Thriving at work: an investigation of the independent and joint effects of vitality and learning on employee health

ABSTRACT Thriving at work has been defined as employees’ joint sense of vitality and learning. Based on the socially embedded model of thriving at work, we examine several competing operationalizations of thriving at work. We hypothesize effects of (a) composite thriving, (b) separate vitality and learning scores, and (c) the interaction between vitality and learning, and we explore effects of (d) the congruence between vitality and learning on self-rated physical and mental health. Data came from n = 1,064 employees who participated in a four-wave study with one-month time lags. Results of multilevel linear and polynomial regression analyses showed that composite thriving was positively related to physical health, and composite thriving and vitality were positively related to mental health at the within-person level. We found no support for interaction or congruence effects. The findings provide limited support for the assumed beneficial health effects of thriving on employees’ health. Implications for theory development include the need to revise the role of vitality and learning aspredictors of physical and mental health in the model of thriving at work.

Thriving at work has been defined as employees' joint sense of vitality and learning (Spreitzer et al., 2005). Accordingly, employees who are thriving feel active and strong (i.e., vitality) and, at the same time, believe that they are acquiring new knowledge and skills (i.e., learning). To experience thriving, both vitality and learning have to be present at high levels (Spreitzer et al., 2005). In developing their socially embedded model of thriving at work, Spreitzer et al. (2005) explain how contextual features of the work environment (e.g., climates of trust and respect) and resources (e.g., knowledge, support) promote agentic employee behaviours (e.g., exploration, task focus) that ultimately lead to thriving. Moreover, the model assumes that physical and mental health, as well as selfdevelopment, are key outcomes of thriving (Spreitzer et al., 2005). Since the introduction of Spreitzer et al.'s (2005) model and the development of self-report measures to assess vitality and learning (e.g., Porath et al., 2012), thriving at work has attracted considerable research interest. Synthesizing this literature, a meta-analysis reported positive relationships between overall thriving at work and various health-related (e.g., subjective health, burnout), attitudinal (e.g., job satisfaction, commitment), and performance-related outcomes (e.g., task performance, organizational citizenship; Kleine et al., 2019).
In empirical studies, thriving at work has been operationalized using either average or sum scores of ratings on separate vitality and learning scales (e.g., Carmeli & Spreitzer, 2009;Porath et al., 2012), or by assessing the effects of vitality and learning as two predictors that each constitute part of the thriving construct (e.g., Marchiondo et al., 2018). The use of average and sum scores may be problematic for the study of thriving at work, because this operationalization does not represent the experience of simultaneously high levels of vitality and learning (Kleine et al., 2019). As per Spreitzer et al.'s (2005) definition: "If individuals see themselves as learning, but depleted, they are not thriving. This might be captured by the experience of an employee who sees that she is learning in significant ways as she masters new technology, but feels burned out in the learning process. Conversely, if an employee experiences vitality at work but has no sense that he is adding to his existing knowledge or skills, he is not thriving" (Spreitzer et al., 2005, p. 538).
By collapsing across these two dimensions, using either an average or sum scoring approach, an employee with a low vitality score and a high learning score, or vice versa (i.e., incongruence between the two dimensions), would be assigned the same overall thriving score as an employee with moderate scores on both vitality and learning (i.e., congruence between the two dimensions). Such an operationalization is indeed inconsistent with Spreitzer et al.'s (2005) definition. A more appropriate operationalization may be the interaction between vitality and learning. Thriving as the interaction of vitality and learning entails that the positive effects of vitality on employee health are enhanced by a high level of learning at work, and vice versa. Additionally, the joint experience of vitality and learning may be operationalized in terms of the congruence between vitality and learning, suggesting that employees benefit most in terms of health if their vitality and learning at work are at similar levels. Interactions of, and congruence between, vitality and learning are more consistent with Spreitzer et al.'s (2005) notion of the joint experience of vitality and learning at work (i.e., in contrast to thriving as a composite of vitality and learning items).
Another limitation of previous research on thriving at work relates to the design of studies. Specifically, the majority of studies included in Kleine et al.'s (2019) meta-analysis used cross-sectional research designs to examine between-person relationships among constructs. One study on thriving with multiple measurement waves has shown a positive effect of thriving on general health (Walumbwa et al., 2018). However, this study did not examine the relative effects of the two thriving dimensions on employee health. Moreover, Walumbwa et al. (2018) assess the effect of thriving on general health without differentiating between physical and mental health (Ware et al., 1996).
The current study aims to contribute to the literature on thriving at work in three important ways. First, we examine the effects of thriving on employee health consistent with previous theory and research (Porath et al., 2012;Spreitzer et al., 2005). Specifically, we assess (a) the effects of thriving as a composite score of vitality and learning, (b) the unique effects of separate vitality and learning scores, (c) the interaction effect of vitality and learning, and (d) the effect of thriving as congruence between vitality and learning on physical and mental health. Modelling these effects separately enables us to investigate thriving at work as the joint experience of vitality and learning as well as the two thriving dimensions as separate predictors of employee health (Spreitzer et al., 2005).
Second, we focus on two health outcomes that have been neglected in previous research on thriving at work (Kleine et al., 2019). Specifically, we investigate the effects of thriving on selfrated physical and mental health. Using these outcomes enables us to comprehensively test the link between thriving and health, which has been proposed as a central path (i.e., conceptualized next to effects on "development") in the model of thriving at work (Spreitzer et al., 2005).
Finally, we contribute to research on thriving at work and health (see, Kleine et al., 2019) by shedding light on possible within-person dynamics. We attempt to answer whether monthly fluctuations in thriving at work are associated with monthly fluctuations in physical and mental health. This approach is motivated by the variations in vitality, learning, and employee health observed over time spans of one day to multiple weeks (e.g., Bensemmane et al., 2018;Jenkinson et al., 1997;Niessen et al., 2012). Accordingly, we examined withinperson relationships using data collected at four measurement points with time lags of one month each. By using multilevel modelling, we are able to account for dependencies that occur due to the nested data structure by separating within-and between-person variance components of the model (Preacher et al., 2011).

Thriving at work and employee health
Next to self-development, health is considered a main outcome of thriving at work in Spreitzer et al.'s (2005) model. Physical health encompasses aspects of physical functioning, the absence of role limitations due to physical health problems, and the absence of bodily pain, whereas mental health refers to social functioning, the absence of role limitations due to emotional problems, and emotional well-being (Ware et al., 1996). Spreitzer et al.'s (2005) model suggests that thriving at work has positive effects on general physical and mental health. That is, employees who feel vital at work derive energy from completing their work tasks. According to Porath et al. (2012), vitality counteracts negative emotions, such as worry and depression, and increases people's resilience towards stressful situations. Indeed, vitality has been shown to prevent or diminish the experience of adverse physical (e.g., coronary heart disease; Kubzansky & Thurston, 2007) and mental states (e.g., depression; Ryan & Frederick, 1997). Accordingly, we argue that vitality at work promotes employees' physical and mental health. The experience of learning at work fulfils the psychological needs for competence, relatedness, and autonomy (Vansteenkiste et al., 2006). Specifically, learning at work broadens employees' possibilities to get involved with various tasks, engage in collaborations where their expertise is needed, and take over responsibility. This need-fulfiling function of learning at work enables employees to be autonomously motivated when completing their work tasks, which, in turn, benefits their physical and mental health (Fernet, 2013;Van Scheppingen et al., 2014). Indeed, a study showed that employees who reported more learning believed that their work affected their physical and mental health positively (Ettner & Grzywacz, 2001). Taken together, these arguments suggest that both vitality and learning at work enhance employees' physical and mental health. Accordingly, we propose that when vitality and learning are higher than on average across all monthly time points, employee physical and mental health will also be higher: Hypothesis 1: There are positive within-person relationships between (a) vitality and physical health, (b) learning and physical health, (c) vitality and mental health, and (d) learning and mental health.
Because we expect both vitality and learning to be positively related to physical and mental health, we also propose a positive relationship between thriving as the composite of vitality and learning (Spreitzer et al., 2005) with physical and mental health: Hypothesis 2: There are positive within-person relationships between thriving at work as the composite of vitality and learning and (a) physical and (b) mental health.
When arguing for the relationships of thriving at work with health and well-being. Outcomes, organizational scholars have typically referred to arguments that suggest unique and linear effects of vitality and learning (see, Kleine et al., 2019). However, according to the definition offered by Spreitzer et al. (2005, p. 548, see above), experiencing vitality without learning and learning without vitality would not constitute thriving. Applied to the current study, Spreitzer et al.'s (2005) argument implies that the positive effects of vitality on health outcomes may be moderated by employees' experience of learning at work, or vice versa. That is, individuals who experience high levels of learning (vitality) may benefit more from the positive effects of vitality (learning) at work. Low levels of learning (vitality), in contrast, may decrease the positive effects of vitality (learning) on employee health. Thus, we propose: Hypothesis 3: The interaction between vitality and learning predicts (a) physical health and (b) mental health, such that the positive within-person relationships between vitality and health outcomes are stronger when learning is high (vs. low), and the positive within-person relationships between learning and health outcomes are stronger when vitality is high (vs. low).
In addition to these main and interaction effects, we explore the relationship of congruent levels of vitality and learning with employee health. When introducing the concept of thriving, Spreitzer et al. (2005) argue that individuals who see themselves as learning but depleted are not thriving and, as a consequence, do not benefit from the positive effects of thriving on their physical and mental health. Similarly, they argue that an employee who feels vital at work but does not learn new things feels stagnated, which, in turn, does not benefit their health. Accordingly, we may ask whether, in addition to positive linear main effects, employees benefit in terms of better health if their vitality and learning at work are at similar levels. For example, do employees benefit more in terms of better health if vitality and learning are experienced at similar moderate levels as compared to when, for example, vitality is high and learning is low? Because Spreitzer et al. (2005) do not make predictions about the effects of congruence between vitality and learning on health outcomes, we propose the following exploratory research question: Research Question 1: Is the congruence between vitality and learning positively related to employees' (a) physical and (b) mental health at the within-person level?

Procedure and participants
Data for this study were collected as part of a larger data collection effort with five measurement waves, and three other studies based on the same dataset, but with completely different research question and completely different substantive variables, have been published (Weiss & Zacher, 2022;Zacher & Rudolph, 2022;Zacher & von Hippel, 2022). The five measurement waves consisted of one initial survey that asked questions regarding demographics (Time [T] 0 = July 2018), and four subsequent surveys one month apart (T1 = August 2018, T2 = September 2018, T3 = October 2018, T4 = November 2018) to assess substantive variables. We used a time lag of one month between surveys because previous research has shown that thriving at work, as an indicator of transient workrelated well-being, may change over time spans of several weeks (Bensemmane et al., 2018). Additionally, both physical and mental health may change over time spans of four weeks (Jenkinson et al., 1997). In their investigation of the effects of thriving at work on employee health, Walumbwa et al. (2018) used time lags of two weeks between the measurement waves.
We commissioned an online panel company to recruit fulltime employees in Germany as participants in this study. Anonymous participant IDs provided by the panel company were used to match the responses over time. For the initial measurement wave (T0), approximately 3,500 potential participants were contacted. Overall, 1,064 participants provided data on the study variables and, thus, constitute the sample for the main analyses of this study. The sample included 594 men (55.8%) and 470 women (44.2%). Participants' ages ranged from 19 to 74 years, with a mean age of 43.1 years (SD = 11.0). In terms of education, 438 (41.2%) had obtained a primary or secondary school degree (no university-entrance diploma), 276 (25.9%) held a technical college or universityentrance diploma (German Abitur), 349 (32.8%) held a university degree (undergraduate and/or postgraduate), and one participant held a doctoral degree. Participants worked across 22 different industries with most working in the public sector (n = 138, 13.0%), manufacturing (n = 131, 12.3%), or healthcare (n = 111, 10.6%). Participants' organizational tenure ranged from below one to 51 years, with a mean of 11.6 years (SD = 10.1).

Measures
The survey items were presented in German. To translate the original items from English to German, we followed the translation/back-translation procedure proposed by Brislin (1970). Cronbach's alpha (α) at the within-and between-person levels are shown in Table 1.

Vitality and learning
We measured vitality and learning using self-reports at each of the four monthly measurement waves (T1-T4). To reduce the burden on participants' time and cognitive capacity, we shortened the vitality and learning scales used in this study. Specifically, we included only three items from each of the original scales that had the highest factor loadings reported for a young professionals sample in the measurement development study (Porath et al., 2012). Before each set of three items, we presented the phrase "During the past month (in the last 4 weeks) at work . . . " Items for vitality are "I have felt alive and vital," "I have had energy and spirit," and "I have been looking forward to each new day." Items for learning are "I have found myself learning often," "I have continued to learn more as time went by," and "I have seen myself continually improving." Participants provided their responses on 5-point scales ranging from 1 = strongly disagree to 5 = strongly agree.
To provide construct validity evidence for our shortened three-item vitality and learning scales with items in German language, we used data from a study published by Weigelt et al. (2019) with N = 472 employees. In this study, the full fiveitem vitality and learning scales by Porath et al. (2012) were measured in German language (using seven-point response scales ranging from 1 = very strongly disagree to 7 = very strongly agree). We calculated the means of the five-item vitality scale (M = 4.22, SD = 1.22, α = .85), and a shortened three-item vitality scale that was also used in our study (M = 4.06, SD = 1.43, α = .93), the five-item learning scale (M = 4.75, SD = 1.34, α = .91), and a shortened 3-item learning scale that was also used in our study (M = 4.53, SD = 1.44, α = .94). The correlations between the full and shortened scales were r = .96 for vitality and r = .97 for learning, providing some evidence for the construct validity of the shortened scales used in the current study.

Physical and mental health
We measured physical and mental health via self-reports at each of the four measurement waves (T1-T4) using the widelyused and well-validated SF-12 survey (Gandek et al., 1998;Ware et al., 1996). The items cover four health domains per component (i.e., general health, bodily pain, physical functioning, and role physical for physical health; and mental health, vitality, social functioning, and role emotional for mental health). An example item for physical health is "In the past four weeks, how much did pain interfere with your normal work (including both work outside the home and housework)?", with answers ranging from 1 = not at all to 5 = extremely. An example item for mental health is "In the past four weeks, how much of the time have you felt calm and peaceful?", with answers ranging from 1 = all of the time to 5 = none of the time. The physical and mental health component scores were computed using a scoring algorithm provided by the scale authors . The selection of the questionnaire items and the scoring method for the SF-12 Health Survey has been cross-validated in multiple countries, including Germany (Gandek et al., 1998). We used the original twelve items (six per component) to calculate Cronbach's alpha for physical and mental health.

Control variables
There is meta-analytic evidence for relationships of age, gender, education, and tenure with physical and mental health, suggesting that younger, male, more highly educated employees, and those with lower tenure have better physical and mental health (Brewer & Shapard, 2004;Cutler & Lleras-Muney, 2006;Ng & Feldman, 2010;Purvanova & Muros, 2010). Thus, we considered age (in years), gender (1 = male, 2 = female), education (1 = primary or secondary school degree (no university-entry diploma), 2 = technical college or university-entrance diploma, 3 = university degree (undergraduate and/or postgraduate), 4 = doctoral degree), and organizational tenure (in years) as control variables at the between-person level.
Both positive and negative affect have consistently been associated with physical and mental health outcomes (e.g., Cohen & Pressman, 2006;Watson, 1988). We controlled for the effects of positive and negative affect at the withinperson level, measured with two sets of five items from Mackinnon et al.'s (1999) short form of the Positive and Negative Affect Schedule (PANAS). Job demands act as stressors, affecting employee health negatively, whereas job autonomy denotes greater decision latitude with positive effects for employee health (Karasek, 1979). Thus, we controlled for the effects of job demands at the within-person level, using three items adapted from Spector and Jex (1998) and for job autonomy using three items from the German version of the work design questionnaire (Morgeson & Humphrey, 2006;Stegmann et al., 2010). Finally, employees who feel supported at work experience less stress (e.g., Frone et al., 1997) and, thus, may feel mentally and physically healthy (Heaney et al., 1995). We therefore controlled for co-worker and supervisor support at the within-person level, measured with four items each (Caplan et al., 1975).

Data analysis
All data analyses were conducted using R (R Core Team, 2019). Because repeated monthly measures (Level-1) are nested within participants (Level-2), the assumption of independent observations is violated. The intraclass correlation coefficient (ICC) for physical health was .83 and the ICC for mental health was .87, indicating that a considerable amount of variance in the dependent variables was explained by group membership 06| are significant at p < .05. ICC = intraclass correlation coefficient (denotes the proportion of total variance explained by the grouping structure); α w = Cronbach's alpha at the within-person level; α b = Cronbach's alpha at the between-person level. Code for Gender: 1 = male, 2 = female. Higher value for education indicates higher educational level.
(i.e., interindividual differences). We accounted for the nested data structure by testing our hypotheses using multilevel regression analyses. Before testing the hypotheses, we examined the measurement model of our core model variables (i.e., vitality and learning, physical and mental health) with multilevel confirmatory factor analysis (MCFA) using the R package {lavaan} (Rosseel, 2012). The four-factor model was assumed to be corroborated at each level of analysis; that is, at the between-person level, which captures variance across participants, and at the within-person level, which describes variance related to monthly fluctuations. We assumed at least reasonable fit for models with CFI and TLI values exceeding .95 (Hu & Bentler, 1999). Root-mean-square error of approximation (RMSEA) values smaller than .05 indicate close fit, and values smaller than .08 are considered acceptable (Browne & Cudeck, 1992). Finally, root-mean-square residual (SRMR) values up to .08 are considered good (Hu & Bentler, 1999; see, also Jackson et al., 2009) at both Level 1 and Level 2. We compared the model fit indices of our theoretical model with the fit indices of two more parsimonious models. First, we let the two thriving dimensions load on one common factor; second, because feeling energized is part of the mental health construct (Ware et al., 1996), we specified a model in which vitality and mental health load on one factor. Only if the theoretically based model (i.e., vitality, learning, physical health, and mental health as separate factors) fits the data better than both more parsimonious models may the variables included in our model be considered sufficiently distinct.
Polynomial regression analysis (PRA) allows testing whether levels of an outcome variable increase if the levels of two independent variables are congruent. PRA includes linear, interaction, and quadratic terms (Edwards, 1995;Edwards & Parry, 1993;Nestler et al., 2019). The results of a PRA may be plotted in a three-dimensional response surface. Response surface analysis (RSA) may be used to interpret the PRA estimates based on a graphical representation of the congruence effects (Edwards, 2002). Because of the nested data structure (i.e., time points are nested within individuals), we applied multilevel PRA (MPRA) to test Hypotheses 1, 3, and Research Question 1 using the R packages {RSA} (Schönbrodt & Humberg, 2018), {lme4} (Bates et al., 2015), and {lmerTest} (Kuznestova et al., 2017). We centred all intraindividual predictor variables at their person-mean. Thus, the reported effects refer to deviations from the average level of each predictor variable over multiple months across participants. The squared and interaction terms of the person-mean centred vitality and learning variables were calculated and added to the dataset as were the Level-2 means of all Level-1 (i.e., time-varying) predictor variables to disentangle within-person change from between-persons effects (Chen et al., 2005). We included time centred around the first month as a predictor variable to account for potential growth effects and to rule out that the within-person estimates may be biased by interindividual patterns of change. Multilevel regression analysis allows for modelling n − 1 random effects (McCoach, 2010), where n refers to the number of time points. Because we used four measurement waves, we may estimate three random effects (i.e., one random intercept and two random slopes).
Explanations for the different terms used in PRA can be found in our Online Appendix: (https://osf.io/mxv4a/). To facilitate easier interpretation of the response surface parameters, we plotted the average response surface for physical and mental health. According to Nestler et al. (2019), a strict congruence effect would be supported if (a) the fixed-effects line of congruence (LOC) is linear with a slope of zero (i.e., â 1 = 0 and â 2 = 0); (b) the fixed-effects line of incongruence (LOIC) follows an inverted U-shape with its maximum above the LOC (â 4 < 0 and â 3 = 0); and (c) the fixed-effects LOC is equal to the fixedeffects first principal axis (FPA) (â 5 = 0). The fixed effects estimated in PRA reflect within-person effects. Finally, to examine composite thriving as a predictor of health and well-being outcomes (Hypothesis 2), we used multilevel regression analysis and regressed each health variable on a composite thriving score (i.e., the person-mean centred composite of the vitality and learning items), control variables, and the Level-2 means.

Results
Descriptive statistics, intraclass correlation coefficients, and correlations are presented in Table 1. To account for the possibility of high levels of collinearity, we examined variance inflation factors (VIF) of each variable included in our regression analyses. Maximum VIF values were 1.58 and 1.86 among the centred predictors of physical and mental health, respectively. Thus, the maximum VIF values do not exceed the critical values of 4 or 10 as signs of severe multicollinearity (O'Brien, 2007).

Hypothesis tests
The variance component of the random slope for learning was near zero in all models, leading to singular model fit. Accordingly, we included a random intercept and a random slope for vitality as random parts of the models. 1 According to Hypothesis 1, vitality and learning are both positively related to physical and mental health. Hypothesis 2 proposes that thriving at work as the composite of vitality and learning is positively related to (a) physical and (b) mental health. Hypotheses 1 (a) and (b) were not supportedneither vitality nor learning significantly predicted employee physical health. Support was found for Hypothesis 1 (c) as vitality at work was positively related to mental health, at the within-person level (b = 0.13, p < .001). Finally, learning was unrelated to mental health, thus not supporting Hypothesis 1 (d). Consistent with Hypothesis 2 (a), composite thriving positively predicted physical health (b = 0.06, p = .046). Finally, we found support for Hypothesis 2 (b): composite thriving positively predicted mental health (b = 0.09, p = .002). 2 According to Hypothesis 3, there is an accentuating interaction effect of vitality and learning predicting (a) physical and (b) mental health. As shown in Tables 2 and 3, neither physical nor mental health were predicted by the interaction of vitality and learning. Accordingly, Hypothesis 3 (a) and (b) were not supported.
Finally, to answer our Research Question 1, we investigated the existence of a congruence effect as indicated by the criteria described by Nestler et al. (2019). Response surface plots are shown in Figure 1 (physical health) and Figue 2 (mental health). To construct these plots, we used the coefficients estimated in the PRA analyses. Interactive plots are presented in the Online Appendix (https://osf.io/ mxv4a/). As can be seen from Table 2, there was no linear main effect of vitality and learning at work on physical health (â 1 is non-significant). Additionally, â 2 was not significantly different from zero, thus meeting condition (a) of a congruence effect (Nestler et al., 2019). However, while â 3 was not different from zero, â 4 was not < 0, thus not fulfiling condition (b) of a congruence effect. Finally, â 5 was not significant, thus fulfiling criterion (c) of a congruence effect. Regarding Research Question 1 (a), we conclude that there was no effect of congruence between vitality and learning on physical health because criterion (b) was not fulfilled. As shown in Table 3, there was a common linear main effect of vitality and learning at work on mental health (â 1 is significant). While â 2 was not different from zero, â 3 was significantly positive. Moreover, â 4 was not < 0 and â 5 was significantly negative. Because none of the criteria (a), (b), or (c) of a congruence effect were fulfilled, we conclude regarding Research Question 1 (b) that there was no effect of congruence between vitality and learning on mental health.  Figure 2. Response surface plot for mental health. Note. The limits of the z-axis were restricted to reflect mental health values present among 90% of the participants (i.e., using the 0.05 and 0.95 quantiles as axis limits). The values on the x and y axes represent the person-mean centred vitality and learning scores, respectively. The inner polygon denotes the area where 50% of the data points reside and the outer polygon denotes the area where the remaining data points reside.
The proportion of variance in physical health explained by the fixed effects alone (marginal R 2 ) was .120 in the model with composite thriving and .126 in the full polynomial model. The amount of variance explained by both fixed and random effects (conditional R 2 ) was .656 and .651, indicating that a large amount of the variability in physical health remained unexplained (Nakagawa & Schielzeth, 2013). In contrast, a relatively large proportion of variance in mental health was explained by fixed effects in the model with composite thriving (marginal R 2 = .605; conditional R 2 = .750) and the full polynomial model (marginal R 2 = .617; conditional R 2 = .751).

Exploratory lagged analyses
The analysis of within-person relationships using a multilevel model does not allow for conclusions regarding lagged effects over time. Accordingly, we conducted exploratory analyses to examine the lagged relationships between thriving at work (time T-1) and subsequent physical and mental health (time T). The results are presented in the Online Appendix (https://osf.io/mxv4a/). Vitality and learning at work were unrelated to physical health. In contrast to the results of the main analyses, learning, but not vitality at work, was positively related to lagged mental health (b = 0.06, p = .020). Composite thriving was N= 3,399; sample size = 1,064. B= unstandardized estimate; SE = standard error; β = standardized estimate. Dummy code for Gender: 1 = male, 2 = female. The marginal R 2 describes the proportion of variance explained by the fixed factors. The conditional R 2 describes the proportion of variance explained by fixed and random factors. a The random slope in Model 2 refers to the slope for vitality at work. unrelated to lagged physical health and mental health. We found no evidence of lagged interaction or congruence relationships.

Discussion
The main goal of the current study was to investigate the within-person relationships between different conceptualizations of thriving at work and employee physical and mental health. Specifically, based on Spreitzer et al.'s (2005) definition of thriving at work and previous research (e.g., Porath et al., 2012), we assessed the unique effects of separate vitality and learning scores, their composite score, their interaction, and their congruence as predictors of employee health outcomes. Previous research indicated that thriving at work and employee physical health are dynamic constructs that may change over the course of one month (e.g., Bensemmane et al., 2018;Jenkinson et al., 1997;Niessen et al., 2012). Accordingly, we collected multi-wave data on employee thriving at work and health to examine within-person relationships among the study variables. Overall, we found mixed support for our hypotheses of positive within-person relationships between composite thriving, vitality and learning, and physical and mental health. There was a positive relationship between composite thriving and physical health, whereas vitality and learning were not significantly related to physical health. In line with Table 3. Results of a polynomial regression model with mental health predicted by composite thriving (Model 1); and vitality and learning main, interaction, and quadratic terms (Model 2).

Model 1
Model 2 the model of thriving at work (Spreitzer et al., 2005), vitality and composite thriving were positively related to employee mental health, suggesting that within-person fluctuations in the experience of thriving and vitality at work were accompanied fluctuations in mental health. However, contrasting the idea that developing skills and acquiring knowledge at work is associated with mental health benefits (Spreitzer et al., 2005), learning was not significantly related to mental health at the within-person level. Interestingly, results of multilevel lagged analyses showed that learning, but not vitality at work, was positively related to mental health, suggesting that withinperson fluctuations in learning at work may need some time to unfold their beneficial effects on mental health. Finally, there was no evidence of interaction or congruence effects, indicating that the relationship between the two variables (vitality and learning) with employee health was not positively reinforced by levels of the other variable.

Theoretical and practical implications
Our findings have several implications for theory development and research on thriving at work. First, they indicate that the notion of health as a central outcome of thriving (Spreitzer et al., 2005) needs further elaboration. Although composite thriving was positively associated with physical health, neither vitality nor learning were significantly associated with employee physical health. The factor analysis results suggested that vitality and learning are distinct factors, making the interpretation of the associations between composite thriving and physical and mental health challenging. For example, an employee with a low vitality score and a high learning score would score similarly on a composite thriving score as an employee with medium vitality and learning scores. Accordingly, the positive relationship between composite thriving and physical health should be interpreted with caution. The small marginal R 2 value compared to the relatively high conditional R 2 suggests that much of the variability in physical health exists between participants (Nakagawa & Schielzeth, 2013). This seems reasonable as several stable predictors of physical health that vary between rather than within persons have been identified in past research (e.g., income, personality, availability of social support; Adler & Snibbe, 2003;Johnson & Krueger, 2005). In contrast, for mental health, a relatively large portion of variance at the within-person level was explained by the fixed effects predictors included in our models. Both vitality at work and composite thriving were positively related to mental health. This finding is partly in line with Spreitzer et al.'s (2005) model of thriving at work: Feeling alive and vital may counteract the emergence of anxiety and depression, thus being positively associated with employee mental health. However, contrary to the model's predictions, fluctuations in employee learning were not significantly associated with fluctuations in mental health.
Our findings highlight the need to distinguish between physical and mental health outcomes in the model of thriving at work. It is possible that vitality positively relates to simultaneously measured mental health because the positive energy derived from feeling vital at work immediately spills over and affects a general feeling of mental well-being. In contrast, elements of work-related wellbeing, such as vitality and learning at work, may need more time to unfold their beneficial effects for physical health. That is, physical health may suffer from the prolonged experience of low vitality and learning at work, but may be less susceptible to short-term changes.
Although Spreitzer et al. (2005) do not make any specific predictions regarding interaction or congruence effects, they propose thriving as the joint sense of vitality and learning to be positively related to good health. However, our findings suggest that it is primarily vitality at work that is positively related to mental health (although we also found a positive association between learning and lagged mental health in supplemental analyses). Accordingly, instead of conceptualizing vitality and learning as forming a unit (i.e., thriving at work), we suggest focusing on the unique contributions of each of these variables for (timelagged) employee mental health.
In summary, our research findings suggest refining the model of thriving at work in three ways. First, by elaborating on the role of physical health as an immediate or a time-lagged outcome of vitality and learning at work. None of the two thriving dimensions were significantly associated with physical health, and, compared to mental health, little variance in physical health was explained. However, we cannot rule out the possibility that vitality and learning at work influence physical health in the long run. Future research needs to take into account that the time lag needed to observe relationships between vitality and learning with employee health may depend on the specific health outcome studied. Second, by considering vitality and learning as separate variables. The results of MCFA suggest that vitality and learning reflect distinct factors. Accordingly, using the average of all vitality and learning items to investigate the effect of a composite thriving score on employee health poses overly strict restrictions on the contribution of vitality and learning to an overall thriving score. Third, by considering vitality as a predictor of concurrent mental health and learning as a predictor of lagged mental health. We found preliminary evidence for distinct relationships of the two thriving dimensions with concurrent and lagged mental health. However, these results need to be replicated in future studies to strengthen the implications for the model of thriving at work (Spreitzer et al., 2005).
In terms of practical implications, organizations and managers could create conditions under which vitality is stimulated to improve employee mental health. Spreitzer et al. (2005) describe thriving at work as being socially embedded. Research has shown that vitality and learning may be enhanced by building up and maintaining relational resources (Kleine et al., 2019). For example, workgroups or teams may seek to establish meaningful connections at work and share common values of workplace civility to foster individual thriving. Additionally, supervisors should seek to support and empower their employees. Furthermore, they may increase their subordinates' thriving at work by acting as role models and actively stimulating employees' personal and professional development (Kleine et al., 2019).

Limitations and implications for future research
This study has some limitations that could be addressed in future research. First, we used shortened scales to measure vitality and learning at work. Although we provided some construct validity evidence for these scales, showed appropriate reliability, and demonstrated that they were sufficiently distinct from other variables included in the analyses, more research is needed to assess the quality of the shortened scales. In the same vein, although the physical and mental health component scores of the SF-12 are well-established indicators of self-reported physical and mental health, future research may test the proposed model using other health-related measures (e.g., work ability).
Second, we used self-report survey measures for thriving and employee health. Future research should make use of more objective measures to operationalize employee health (e.g., physiological health indicators; days absent from work due to health issues) or peer measures (e.g., ratings by colleagues or supervisors; Danna & Griffin, 1999). Third, we did not examine potential mediators and moderators of the relationship between thriving and health. For example, personality traits, such as learning goal orientation, may influence whether individuals benefit from learning experiences in terms of enhanced health (Hurtz & Williams, 2009). Thus, future research may foster an understanding of the learning-health relationship by considering mediator and moderator variables.
Fourth, the goal of the current study was to shed light on the within-person associations between vitality and learning and physical and mental health using data collected across four months. Although we examined both concurrent and lagged relationships, our results do not allow drawing causal inferences and do not rule out the possibility of reverse effects. Previous research has shown that work engagement, in terms of feeling enthusiastic and energetic at work, which is conceptually similar to vitality, is reciprocally related to mental health over time (Reis et al., 2015). According to the broaden-andbuild theory (Fredrickson, 2001), positive energy derived from one's work tasks may build up over time and positively influence employee mental health. Increased mental health, in turn, may serve as a resource that supports employees in immersing themselves in their work tasks, consequently deriving more positive energy from what they do. Similarly, Shirom et al. 2008) found both a time-lagged effect of vigour on employee health and the reverse effect of health on vigour. The investigation of a reciprocal relationship between vitality and mental health requires researchers to assess data at multiple time points and to include bidirectional paths in the model estimation.
Finally, we suggest two approaches for future research on outcomes (and predictors) of thriving at work. First, researchers may examine whether our findings based on linear, interaction, and congruence effects of vitality and learning generalize beyond the type of health indicators considered here as well as beyond other theoretically-based (i.e., personal growth and development; Spreitzer et al., 2005) and empirically suggested outcomes (e.g., employee performance and attitudes; Kleine et al., 2019) of thriving at work. Second, our findings call for more research that illuminates the longitudinal (and possibly dynamic) relationship between thriving and employee health. Future research may combine the use of short-term (e.g., daily or weekly), mid-term (e.g., monthly), and long-term assessments (e.g., yearly) to gain a better understanding of the temporal dependency of the effects of thriving on mental and physical health, and vice versa.

Conclusion
This study contributes to theory and research on thriving at work by considering composite, linear, interaction, and congruence effects of vitality and learning on employee physical and mental health. These effects represent a range of possible operationalizations of thriving that are in line with its definition as the joint experience of vitality and learning (Spreitzer et al., 2005). Only composite thriving, but not vitality and learning as separate variables, was positively related to employee physical health at the within-person level. Both vitality and composite thriving were positively related to employee mental health, whereas learning was not associated with mental health at the within-person level. There was no evidence of an interaction or congruence effect of vitality and learning on physical or mental health. Results of exploratory lagged relationships showed that learning, but not vitality at work, was positively related to mental health measured one month later. More research on thriving at work is needed to examine mechanisms and boundary conditions of the relationships between the two thriving dimensions with employee health.

Notes
1. The results did not change if the random slope for learning was added to the model. 2. The results did not change if only vitality or only learning were added as predictors of physical or mental health (i.e., without the interaction and squared terms).

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
The study reported in this article is funded by Volkswagen Foundation [Az.
96 849], The Role of Work in the Development of Civilization Diseases