To be included, the studies should meet several criteria.

Regarding the population, all studies are included in which the intervention group (IG) consisted of individuals working in the healthcare sector, comprising physicians, nurses, paramedics, physiotherapists, and other unspecified healthcare professionals, along with students of healthcare sciences.

All studies incorporating an intervention, regardless of the presence of a control group (CG), whether randomized or non-randomized are included. The requirement is that both baseline and post-intervention data are collected.

The studies must be written in English or German.

Another criterion is that the studies involve interventions aiming at maintaining or promoting resilience or related factors. The specific content of the intervention is not predetermined.

All studies offering resilience interventions digitally or in a blended format incorporating both online and in-person aspects are taken into consideration.

An additional inclusion criterion concerns the measurements. Included are studies comprising quantitative measurements of resilience or related outcomes such as stress, burnout, depression, anxiety, or well-being.

The eligibility criteria are summarized below according to the PICOS scheme ( Table 1 ).

All studies that do not meet the above criteria are excluded. This encompasses studies with a target population that is either not further specified or not employed in the healthcare sector; all reviews, study protocols, or similar works lacking an intervention design; and studies published in languages other than English or German. Additionally, interventions that are unrelated to resilience or related symptoms, interventions exclusively offering in-person or remote training, and studies with only qualitative outcome assessments and lacking a quantitative measurement method are also excluded.

For the systematic literature search, three databases were explored: PubMed, Embase, and Web of Science. The search strategy was based on three key terms: health care, resilience, and online or blended interventions. All articles published until the end of April 2024 were considered. In total, the search yielded 8271 search results.

All articles found from the search were extracted to Endnote and duplicates were deleted. The remaining studies were classified by relevance using an algorithm and sorted accordingly in an Excel spreadsheet. The AI (artificial intelligence) algorithm is based on the AS Review program. It is a language model that categorizes and prioritizes studies in the dataset under investigation based on their likely relevance using titles, sentences, and words. Hand-selected studies meeting the inclusion criteria serve as the reference (28).

Based on title and abstract, the majority of irrelevant studies were excluded in a first step. In a second step, the remaining studies were screened using the full text and ultimately included if the inclusion criteria were met. The detailed protocol can be found in the appendix (screening protocol).

More studies that met the inclusion criteria were identified from the reference lists of relevant studies and were added to the study pool.

This analysis was performed by two independent authors (ML, AM) and subsequently discussed until a final list of included studies was generated on Excel by joint consensus.

During the data extraction process, included studies were divided into two groups. The first group consisted of studies with purely digital, asynchronous interventions. The second group comprised studies that offered interventions in blended formats, combining both digital and in-person components.

From all included studies, the key information was extracted and presented in an Excel spreadsheet. This included the publication details (title, authors, year of publication, journal, location), population (number of participants, profession, age, proportion of female participants, attrition), intervention (content, delivery, duration of the intervention, frequency), outcome (what are the outcomes, how and when are they measured), study design and control group, results (resilience, stress, burnout, depression, anxiety, well-being), and correlations if investigated.

Given the significant heterogeneity between the two groups (online vs. blended), a subgroup analysis was conducted for better comparability, focusing on study characteristics, sample size and population characteristics, and intervention characteristics. A descriptive analysis was then performed to summarize the effectiveness and significate of the measured outcomes.

To assess the risk of bias, The Effect Public Health Practice Project (EPHPP) has been utilized. This involved considering selection bias, the study design, confounders, the blinding, the data collection method, and withdrawals and dropouts.

After capturing all the aforementioned factors, a global rating was created to illustrate the risk of bias.

Ethics approval was not required for this study, as no primary data were utilized, and neither humans nor animals were directly involved.

See Figure 1 for flow chart.

In total, 39 studies were assigned to the online intervention group (29–67).

A total of 16 studies were allocated to the blended group (68–83).

The Effect Public Health Practice Project (EPHPP) was used to assess the risk of bias.

Among the 55 included studies, only three studies achieved a strong global rating, while at least 12 attained a moderate global rating.

Regarding Selection Bias, all studies included a population likely to be representative of the target population, as this was an inclusion criterion from the outset. However, certain studies received a weak rating since only a specific portion of the initially approached individuals ultimately participated.

According to the inclusion criteria, all studies with an intervention design were included, regardless of whether they were randomized controlled trials (RCTs), uncontrolled trials, or pre-post intervention studies. As a result, only 23 studies received a strong rating in the study design section, as they were the only ones conducted as RCTs. The remaining studies, being either uncontrolled trials or using pre-post intervention designs, did not achieve the same rating.

Several studies were assessed as weak in the Confounders section, particularly if the study was designed with only one group, making an assessment in this regard impossible. The same applied to the Blinding section. In the Data Collection Methods, all studies demonstrated strength, as the use of a valid and quantitative measurement method was a criterion for inclusion. In the Dropouts section, one-third of all included studies had to be rated as weak.

Detailed results of the EPHPP are provided in the appendix (Effective Public Health Practice Project (EPHPP) Risk of bias assessment).

The three studies that achieved a strong global rating are summarized in the following table ( Table 10 ).

All three studies were conducted as randomized controlled trials and performed moderately to strongly in the blinding section due to single or double blinding. A large proportion of the initially approached participants were recruited for the intervention and were likely to be representative of the target population, leading to strong performance in the selection bias section. Additionally, the attrition rate was low in all three studies, which was methodologically solid. Due to the relatively high number of participants in two out of these three studies, their findings are more robust and generalizable. In terms of content, the CBT intervention achieved a significant reduction in stress, while one of the two mindfulness interventions led to improvements in resilience and burnout.

This systematic review screened existing literature on how to foster resilience in healthcare and finally included 55 studies that examined the effects of various online and blended interventions on resilience and related outcomes, designed as (randomized) controlled trials and cohort studies.

Overall, 7898 healthcare workers comprised of doctors, nurses, allied health personnel and healthcare students were able to receive interventions incorporating cognitive-behavioral therapy, acceptance and commitment therapy, mindfulness, mind-body skills, gratitude, meditation, leadership, support, positive psychology, guided imagery, and physical activities. Most studies reported immediate and significant improvements in various mental health outcomes, indicating that these resilience interventions have meaningful and lasting effects on well-being.

Some studies highlighted correlations between the frequency and time invested in training, and the extent of improvement in well-being, resilience, stress, and burnout. Furthermore, a relationship emerged between resilience and emotional distress, well-being, and empathy. This demonstrates how interconnected resilience is and how an improvement in it can influence other outcomes as well.

One strength of this review is the inclusion of student populations from diverse health disciplines. Findings suggest that fostering resilience during academic years may improve coping with stress in professional life, highlighting the interventions potential long-term benefits.

Both studies that conducted interventions purely online and those that used a blended format were able to achieve significant results in the five outcomes examined.

A significant advantage of conducting the intervention purely online is that it allows participants to engage at their own pace and at a place and time of their choosing, seamlessly integrating into their individual schedules. Furthermore, there is no need to provide a location or instructors, which represents a financial advantage.

However, the added personal component in blended interventions was appreciated. For instance, participants appreciated the personal interaction and exchange with other participants. Furthermore, the instruction of interventions could be better explained, and any open questions were promptly addressed.

Generally, the comparatively higher attrition rate in pure online intervention studies (84–88) was also evident in this review where the blended group scored higher in the risk of bias assessment regarding withdrawals and dropouts compared to the online group.

In conclusion, both online and blended interventions significantly improve psychological outcomes, each offering distinct advantages. The choice of format should depend on factors like time constraints, location, needs, and costs.

There has been previous work conducted regarding web-based resilience interventions in healthcare professions.

Three other systematic reviews were found that examined the effects of online interventions to enhance mental health in physicians, nurses, and allied healthcare staff.

Ladino et al. (20) conducted a systematic review involving four articles to examine the effects of internet-based psychosocial interventions on professional burnout. Despite a meta-analysis of two articles suggesting no significant reduction in professional burnout, and even though post-interventionally no significant differences compared to the control group could be observed, the review highlighted psychoeducation as a promising intervention for addressing this issue. It was also demonstrated in this review that psychoeducation is a way to improve mental health (29, 34, 41, 44, 45, 51, 55).

In a similar vein, Henshall et al. (26) conducted a systematic review encompassing eight studies, revealing a positive impact of web-based interventions on resilience and associated symptoms in health professionals in clinical practice settings. These results largely align with the findings of this review and illustrate the impact that online interventions can have not only on resilience, but on various psychological outcomes.

Further insights were gained from Lopez-Del-Hoyo et al. (27), who included 27 studies in their review to assess the effectiveness of eHealth interventions in reducing stress and promoting mental health. Thirteen studies reported a significant post-intervention reduction in stress, along with improvements in depression, anxiety, burnout, resilience, and mindfulness. Nevertheless, the study’s conclusions were constrained by the observed heterogeneity in the interventions. This study, which encompassed a similar volume of research, yielded comparable results to this review. However, what stood out with the inclusion of a larger number of studies, as observed in this review, is the challenge posed by the heterogeneity of interventions and measurement methods, highlighting the need for standardization to enhance comparability.

The present review contributes to the current research on interventions aimed at preserving psychological health across various healthcare professions to further progress in this rapidly evolving subject. The potential of these interventions was demonstrated through significant improvements in resilience, well-being, and the reduction of symptoms related to burnout, stress, depression, and anxiety.

In the risk of bias assessment, only three out of the 55 included studies received a strong global rating. The three strong studies demonstrated that CBT and mindfulness interventions could improve resilience, stress, and burnout. Twelve studies achieved a moderate rating, while the remaining studies were rated as weak. Consequently, the findings from the majority of studies cannot be reliably analyzed, and even statistically significant results should be interpreted with serious caution.

Notably, all studies that received a strong rating were in the online group, suggesting that the results from the blended group are only partially reliable. This reduced reliability may stem from greater difficulties in implementing blinding for in-person components, the presence of confounding factors between groups, and selection bias.

The three studies with strong ratings achieved high methodological quality using randomized controlled trial (RCT) designs, single or double blinding, low attrition rates, and minimal selection bias. In contrast, the remaining studies faced various methodological issues. Although the study populations across all studies were representative of the target population according to the inclusion criteria, some studies struggled to recruit sufficient participants, with only a fraction of the initially approached individuals participating in the intervention and consequently received lower ratings. Additionally, some studies experienced higher attrition rates, further affecting their reliability.

The data collection methods were consistently rated as strong, as they met the criteria for inclusion. Regarding study designs, all intervention studies were included, with RCTs performing better than pre-post intervention designs. Blinding, particularly in psychological or behavioral interventions, was often not feasible due to the nature of the interventions, resulting in many studies receiving a weak rating in the blinding section. The lack of effective blinding may have contributed to performance bias, where various factors, such as participants’ awareness of the intervention or the mere attention from a professional, may influenced the participants’ behavior or reported outcomes. As a result, improvements cannot be clearly attributed to the intervention itself, complicating the assessment of its true efficacy. This raises the possibility that the strong effect sized reported in some studies may be inflated by placebo effects or participant expectations rather than reflecting the genuine impact of the intervention.

Until today, there is no uniform definition of resilience. As a result, there are still no universally valid measurements methods for this. The included studies have used various tools to objectify resilience. Related symptoms such as stress, burnout, anxiety, and depression have been used as proxy indicators for resilience, which further complicates the interpretation.

This problem is accompanied by a further limitation. Due to the versatility of resilience, the included studies have implemented a variety of interventions with diverse content.

Additionally, the timing of measured changes following interventions varied significantly across the studies (post-intervention vs. follow-up). This variability suggests the possibility that some changes may have been missed if measurements were taken only after a certain period. While other studies may not provide insights into the long-term effects of the interventions. All this leads to challenges in the general interpretation and comparability of the studies. In the future, efforts should be made to establish a universally accepted definition of resilience and develop standardized, objective measurement tools. Such consensus is crucial to enhance the comparability of studies, particularly regarding the timing of assessments, which currently varies widely. A unified conceptual framework and consistent measurement would allow researchers to more accurately capture changes in resilience, compare intervention outcomes across different studies, and identify which interventions are most effective. Furthermore, standardized tools would facilitate meta-analyses and evidence synthesis, ultimately supporting the optimal design and tailoring of interventions specifically aimed at improving resilience in diverse populations.

An additional constraint to generalizability is the composition of the study populations. In the vast majority of studies, a significant proportion of participants are female. Given the findings of a recent study that approximately 77% of nurses worldwide are female (89), this is not surprising, considering that nurses constitute the largest proportion of all professions in this review. However, this gender distribution does not reflect all healthcare professions, many of which have a more balanced or different gender ratio. As a result, the findings of this review have limited generalizability across the broader range of healthcare workers and are even less applicable to the general population. To draw more robust and widely applicable conclusions about the effectiveness of resilience interventions, future studies should include samples that are more balanced and diverse in terms of gender and professional background, since intervention effects observed primarily in predominantly female nursing populations may not apply to other healthcare professions or demographic groups. Differences in occupational roles, stressors, and coping strategies could lead to varying outcomes, so caution is necessary when extrapolating these results beyond the studies groups.

Furthermore, several studies lacked a control group, which limits the conclusiveness regarding intervention effectiveness. Future studies should therefore be designed and conducted as randomized controlled trials.

The relatively high attrition rate is another limitation. Although many individuals could be initially reached, some were lost until follow-up surveys. Attrition is a well-known phenomenon especially in online interventions (90) and should be minimized in future studies.

Lastly, a significant limitation is the methodological quality constraints of many studies. Due to high attrition rates, selection bias, study designs, and challenges with blinding, only three studies achieved a strong global rating in the risk of bias assessment. Consequently, only the findings from these studies are considered decisive.

Overall, the previously mentioned heterogeneity among the studies continues to be a challenge, limiting full comparability.

To improve the validity of future studies, it is critical to address these methodological weaknesses to enhance the reliability and significance of their findings. Ensuring higher recruitment rates and lower attrition would help provide more representative results. Incorporating blinding techniques where feasible could reduce bias and lead to more reliable findings. Without these improvements, even statistically significant outcomes must be interpreted with caution, as they may be influenced by uncontrolled biases and limitations inherent in the study design.

Due to mentioned limitations, the results should be interpreted with caution until replicated.

As demonstrated in this review, numerous resilience interventions have shown a significant improvement in multiple mental health outcomes. This contributes to strengthening individual resilience and acts as a strategy against burnout symptoms and other manifestations related to stress.

Beyond structured training programs, there are various other individual-level workplace interventions that may contribute to strengthening resilience (91–96). These include, for example, employer-provided access to physical activity facilities such as on-site gyms, which may help employees enhance their mental well-being through exercise. Similarly, the availability of quiet or relaxation rooms can offer staff opportunities for mental rest and recovery during demanding workdays. Other supportive measures include designated contact persons or anonymous hotlines, which allow employees to express concerns and receive guidance in times of distress. However, such interventions tend to be reactive rather than preventive in nature and may lack the lasting effects that targeted resilience interventions aim to achieve. Moreover, their overall impact is often limited to the individual level and does not necessarily translate into broader organizational change.

However, as mentioned at the outset, there are organizational factors that cannot be solely influenced by individual actions and necessitate a broader approach. Employers and large institutions are thus encouraged to take measures, such as avoiding prolonged working hours, offering psychological support when needed, and emphasizing overall employee well-being as part of the institutional culture to contribute further to the preservation of the mental health of their workforce.

For future considerations, it is advisable to address the issue in two ways: firstly, through individual training sessions aimed at strengthening the resilience of individual persons, as seen in the included studies, and secondly, through organizational changes on the part of employers that can further positively influence mental health at the organizational level. Such measures may include implementing clear and manageable working hours, ensuring reasonable workload distribution, promoting supportive leadership and effective communication, providing access to psychological support services, and fostering a workplace culture that prioritizes employee well-being and work-life balance.

This review focused exclusively on assessing the impact of resilience interventions within healthcare personnel. It would be intriguing for future research to expand the investigation to the entire workforce, thereby revealing the potential for preserving and enhancing resilience and mental health, particularly in today’s world marked by pervasive shortages of skilled personal in various fields.