Previous research indicates that companies investing in improving job resources (e.g., good team spirit and high level of coordination; Salanova et al., 2012) impact positively: both on levels of resilience, efficacy, and engagement as on organizational outcomes. The same relationship applies to personal resources, both on healthy employees (e.g., mental and emotional competence; Tripiana and Llorens, 2015) as on organizational outcomes (e.g., vertical and horizontal trust; Guinot and Chiva, 2019; Salanova et al., 2021). Moreover, both types of resources (personal and job) can act as mediators of various indicators related with occupational well-being (Gómez-Borges et al., 2022, 2023). HOPs (i.e., work-family balance, mobbing prevention, psychosocial health and organizational communication) have demonstrated that they improve the productivity of the teams (Salanova et al., 2021). On the other hand, specific HOPs, aiming at mobbing prevention, promotion of communication and information, and skills and career development, allow the growth of a higher positive group affection (Peñalver et al., 2023) and collective empathy (Solares et al., 2016). However, hardly any studies are available on HOPs, specifically focusing on equality and gender diversity management. Therefore, these HOPs have been included in this study as a differentiated resource factor, putting to the test an extension of the model.

Previous research has provided evidence that both employees’ health as their organizational outcomes are affected by the combination of high demands and lack of resources to meet them (Bakker and Demerouti, 2024), both at the individual (Salanova et al., 2019) and at the organizational level (Llorens et al., 2022). In particular, the perception of workers’ specific demands (i.e., mental overload) can determine the resources they will seek and use to meet them (i.e., mental competence). Possessing the necessary resources to meet a moderate demand allows to perceive this demand as a challenge instead of a threat, leading to better coping with it which, in turn, will result in healthier employees and organizational outcomes (Llorens et al., 2022; Salanova et al., 2019; Ventura et al., 2015). This is important since, as can be seen below, women tend to have a higher capacity to use resources, enabling them to better cope with demands (Cifre et al., 2011).

Previous literature exists, indicating specific gender-related differences in the variables that constitute the HERO model. In terms of resources and HOPs, some may be more necessary for a specific gender. For example, HOPs promoting autonomy have shown higher efficacy when it comes to promoting organizational outcomes in women (Gervais and Millear, 2014, 2024). With regard to demands, women and men may experience differences in demands, due to gender roles or to the position they occupy within the company (Cifre et al., 2000). For example, moderate role ambiguity can be positive for men’s job satisfaction, but not for women’s, due to their different motivations. In particular, role ambiguity only seems to be a predictor of job satisfaction in the male sample. The fact that role conflict also has a positive impact suggests that being submitted to role stress (role conflict and ambiguity) is motivating for men, since it brings along increased job activity (Cifre et al., 2000). On the other hand, men tend to perceive more mobbing and, since they possess a lower capacity to cope with it, this can have more negative consequences than in women (Cifre et al., 2011). Another example is the work–family conflict that men and women perceive differently, as the patriarchal system leads women to face more pressure due to caregiving duties, while men experience stress related to breadwinner expectations (Alcañiz, 2017). In terms of healthy employees, stress-derived symptoms are different between men and women in the same position (Ornek and Esin, 2020), since the job-related demands and the resources to meet them are different as well (Ornek and Esin, 2020). Finally, with regard to organizational outcomes, it must be noted that women tend to achieve higher scores in “job performance measures,” however, “ratings of promotion” are higher in men (Roth et al., 2012). Therefore, any measure aimed at improving organizational outcomes can be more effective if it is implemented from a gender perspective.

Despite recent studies concluding the impact of HERO variables such as resources (Gómez-Borges et al., 2022) and Healthy Organizational Practices (HOP) (Peñalver et al., 2023) across various contexts, including the validation of the HERO model from a multilevel perspective (Salanova et al., 2021), which highlight the effects on well-being and performance, very few studies have considered the gender perspective. This study allows putting to the test an extension of the HERO model by including this perspective. In this manner, it will constitute a key tool in the management of health and outcomes in organizations, especially those with high gender diversity, since it allows the study of gender-specific predictors for well-being and outcomes in organizations. In particular, the objective of this study is to identify the main predictors (resources, healthy organizational practices, and demands as novelty) for healthy employees and healthy organizational outcomes from a gender perspective. Considering this objective, the following hypothesis are expected:

The general study sample consisted of 3,863 employees from 8 Spanish companies (57% women and 43% men) from two different sectors (90% from the tertiary and 10% from the secondary sector). Missing value statistics were explored showing a significatively higher proportion of participants with missing data among men [proportion of men with missing data: 54.4%; proportion of women with missing data: 28.1%; χ²(1) = 132.97, p < 0.000]. The average of missing items per person was 3 for men against 1 for women. In order to keep a balanced sample between men and women, a multiple imputation method was performed through chained equations (Wulff and Jeppesen, 2017) using the R Studio MICE package (Van Buuren and Groothuis-Oudshoorn, 2011). After the imputation, the final sample consisted of 2,128 employees (70.3% women, 30.5% between 36 and 45 years old) representing various economic sectors, with 84% affiliated with the tertiary sector (specifically, hospitals; 80.4% women) and the remaining 16% with the secondary sector (100% men). Table 1 shows the results of the description of women and men taking into account the age.

Data were collected from participants through an online survey uploaded onto the Qualtrics platform. A survey invitation was sent via email. The information about the project was presented in three stages for all organizations. First, an introductory meeting took place with the CEOs; second, a meeting involving managers and the HR department. In the third meeting, the data collection process was planned. The HR departments arranged suitable locations and groups for employees to complete the questionnaires in shifts. Prior to answering the questionnaires, anonymity was ensured for all participants and the informant consent agreement was provided for signature on voluntary basis.

Healthy Organizational Practices were assessed at the organizational level with 19 items included in the HERO-CHECK questionnaire (Salanova et al., 2012, 2019). Subsequently, 8 healthy organizational practices (HOPs) —related to gender and other diversities— were considered in the analyses, each with one item: work-family balance, equity, equal opportunities, maternity/paternity, skills and talent, leadership equality, culture diversity, and training diversity. We used a 7-point Likert-type scale ranging from 0 ‘never’ to 6 ‘always’. An example of item is ‘[In the last year, practices and strategies have been introduced in this organization in order to expand the requirements of current legislation to enjoy maternity/paternity benefits (e.g., low extension, flexible hours)]’ (𝞪 = 0.95; 𝟂 = 0.95).

Job and personal resources were assessed with 9 items included in the HERO-CHECK questionnaire (Salanova et al., 2012, 2019). These items were: autonomy, feedback, supportive climate, coordination, mental competence, emotional competence, positive leadership, vertical trust, and horizontal trust. Respondents answered using a 7-point Likert-type scale (from 0 ‘strongly disagree’ to 6 ‘strongly agree’). An example of item is ‘Degree to which they have sufficient control to decide the tasks they will carry out during the day, the number of tasks, the order in which they will be carried out, and the moment they will start and/or finish them’ (𝞪 = 0.88; 𝟂 = 0.91).

Job demands were assessed with eight items, based on the HERO-CHECK questionnaire (Salanova et al., 2019; Villarroel-Núñez et al., 2024). These items were the following: quantitative overload, mental overload, emotional overload, role ambiguity, role conflict, routine, mobbing, and emotional dissonance. Respondents answered using a 7-point Likert-type scale (from 0 ‘strongly disagree’ to 6 ‘strongly agree’). An example of item is ‘Degree to which the amount of work to be done “overwhelms” them, either due to lack of time or due to excess tasks’ (𝞪 = 0.81; 𝟂 = 0.81).

Healthy employees were assessed with 4 items, based on the HERO-CHECK questionnaire (Salanova et al., 2012, 2019). These items were: self-efficacy, work engagement, resilience, and burnout. Respondents used a 7-point Likert-type scale (from 0 ‘strongly disagree’ to 6 ‘strongly agree’). An example of item is ‘Degree to which they believe in the abilities of their work team to carry out tasks successfully despite obstacles’ (𝞪 = 0.77; 𝟂 = 0.79).

Healthy organizational outcomes were assessed using three items from the HERO-CHECK questionnaire (Salanova et al., 2012, 2019). These items were: in-role performance, extra-role performance, and commitment. Respondents used a 7-point Likert-type scale (from 0 ‘strongly disagree’ to 6 ‘strongly agree’). An example of item is ‘Degree to which the tasks of their work are carried out and fulfilled, those that are prescribed in their employment contract’ (𝞪 = 0.46; 𝟂 = 0.54). The item of extra-role performance was deleted in order to improve the internal consistency of the scale (𝞺 = 0.44, p < 0.001).

Descriptive data analysis (means, standard deviations per gender) was performed using R 1.4.2 (R Core Team, 2021). Scale reliability was estimated with Cronbach’s alpha (𝞪) and MacDonald’s Omega (𝟂) total using Psych R package (Revelle, 2017). Gender differences were tested on each item. First, the F-test for equality of variances for checking homoscedasticity among genders was used. Depending on the homoscedasticity results, t test analysis was performed for gender groups (with Welch approximation to the degrees of freedom if variances are not equal). Due to the large sample size, normality was assumed on the distribution without compromising results (Pardo Merino and San Martín, 2010).

To ensure the robustness and validity of the theoretical model, a Confirmatory Factor Analysis (CFA) was conducted using the lavaan R package (Rosseel, 2012). The CFA was employed to verify the factor structure of the proposed model and assess the potential influence of common method variance through Harman’s single-factor approach (Podsakoff et al., 2003). The estimation method utilized was Maximum Likelihood with Satorra-Bentler correction (MLMV), which provides more reliable results, especially in situations where data may not perfectly conform to normality (Maydeu-Olivares, 2017). Additionally, to ensure the independence of variables, collinearity among items was assessed using the Variance Inflation Factor (VIF), following guidelines by Kock (2015).

The CFA results were tested by considering three absolute goodness-of-fit indices to evaluate the goodness-of-fit of the models: (1) the chi-squared (χ²) goodness-of-fit statistic, (2) the Root Mean Square Error of Approximation (RMSEA), and (3) the standardized root mean square residual (SRMR); and two relative goodness-of-fit indices: (1) Comparative Fit Index (CFI) and (2) Tucker-Lewis Index (TLI, also called the Non-Normed Fit Index). For RMSEA and SRMR, values smaller than 0.05 indicated an excellent fit, 0.08 indicated an acceptable fit, and values greater than 0.10 led to model rejection (Browne and Cudeck, 1993). Regarding the relative fit indices, values greater than 0.90 indicated a good fit (Hu and Bentler, 1999).

Finally, a multigroup structural equation model (SEM) was tested using Lavaan R package (Rosseel, 2012) to determine which predictors (i.e., Healthy Organizational Practices, Job demands and Job Resources) are related to Healthy employees and Healthy organizational outcomes. Healthy employees and Healthy organizational outcomes were included as latent factors, operationalized by four (self-efficacy, engagement, resilience, and burnout) and two (i.e., in-role performance, and commitment) items, respectively (Salanova et al., 2012; Villarroel-Núñez et al., 2024). Predictor items were included as exogenous observed variables in the model. Maximum likelihood with Satorra-Bentler correction (MLMV) was used as an estimation method (Maydeu-Olivares, 2017). Absolute and relative indexes of goodness-of-fit (GoF) (Marsh et al., 1996) were used to test the adequacy of the model: chi square statistic, Root Mean Square Error of Approximation (RMSEA), Standardized Root Mean Square Residual (SRMR), Tucker-Lewis Index (TLI), and Comparative Fit Index (CFI). Values smaller than 0.08 for RMSEA and SRM, and greater than 0.90 for TLI and CFI indicate an acceptable fit (Marsh et al., 2004).

To test the invariance of the proposed model across genders, we followed the free baseline approach (Stark et al., 2006) that compares models with different levels of constraints for parameter estimation between groups and a baseline model with no constraints (Bentler, 2000). First, we tested measurement invariance following Meredith’s classification (1993) (Meredith, 1993). Subsequently, we compared the unconstrained baseline model (M1) with (a) a model with restricted factor loadings (M2) to test weak factorial invariance; (b) a model with constrained factor loadings and intercepts (M3) that would support strong factorial invariance; (c) a model with constraints in factor loadings, intercepts, and measure residuals (M4), testing strict invariance; (d) a model with constraints in factor loadings, intercepts and measure residuals, and factor means (M5) for testing structural invariance and finally, (e) a model with constraints in factor loadings, intercepts, residuals and regressions across gender groups. Strong factorial invariance is considered sufficient for testing predictions in latent variables across groups (Newson, 2023).

To evaluate the magnitude of change between the baseline model and the constrained ones, the chi-square difference test between reproduced models was estimated, as well as the increments in RMSEA and CFI indexes (Chen, 2007; Hu and Bentler, 1999). Low changes in CFI, TLI, RMSEA, and SRMR indexes (usually accepted |Δ < 0.01|) indicate parameter invariance among groups (Cheung and Rensvold, 2002).

Descriptive statistics, statistical differences among genders, and reliability analysis results are displayed in Table 2 to test Hypothesis 1. Cronbach’s alpha and McDonalds’ omega values show excellent results (above 0.70) (Campo-Arias and Oviedo, 2008).

Significant differences between genders were found for most items. In terms of demands, women showed significantly higher emotional overload, while men experienced higher role ambiguity and routine. Women showed statistically higher levels of resources and organizational practices, except for emotional competence. In addition, scores for wellness and performance items were higher for women than for men. In general terms, these results confirm our Hypothesis 1.

Harman’s single-factor test result indicates that the single-factor solution does not explain the variance in the data [χ²(495) = 18106.39, p < 0.000; CFI = 0.59; TLI = 0.56; RMSEA = 0.13; SRMR = 0.13], so common method variance is not a serious problem in this study [∆χ² (10) = 2727.6, p < 0.01; ∆CFI = 0.28; ∆TLI = −0.30; ∆RMSEA = −0.06; ∆SRMR = −0.06]. Additionally, VIF results were below the 3.3 cut-off value (Kock, 2015), except for the items evaluating practices in equal opportunities (VIF = 4.295), maternity/paternity (VIF = 3.547), skill and talent (VIF = 5.223), leadership equity (VIF = 4.476), and culture diversity (VIF = 5.091). However, only culture and diversity slightly exceeded the alternative threshold of 5 proposed by Ringle et al. (2015).

The results of the CFA factor model show that the model with 5 dimensions (i.e., job demands, job and personal resources, organizational practices, healthy employees and healthy organizational outcomes) has an acceptable fit (RMSEA and SRMR values <0.08) although the relative fit indices did not reach the 0.9 cutoff value, but were above 0.85 (CFI = 0.88; TLI = 0.87; RMSEA = 0.07). All items showed loadings over 0.40, which has been considered the cutoff value of the contribution of each item to the latent factor variability (Guadagnoli and Velicer, 1988; see Figure 1).

The GoF for models with different constraints are displayed in Table 3. The baseline model (M1) showed a good fit (CFI = 0.874; RMSEA =0.060) that indicates the adequacy of the model. All items considered for the measurement model had significant factor loading in their factor (i.e., healthy employee and healthy organizational outcomes) in both groups. Standardized factor loadings were above 0.4 —which is generally considered a cut-off value (Guadagnoli and Velicer, 1988)— for all items (Figure 2).

Regarding measurement invariance across genders, the results confirm strict invariance, that is, there is no variability in factor loadings, intercepts, and residuals among gender groups. Results of the chi-square test show that all models differ significantly from the baseline model. Absolute increments in GoF between M1 and M2, M3 and M4 do not reach the 0.01 cut-off value. Regarding the difference between M1 and M5, CFI decrease is above 0.01. Considering the sensitivity of chi-square tests, we conclude that the proposed model has strict invariance. This level of invariance allows comparisons between regression coefficients (see Table 3).

Standardized beta scores for SEM paths between associated z-scores and p-values for both latent factors are displayed in Table 4 to test Hypothesis 2. Significant regressions in the women’s group for healthy employee were found in role ambiguity (β = −0.063, p < 0.01), routine (β = −0.076, p < 0.001), emotional dissonance (β = −0.056, p < 0.01), coordination (β = 0.114, p < 0.001), horizontal trust (β = 0.295, p < 0.001), vertical trust (β = 0.326, p < 0.001), mental competence (β = 0.057, p < 0.01), and emotional competence (β = 0.197, p < 0.001). For men, significative regressions were found in routine (β = −0.08, p < 0.01), mobbing (β = −0.076, p < 0.05), coordination (β = 0.133, p < 0.001), horizontal trust (β = 0.302, p < 0.001), vertical trust (β = 0.358, p < 0.001), mental competence (β = 0.092, p < 0.05), and emotional competence (β = 0.19, p < 0.001).

These results suggest that routine, coordination, horizontal trust, vertical trust, mental competence, and emotional competence have a significant influence on healthy employees for both men and women. As expected, all significant regressions related to job demands (i.e., role ambiguity, routine, mobbing, and emotional dissonance) were negatively related to healthy employee, whereas the variables related to job and personal resources (i.e., coordination, horizontal trust, vertical trust, mental competence, and emotional competence) had a positive value. Diversity-related healthy organizational practices were not significantly linked to healthy employees in any group.

Regarding significant beta scores in healthy organizational outcomes in the women’s group, significant values were found for mental overload (β = 0.065, p < 0.05), autonomy (β = 0.078, p < 0.01), supportive climate (β = − 0.115, p < 0.01), coordination (β = 0.101, p < 0.01), horizontal trust (β = 0.139, p < 0.001), vertical trust (β = 0.504, p < 0.001), mental competence (β = 0.122, p < 0.05), and emotional competence (β = 0.166, p < 0.001). For the men’s group, significant regressions were found in mobbing (β = − 0.129, p < 0.01), coordination (β = 0.161, p < 0.01), horizontal trust (β = 0.111, p < 0.05), vertical trust (β = 0.382, p < 0.001), emotional competence (β = 0.271, p < 0.001), and work-family balance practices (β = 0.183, p < 0.001).

For both men and women, coordination, horizontal trust, vertical trust, and emotional competence had a significant weight in healthy organizational outcomes. Against expectations, mental overload was positively related with healthy organizational outcomes in the women’s group, whereas supportive climate was negatively related. Coordination, horizontal trust, vertical trust, and emotional competence are significant and positively related for both men and women as for healthy employees and healthy organizational outcomes. For the women’s group, mental competence showed significant scores in both latent factors. For the men’s group, mobbing was significant and negatively related with healthy employees and healthy organizational outcomes. These results partially support Hypothesis 2.

In conclusion, our results support the effectiveness of promoting specific resources and prevent specific demands to build positive organizations among genders. Concretely, horizontal and vertical trust, emotional and mental competence, coordination, and routine predict healthy employees and/or results for both men and women. In terms of demands, only mobbing is negatively related to healthy employees and results in men, while only role ambiguity and emotional dissonance is negatively related with healthy employees in women. Furthermore, autonomy and mental competence predict healthier organizational outcomes in women (not in men), while they also perceive more resources (except emotional competence) and HOPs.

In addition, the study unveils three surprising findings:

Firstly, mental overload is found to be positively associated with healthy organizational outcomes in women. This underscores the importance of perceived competence in managing challenging demands, which may motivate employees to excel, potentially enhancing commitment and organizational performance. Secondly, supportive climate shows gender differences in the perception of support and the prevalence of “mansplaining” highlights the importance of fostering a real supportive environment —especially for women— to mitigate adverse effects on performance and well-being. Thirdly, only work-family balance practices predict healthy organizational outcomes, and only in men. This underscores the need for gender-sensitive practices that address work-family dynamics and institutional barriers to foster equity and productivity. One key lesson is that being competitive and surviving successfully in a VUCA context could not only depend on promoting resources to workers to cope with the demands, but also on doing so in a gender-sensitive manner.

The practical takeaway from the research for sample firms revolves around the notion of investing in practices to promote job resources—such as coordination—and personal resources—including horizontal and vertical trust, as well as emotional and mental competence to foster healthier employees and yield positive outcomes for both men and women. Besides this, organizations must consider the more complex theoretical implications of this research, which show the different importance of demands and resources for the well-being and performance of each gender. Given that women experience greater emotional dissonance and perceive less emotional competence, interventions are urgently needed to provide them with the ability to cope with the specific burdens they may feel, such as emotional intelligence training sessions. Additionally, providing women with greater autonomy could help them address demands that affect their health and well-being, such as role ambiguity. On the other hand, HRM must pay more attention to prevent mobbing in men, gender-sensitive interventions breaking down gender stereotypes to stimulate men in the communication of vulnerability can help (Staiger et al., 2020).

In addition, the fact that work-family balance practices only positively affect the performance of men can be addressed by emphasizing the need for effective supervisor support and organizational cultures that genuinely value and support the use of work-family balance practices without gender bias (McKinsey and Company, 2023). In other respects, the fact that other HRM practices related to gender equality and diversity are not resulting in significant effects on well-being and labor productivity brings along the urgent need for CEOs and politicians to test the effectiveness of the practices implemented and review the sums invested in them to improve their effects, especially in the case of women.

The present study has some limitations. The first is that all data were obtained through cross-sectional self-reports. Future studies could focus on longitudinal measures to test their stability. Secondly, an important limitation of this study is related to the significant number of missing values in our data set. A notably higher proportion of missing data was observed among male participants (54.4%) compared to female participants (28.1%), as indicated by a significant chi-square test result [χ²(1) = 132.97, p < 0.000]. This disparity in missing data could introduce bias and affect the reliability of the findings. Although a multiple imputation method was employed to address this issue and maintain a balanced sample between genders, the potential impact of the missing data on our results cannot be entirely ruled out. However, the use of advanced imputation techniques has likely minimized any distortions, allowing for more robust and reliable conclusions despite these challenges.