Psychological resilience factors in intensive care nursing: a hybrid multi-criteria decision-making model

Initially, the research system on psychological resilience was based on a study involving hospital nurses during the COVID-19 pandemic. Next, questionnaires were designed and collected from ICU nurses in the case hospitals, taking into account the DEMATEL, DEMATEL-based DANP, and IPA methods. These questionnaires aimed to assess the mutual influence between factors, as well as the self-assessment of psychological resilience based on the clinical experience of ICU nurses. Lastly, the key influencing, promoting, and improvement factors of ICU nurses' psychological resilience during COVID-19 were analyzed using the aforementioned methods. The results can be utilized by hospital decision-makers to propose strategies for improvement. Please refer to Fig. 1 for an illustration of the decision modeling and analysis processes of this study.

In early 2022, an outbreak of COVID-19 occurred in Shanghai, China. During the outbreak, Jiang, Liu [30] employed phenomenological research methods from qualitative studies to explore the psychological resilience of emergency nurses in Shanghai. They conducted in-depth semi-structured interviews and used Colaizzi's seven-step method for data collection and analysis. Finally, they constructed the psychological resilience factor model. The model of this study is shown in Table 1.

In this study, the DEMATEL method was used to more clearly demonstrate the dependency relationship among various factors. The DEMATEL method was developed by the Battelle Memorial Institute from 1972 to 1976, wherein the relationships between various factors in complex systems were analyzed using charts and rectangular tools [31]. It can transform the dependencies between the variables in a system into specific numbers and present the results in the form of charts. This can help decision-makers focus on core impact dimensions or criteria [32]. Researchers can make better decisions by using this method to determine the interactions and interdependence between risk factors [33]. For example, Zhu, Hu [34] developed an international construction project risk assessment model that used the DEMATEL method to determine the interdependence of risks within the overall project. Similarly, Dehghani, Hormozi [35] used the DEMATEL method to determine the key risk factors associated with the construction process, adding a description of the degree of interdependence compared with previous research methods.

The detailed calculation steps of the DEMATEL method are as follows [32, 36, 37]:

Step 1: Quantification of the level of influence between the criteria. A five-point Likert scale is used, ranging from no influence (0) to very high influence (4). All ICU nurses quantify the influence level of interactions between all criteria based on their clinical practice experience. Finally, the individual experiences of all ICU nurses are averaged to form an influential matrix (J) representative of the overall sample, as shown in Eq. 1.

The matrix p _ij represents the empirical matrix of the degree of influence criteria interaction for a given ICU nurse. The parameter ij indicates the degree of influence of criterion i on criterion j. The variables n and z represent the total number of criteria and respondents, respectively, for all ICU nurses.

Step 2: Conversion to proportional form between 0 and 1. The matrix J is transformed into proportional form using Eq. (3), after determining the maximum influence boundary value as the threshold value using Eq. (2).

Step 3: Step 3: Derivation of the overall level of influence relationship. The matrix K in Eq. (4) is utilized to derive the overall level of influence relationship between all criteria, encompassing both direct and indirect influence levels.

Step 4: Construction of the cause-effect impact network diagram (CEIND). The matrix s is computed using Eqs. (5) through (8), incorporating impact (\({\textit{r}}_i\)), affected (\({\textit{u}}_\textit{i}\)), centrality (\({\textit{r}}_i+{\textit{u}}_\textit{i}\)), and causality (\({\textit{r}}_i-{\textit{u}}_\textit{i}\)) for each criterion.

Based on the influence and affected viewpoints, four influence indicators are derived from the total influence relation matrix: influence, affected, influence centrality, and influence causality, as demonstrated in Eqs. (5, 6, 7 and 8).

Impact:

Affected:

Centrality:

Causality:

In the centrality (\({\textit{r}}_i+{\textit{u}}_\textit{i}\)) matrix, the larger the value, the more relevant the criterion is in the overall interaction and the greater the degree of influence on the system. In addition, in causality (\({\textit{r}}_i-{\textit{u}}_\textit{i}\)) matrix, a positive value indicates that the influencing attribute is more pronounced and is referred to as the “cause”; conversely, a negative value indicates that the influenced attribute is more pronounced and is referred to as the “effect.”

The DANP method is a novel weighting method that combines the DEMATEL method with the concept and calculation of the ANP. Unlike the ANP method, the DANP method considers the interaction among all indicators and avoids assuming equal weights among the dimensions/clusters [38, 39]. This method assists decision-makers in effectively reducing decision complexity and has been successfully applied in various fields. For instance, Liu, Lin [40] utilized a new hybrid modified multiple criteria decision-making model to enhance the selection of optimal disposal facility locations. Similarly, Zhang, Qin [41] identified the key influencing factors of public-public cooperation in food safety risk management using the DANP method. Additionally, Duman, Taskaynatan [42] applied a gray-DANP model within a balanced scorecard framework to evaluate the environmental and social sustainability of the food industry. The detailed calculation steps of the DANP method are as follows [43]:

Step 5: Construction of an unweighted supermatrix.

The matrix \(\textit{S}\) can be distinguished into criterion level \({\textit{S}}_\text{c}\) and dimension level \({\textit{S}}_\textit{D}\), where \({\textit{S}}_\textit{c}\) is converted into a proportional matrix by Eqs. (9, 10). Then \(\textit{S}\) is transposed to obtain an unweighted supermatrix \(\textit{A}=\left(\textit{S}_C^\varepsilon\right)\),

where \(S_{{}_c}^{\varepsilon 11}\) can be obtained by Eq. (10). Then, by repeating the same calculation method, other sub-matrices (i.e. \(S_{{}_c}^{\varepsilon ij}\)) can be obtained,where \(x_i^{11} = \sum\nolimits_{j = 1}^{m_1} {s_{_Cij}^{{}11}} , \, i = 1,2,...,{m_1}\) 

Step 6: Establish a weighted supermatrix. A weighted supermatrix \(W\) can be obtained after the matrix \(S_D\) is passed through Eqs. (11, 12),where \({o_i} = \sum\nolimits_{j = 1}^m {s_D^{ij}} {, }i = 1,2,...,m\) and \(s_D^{ \otimes ij} = s_D^{ij}/{o_i}, \, j = 1,2,...,m\) .

Step 7: Determine the influential weights of all criteria. After the matrix \(W\) is limited, the weight value of each criterion can be obtained using Eq. (13).

The IPA method was used in this study to estimate the influence of individual resilience factors on nurses' importance and productivity outcomes. This method helps identify priorities for managerial intervention and optimize resource utilization. The IPA method was first proposed by Martilla and James in 1977 [44]. It is a classic performance evaluation and analysis technique used in management decision-making. The steps of the method are as follows:

Here is the basic description of the IPA quadrant diagram used in this study [44, 45]:

The questionnaire consisted of two parts, both based on the psychological resilience factor model (Table 1) and containing different content. The purpose of the first part of the questionnaire is to evaluate the interplay and impact of external factors, while the purpose of the second part is to assess individuals' intrinsic reactions and adaptive capacities to these factors. In the first part, each participant scored the degree of interaction between each factor/criterion, ranging from very low impact (0) to very high impact (4). The scores were then calculated using the DEMATEL method to obtain the cause-effect impact network diagram (CEIND), and the weight of the influence relationship was obtained using the DANP method. In the second part, each participant was asked to assess the resilience of each factor/criterion in their self-assessment, and self-cognition was scored from always (1) to never (5). The questionnaire survey was completed between January and February 2023.

This study was approved by the ethics committee of Taizhou Hospital of Zhejiang Province Affiliated to Wenzhou Medical University (Approval number: K20230215). All participants agreed to participate in the study, and their relevant personal information remained anonymous.

Data were collected from 35 ICU nurses. Thirty-two of the ICU nurses were women (91%). Those with an undergraduate education accounted for 97% of the total population. Most of the study participants fell in the age range of 30–39 years, totaling 19 individuals (54%). Ten individuals were under the age of 30 years, accounting for 29% of the total. The least represented age group was 40–49 years, with six participants (17%). Regarding clinical nursing experience, 49% of the nurses had less than 10 years of experience, whereas 51% had more than 10 years. Notably, all of the ICU nurses gained clinical nursing experience managing COVID-19 during the pandemic. Of these, 26 ICU nurses (77%) had more than 37 months of clinical experience in the ICU since the COVID-19 pandemic. The remaining eight ICU nurses had six to 36 months of clinical nursing experience since the onset of the COVID-19 pandemic. Information regarding the participants' backgrounds and personal characteristics is provided in Table 2.

This research utilized the DEMATEL method to construct an average interaction relationship matrix (Table 3). The matrix was developed based on input from 35 clinical ICU nurses who have experienced the COVID-19 pandemic. The confidence level of the obtained result was 99.6%, with a gap of 0.4%, exceeding the 95% threshold and indicating good stability. Tables 4 and 5 present the four influence indices and corresponding influence weights for each factor/criterion, as determined using the DEMATEL and DANP methods.

The “centrality (r _i + u _i)” index displays the intensity of influence of each factor/criterion within the entire model. According to this analysis, (C ₁) has the highest centrality value compared to the other two dimensions. Under their respective dimensions, (C ₁₁), (C ₂₂), and (C ₃₂) were the factors/criteria with the highest centrality values. In addition, the “cause (r _i-u _i)” index indicates the relative influence of each factor/dimension/criterion. A positive value indicates that the factor or criterion exerts influence on others, while a negative value shows that it is influenced by others. Notably, (C ₂) shows a positive value in its dimension compared to the other dimensions. Within other dimensions, (C ₁₁), (C ₂₁), (C ₂₂), and (C ₂₃) also show positive values, while the remaining criteria display negative values.

“Local Weight” refers to the relative importance of an individual factor within its directly related factors. It reflects the influence and importance of a factor within a specific local environment or context. On the other hand, “Global Weight” represents the comprehensive importance of a factor across the entire system. It takes into account not only the direct influences between factors but also the indirect effects produced through other factors [46].

Table 5 presents the influence weights of all the psychological resilience factors determined by the DANP method. At the dimension level, (C ₁) is the highest-ranking factor, followed by followed by (C ₂) and (C ₃). At the local level, (C ₁₁), (C ₁₂), and (C ₃₂) hold the largest weights within their respective dimensions. At the global level, the top three factors are (C ₃₂), (C ₃₁), and (C ₁₁). Therefore, it can be inferred that (C ₁), (C ₁₁), (C ₂₂), and (C ₃₂) are the key influencing factors and directions that affect the resilience of ICU nurses in the context of COVID-19.

The IPA location map is divided into four quadrants by two perpendicular lines intersecting at the coordinates x = 0.111 and y = 1.964, which are calculated by averaging the Performance and Importance values. Using SPSS software (version 26), the coordinates of each indicator were imported into the coordinate system and assigned to one of the four quadrants. The results showed that (C ₁₂), (C ₃₁), and (C ₃₂) were located in the first quadrant; (C ₂₁), (C ₂₂), (C ₂₃), and (C ₂₄) were in the second quadrant; no indicators fell into the third quadrant, and (C ₁₁) and (C ₁₃) were positioned in the fourth quadrant. The detailed results are presented in Table 5.

This study established an interaction structure between dimensions and criteria of the resilience evaluation system and quantified the key factors influencing the resilience of ICU nurses during the COVID-19 epidemic. Figure 2 highlights that (C ₂), (C ₁₁), (C ₂₂), and (C ₃₂) are the key promoting factors for improving the psychological resilience of ICU nurses.

Patients with severe COVID-19 often require various nursing procedures and life support technologies, including disease monitoring, specimen collection, treatment plan implementation, and meticulous care. These interventions may involve artificial airways, prone ventilation, and extracorporeal membrane oxygenation (ECMO) [47]. Given their essential role on the frontline, ICU nurses play a vital role in treating patients with severe cases of public health events such as COVID-19 [48]. During the COVID-19 pandemic, ICU nurses are responsible for closely monitoring patients' vital signs, observing their respiratory and circulatory status, managing artificial airways, conducting renal replacement therapy, caring for patients on ECMO, providing prone ventilation, and preventing a range of complications such as ventilator-associated pneumonia, bloodstream infections, vasculitis, deep vein thrombosis, and pressure ulcers [49]. In medical these medical contexts, the unique demands of COVID-19 have increased medical demands, and new policies and care procedures have also multiplied the workload of medical staff [50].

In terms of (C ₁₁), the increased workload on medical staff often leads to distractions, resulting in a higher likelihood of errors, such as drug delivery mistakes. This elevated workload also exerts additional mental pressure on medical staff [51]. Regarding (C ₂₂), research by Vilete, Figueira [52] highlights the critical importance of social support for nurses' resilience. This study aligns with previous findings and provides further quantitative insights into the significance of social support. Moreover, concerning (C ₃₂), Tomietto, Paro [53] demonstrate that the realization of self-worth can increase nurses' well-being and work engagement.

Nursing managers should recognize the importance of reducing unnecessary workloads and providing special periods focused on ICU nurses' realization of self-worth. These measures are crucial for promoting both well-being and resilience. Such support includes, but is not limited to, unified training on the correct wearing of protective measures by nurses, supporting ICU nurses in developing personal support networks, and encouraging them to participate in continuing education.

In Table 5, (C ₁), (C ₃₂), (C ₃₁), and (C ₁₁) were identified as the key influencing factors of resilience and are also the key influencing factors of ICU nurses’ overall psychological resilience. During the COVID-19 pandemic, many nurses involved in treatment lacked coping experience or psychological preparation. Regarding (C ₁₁), several nurses had insufficient knowledge of professional isolation and protection measures, which put them at risk of exposure or infection. Additionally, strict protective measures such as wearing N95 masks can affect breathing and communication. Goggles and protective covers may also limit nurses' field of vision to some extent, and protective clothing can reduce their flexibility and increase mobility burdens. In their research, Huang, Lin [50] and Galletta, Piras [54] demonstrate that providing intensive self-protection education and training for nurses, implementing a scientific and reasonable scheduling system, and establishing more flexible and adjustable policies can reduce hospital infections, thereby alleviating nurses' psychological pressures.

Regarding (C ₃₁), many individuals are drawn to the nursing profession out of a desire to help others and promote healing. However, research conducted by Aiken, Clarke [55] reveals that hospitals with higher nurse-to-patient ratios have a greater risk-adjusted 30-day mortality and failure-to-rescue rates. This situation makes nurses more susceptible to job burnout and dissatisfaction. When nurses maintain a connection between their intrinsic motivation and altruistic values, they are better equipped to handle the challenges of their profession.

Regarding (C ₃₂), a high-pressure work environment is more likely to impact an individual's mood [56]. In high-stress, high-workload settings such as the ICU, a strong sense of self-worth can help mitigate the impacts of stress and burnout. Orth, Robins [57] confirmed the significance of self-worth, as their findings indicated that it influences an individual's lifelong happiness, health, and social interactions. Moreover, self-worth is positively associated with psychological resilience, and a strong sense of self-worth aids in the recovery of psychological resilience.

Based on the analysis results in the IPA quadrant diagram (Fig. 3), it can be identified that (C ₁₁) and (C ₁₃) are the key improvement factors reflecting ICU nurses’ psychological stress resistance ability in case hospitals. These factors are also important for ICU nursing managers to focus on. However, it should be noted that while these factors are important, they may not be the only key factors in improving psychological resilience. Therefore, from CEIND's perspective, we find that (C ₁₃), as a key improvement factor, is also affected by (C ₁₁). Therefore, nursing managers should give priority to (C ₁₁).

Due to the unique circumstances of COVID-19, many patients lack family members and nursing workers to provide basic life care. In addition to their necessary rescue and treatment work, nurses also have the added responsibility of providing life care for patients, which significantly increases their workload. Hospital managers should take proactive measures to create a supportive work environment throughout the hospital, increase social support for nurses, and reduce stress and job burnout. Furthermore, policies and practical actions should be implemented to provide psychological and physiological support to staff on the frontlines of anti-epidemic clinical work.

Firstly, hospital managers should provide support to frontline clinical staff in terms of resources, workforce, and psychological support. It is crucial to conduct comprehensive and strict training on protective measures to minimize risks and pressures faced by nurses due to errors in protective operation [50]. Secondly, doctors and nurses from other departments who are dispatched to assist should receive appropriate training in advance to reduce the complexity of their work and alleviate additional work pressure on the original department staff who may be unfamiliar with operating procedures. Additionally, hospital managers should establish a more scientific and reasonable scheduling system for special periods, formulate flexible and adjustable policies, ensure that ICU nurses receive sufficient rest, reduce work pressure and job burnout, and enhance the psychological resilience of ICU nurses.