There has been substantial research on how satisfaction of basic needs enhances employee motivation since Maslow’s seminal work on the hierarchy of needs (Maslow, 1943). One frequently cited framework on psychological needs is the Self-Determination Theory (SDT) (Deci & Ryan, 2000), which identifies three basic psychological needs: competence, relatedness, and autonomy. According to this theory and widely supported, satisfaction of these needs leads to ongoing psychological growth, integrity, and well-being, and conditions that thwart need satisfaction can lead to degradation of well-being or ill-being (Deci & Ryan, 2000; Olafsen, Niemiec, Halvari, Deci, & Williams, 2017; Van den Broeck, Ferris, Chang, & Rosen, 2016). Intrinsic motivation and basic need satisfaction also facilitates effective performance and well-being in particular if the tasks are more heuristic and require creativity, cognitive flexibility, or deep processing of information (Gagne & Deci, 2005) and thus, is directly relevant to to knowledge work and hence, academia.

Although the SDT has a strong empirical foundation, there is a lack of studies that have tested the theory within different organizational settings (Gagne & Deci, 2005) and university settings in particular. In a meta-analytic review of studies on basic need satisfaction, Van den Broeck et al. (2016) called for more exploratory research comparing SDT’s basic psychological needs to other potential needs and motivational constructs. One such study identified seven different needs, or states, that have the potential to boost work motivation and well-being: autonomy, belongingness, competence, psychological safety, positive emotions, fairness, and meaning (Maslach & Banks, 2017).

Gappa and Austin (2010) described different challenges facing higher education institutions today and, argued that faculty members’ essential needs go beyond academic freedom, shared governance, and job security. They added equity, collegiality, flexibility, and professional growth. Joined by respect, which is seen as the core value on which all others rest, this extended list constitutes the “Essential Elements” of twenty-first-century faculty work and pertain to all faculty members regardless of their appointment type. They argue that these elements are significant for attracting and retaining excellent academic employees in times of change.

Informed by the seven needs identified by Maslach and Banks (2017) and by the essential elements identified for faculty members in the twenty-first century (Gappa & Austin, 2010), we explored how different needs related to health and well-being in academia. Specifically, we explored the predictive value of job autonomy, social community, task completion clarity, work engagement, trust in unit management, recognition, and meaning on work-related positive and negative health among university employees.

Autonomy and academic freedom are seen as the core of traditional academic value (Aberbach & Christensen, 2018), a primary intrinsic motivator among academics (Bellamy, Morley, & Watty, 2003), and a basic need for health and well-being (Maslach & Banks, 2017). In addition, Gappa and Austin (2010) argue that changes in societal and faculty expectations of work and worklife issues compel the need for balance and flexibility in their career. Whereas job autonomy refers to perceived control over schedules and the organization of work, academic freedom refers to academic staff’s empowerment to make decisions regarding teaching and research, the core activities and tasks of the university (Aberbach & Christensen, 2018). In the present study, we measure autonomy by respondents’ influence over how the work is carried out and academic freedom by their ability to determine when their tasks are completed (task completion clarity). Thus, we hypothesized that;

It has been argued that psychological safety in teams (PST), characterized by interpersonal trust, respect, and caring within the work team, is a prerequisite for learning (Edmondson, 1999) and, health and well-being (Maslach & Banks, 2017). Because of their relevance to academics, we explored how these three characteristics relate to work-related health in academic settings. First, increasing diversity in employee demographics, such as gender, ethnicity and race, family status, and age, demand the need for trust (Gappa & Austin, 2010). For example, Downey, van der Werff, Thomas, and Plaut (2015) found that a trusting climate directly related to inclusion and mediated the positive effects of diversity practices on work engagement. Based on their findings, they call for more studies exploring how trust at different levels (e.g., trust in supervisor, team trust, trust in organization) are related to employee well-being. Therefore, we explored how psychological safety, as measured by the variable, trust in unit management, relates to employee work-related health.

Second, fairness or equity is considered highly significant for employee health and well-being in general (Maslach & Banks, 2017) and among academics in particular (Gappa & Austin, 2010). Fairness relates to the extent that decisions at work are perceived as just, and people are treated with respect. In the present study the need for fairness and respect (i.e., I am respected by the unit management, I am treated fairly by the unit management) are measured by the variable, recognition. A Norwegian study among academics found that recognition is particularly important among older workers, predicting meaning and commitment at work (Anthun & Innstrand, 2016). However, many academics report that they are not valued or recognized for their work (Bellamy et al., 2003). Thus, lack of satisfaction with this need could generate a possible hazard for academics’ health and well-being as well as their organizational commitment.

Third, caring within the work teams, or collegiality, is also identified as one of the essential needs for faculty members (Gappa & Austin, 2010) and is closely related to the SDT need, relatedness, or alternately labeled belongingness (Maslach & Banks, 2017). Relatedness is the need to feel connected to others and is satisfied when people see themselves as a member of a group, experience a sense of communion, and develop close relations. Although past work on SDT argued that relatedness plays a more distal role than autonomy or competence, a recent meta-analysis showed that relatedness plays a more significant role than previously expected and is strongly related to health and well-being (Van den Broeck et al., 2016). In the present study, this need is operationalized and measured by social community at work.

Lastly, professional growth is suggested to be an essential element that needs to be present in faculty members’ working environment to recruit and retain faculty members (Gappa & Austin, 2010). Within the SDT, psychological growth is typically manifested by intrinsic motivation where engagement in work is considered enjoyable in and of itself (Van den Broeck et al., 2016). Work engagement, which is defined as a positive work-related state of mind characterized by the feelings of vigor, dedication, and absorption (Schaufeli, Salanova, Gonzalez-Romá, & Bakker, 2002), can be interpreted as intrinsic motivation. Work engagement is particularly important in knowledge intensive work-places, such as universities and colleges, where the employees are the primary bearers of knowledge and thus, the competitive parameter in those organizations (Ipsen & Jensen, 2012). Several studies have found that work engagement is linked to both improved job performances, psychological well-being, and health in academia (Christensen, Dyrstad, & Innstrand, 2020; Innstrand, Christensen, & Helland, 2022).

Work engagement is also closely related to meaning, which has positive associations with health and well-being (Maslach & Banks, 2017). However, meaning can differ from work engagement by arising in “transcendent moments in time,” rather than as a sustained state of being (Bailey & Madden, 2017). According to Bailey and Madden (2017) are these moments “imbued with a sense of the coming together of the practice of work, reflection on that practice and the sense of a job well done, connection with others, coupled with an awareness of the significance of work in its historical and future context” (p. 15). Further they claim that meaning can arise “when an individual perceives an authentic connection between their work and a broader transcendent life purpose beyond the self” (p. 4). Using SDT terminology, meaning might be defined as a motivation in which behavior is more self-endorsed and viewed as important and/or in line with one’s closely held values (Van den Broeck et al., 2016). Meaning in work is highly significant for academics, and despite worsening work conditions in the academic sector, many academics consider their position as a calling, not just a job (Bellamy et al., 2003). Therefore, meaning in work is not only positively associated with health, but may act as a buffer against stress in one’s work and work life. In the present study, both work engagement and meaning are hypothesized to relate to university employees’ work-related health:

Individual differences in psychosocial needs can be found in mean levels and in the strength of relationships in the hypothesized model presented here. Although the benefits associated with need satisfaction are universal, the emphasis on these needs might differ in accordance with the values and practices of different cultural climates, as suggested by Vallerand (2000). Differences in the cultural climate and satisfaction of needs does not only relate to cultural diversity across countries but might also exist across work environments due to differences in opportunities or resources for need satisfaction, variation in the emphasis placed on the needs, or in the way these needs are met (Chen et al., 2015). As such, the level of needs might vary across different groups.

Chen et al. (2015), however, found the relation between need satisfaction and well-being to be invariant across cultures and not moderated by need strength. This suggests that basic needs contribute to well-being regardless of the extent to which needs are valued. If so, this supports the universality assumption central to SDT, conceptualizing needs as necessities for psychological well-being rather than socially constructed preferences (Gagne & Deci, 2005). The present study explores both the invariance of the level of the needs as well as in its relationship with positive/negative work-related health by means of multi-group analyses across academic groups (administrative staff, PhD students, and faculty), and gender. Based on the assumption by SDT and previous findings, we hypothesized that:

Data for the present study were gathered from ARK,¹ a healthy university initiative originated in Norway, that involved the administration of the KIWEST (Knowledge Intensive Work Environment Survey) to all faculty, administrative staff, and production workers in some of the largest Norwegian colleges and universities (Innstrand & Christensen, 2020). From autumn 2013 to spring 2015, all employees from the participating universities in ARK at that time, and with a regular payroll for a minimum of 20% position, were invited to answer the survey (N = 18,599). In total 12,170 responded, providing a response rate of 65%. In the present study, leaders were omitted from the analyses due to leadership questions. Thus, the final sample consisted of 4562 faculty (research and teaching), 1452 PhD students, and 5519 administrative and technical staff (n = 11,533). The sample was equally distributed with 54% women and 46% men. Age distributed as follows; under 30 years (9.8%), 30–39 years (23.2%), 40–49 years (27.2%), 50–59 years (24.3%), and 60 years or older (15.5%). Most were permanent employees (75%) and had a full-time position (86%). The demographics across the three employee categories are provided in Table 1. Due to concerns for anonymity, the universities did not provide data that identified to which faculties and departments the individual respondents belonged. The ARK study is approved by the Norwegian Center for Research Data (NSD) and informed consent was obtained by participants ticking “submit” at the end of the survey.

All variables were from the KIWEST questionnaire, which was based on previously validated and tested scales measuring the psychosocial work environment (Innstrand, Christensen, Undebakke, & Svarva, 2015). All but the engagement and work-related health survey items had a response format ranging from 1 (Strongly disagree) to 5 (Strongly agree). In the engagement scale, the response alternatives ranged from 0 (Never), 1 (A few times a year or less), 2 (Once a month or less), 3 (A few times a month), 4 (Once a week), 5 (A few times a week), 6 (Every day), and in work-related health, the response alternatives ranged from 1 (To a very small extent) to 5 (A very large extent).

Job autonomy was measured by a four-item scale from Näswall et al. (2010) and measures the extent of autonomy and influence over how the work is carried out. Sample item: “I have a sufficient degree of influence in my work.”

Task completion clarity was measured by three items from Näswall et al. (2010) which aim to capture the extent to which the employees themselves can, or have to, determine when their tasks are completed. Sample item: “I determine when my work assignments are completed.”

Trust in unit management was measured by five items from Näswall et al. (2010) and reflects perceptions of the employers’ reliability and trust-worthiness. Sample item: “My unit management is always reliable.”

Recognition was measured by a three-item scale from COPSOC II (Pejtersen, Kristensen, Borg, & Bjorner, 2010), which assessed the extent respondents feel they are recognized and appreciated for their efforts. Sample item: “My work is recognized and appreciated by the unit management.”

Social community was measured by a three-item scale from COPSOQ II (Pejtersen et al., 2010), which assessed the extent respondents experience a strong degree of social community with colleagues in their own unit. Sample item: “There is a good atmosphere between me and my colleagues.”

Work engagement was measured by the shortened, nine-item version of the Utrecht Work Engagement Scale (UWES). This shortened version has shown satisfactory psychometric properties (Schaufeli, Bakker, & Salanova, 2006), and has been validated in many countries. Sample items: “At my work, I feel bursting with energy” (vigor); “I am proud on the work that I do” (dedication); and “I get carried away when I’m working” (absorption).

Meaning of work was measured by three items from COPSOQ I and II (Pejtersen et al., 2010) and captures respondents’ experience of having a meaningful job. Sample item: “My work is meaningful.”

Work-related health was measured by two single items assessing respondents’ experience of how the work impacts their health. Sample items: “My work has a positive influence on my health” and “My work has a negative influence on my health.” The items were made for the ARK study and have been published elsewhere (Innstrand et al., 2022; Langseth-Eide, 2019).

All statistical analyses were performed using the structural equation modeling (SEM) package on Stata statistical software version 14.1 (Acock, 2013), with maximum likelihood (ML) and listwise deletion. Such restriction to complete cases only reduced the final analytic sample size to 10,096. We perform a Doornik-Hansen chi-square test to assess the joint normality of the observed variables (Doornik & Hansen, 2008), in which we reject the null hypothesis of joint normality (p < 0.001). To mitigate Type I error when using the ML estimator with non-normal data, we estimate all models using Satorra-Bentler robust standard errors and report both the ML and Satorra-Bentler chi-squared test statistics (Curran, West, & Finch, 1996; Maydeu-Olivares, 2017).

All the latent variables were allowed to correlate, as they all represent psychosocial work variables assumed to be related. As the chi-square statistic is found to be sensitive to large sample size (Mehmetoglu & Jakobsen, 2016) the SEM model was considered by these additional fit indices: Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), Root Mean Square Error of Approximation (RMSEA) and the Standardized Root Mean Square Residual (SRMR). A value of 0.90 or above is considered an acceptable fit for the CFI and the TLI, and a value of 0.08 or lower for the RMSEA and the SRMR (Acock, 2013). The statistical analyses were performed in three steps.

First, we performed confirmatory factor analysis (CFA) to ensure that the measurement model appropriately fit the data. Fit indices were first computed for the measurement model under the assumption that all latent constructs (but no item-specific error terms) were correlated. To strengthen the model fit, we modified the measurement model consistent with the estimated modification indices to allow for correlation in the error terms between two pairs of items in the work engagement scale: First, item 1 (“At my work, I feel bursting with energy”) and item 2 (At my job, I feel strong and vigorous”), and second, item 4 (“I am immersed in my work”) and item 5 (“I get carried away when I’m working”).

To confirm each item predicted adequate variation in the intended underlying construct, we examined standardized factor loadings and average variances extracted (AVE) for each latent variable using a cutoff point of 0.5 for both (Mehmetoglu & Jakobsen, 2016). The unique contribution of each underlying construct (discriminant validity) was assessed by the magnitude of each factor’s AVE relative to the squared correlation (SC) of the other factors. By convention, if the AVE is greater than the SC for all other factors, we conclude that the factor is distinct from other constructs (Mehmetoglu & Jakobsen, 2016).

All standardized factor loadings were sufficiently large (>0.5) except for one task completion clarity item (“I know when a task is completed”), which was estimated to be 0.30. In an assessment of equation-level goodness of fit, this item had a corresponding R-squared value of <0.20; these two estimates together led us to remove this item.

Next, to assess which of the suggested psychosocial need variables are the strongest predictors for health, we estimated a full SEM model in which all seven latent constructs were regressed on positive and negative work-related health for the whole sample (see Figure 1). Including both outcomes in the same model allowed for simultaneous estimation of the pathways between the latent variables and positive and negative work-related health. As two distinct but closely related parts of work-related health, the error terms of the two outcome measures were allowed to correlate.

Finally, we performed multi-group CFA (MG-CFA) and multi-group SEM (MG-SEM) to examine potential subgroup differences. Specifically, we assessed whether the mean value of each of the seven needs, as well as its predictive value on the two health outcomes, differed significantly between gender and across the three employee groups: faculty, doctoral students (PhD), and technical/administrative staff. For each of the two sets of analyses, we estimated three subgroup models in which all parameters were allowed to vary (model 1), factor loadings were invariant (model 2), and both factor loadings and intercepts were invariant (model 3). In each model with imposed constraints, we performed Wald tests to assess whether the assumption of invariance is valid (StataCorp, 2017, p. 71) and likelihood-ratio tests to compare nested models.

The goodness-of-fit estimates for the final measurement model all suggest a strong model fit: CFI = 0.96, TLI = 0.95, RMSEA = 0.05, and SRMR = 0.03. The standardized factor loadings were sufficiently large (>0.5) ranging from 0.60 to 0.93. As shown in Table 2, estimated AVEs for the seven factors ranged from 0.49 to 0.80. However, the comparison of the magnitude of each factor’s AVE relative to the squared correlation (SC) suggests a lack of discriminant validity between meaning and work engagement as well as between trust in unit management and recognition. As these factors were representing different aspects of the same need, the need for professional growth or positive emotions, and psychological safety, respectively, such a lack of discrepancy was somewhat expected. The estimated composite reliability (CR) measures ranged from 0.78 to 0.95, suggesting adequate reliability of the final measurement model (Mehmetoglu & Jakobsen, 2016).

A full structural equation (SEM) model with all seven latent constructs were regressed on each of our outcome measures, positive and negative work-related health, provided a good fit: RMSEA = 0.05, SRMR = 0.03, CFI = 0.96, TLI = 0.95. The Satorra-Benlter-corrected fit statistics are nearly identical (RMSEA = 0.04, SRMR = 0.04, CFI = 0.96, TLI = 0.95), and the p-values for the ML and Satorra-Bentler likelihood ratio tests are both <0.001. As shown in Figure 1, all factor loadings, except task completion and trust in unit management (ns), were significantly related, and in the intended direction to positive and negative health.² The three strongest relationships to positive health were work engagement (β = .25, p < 0.001), meaning (β = .23, p < 0.001), and social community (β = .19, p < 0.001). The three strongest relationships to negative work-related health were work engagement (β = −.23, p < 0.001), social community (β = −.23, p < 0.001), and autonomy (β = −.19, p < 0.001). In addition, recognition was positively related to positive work-related health (β = .11, p < 0.05) and negatively related to negative work-related health (β = −.13, p < 0.05). It should be noted that some of the antecedent variables were differentially related to positive and negative work-related health. For example, autonomy was weaker related to positive work-related health (β = .07, p < 0.05) as compared to negative work-related health (β = −.19, p < 0.001). Conversely, the meaning was more strongly related to positive work-related health (β = .23, p < 0.001) as compared to negative work-related health (β = −.09, p < 0.05). Thus, although these two health constructs were significantly related (β = −.66, p < 0.001), they are differentially related to psychosocial work variables.

The study’s strength is its large sample and the use of sophisticated statistical analyses, which enable more sensitive analyses of group differences compared to traditional statistical techniques like ANOVAs and t-tests (Hong, Malik, & Lee, 2003). The findings also need to be interpreted with limitations. First, the study was based on the needs theory (Maslach & Banks, 2017) and previous findings of “Essential Elements” for the twenty-first century faculty work (Gappa & Austin, 2010). Although our seven psychosocial variables related to these findings, they do not directly represent need satisfaction in the same way as variables studied in SDT (Deci & Ryan, 2000). Rather, our variables were proxies for the psychosocial needs presumed to be important for university employees. Additionally, we were not able to measure competence, one of the three basic needs assumed by SDT, within our dataset.

Our measure of health may also restrict a thorough examination of the relationship between need satisfaction and health. Although self-rated, single item measure of health has proved to be a reliable measure for health (DeSalvo, Bloser, Reynolds, He, & Muntner, 2006), and is judged to be appropriate for use in population surveys in general, and in particular, when used as an outcome variable to avoid overlap with different multi-item predictors (Bowling, 2005), multi-item measures are less prone to socio-psychological biases (Bowling, 2005), and the study results should be interpreted with this in mind. Future studies could benefit from using clinical measures of health. It should be noted that although job and life satisfaction is closely related, the health measure used in this study was restricted to factors at work. The rationale for this was to sort out factors like a broken leg due to skiing or other non-work-related factors. However, we recognize the interdependent relationship between well-being at work and well-being at life in terms of work meeting psychosocial and other needs and as a potential explanation for gender differences.

A second issue relates to cross-sectional designs as they preclude the possibility of drawing causal conclusions. Although SEM analysis and large samples compensate for much of the concerns related to these designs, the present findings can only be seen as significant relationships. Finally, the present sample consists of university employees only. Hence, the present findings can only be generalized to this occupational group.

The findings in the present study confirm the universality of the effect of needs on health across different employee groups despite differences in the level of the needs and are consistent with SDT (Deci & Ryan, 2000), and previous findings (Chen et al., 2015). Moreover, Essential Elements seem to generalize across university employees as suggested by Gappa and Austin (2010). Males and females, on the other hand, appear to differ both in how they perceive the influence of these needs on their health, as well as in the character of these needs. Future studies should confirm these findings by testing these assumptions on different samples and with a longitudinal design, more clinical measures on health, and SDT approved variables. Nevertheless, as suggested by the present study, universities may achieve greater equality by considering individual variances in the sources of need satisfaction in constructing health-promoting initiatives to create a healthy university for all.