Workplace mental health screening: a systematic review and meta-analysis

Workplaces are an important location for population mental health interventions. Screening to detect employees at risk of or experiencing mental ill health is increasingly common. This systematic review and meta-analysis examined the efficacy of workplace mental health screening programmes on employee mental health, work outcomes, user satisfaction, positive mental health, quality of life, help-seeking and adverse effects. PubMed, PsycINFO, EMBASE, CENTRAL, Global Index Medicus, Global Health and SciELO were searched (database inception–10 November 2022) and results screened by two independent reviewers. Controlled trials evaluating screening of workers’ mental health as related to their employment were included. Random effects meta-analysis was performed to calculate pooled effect sizes for each outcome of interest. Grading of Recommendations Assessment, Development and Evaluation was conducted to evaluate the certainty of findings. Of the 12 328 records screened, 11 were included. These reported 8 independent trials collectively assessing 2940 employees. Results indicated screening followed by advice or referral was ineffective in improving employee mental health symptoms (n=3; d=−0.07 (95% CI −0.29 to 0.15)). Screening followed by facilitated access to treatment interventions demonstrated a small improvement in mental health (n=4; d=−0.22 (95% CI −0.42 to –0.02)). Limited effects were observed for other outcomes. Certainty ranged from low to very low. The evidence supporting workplace mental health screening programmes is limited and available data suggest mental health screening alone does not improve worker mental health. Substantial variation in the implementation of screening was observed. Further research disentangling the independent effect of screening alongside the efficacy of other interventions to prevent mental ill health at work is required.

At 5-mo follow up, one trial reports the odds of sickness absence in the intervention group being 1.40 times greater than the control group. Another trial found no differences between groups on sickness absence duration at 12-month follow up. Combining these trials resulted in a Pooled SMD = 0.06 [-0.22 to 0.34]. One trial found borderline statistically significantly difference in mean days of sickness absence at long-term follow up (5-years) with a trend favouring the intervention.

VERY LOW
CRITICAL 7 3 different articles, 2 reporting data from the same trial (short (12-mo) and long term (2-5 year) follow-up). 8 This has been rated as serious as trials scored high on risk of bias assessment. 9 This has been rated as serious as one trial found no effect and another found a negative effect of intervention on sickness absence, and I 2 = 75.88, indicating substantial heterogeneity. 10 This has been rated as serious due to wide confidence intervals that include the null hypothesis.

VERY LOW
CRITICAL 11 This has been rated as serious as two out of four trials had high risk of bias. 12 This has been rated as very serious as one trial showed a positive effect, two trials showed no effect, one trial showed a negative effect. 13 This has been rated as serious as all data were self-reporting of different outcomes. 14 This has been rated as serious due to wide confidence intervals and effect sizes range from small to moderate effect.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)

IMPORTANT
19 This has been rated as serious due to self-report assessment of visiting at least 1 of 11 caregivers (ranging from formal sources i.e., psychologists, to a supervisor or coach). 20 This has been rated as serious due to wide confidence interval calculated using raw data.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance

VERY LOW
CRITICAL 3 Trial used a study specific survey that was not thoroughly explained, not a validated user satisfaction measure. Also limited number of participants who gave user satisfaction data. 4 Results not able to be quantitatively assessed between groups. Only proportions from limited samples provided. 5 Trial gathered user satisfaction data from intervention group only, thus the imbalance in N between intervention and control.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)

VERY LOW
CRITICAL 6 2 different articles but 1 trial, each paper reporting data from the same trial (short (12-mo) and long term (2-5 year) follow-up). 7 Trial scored high on risk of bias assessment. 8 This has been rated as very serious as there was only one small trial with small sample size (N<200), and wide confidence intervals observed.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance

VERY LOW
CRITICAL 9 One from three trials had high risk of bias, the other two had low risk of bias. 10 One trial found no effect and two found positive effects and I 2 = 75.88, indicating substantial heterogeneity. 11 All self-report data assessing different outcomes. 12 Wide confidence intervals, effect sizes range from null to moderate effect.