailoring used by the individual trials, in the results section. Tailoring methods were extracted from each article, using the original authors language and terminology. Commonalities and overlaps were examined, this process was done inductively based upon the language used in the articles, and five categories of tailoring emerged. They are: 1) Algorithm used to triage based upon assessment or questionnaire scores, 2) Algorithm used for tailored Summarised Feedback, 3) Participant choice, 4) Use of automated messages to users, 5) Interventions blended with in-person support. Table 1 shows each of the five categories.

We considered the second round of 10% double screening if the kappa score was less than 0.8. However, we have achieved an excellent inter-rater agreement in our first round (Cohen’s Kappa score 0.96), and no potentially eligible studies were missed. Hence, two reviewers continued screening half of each of the remaining studies. Full texts of potentially eligible studies were then screened independently and in duplicate by two reviewers (TMB and JS). Disagreements were resolved through discussion to achieve consensus, and if necessary, a third reviewer (CFC) was consulted. In all cases, a consensus was reached.

The studies retrieved from the searches were exported into EndNote. After deduplication, the rest of the studies were exported into Rayyan (Rayyan QCRI software for screening) [ 38 ]. The first 10% of the titles and abstracts of the studies were screened in duplicate by two independent researchers (TMB and JS) against the inclusion/exclusion criteria. We coded either 1 for inclusion or 0 for exclusion. We used Cohen’s Kappa statistic to calculate the percentage of positive agreement and negative agreement between the reserachers.

Relevant studies were identified by searching the following electronic databases: MEDLINE, EMBASE, PsycINFO, and The Cochrane Central Register of Controlled Trials (CENTRAL) published from 2000 onwards. We searched Google Scholar for grey literatures and the reference lists of all included articles and relevant systematic reviews were also checked to identify potentially eligible studies. Experts who published in the domain of workplace stress, mental health and digital interventions were contacted to suggest any articles or grey literature that they deemed relevant. We did not apply any language restrictions (See S1 Appendix for a complete list of search terms and combinations). The search was supported by the Centre for reviews and Dissemination (CRD) at the University of York.

The studies were also excluded if the interventions solely addressed addiction or consisted of online training/teaching or online education without tailoring. Interventions aimed at physical outcomes, productivity, performance, presenteeism, absenteeism or lifestyle interventions without mental health or work-stress outcomes were also excluded. Studies reporting on cost-effectiveness were out of scope.

Studies were excluded if they reported interventions that were not tailored or personalised in some way to the individual employee; if the feedback provided to the participants was not tailored or individualised in some way; or if the participants were provided with material to practice on their own, such as mindful breathing or meditation, with no tailoring based upon client assessment and symptomatology. Feedback from an outside therapist or group, not included in the digital intervention itself, did not count as relevant here.

Included studies had to report interventions that were delivered using digital technology. Digital technology includes digital decision aids or materials delivered through a computer, tablet, smartphone or email. This material could be delivered as a website, app or downloadable software. This digital intervention could be combined with further support such as group sessions, individual counselling, or direct feedback from a health professional (such as an occupational physician, a psychotherapist, a coach in the workplace, or psychiatric consultation), so-called blended e-Health interventions. This further support was not necessary for inclusion, however. The studies had to report on mental health outcomes or work-stress. Additionally, they could report on physical outcomes, presenteeism or absenteeism. We have also included studies related to insomnia as sleep-wake disorders are a classification in the Diagnostic Statistical Manual of Mental disorders (DSM-5) [ 37 ]. To be included, a digital intervention needed to have some tailored feedback and content for the individual employee. Feedback from an outside therapist or group, not included in the digital intervention, did not meet this criterion. Also, online training or monitoring without tailoring and personalisation would not meet this criterion.

RCTs, including pilot and feasibility trials, with individual or cluster randomisation and stepped wedge trial designs (SWTD), were eligible if they assessed the impact of digital health interventions compared with standard guidance or waiting list control. The participants had to be employees aged 16 years and above, and the interventions had to be provided in or via the workplace setting.

The review protocol has been registered in the PROSPERO database under ID CRD 42021213292 [ 35 ]. This review complies with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines (See S3 Appendix ) [ 36 ]. This study was funded by the European Union’s Horizon 2020 research and innovation program under grant agreement number 848180.

Results

Study selection The study selection is shown in Fig 1 below. A total of 5992 publications were identified through the initial database search, and an additional 49 studies were identified through google scholar and citation search. After the removal of duplicates, 4774 publication titles and abstracts were screened. Fifty studies were eligible for a full-text review. We could not retrieve a full-text review of 1 study as the authors did not reply to our email [42]. Out of 50 studies, 42 were excluded because the interventions were not tailored to the individuals, reported no mental health outcomes, addressed a mixed or wrong population, had no waiting list control, the full text was not available, and some were ongoing trials or a cost-effectiveness study which was out of scope. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. PRISMA 2009 Flow Diagram. https://doi.org/10.1371/journal.pdig.0000123.g001

Study characteristics Eight publications (7 studies) met the inclusion criteria: five RCTs [43–46], one crossover RCT [47] and two cluster RCTs [48–50]. Two studies [48,49] were part of a larger trial [51] which used the same data to report different outcomes. One study was conducted in the USA [43], two studies (three publications) in the Netherlands [48–50], two in Germany [44,46] and two in the UK [45,47]. The total number of participants included in this review was 2008, with study sample sizes ranging from 48 to 532. The studies were conducted between 2004 and 2019.

Participant characteristics Four studies only included participants with particular characteristics. One study included only participants that screened positive for depression, somatisation, or anxiety based on the following assessment tools: the Patient Health Questionnaire for depression (PHQ-9), for somatisation (PHQ-15), and for Generalised Anxiety Disorder (GAD-7); this study also required that participants were on sick-leave between 4 and 26 weeks [50]. One study required participants to screen positive on the General Health Questionnaire (GHQ 12) [45] and to be sick-listed at least 4 weeks; and one required participants screening positive on the Perceived Stress Scale (PSS-10) [44]. A fourth study only included participants who self-identified as having poor sleep [47]. One study excluded participants if they were on sick leave for more than two weeks [48,49]. The remaining two studies did not report any specific inclusion/exclusion criteria [43,46]. Participants were recruited from a variety of workplaces. Four studies recruited the employees from organisations such as an insurance company, a technology company, or large corporations [43,44,47,50]. Two studies recruited healthcare professionals from local hospitals [45,48,49]. One study recruited employees from the private and public sector companies in three countries [46].

Intervention characteristics Three studies (four publications as mentioned above) included in this review designed their digital interventions to increase wellbeing and act as preventative interventions for mental health problems such as anxiety and depression, rather than treating these conditions [43,46,48,49]. One focused on work performance and provided guidance to the Occupational Physician to address any mental wellbeing issues together with the GP, as a blended eHealth model [50]. The length of the administration of the interventions ranged from four weeks to three months. All studies included in this review compared a web-based intervention, [43,44,48–50] a smartphone app intervention [46] and a combined web based and smartphone app intervention [47] with waiting list or usual guidance control. All the studies stated the use of Cognitive Behavioural Therapy (CBT) techniques as theoretical background for the materials chosen and content uploaded into the digital intervention. Additionally, two stated the use of Mindfulness-based content and materials [46,48,49]. Other studies used the transactional stress model or the job demands-resources (JDR) model as background [44,46]. Triage used in the application of digital support systems refers to the assessment by questionnaires to gauge the mental and or physical state of the participant, which informs the tailoring algorithm. Table 1 below shows each of the five categories and which articles used which tailoring tool. The use of an Algorithm refers to programmed instructions in the digital intervention that guide the assignment of materials or the generation of feedback. Digital interventions that use algorithms, automate, and personalise the feedback and content that a user sees. The algorithm guides content assignment based on the users’ feedback, assessment scores, actions, or behaviour within the digital intervention. Guidance can differ depending on the focus of the digital intervention. For example, with the Kelaa Mental Resilience App [46], the digital intervention used sensors built in users smartphones to track the users’ sleep and then provide detailed feedback. One study provided support from participants by occupational physicians who received instructions and decision support from the digital intervention, based upon the tailoring [50]. This was the only study evaluating a blended digital intervention. Two studies allowed digital intervention to be combined with regular counselling and medication, or “treatment as usual” [44,45]; this is not considered blended care as the real life treatment is not influenced by the digital intervention. The characteristics for tailoring are shown in Table 1 and S2 Appendix.

Outcome measures Most studies reported a mix of psychological (such as depression, anxiety, and stress) and work-related outcome measures (such as absenteeism and presenteeism). Several studies reported physical measures: two studies reported outcomes relevant to sleeping problems [44,46] and one measured the amount of sleep and workplace performance [47]. One study reported physical symptoms related to somatisation [50] and the other two reported levels of physical health impairment using self-reported SF12 and SF 36 questionnaires [44,46]. However, none of the studies reported on long-term medical conditions such as diabetes or cardiovascular disease. Regarding the crossover trial, we only used first period data prior to crossover [47].

Physical outcomes As physical outcomes, no study included in this review published actual measurements of physical conditions, such as cardiovascular disease. However, stress-related physical symptoms related to somatisation and sleep impairment as indicated by study participants were reported.

Sleep impairment Three studies reported sleep impairment using the following assessment: Sleep Condition Indicator (SCI) [47], Insomnia Severity Index (ISI) [44], and the Sleeping Troubles Copenhagen Psychosocial Questionnaire-Revised Version (COPSOQ II) [46]. One study which included self-identified “poor” sleep participants, reported statistically significant differences in the intervention group compared to the control at eight weeks [47]. The other study, which included participants with high stress scores, showed significant effect in favour of the intervention at seven weeks and six months [44]. The third study, which measured sleep impairment, did not significantly differ between the control and intervention groups [46].

Physical health impairment Two studies reported physical health levels, using self-reported SF 12 and SF 36 v2 [52–54]. Both studies found no statistically significant differences between the intervention and control [44,46].

Somatisation One study reported physical symptoms related to somatisation using the Patient Health Questionnaire-15 (PHQ-15) screening tool [55]. There were no statistically significant differences between the eHealth intervention group and the control at one- and three-months post-intervention. But there was a significant improvement of somatisation related physical symptoms in the eHealth intervention group compared to control at 6 and 9 months (p = 0.017 and p = 0.039) [50].

Psychological outcomes All seven studies reported psychological outcomes [43–50]. Six studies reported measures of anxiety and depression symptoms, and four studies reported on stress levels, which also included posttraumatic stress [43,44,46,49].

Anxiety and depression Six studies measured anxiety and depression outcomes using various assessment tools: the Beck Anxiety Inventory (BAI) [43], the Brief Symptom Inventory (BSI) [48], the Hospital Anxiety and Depression Scale (HADS) [44,45], the Generalised Anxiety Disorder-2(GAD-2) [47], the Patient Health Questionnaire-9 (PHQ-9), the Generalised Anxiety Disorder-7 (GAD-7) [50], and the Centre for Epidemiologic Studies Depression Scale-Revised (CES-D) [43,44]. Out of six, four studies did not find a statistically significant difference between the digital intervention and control condition in anxiety and depression [43,47,48,50]. In contrast, two studies which included employees with higher levels of psychological distress showed significant improvement with digital intervention compared to control [44,45]. The study which worked following the Lazarus and Folkman’s ’Transactional Model of Stress’ [44] included only higher stress level participants in their sample (22 or more on the Perceived Stress Scale (PSS-10)). It showed a significant reduction of anxiety and depression at seven weeks and six months (p<0.001 for both anxiety and depression) [44]. Another study [45] which included only sick listed employees of a UK hospital (4 or more on the General Health Questionnaire 12 (GHQ-12)) showed significantly lower depression scores in the intervention group at the end of the treatment and one-month post-treatment (p = 0.028 and p = 0.04), and lower anxiety scores at one-month post-treatment (p = 0.021), compared to wait-list control. However, at 3- and 6-months post-treatment, these differences were no longer significant.

Stress Four studies reported stress outcomes using the different assessment tools (the symptoms of distress scale, the Perceived Stress Score (PSS-10) [56], the Copenhagen Psychosocial Questionnaire-Revised Version (COPSOQ II) [57], the posttraumatic Impact of Event Scale (IES) [58] and the Four Dimensional Symptoms Questionnaire (4DSQ) [43,44,46,49]). Three out of four studies in the intervention group showed significantly lower stress scores than the wait-list control at different time points [43,44,46]. One study showed a significant reduction in stress score at three months (p<0.023) [43] and the other study at seven weeks and six months (p<0.001) [44]. One study which used the JDR model of burnout and measured both general and cognitive stress between the two groups (significant group*time interactions) also showed a significant reduction in stress over time in the intervention group (p<0.001 general stress, p<0.01 cognitive stress) [46]. In contrast, a study [48,49] which included a subgroup analysis of participants with high stress levels, measured both posttraumatic stress and distress did not show significant improvement between the digital intervention group and the control group at six months follow up.

Wellbeing Two studies reported employees’ wellbeing using the Warwick- Edinburgh Mental Wellbeing Scale [46] and WHO-5 wellbeing index [48] with mixed results. One study reported that participants in the intervention group showed significantly more wellbeing over time (significant group*time interaction) than the wait-list control group at six weeks follow up [46]. However, the other did not find a statistically significant positive effect with digital intervention on wellbeing at six months [48].

Presenteeism Five studies reported some form of work performance that in this review were taken as an expression of presenteeism, which illustrates that the definition and operationalisation of presenteeism is not comparable between the instruments and consequently studies. Some studies used two instruments [44,48,49]. Two studies using the Utrecht Work Engagement Scale (UWES) [44,48] reported a small positive effect on work engagement in the intervention group (F = 3.44, p = 0.03, Cohen’s d = 0.25 three months follow-up, 0.15 six months follow-up) [48] and a statistically significant difference between groups in favour of the intervention at the seven week follow up (F = 5.4, p<0.05, Cohen’s d = 0.17) but not at six months [44]. One study [49] used an item of the Work Ability Index in which an individual assesses their own workability by comparing their actual one with the highest ever; and another instrument, the Nurses Work Functioning Questionnaire (NWFQ). This consists of 7 subscales: (1) cognitive aspects of task execution and general incidents, (2) impaired decision making, (3) causing incidents at work, (4) avoidance behaviour, (5) conflicts and annoyances with colleagues, (6) impaired contact with patients and their family and (7) lack of energy and motivation. This study did not find any significant differences between the intervention and control group [49]. Three other studies collected the data for presenteeism using the WPAI and the Work Limitations Questionnaire [43,46]; one did not analyse the data [46], one found a statistically significant difference in presenteeism (p = 0.001) [47] and the other found no significant difference in work productivity between the web-based intervention and wait-list control at three months [43]. One study that assessed presenteeism with a dedicated instrument at 6 months (TiC-P) [44] found a statistically significant improvement in presenteeism (p<0.01).

Absenteeism One study used the Treatment Inventory Cost in Psychiatric Patients (TIC-P) combined with social insurance data to assess Return to Work (RTW) on sick-listed employees with common mental disorders. They reported absenteeism as duration until first return to work and complete return to work after a long-term sickness. The duration until the first RTW in the intervention group was significantly shorter than in the control group (P = .03). The median duration was 77.0 (IQR 29.0–152.3) days in the control group and 50.0 (IQR 20.8–99.0) days in the intervention group, a difference of 27 days (hazard ratio [HR] 1.390, 95% CI 1.034–1.870, P = .03) at 12 months [50]. Two studies assessing absenteeism with a dedicated instrument at 8 weeks (WPAI) [47] and 6 months (TiC-P) [44] found no statistically significant difference.

[END]
---
[1] Url: https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0000123

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/