Skip to main content
. 2022 Aug 5;19(15):9653. doi: 10.3390/ijerph19159653

Table 2.

Summary of intervention effects on mental health or psychosocial work exposure indicators and quality of the evidence on intervention effectiveness according to the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach.

Intervention 1 Intervention Effect 1 on Mental Health or Psychosocial Work Exposure Indicators Quality of the Evidence 2 on Intervention Effectiveness and Justification of Rating
Simulation-based teamwork training program (25 h)
(Beneria et al. 2020) [42]		 ↑ likelihood of anxiety and depression (HADS > 12) post-program in workers having had contact with COVID-19 patients: AOR 2.56, 95% CI: 1.03–6.36; p = 0.043. AOR not reported for all workers who received the training program or for those who received it but had had no contact with COVID-19 patients. Very low
Observational design, serious risks of selection and confounding bias (48% participation rate for control group, important confounders omitted, i.e., history of mental illness, psychosocial work exposures)
Wellbeing centers supported by wellbeing buddies (4–5 months)
(Blake et al. 2020) [43]		 ↑ mental wellbeing of 1.93 points on WEMWBS scale that ranges from 14 to 70 points: mean WEMWBS score (SD) is 47.04 (9.49) for center users and 45.11 (9.35) for non-users; p = 0.02 Very low
Observational design, serious risks of selection and confounding bias (<5% participation rate, confounding factors not considered in analysis)
↑ work engagement of 0.19 points on dedication subscale of UWES-9 ranging from 0 to 6 points: mean score (SD) is 5.02 (1.38) for center users and 4.83 (1.15) for non-users; p = 0.08 Very low
Observational design, serious risks of selection and confounding bias (<5% participation rate, confounding factors not considered in analysis)
≠ % presenteeism past 12 months among center users vs. non-users:
no, never: 16.31 vs. 14.97
yes, once: 17.05 vs. 12.76
yes, 2 to 5 times: 16.92 vs. 12.64
yes, >5 times: 4.53 vs. 4.41
p = 0.28		 Very low
Observational design, serious risks of selection and confounding bias (<5% participation rate, confounding factors not considered in analysis)
≠ % with intention to quit among center users (16.31%) and non-users (15.09%); p = 0.25 Very low
Observational design, serious risks of selection and confounding bias (<5% participation rate, confounding factors not considered in analysis)
Multi-component SARS prevention program: scheduling and staffing adjustments, IPC measures and protocols, latest PPE, daily information, training, mental health team and clinic for workers (3 months) (Chen et al. 2006) [19] ↓ anxiety level from moderate before SARS patient care (T0) to mild two weeks (T1) and one month (T2) under prevention program to no anxiety at final time point (T3), Zung’s self-rating anxiety scale:
Mean anxiety score (SD)

  • T0: 60 (9.28)

  • T1: 51 (10.32)

  • T2: 50 (9.84)

  • T3: 46 (7.48)

Change (improvement)
  • ○

    T0 vs. T1: Z = −2.68; p = 0.0075

  • ○

    T0 vs. T2: Z = −4.45; p < 0.0001

  • ○

    T0 vs. T3: Z = −6.58; p < 0.0001

 Very low
Observational design, serious risks of selection and confounding bias (participation rate ND, unclear if all measured covariables were included in models, other potential confounders not measured, i.e., program compliance, medication use for anxiety, work exposures, e.g., changing work schedules mentioned in discussion, other factors outside work)
↓ depression level from moderate before SARS patient care (T0) to mild two weeks (T1) and one month (T2) under prevention program to no depression at final time point (T3), Zung’s self-rating depression scale:
Mean depression score (SD)

  • T0: 61 (12.62)

  • T1: 51 (11.94)

  • T2: 50 (10.60)

  • T3: 48 (10.76)

Change (improvement)
  • ○

    T0 vs. T1: Z = −4.58; p < 0.0001

  • ○

    T0 vs. T2: Z = −4.80; p < 0.0001

  • ○

    T0 vs. T3: Z = −6.37; p < 0.0001

 Very low
Observational design, serious risks of selection and confounding bias (participation rate ND, unclear if all measured covariables were included in models, other potential confounders not measured, i.e., program compliance, medication use for depression, work exposures, e.g., changing work schedules mentioned in discussion, other factors outside work)
↑ sleep quality under prevention program, but sleep quality remains poor, i.e., PSQI > 5, at all time points
Mean sleep quality score (SD)

  • T0: 12 (3.83)

  • T1: 10 (3.43)

  • T2: 10 (3.77)

  • T3: 8 (2.75)

Change (improvement)
  • ○

    T0 vs. T1: Z = −2.79; p = 0.0053

  • ○

    T0 vs. T2: Z = −3.14; p = 0.0017

  • ○

    T0 vs. T3: Z = −3.37; p = 0.0008

 Very low
Observational design, serious risks of selection and confounding bias (participation rate ND, unclear if all measured covariables were included in models, other potential confounders not measured, i.e., program compliance, work exposures, e.g., changing work schedules mentioned in discussion, other factors outside work)
Multi-component COVID-19 prevention program: recognition measures (2 weeks)
(Zhu et al. 2020) [46]		 Recognition measures are associated with 24% ↓ likelihood of anxiety (GAD-7 ≥ 8) compared to not having received recognition measures: AOR (95% CI): 0.76 (0.60–0.97); p = 0.03 Low
Observational design
Recognition measures are associated with 31% ↓ likelihood of depression (PHQ-9 ≥ 10) compared to not having received recognition measures: AOR (95% CI): 0.69 (0.52–0.90); p = 0.007 Low
Observational design
Recognition measures are associated with 24% ↓ likelihood of acute stress in the past 7 days caused by a traumatic event, COVID-19 being the specific event (IES-R > 33), compared to not having received recognition measures: AOR (95% CI): 0.76 (0.60–0.97); p = 0.024 Low
Observational design
Multi-component COVID-19 prevention program: satisfaction with IPC measures (2 weeks)
(Zhu et al. 2020) [46]		 Satisfaction with IPC measures is associated with 35% ↓ likelihood of anxiety (GAD-7 ≥ 8) compared to not being satisfied: AOR (95% CI): 0.65 (0.50–0.85); p = 0.002 Low
Observational design
Satisfaction with IPC measures is associated with 30% ↓ likelihood of depression (PHQ-9 ≥ 10) compared to not being satisfied: AOR (95% CI): 0.70 (0.51–0.95); p = 0.02 Low
Observational design
Satisfaction with IPC measures is associated with 31% ↓ likelihood of acute stress in the past 7 days caused by a traumatic event, COVID-19 being the specific event (IES-R > 33) compared to not being satisfied: AOR (95% CI): 0.69 (0.53–0.89); p = 0.004 Low
Observational design
Multi-component COVID-19 prevention program: satisfaction with logistic support (shuttle service, meals/hydration and accommodation) (2 weeks)
(Zhu et al. 2020) [46]		 Satisfaction with logistic support is associated with 31% ↓ likelihood of anxiety (GAD-7 ≥ 8) compared to not being satisfied: AOR (95% CI): 0.69 (0.50–0.96); p = 0.03 Low
Observational design
Satisfaction with logistic support is associated with 33% ↓ likelihood of depression (PHQ-9 ≥ 10) compared to not being satisfied: AOR (95% CI): 0.67 (0.47–0.97); p = 0.03 Low
Observational design
Effect of satisfaction with logistic support on likelihood of acute stress in the past 7 days caused by a traumatic event, COVID-19 being the specific event (IES-R > 33), not reported because not significant in univariate analysis Low
Observational design
Multi-component COVID-19 prevention program: satisfaction with work shift arrangements (2 weeks)
(Zhu et al. 2020) [46]		 Effect of satisfaction with work shift arrangements on likelihood of anxiety not reported because not significant in univariate analysis Low
Observational design
Satisfaction with work shift arrangements is associated with 52% ↓ likelihood of depression (PHQ-9 ≥ 10) compared to not being satisfied: AOR (95% CI): 0.48 (0.34–0.67); p < 0.001 Low
Observational design
Satisfaction with work shift arrangements is associated with 55% ↓ likelihood of acute stress in the past 7 days caused by a traumatic event, COVID-19 being the specific event (IES-R > 33), compared to not being satisfied: AOR (95% CI): 0.45 (0.33–0.63); p < 0.001 Low
Observational design
“R2 for Leaders” resilience training program intended to equip healthcare leaders to better lead their staff and organization by identifying and implementing individual resilience and organization-level prevention programs (12 virtual 2-h weekly sessions over 3 months)
(Giordano et al. 2021) [44]		 ↓ emotional exhaustion level in healthcare leaders post-program: mean MBI-EE score (SD): T1: 6.31 (1.35) vs. T2: 5.37 (1.20); p = 0.020; Hedge’s g (corrected Cohen’s d for small samples < 50) = −0.30 Very low
Observational design, serious risks of selection and confounding bias (participation rate ND, potential confounders not considered in analyses), potentially inadequate power
≠ emotional exhaustion level in staff post-program (no clinically or statistically significant difference): mean MBI-EE score (SD): T1: 4.70 (1.63) vs. T2: 4.35 (1.64); p = 0.098 Very low
Observational design, serious risks of selection and confounding bias (participation rate ND, potential confounders not considered in analyses), potentially inadequate power
≠ quality of leaders’ psychosocial work environment post-program: mean HSE-MSIT score (SD) on scale of 22 to 110:
T1: 50.50 (15.33) vs. T2: 50.56 (15.17); p = 0.966		 Very low
Observational design, serious risks of selection and confounding bias (participation rate ND, potential confounders not considered in analyses), potentially inadequate power
↑ quality of staff’s psychosocial work environment post-program: mean HSE-MSIT score (SD) on scale of 22 to 110:
T1: 50.18 (10.56) vs. T2: 46.93 (10.75); p = 0.028; Cohen’s d = −0.29		 Very low
Observational design, serious risks of selection and confounding bias (participation rate ND, potential confounders not considered in analyses), potentially inadequate power
Multi-component COVID-19 prevention program: reorganized wards (e.g., increased ICU beds), procedures (e.g., cleaning and disinfection) and internal paths, increased nurse-to-patient ratios in COVID units, PPE training, other training, promoted participatory approach, autonomy and conscientiousness through continuous clinical and organizational audits, lectures, workshops and meetings, psychological help desk for staff, staff COVID-19 testing (4 months)
(Zaghini et al. 2021) [45]		 ≠ quality of emotional life post-program: mean score on emotional subscale of NQoL-SAT-P (SD) that ranges from 1 to 4:
T0: 3.13 (.49) vs. T1: 3.16 (.52); p = 0.334		 Low
Observational design, risk of confounding bias (several potential confounders ND (i.e., level of adherence to the intervention) or not integrated in analyses (i.e., age, having children))
↑ quality of the psychosocial work environment post-program: mean HSE-MSIT score (SD) on scale of 1 to 5:
T0: 2.46 (0.40) vs. T1: 2.32 (0.50); p < 0.001		 Low
Observational design, risk of confounding bias (several potential confounders ND (i.e., level of adherence to the intervention) or not integrated in analyses (i.e., age, having children))
≠ work demands (workload, time pressure) post-program: mean HSE-MSIT subscale score (SD) on scale of 1 to 5:
T0: 2.81 (0.48) vs. T1: 2.79 (0.58); p = 0.601		 Low
Observational design, risk of confounding bias (several potential confounders ND (i.e., level of adherence to the intervention) or not integrated in analyses (i.e., age, having children))
↑ job control post-program: mean HSE-MSIT subscale score (SD) on scale of 1 to 5:
T0: 2.76 (0.67) vs. T1: 2.65 (0.65); p = 0.020		 Low
Observational design, risk of confounding bias (several potential confounders ND (i.e., level of adherence to the intervention) or not integrated in analyses (i.e., age, having children))
↑ managerial support post-program: mean HSE-MSIT subscale score (SD) on scale of 1 to 5:
T0: 2.34 (0.88) vs. T1: 2.17 (0.98); p = 0.020 		Low
Observational design, risk of confounding bias (several potential confounders ND (i.e., level of adherence to the intervention) or not integrated in analyses (i.e., age, having children))
↑ peer support post-program: mean HSE-MSIT subscale score (SD) on scale of 1 to 5:
T0: 2.12 (0.67) vs. T1: 1.93 (0.69); p = 0.001 		Low
Observational design, risk of confounding bias (several potential confounders ND (i.e., level of adherence to the intervention) or not integrated in analyses (i.e., age, having children))
↑ quality of relationships at work (harassment, tension, bullying) post-program: mean HSE-MSIT subscale score (SD) on scale of 1 to 5:
T0: 2.23 (0.88) vs. T1: 2.04 (0.68); p = 0.001 		Low
Observational design, risk of confounding bias (several potential confounders ND (i.e., level of adherence to the intervention) or not integrated in analyses (i.e., age, having children))
≠ role clarity at work post-program: mean HSE-MSIT subscale score (SD) on scale of 1 to 5:
T0: 1.71 (0.52) vs. T1: 1.69 (0.60); p = 0.798 		Low
Observational design, risk of confounding bias (several potential confounders ND (i.e., level of adherence to the intervention) or not integrated in analyses (i.e., age, having children))
Improvement in how organizational change is managed and communicated at work post-program: mean HSE-MSIT subscale score (SD) on scale of 1 to 5:
T0: 2.98 (0.49) vs. T1: 2.46 (0.79); p < 0.001		 Low
Observational design, risk of confounding bias (several potential confounders ND (i.e., level of adherence to the intervention) or not integrated in analyses (i.e., age, having children))
Study reporting on the association between perception of adequate PPE, training and support and mental health indicators 13–25 months after SARS outbreak (no intervention described per se) (Maunder et al. 2006) [47] 20% ↓ likelihood of post-traumatic stress (IES-R ≥ 26) post-outbreak, multivariate logistic regression model: β = −0.22; p = 0.01 Low
Observational design
24% ↓ likelihood of emotional exhaustion (MBI-EE ≥ 27) post-outbreak, multivariate logistic regression model: β = −0.27; p = 0.002 Low
Observational design
Likelihood of psychological distress (K10 ≥ 16) not reported because the “Training, protection and support” indicator was not significant in univariate models Low
Observational design

1 Detailed descriptions of intervention content and effectiveness are provided in Supplementary Table S1. 2 Low: our level of confidence in effect estimates is low, the true effect could be very different from that estimated in the studies; very low: our level of confidence in effect estimates is very low, the true effect is probably very different from that estimated in the studies. ↑ higher; ↓ lower; ≠ no change. AOR: adjusted odds ratio; CI: confidence interval; GAD: Generalized Anxiety Disorder; HADS: Hospital Anxiety and Depression Scale; HSE-MSIT: health and safety executive management standards indicator tool; ICU: intensive care unit; IES-R: impact of event scale-revised; IPC: infection prevention and control; K10: Kessler 10-item psychological distress scale; MBI-EE: Maslach burnout inventory−emotional exhaustion subscale; ND: not documented; NQoL-SAT-P: Nurses Quality of Life Scale−Satisfaction Profile; PHQ-9: Patient Health Questionnaire; PPE: personal protective equipment; PSQI: Pittsburgh sleep quality index; SARS: severe acute respiratory syndrome; SD: standard deviation; UWES-9: Utrecht Work Engagement scale; WEMWBS: Warwick—Edinburgh Mental Wellbeing Scale