Resilience training for leaders shows promise but strains under polycrisis pressure

When leaders in two Danish organizations were put through a structured psychosocial resilience training program while simultaneously steering their workplaces through the COVID-19 pandemic, labor actions, supply-chain disruptions and demand shocks, the results revealed an uncomfortable paradox: the training worked best as a learning experience, but the very crisis conditions it was designed for made it hardest to sustain. That is the central finding of a multi-method process evaluation published in Frontiers in Psychology, which followed 48 public- and private-sector leaders through twenty weekly “resilience sprints” over nearly six months of compounding adversity.

The study, conducted by Christina Hersom of the Danish School of Education at Aarhus University and André Torre of VeldtofteStatistics in Copenhagen, addresses a striking gap in the research literature. A recent empirical review of 48 studies found no resilience training program that had been developed and evaluated specifically for professional leaders, even as organizations increasingly treat resilience as a strategic human resource management priority. The Danish researchers set out to examine not whether such training works in the narrow sense of a randomized trial, but how it is actually implemented, how its change mechanisms unfold, and which contextual factors help or hinder leaders engaging with it while crises rage around them.

The two participating organizations were chosen for the intensity of the pressures their leaders faced. The public-sector organization delivers acute, round-the-clock human-facing services, and its leaders contended with pandemic restrictions, sector-wide labor actions and mandatory compensatory time-off obligations that, while necessary given staff exhaustion, simultaneously drained staffing and increased the load on those who remained on duty. The private-sector organization produces a critical, globally demanded product in a supply-chain-dependent industry, and its leaders navigated sudden demand surges driven by international stockpiling, disrupted freight and logistics capacity, shortages of essential production inputs, and the logistical challenge of keeping production running through strict infection-control measures such as isolated shift teams. The authors describe this constellation as a polycrisis: simultaneous, entangled crises with no single causes and no simple fixes, in which decisions made to manage one pressure directly constrain the ability to respond to another.

The training itself was built on experiential learning principles. Rather than lecturing leaders about resilience, the program organized twenty composed “sprints” — weekly cycles combining a joint review of the past week’s adversities, new theoretical input, individual work with a training book, and dialogue with peers. Both organizations completed six strength-based sprints; cognitive resilience training comprised six sprints in the public-sector organization and ten in the private sector, while emotional and relational training comprised eight in the public sector and four in the private sector. The sprints drew on adversity leaders had directly experienced in the preceding week as well as anticipated “what if” scenarios, deliberately pushing participants beyond linear thinking. Internal HR resilience trainers were educated in advance on the locally designed curricula, an approach the authors link to evidence that trainers who use resilience skills themselves foster trust and safety among participants.

To evaluate the program, the researchers applied the Medical Research Council framework for complex interventions, breaking the assessment into three interdependent components: implementation, mechanisms of impact, and context. Implementation was measured with a tailored four-item questionnaire targeting the structural parameters of the sprint format. The results were telling. Nearly half of the 35 leaders who completed the post-evaluation — a 97 percent response rate — agreed that the internal distribution of sprint elements was appropriate, making the session structure the strongest component of the design. But agreement collapsed on the outer parameters: only 20 percent found the total number of twenty sprints appropriate, and just 31 percent considered the roughly one-week interval and the 60-minute session duration suitable. A substantial share of respondents, between 20 and 37 percent across items, chose the neutral midpoint, suggesting genuine uncertainty rather than outright rejection of the format.

The mechanisms of impact were assessed weekly using an adapted version of the Kirkpatrick evaluation model, which distinguishes effective training — immediate reaction and learning — from training effectiveness, meaning observable behavior change and results. Across 42 leaders who completed the twenty sprints, the pattern was lopsided: 78 percent of evaluations fell at the reaction and learning levels, while only 21 percent reached the behavior and results levels. The two organizations diverged in how they experienced the training. Public-sector leaders’ evaluations clustered at 31 percent reaction and 54 percent learning, with just 4 percent at the results level, whereas private-sector leaders recorded 46 percent reaction, 29 percent learning, and a comparatively high 15 percent results. A chi-square test confirmed that evaluations of mental resilience differed significantly between the organizations (p < 0.0001), while strength resilience showed no significant difference (p = 0.3169) and relational resilience approached but did not reach conventional significance (p = 0.0582).

The authors offer a candid interpretation of this imbalance. Several participants struggled to recognize or self-evaluate their own resilience development even when they could articulate newly applied strategies in conversation with peers. One public-sector leader, hesitating in front of an evaluation poster, asked: “But… how do I dare to score it? How do I know if I have developed my resilience?” The researchers suggest that the concentration of evaluations at the lower Kirkpatrick levels may partly reflect the difficulty of understanding and self-assessing psychosocial resilience, not merely the absence of behavior change. Reaction and learning can be reported relatively directly, they note, whereas judging one’s own resilience development demands a more reflective form of self-assessment that does not come easily to everyone — a challenge compounded because resilience remains a relatively new concept in leadership development.

To probe what might happen if the training were extended, the team applied a Markov chain model, treating the four Kirkpatrick levels as sequential states and estimating transition probabilities between them. The long-run projections, which the authors stress should be read as indicative rather than robust forecasts, remained concentrated at the reaction and learning levels even under indefinite continuation of the training. For strength resilience, the stationary distribution settled at roughly 32 percent reaction, 41 percent learning, 13 percent behavior and 14 percent results; for mental resilience, 52 percent reaction, 32 percent learning, 7 percent behavior and 9 percent results. For relational resilience, the model produced what appeared to be an absorbing state at the results level, but the authors flag this explicitly as a modeling artifact: no transitions out of Level 4 were observed because of missing data, and the pattern should not be interpreted as evidence that participants remained at that level. The analysis relied on a 38 percent response rate, with transitions involving missing responses excluded, and the assumptions of stationarity and the first-order Markov property could not be formally tested with the available data.

The contextual findings may be the most consequential for human resource practice. Agreement was consistently high that being gathered as leaders for training worked well (63 percent at both measurement points), that colleagues accepted leaders leaving for sprints (60 percent, falling only slightly to 57 percent), and that the atmosphere for sharing knowledge was free and open (74 percent rising to 77 percent). Trainer commitment emerged as a key facilitator: the external trainers who onboarded the internal HR trainers were rated as engaged by 85 percent of participants, and internal trainer commitment rose from 60 percent to 77 percent agreement over time. Against these positives stood a persistent barrier: only about one third of leaders agreed they could prioritize the training without significantly disrupting their work, and this did not improve over the program. Motivation to develop resilience, initially reported by 60 percent of participants, dropped to 40 percent by the end. The perceived suitability of the training’s timing also eroded, from 51 percent agreement to 34 percent, with 43 percent of participants shifting to a neutral position.

The authors interpret this erosion through the lens of an experiential learning paradox they call “training to work” versus “working to train”: leaders were asked to practice resilience in real time while simultaneously upholding operational responsibilities, and when the two collided, the potential of experiential resilience learning was constrained. From a Self-Determination Theory perspective, voluntary participation and peer-based learning supported autonomous motivation, but competing demands and the pressure to train may have undermined it — a pattern with parallels in longer interventions generally. The researchers also point to broader evidence that those who would benefit most from resilience training are often the least able to engage consistently because of workload and stress. Notably, they caution against a simple fix: mandating training, as in military programs such as the U.S. Comprehensive Soldier Fitness initiative, could undermine autonomous motivation in civilian contexts and deepen the very tension the training is meant to relieve.

The study’s limitations are acknowledged plainly. It included leaders from only one private-sector and one public-sector organization in a single country; the analytic samples of 35 to 42 participants were small relative to the number of subgroup comparisons; the tailored questionnaires were face-validated but not psychometrically validated; and Kirkpatrick Level 4 rested on leaders’ self-report rather than independently observed behavior or a validated resilience measure. Attrition was also substantial, with participation declining from 48 leaders at pre-intervention to 41 during and 36 after the program, an overall rate of 25 percent. Within those constraints, the authors distill seven practice considerations for HR partners, including building adaptive slack into sprint content and cadence, protecting time for training rather than individualizing a shared organizational vulnerability, preserving the psychologically safe peer forum as a resilience resource in its own right, lowering reflective demands during crisis peaks, setting realistic expectations for when behavior change becomes visible, offering brief resilience psychoeducation before the first sprint to strengthen self-report validity, and — given that 49 percent endorsed the internal structure of sprints but only 20 percent endorsed their number — reducing sprint frequency without sacrificing the content leaders valued most. The overarching message is that a perfect-fit resilience design may be unrealistic in high-demand, continuously changing leader contexts; instead, resilience interventions must themselves function as adaptive systems capable of reorganizing in rapid response to emerging challenges, much like the leaders they aim to strengthen.

Subject of Research: Process evaluation of HR-driven psychosocial resilience training for leaders in two Danish organizations during polycrisis conditions

Article Title: Preparing to lead in polycrisis: a multi-method process evaluation of HR-driven psychosocial resilience training for leaders

Article References: Hersom, C., & Torre, A. (2026). Preparing to lead in polycrisis: a multi-method process evaluation of HR-driven psychosocial resilience training for leaders. Frontiers in Psychology, 17, Article 1900181. https://doi.org/10.3389/fpsyg.2026.1900181

Image Credits: AI Generated

DOI: 10.3389/fpsyg.2026.1900181

Keywords: leader resilience, psychosocial resilience training, polycrisis, human resource management, Kirkpatrick model, process evaluation, experiential learning, crisis leadership, organizational resilience, Markov chain model, workplace training, Denmark