Work analysis tool for higher education: Development and validation of the german student measure WA-S Screening.

BACKGROUND: Work demands, resources and stressors affecting health, well-being and motivation also exist in the work of university students. There is a shortage of measures for analyzing work characteristics in this setting.
OBJECTIVE: This article addresses that shortage of measures and describes the development and the validation of the short Work Analysis Measure for Students (WA-S Screening).
METHODS: In study 1 (N = 422 students in Austria) the final version of the measure was developed based on analyzing the factor structure and psychometric properties of items and scales. Study 2 (N = 333 German-speaking students in Germany, Austria and Switzerland) was conducted for a cross validation and analyzing the criterion validity.
RESULTS: An eight-scale structure of the WA-S Screening was supported in study 1 and 2. The scales have shown to be significantly associated with burnout and work engagement in study 2.
CONCLUSIONS: The examinations indicate that the WA-S Screening is a short, reliable and valid instrument to identify critical, health-promoting work characteristics in the context of studying at university.

Introduction

Due to the transnational change to bachelor and master degrees in Europe (Bologna process) accompanied by a substantial change of demands, university students faced some new conditions for their study work in the last decade [1]. Examples relate to time constraints, higher frequency of exams, less scope for action or less training of “thinking-out-of-the-box skills” and less quality of teaching [1-3]. Meanwhile, the numbers of students in Europe (EU) were constantly increasing through the last decades, i.e. from about 17 million to over 20 million students between 2002 and 2012 [4], remaining at a high level between 19 and 20 million in total [5] and still slightly increasing in some countries like Germany [6]. In addition, students have become one of the groups being studied and associated with the burnout syndrome [7, 8]. International research of student burnout mostly fo-cuses on medical students so far –in particular to find early causes of recurrent physician burnout [9, 10], which affects a society‘s health care system by risking a lower quality of patient care [11]. Studies examining (medical) students revealed a burnout prevalence –mostly regarding the burnout dimension exhaustion –about 50 up to over 70 percent with having at least once burnout symptoms during medical school [12-16]. Robins, Roberts, and Sarris [17] showed, regarding exhaustion and cynicism in a longitudinal study with different health professions, that student burnout directly predicted future burnout in the workplace. Nearly all of these studies recommended to increase the awareness of the phenomenon “burnout” in university students and to im-plement interventions in the (medical) curriculum, such as educational and individual interventions to increase general student well-being [12]. However, they rarely suggested looking further at condition-related demands, stressors and resources that could cause or prevent student burnout as well as other mental health outcomes. We know that students – like employees – are able to experience burnout as well as work engagement. Consequently, an examination of work characteristics promoting or preventing those outcomes appears to be important for the identifica-tion of preventive measures within the study conditions. Work and Organizational Psychology provides well-established models and instruments for this purpose in the “normal” gainful work context. Apart from the individual perspective (e.g., capacity to deal with job demands), corresponding studies include an integrated organizational perspective, where working conditions as well as person-organization fit come to the fore [18, 19]. Work-role fit for example is not only associated with burnout and work engagement [20] but also with meaning in work [21]. In addition, many scales have been developed for describing and analyzing work characteristics in different occupational fields [e.g., Six Areas of Worklife Scale (AWS) [18], Job Diagnostic Survey (JDS) [22], Stress Oriented Task Analysis (ISTA) [23], Screening for Work and Task Analysis (Screening TAA) [24]. Most of them are capable to predict a certain risk of burnout and/or are related to other work-related aspects of mental health, motivation and well-being (e.g., work engagement) and thus can deliver important infor-mation for work design. In the last years, a few studies applied work-related models in the university student context, however without deeply arguing the adequacy for this special work context from a theoretical perspective [25, 26]. Therefore, there is a shortage of theory-driven condition-related instruments to examine the study context properly. So far, only one German student measure (BARI-S) [27] has been described, which analyzes three demands (time pressure, overtaxing demands, conflict between study and private life) and three resources (support by fellow students, support by lecturers, potential for qualification), applying the Job Demands-Resources (JD-R) model to university students [28]. The authors of BARI-S [27] compared some aspects of student and employment working conditions and concluded that the JD-R model could be applied successfully in a student sample. However, there are further developments of the JD-R model [29] and the question of the comparability of “employment” versus “study work” regarding basic work characteristics has not been comprehensively discussed on a theoretically grounded level.

Thus, our aim was to go further to existing theories and concepts of work characteristics and their positive or negative effects on the individual, developing a condition-related instrument for measuring work characteristics of students that distinguishes not only between demands and resources (like the JD-R) but between learning demands, stressors and resources [29]. In the following, we first elaborate the basic assumption: The comparability of employment versus study work regarding basic work characteristics. Then, we illuminate student-relevant theories and re-search findings about working conditions predicting burnout and other health and well-being variables. Subsequently, we describe the development and the validation of a measure for analyzing work characteristics of students in German speaking countries.

Do students at university “work”?

Is the work in employment structurally comparable to the activities a student is involved in for her/his studies at university? The Online Oxford Dictionary [30] defines work as (1) an “activity involving mental or physical effort done in order to achieve a purpose or result”, (2) “a task or tasks to be undertaken” and (3) “a thing or things done or made; the result of an action”. Work therefore consists of many targeted actions “one has to do in order to earn a living or to achieve a particular aim”. In the last citation, “work” refers to typical labor in employment as well as in a broader formulation (“achieving an aim”) “work” possibly includes e.g. domestic work, working as parents, voluntary work or study work. Both cases refer to the same work construct, therefore the answer can be yes – students work at university for the aims of obtaining professional skills and a university degree. However, at this stage it is still unclear if there are critical differences or not in terms of measureable work characteristics. The fundamental perspective of the Action Regulation Theory [31, 32] may provide a framework for answering this question. According to the Action Regulation Theory, goal-oriented actions regulated by the individual are the core of work, regardless whether these actions are paid or unpaid [33]. In employment, the goals of actions are usually redefined from formal or informal job descriptions for the specific position within the enterprise or organization and formulated tasks by supervisors. The situation in university studies is practically the same: Students have to show goal-oriented actions (e.g., preparing and presenting a paper, learning for an exam, conducting an experiment or fulfilling tasks in the laboratory). The goals of such actions are also redefined from institutionally specified higher-level goals described in curricula or formulated by professors and lecturers. Both, the employees and the students, have to regulate their actions mentally to achieve their redefined goals. Thus, the psychological structure of activities in employment and studying at a university appears widely similar. The only difference is that the final “object” or “product” of goal-oriented actions in employment is usually located outside the working person, whereas the student finally works on her- or himself by developing her or his own knowledge and competences. However, also the student has to accomplish external “products”, like exams, essays or presentations, to achieve defined sub goals. Finally, the highly competent student –developed by an institution of higher education –serves the society as a whole [34]. Looking at the illustrated basic psychological similarities between activities in employment and studying, we propose that there are parallels of relevant work characteristics affecting the psychological regulation and the probability of positive (e.g., work engagement) or negative (e.g., burnout) consequences for the individual.

Theories and concepts of work characteristics affecting mental health and well-being

Addressing the mentioned concepts of work characteristics beyond the JD-R, this section aims to provide more insights for the evidence of the proposed comparability of employment and study work.

According to the above-mentioned Action Regulation Theory, work and task characteristics can be categorized and evaluated based on their role for the regulation of goal-oriented actions. Frese and Zapf [31] distinguished between (1) regulation requirements (e.g., task complexity), (2) helpful resources for this regulation (e.g., decision latitude) and (3) conditions causing regulation problems (e.g., quantitative work overload, external interruptions). Against this theoretical background, elicited regulation problems are the “defining feature” for classifying a work characteristic as “stressor” impairing health and well-being and causing stress reactions like burnout. Con-sidering the conditions of work – no matter if labor or study work – these basic processes should comprehensively be valid in every kind of goal-oriented mental or physical activity [31]. Equivalent effects apply to other classifications of work characteristics. For example, the well-established JD-R model [28] distinguishes between (1) demands and (2) resources. In the JD-R model, demands – “associated with certain physiological and/or psychological costs” [28 p. 312] – are conceptualized as potential stressors. If not buffered, they lead to strain and subsequen-tly to health impairment and burnout, among others. Resources refer to “physical, psychological, social, or organizational aspects of the job that are either/or: Functional in achieving work goals; Reduce job de-mands and the associated physiological and psychological costs; Stimulate personal growth, learning, and development” [28 p. 312]. According to the JD-R model, resources can increase motivation (e.g., work engagement) and subsequently may lead to better health, as they are able to buffer the costs of health impairing demands [28]. Nevertheless, the (direct) effects in each case between resources affec-ting work engagement (positive connotation) and de-mands affecting burnout (negative connotation) seem to be the strongest [35]. Since empirical results have shown that job demands do not always impair performance and job satisfaction [36], LePine, Podsakoff and LePine [37] distinguished between challenge and hindrance stressors. In their definition, challe-nge stressors are able to elicit learning and perso-nal growth, whereas hindrance stressors threaten regulation capacities and health. Whether a stressor functions as challenge or hindrance depends on subjective appraisal processes [38]. In this case, the term stressor is equivalent to the term demand because of their “cost-character” as they are conceptualized in the JD-R model [39]. In a recent paper, Glaser et al. [29] theoretically integrated the classification of work characteristics based on the Action Regulation Theory, JD-R model, and the challenge-hindrance-demands distinction into one Model of Learning De-mands, Work-related Resources, and Stressors. This is the specific model we refer to in our study. Learning demands (e.g., cognitive demands) in this case correspond to challenge stressors/demands, show overlap with regulation requirements and have a positive, motivational and developing character. Re-sources (e.g., autonomy) play a similar but rather supportive role in this model in order to enhance motivation, well-being, engagement and performance. The stressors (e.g., work overload) on the other hand clearly correspond to hindrance stressors and regulation problems. They can directly lead to irritation, mental/physical health problems and absenteeism, possibly being buffered by the impact of resources. This model should consequently be a standard reference, as it integrates many relevant aspects from the research field of work characteristics and extends the well-established JD-R.

Regardless of which taxonomy of work characteristics is referred to, none proposes basic elements or mechanisms restricted to work in employment relationships (e.g., receiving a salary) and therefore in principle they have to be valid in students’ work. As it is evident that employment and study work are both “work” with the same underlying action regulation requirements (see 1.1), established concepts of work analysis from the field of employment should also be applicable in the study context and contribute to a better understanding and prediction of students' work-related health and well-being.

Present research in the European university context

There are several studies analyzing potential negative (mental) health effects of study conditions confirming the conclusions above [2]. In this regard, work characteristics like (perceived) workload, low social support, information problems or a conflict be-tween study and private life were often associated with burnout [40, 41]. The studies refer in their terms and theories to the Demand-Control Model (DCM) [42] or the JD-R model [27, 17]. Wörfel et al. [2] showed an existing relation of increasing demands at university and physical complaints. Concerning burnout, Gusy, Lohmann, and Drewes [40] found the highest burnout rates in German students, compared to student groups from the Netherlands and Spain. Organizational factors were examined and authors recommended more autonomy as well as better support by supervisors in order to create healthier study conditions [43]. Bachmann, Berta, Eggli, and Hornung [44] showed that structured and transparent study conditions lead to lower strain at universities. Also, in connection with organizational conditions, the longitudinal results of Dahlin and Runeson [45] showed that it is workload that predicts high bur-nout of medical students. Olwage and Mostert [7] identified inconsistent information as predictor for cynicism and Robins, Robert, and Sarris [17] a com-bination of high job demands (i.e. stressors: workload, lack of organizational structure etc.) and low job resources (feedback, skill variety, autonomy). Gusy et al. [27] identified social support by colleagues and teachers as well as the anticipated potential for qualification as significantly related to students’ work engagement. Work-family conflict and job demands were the most important predictors of psychological distress and risk of a psychiatric disorder in PhD students in Flanders [46]. Potentially positive learning demands according to the demand definition of the Action-Regulation Theory have not been analyzed in the study context so far.

Adaptation of the Screening TAA and pre-test

The TAA measure

To develop a screening measure of relevant work characteristics in the study setting, we decided to adapt a well-established work analysis screening measure from the field of employment: The Screening for Work and Task Analysis (Screening TAA) [47] is a validated measure, which distinguishes between learning demands, resources and stressors in different kinds of work settings (universal screening). It arose from the self-report version of the Activity and Work Analysis in Hospitals (TAA-KH-S) [24], being originally based on the Action Regulation Theory and the concept of the completeness of actions [31, 48, 49]. The newer and universal Screening TAA scales also fit to the state-of-the-art Model of Learning Demands, Work-related Resources, and Stressors by Glaser and colleagues [29], both theoretically and practically [50]. The Screening TAA consists of 21 subscales (80 items) representing five domains: Learning demands (e.g., cognitive demands), task-related and social re-sources (e.g., supervisor feedback), organizational and social stressors (e.g., job insecurity), task-related stressors (e.g., work overload) and physical stressors (e.g., physical workload). The instrument or selected subscales has/have been applied to many studies in different work settings proving good psychometric properties as well as good prediction capabilities [29, 47, 51–54]: Stressors and low work-related resources predicted e.g. health impairment and emotional exhaustion, whereas learning demands and work-re-lated resources predicted e.g. work engagement, int-rinsic motivation and (creative) performance.

Adaptation procedure and pre-test

To develop a sound and economic measure focusing on the central aspects of work characteristics in the context of studying at university, the following steps were taken.

First, we began to adapt and shorten the Scr-eening TAA [47] in detail for study contexts, with permission of and in cooperation with one of the original authors of the measure. We reduced the pool of the comprehensive inst-rument of 80 items/21 subscales to 40 items/ 10 subscales further on representing learning demands, resources and stressors as intended by the Model of Learning Demands, Work-related Resources, and Stressors [29]. Obviously inappropriate scales for the study context were removed after consensual group discussions between the authors of this article and in consultation with the researchers involved in the development of the original Screening TAA and TAA-KH-S. For example: Unfavorable Work Environment (e.g., noise, lighting, climatic conditions) and Physical Workload (e.g., long distances, hefting) due to not (str-ictly) being mental aspects; Quality Impairments (e.g. accumulation of errors due to unfavorable conditions), Work Interruptions (e.g., by calls, missing work equipment), Task Predictability (e.g., knowing the chronological sequences of the tasks), Skill Applicability (e.g., applicability of practical skills learnt) and five other subscales due to focusing too much on a fixed office/employment setting and permanent job-inherent tasks. For adaptation process including all previous subscales, see also Table A.1. In every item of the remaining subscales, specific employment-related wording (e.g., concerning the location “at work”) was reformulated (e.g., into “in my studies”).

To ensure content validity of the shortened and adapted measure (version 1), we conduc-ted a cognitive debriefing, which is a cognitive questionnaire pre-testing method, where target group representatives evaluate a new (adapted/translated) measure [55]. Therefore, we first generally discussed the main topics (subscales) of the Screening TAA with three students via semi-structured interviews regar-ding their incidence and spontaneous associ-ations in the students‘ context. Secondly, the students were asked to verbally evaluate the items of the adapted measure regarding comprehensibility and appropriateness for the uni-versity context in general.

Based on the qualitative pre-test, we added a few content-related supplements by adaptation of wording in existing items and adding five new items, which were named to be necessary in study context (e.g., sufficient self-regulation capability in the subscale Skill Adequacy). This slightly adapted version (version 2) was then tested in a quantitative online pre-test survey (N = 125).

The quantitative pre-test survey led to a re-moval of another five items including one subscale (Task Transparency) due to poor reliability (e.g., Omega total far below recommen-ded cut-off ωt>.70 and item factor loadings < .50).

The final version for study 1 (version 3) consisted of 40 items and nine subscales.

Study 1

The purpose of study 1 was to examine the factorial construct validity and the psychometric properties of the Work Analysis Measure for Students (WA-S Screening), analyzed in the pre-tests.

Hypothesis 1: We hypothesize a nine-factor-str-ucture representing the nine subscales identified in the adaptation process and the pre-tests.

Methods

Participants and procedure

Cross-sectional data were collected in a university town in Austria addressing university students from a wide range of disciplines at one university. The link to the survey was sent out by mail to all enrolled students at this university, which did not explicitly deregister from a mailing list of academic surveys (N > 10,000). An additional incentive was a raffle, in total 50,- € worth of money for five prizewinners. N = 422 students of 71 different studies fully completed the questionnaire. 67 % of the students were female. The age ranged from 18 to 50 years (M = 23.5; SD = 4.6), with 94 % ranging between 18 and 30 years.

Measures

We tested the adapted WA-S Screening (version 3) to assess the study work characteristics as described before in section 2. Nine subscales with 40 items were applied on a 5-point Likert-scale (1 = no, not at all to 5 = yes, exactly). Missing-values were allowed. The nine subscales were: (1) Skill Adequacy (e.g., “My self-regulation skills correspond with the demands of my study”), (2) Skill Acquisition (e.g., “My studies offer opportunity to expand my theoretical knowledge”), (3) Cognitive Demands (e.g., “My studies require to continually weigh various topics and to set priorities before I can get things done”), (4) Lecturer Feedback (e.g., “My lecturers provide clear feedback on my study performance”), (5) Autonomy (e.g., “I am free to determine how I do my work”), (6) Participation (e.g., “In these studies, one can participate in decisions on process and organization of courses”), (7) Organizational Stressors (e.g., “In these studies, one is frequently confronted with ambiguous information or rumors”), (8) Work Overload (e.g., “I often have too much work to do at once”) and (9) Information Problems (e.g., “Information needed for my studies is frequently not available”).

Statistical analysis

The psychometric scale properties were analyzed using SPSS 21.0.0.2. For assessing reliability/inter-nal consistency of scales, we calculated Omega total as state-of-the-art measure instead of Cronbach’s alpha [56]. To confirm the nine-factor structure (construct validity regarding the dimensions) of the WA-S Screening we first conducted a confirmatory factor analysis (CFA) using the maximum likelihood algorithm (Amos 21.0.0). The theoretical foundation of the adapted measure, the given sample size (>300) [57], and the level of measurement indicated that our sample properties met the requirements for per-forming a CFA. Further requirements (lack of collinearity, multivariate normal distribution) [57] were tested. All models included the manifest variables as item scores (and not as sum scores or item parcels) and correlations among the latent variables (subscales).

The following conventional fit indices were calculated to assess the model fit according to the cut-offs [58, 59]: For comparative fit index (CFI), values above .90 are satisfactory, around .95 they are good. A root mean square error of approximation (RMSEA) of .06 or below and a standardized root mean square residual (SRMR) of.08 and below indicate good fit.

Results

Psychometric properties

Analyses showed that critical collinearity was not given as the item-inter-correlations never exceeded r > .80, which allows performing a CFA. The Mardia test for multivariate normal distribution revealed adequate kurtosis for all items (<7) and still high multivariate kurtosis (182.87; z = 32.42). Therefore, we first corrected the p-Value for the χ2-test with Bollen-Stine-Bootstrapping (N = 1000). Internal consistencies (Omega total, ωt) of eight of the nine subscales ranged from ωt = .70 to ωt = .89, which dis-plays sufficient reliability. The subscale Skill Acquisition showed a questionable internal consistency of ωt = 63. Furthermore, we recorded a small number of 23 missing values out of 16,880 answers to be indicated (40 items×422 participants).

Construct validity

The first CFA (Model 1, nine factors, 40 items; WA-S Screening version 3; see Table 1) revealed an insufficient model fit [χ2(704) = 1918.45, p < .001; χ2/df = 2.73, RMSEA = .064 (CI90: .061–.067), SRMR = .081, CFI = .838]. We eliminated one item each of the subscales Cognitive Demands, Skill Adequacy and Participation due to poor factor loadings (cut-off <.50; CFA). To identify potential double and multiple loadings, we additionally performed an exploratory factor analysis (EFA; Kaiser’s eigenvalue criterion > 1). This led to the elimination of four items in the subscale Autonomy due to double or multiple loadings (EFA). Furthermore, both the factor loadings and the internal consistency (ωt = 63) of the sub-scale Skill Acquisition were poor. Also, the content validity and value of this scale appeared more and more questionable as the characterization of studying at university itself is defined by skill acquisition on a global level and no acquired skill beyond the study tasks seems to be typical of this specific work anymore. Thus, statistical and content arguments just-ified removing the subscale Skill Acquisition as a whole. Afterwards a second model of eight subscales and 28 items remained (Model 2, eight factors–28 items: WA-S Screening version 4, see Table 2). The CFA for Model 2 confirmed a good model fit [χ2(322) = 664.9, p < .001; χ2/df = 2.06, RMSEA =.050 (CI90: .045 –.056), SRMR = .050, CFI = .935]. Concerning the internal consistencies, only the subscale Skill Adequacy (ωt = .67) did not meet the minimum standards perfectly. Due to the fact that this subscale assessing the adequacy of theoretical, social and practical skills is more formative than reflexive [60], a comparatively lower internal consistency can be justified. Moreover, we assessed an alternative four-factor model (Model 3), which integrates second order factors of resources and stressors including three subscales each, keeping the two subscales Skill Adequacy and Cognitive Demands as separate factors (see Table 1). According to the theory, this was the most reasonable way to group the eight subscales left. Three of the subscales each are clearly defined as resources and stressors, respecti-vely [24]. Cognitive Demands are defined as so-called learning demands [29], but strictly speaking, Skill Adequacy is not a demand within the work activities but an important conditional qualification aspect [24]. Therefore, we kept the two scales separately.

Table 1

Fit indices for models 1, 2 and 3 | study 1

Study 1		Fit indices
(N = 422)
	Version	Factors	Items	χ2	df	χ2/df	CFI	RMSEA [CI]	SRMR
Model 1	V3	9	40	1918.5	704	2.73	.838	.064 [.061 –.067]	.081
Model 2	V4	8	28	664.9	322	2.06	.935	.050 [.045 –.056]	.050
Model 3*	V4	4	28	756.98	338	2.24	.921	.054 [.049 –.059]	.069

Note. χ2= chi-square discrepancy; df = degrees of freedom; χ2/df = relative chi-square; CFI = comparative fit index; SRMR = standardized root mean square residual; RMSEA = root mean square error of approximation; CI = 90% confidence interval for population RMSEA; *=(alternative model: 2nd order).

Table 2

Intercorrelations of the WA-S Screening subscales (version 4) | study 1

Scale		Variables indicated by numbers
		1	2	3	4	5	6	7
1	Skill Adequacy	1
2	Cognitive Demands	–.03
3	Lecturer Feedback	.22**	.12*
4	Autonomy	.19**	–.07	.16**
5	Participation	.16**	–.04	.41**	.18**
6	Organizational Stressors	–.17**	.12*	–.24**	–.24**	–.11*
7	Work Overload	–.24**	.33**	–.23**	–.22**	–.20**	.49**
8	Information Problems	–.24**	.11*	–.23**	–.11*	–.07	.64**	.48**

Note. *p < .05 (2-tailed), **p < .01 (2-tailed); N = 422.

The model fit was still satisfactory but showed less model fit than for Model 2 [χ2(338) = 756.98, p < .001;χ2/df = 2.24, RMSEA = .054 (CI90: .049 –.059), SRMR = .069, CFI = .921] (see Table 1).

Table 2 displays the intercorrelations of the eight final subscales in Model 2 and 3 (both version 4). 24 of the 28 correlations are significant at least at the significance level of 5% (2-tailed) and lead into the expected direction. For example, the three re-sources were significantly intercorrelated ranging from r = .16 to.41, the stressors from r = .48 to.64, and all stressors were negatively intercorrelated with all resources.

In sum, results of study 1 revealed good psychometric properties for the WA-S Screening (version 4). However, the model fit of the original nine-factor model (Model 1) and some factor loadings did not meet the necessary criteria. Therefore, our initial Hypothesis 1 had to be rejected. After excluding in total twelve items due to poor and multiple factor loadings, the model fit improved clearly (Model 2) without losing essential content. 28 items and eight reliable subscales remained.

Study 2

For a cross-validation of the results of study 1 and analyzing the relations of the subscales of the WA-S Screening with the three dimensions of bur-nout (emotional exhaustion, cynicism, inefficacy) and work engagement (criterion validity), we conducted a second, transnational study including various uni-versities in Germany, Austria and Switzerland. Re-capitulating mentioned theories and studies in employment and study setting, resources at work are positively related to work engagement and negatively to the three burnout dimensions emotional exhaustion, cynicism and inefficacy [28, 29, 31, 40, 43]. Skill adequacy –in terms of enabling the applicability of personal skills and therefore the fit between a person’s skills and the work tasks –is a basic work characteristic of an organization affecting action regulation requirements. In accordance with the theory, it is also positively related to work engagement and negatively to the three burnout dimensions [20, 31]. Cognitive demands are a prototype of challenge demands [37] and are therefore expected to be positively related to work engagement and negatively to the three burnout dimensions [29]. Stressors at work are negatively related to work engagement and positively related to the three burnout dimensions [2, 7, 28, 31, 45].

According to stated findings and theories, we set the following hypotheses in study 2:

Hypothesis 1: We hypothesize an eight-factor-structure of the WA-S Screening representing the eight subscales identified in study 1.

Hypothesis 2: Resources measured by the WA-S Screening (Autonomy, Lecturer Feedback and Participation) are positively related to work eng-agement and negatively related to burnout (emotional exhaustion, cynicism and inefficacy).

Hypothesis 3: Skill Adequacy measured by the WA-S Screening is positively related to work eng-agement and negatively related to burnout (emotional exhaustion, cynicism and inefficacy).

Hypothesis 4: Cognitive Demands measured by the WA-S Screening –as learning demands –are positively related to work engagement and negatively related to burnout (emotional exhaustion, cynicism and inefficacy).

Hypothesis 5: Stressors measured by the WA-S Screening (Organizational Stressors, Work Over-load and Information Problems) are negatively related to work engagement, and positively re-lated to burnout (emotional exhaustion, cynicism and inefficacy).

The cross-sectional data for study 2 were collected from 19 different universities in Germany (N = 11), Switzerland (N = 2) and Austria (N = 6) via online survey (N = 333 students). The link to the survey had been sent out by mail to available distributors at the universities and had been posted in social networks. A special incentive was the offer for individual feedback, for example on their current burnout risk status. Students of all stages, from the first to the seventh year of study, were represented from a broad range of academic disciplines. The two largest groups were psychology students (N = 118) and medical university students (N = 172). 71 % were female and the age ranged from 18 to 52 years (M = 24.3; SD = 6.4). Like in study 1, the majority of students (89%) ranged between 18 and 30 years. The psychology students were the “oldest” group (32% >30 years) and the medical students appeared the “youngest” group (0% >30 years).

In order to assess the student work characteristics, we applied the final 28-items WA-S Screening version as described before (version 4). Responses had to be given on a 5-point Likert-scale (1 = no, not at all; 5 = yes, exactly). This time, no missing-values were allowed (due to the extremely small number of missing-values in study 1). For analyzing criter-ion validity, we assessed burnout and work engagement with two widely established valid and reliable measures. Burnout with its three dimensions Emo-tional Exhaustion (EE), Cynicism (CY) and reversed Efficacy (Inefficacy; IE) was measured by the 15-item German Version of the Maslach Burnout Inve-ntory-Student Scale (MBI-SS-GV) by Gumz, Erices, Brähler, and Zenger [61] originally by Schaufeli et al. [8]. Item-examples for the subscales are “I feel emotionally drained from my studies” (EE), “I doubt the significance of my studies” (CY) and “I feel stimulated when I achieve my study goals” (IE - reverse). Responses had to be given on a 7-point frequency scale (0 = never; 6 = daily). The three scales showed acceptable and good reliability in our study (EE: ωt = .87, CY: ωt = .88 and IE: ωt = .76). Work engagement was measured with the German 9-item stu-dent-version of the Utrecht Work Engagement Scale (UWES) [62], including the three dimensions vigor, dedication and absorption. Following the recomm-endations for the short 9-item version, one total score was used as indicator of work engagement without distinguishing the three dimensions [63]. An item example is “When I’m doing my work as a student, I feel bursting with energy”. Responses had to be given on a 7-point frequency scale (0 = never; 6 = always/every day). The scale showed a good reliability in our study (ωt = .93).

Statistical analyses

To confirm the eight-factor structure (construct/cross validity) of the WA-S Screening, we conducted a confirmatory factor analysis (CFA) using maximum likelihood estimation in Amos 21.0.0. All models included the manifest variables as item scores (and not as sum scores or item parcels) and correlations among the latent variables (subscales). Item parameters for both studies were analyzed via SPSS 21.0. Configural invariance regarding the version 4 measurement models in both studies was tested by comparing the different measurement models in Amos 21.0.0. Metric invariance was tested by stepwise placing equality constraints on factor loadings [64]. This process further assesses the psychometric validity of the instrument.

To assess criterion validity, we inspected the bivariate correlations and performed multiple linear regression analyses using SPSS 21.0, with the three Burnout dimensions and Work Engagement as dependent variables.

Cross-validation (construct validity)

The CFA of the model with the best psychome-tric properties in study 1 (Model 2; eight factors, 28 items; version 4; see Table 1) revealed satisfactory standardized factor loadings from .57 to .92 and a very good model fit [χ2(322) = 600.2, p < .000; χ2/df = 1.86, RMSEA = .048 (CI90: .042 –.053), SRMR = .049, CFI = .939]. The second-order factor model (Model 3) was slightly poorer but also suf-ficient [χ2(338)=710.22, p < .000; χ2/df = 2.13, RMSEA = .054 (CI90: .048–.059), SRMR = .066, CFI =.919]. Internal consistencies for seven scales were acceptable to good (ωt = .75 to ωt = .88). Like in study 1 the subscale Skill Adequacy showed a lower internal consistency (ωt = .66). Due to its formative scale-properties (see study 1), again the comparative lower reliability can be justified. Cross-validation for Model 2 and 3 was therefore confirmed. Means, standard deviations, internal consistencies and correlations are depicted in Table 5. Analyses of scale-scores between the largest groups of study (medicine, psychology and others) showed that medical students reported comparatively high scores on Cognitive Demands (M = 4.11; SD = .72) and low scores on Participation (M = 1.77; SD = .90). Whereas psychology-students scored high on Skill Adequacy (M = 4.19; SD = .61) and lowest on the three stressors (M = 2.23 –2.66; SD = .81 –.92).

Table 5

Omega total, means, standard deviations and correlations | study 2 | version 4

		Omega total [CI]	M (SD)	Min-max	Skewness	1	2	3	4	5	6	7	8	9	10	11
Subscales
1	Skill Adequacy	.66 [.60;.72]	3.99 (.66)	1 – 5	–.49
2	Cognitive Demands	.82 [.79;.85]	3.79 (.87)	1 – 5	–.46	–.11*
3	Lecturer Feedback	.86 [.83;.89]	2.29 (.88)	1 – 5	.37	.07	.03
4	Autonomy	.84 [.81; 87]	3.48 (.89)	1 – 5	–.41	.10	.09	.06
5	Participation	.75a	1.96 (.89)	1 – 5	.81	.05	–.11*	.39**	.15**
6	Organizational Stressors	.78 [.74;.82]	2.78 (.82)	1 – 5	.29	–.13*	.18**	–.16**	–.14**	–.04
7	Work Overload	.87 [.84;.89]	3.14 (.99)	1 – 5	.03	–.37**	.39**	–.21**	–.17**	–.19**	.42**
8	Information Problems Criteria	.88 [.86;.90]	2.52 (.99)	1 – 5	.48	–.22**	.07	–.11*	–.18**	–.09	.57**	.43**
9	Exhaustion	.87 [.85;.90]	2.60 (1.16)	0 – 6	.37	–.41**	.23**	–.20**	–.24**	–.12*	.31**	.59**	.34**
10	Cynicism	.88 [.85;.90]	1.64 (1.47)	0 – 6	.87	–.22**	–.20**	–.14*	–.26**	.08	.31**	.12*	.23**	.38**
11	Inefficacy	.76 [.71;.80]	4.08 (.84)	0 – 6	.14	–.59**	–.07	–.21**	–.16**	–.15**	.16**	.35**	.24**	.44**	.38**
12	Work Engagement	.93 [.92;.94]	3.94 (1.15)	0 – 6	–.52	.34**	.28**	.21**	.24**	.05	–.21**	–.14*	–.24**	–.37**	–.70**	.56**

Note. a Spearmann correlation for 2-item-scale (Eisinga, Grotenhuis, & Pelzer, 2013);  < .05; ** p < .001; N = 333.

Analyses of invariance

The analyses of configural invariance including both studies (measurement model version 4) again confirmed a very good model fit [χ2(644)=1216.4, p < .000; χ2/df = 1.89, RMSEA = .034 (CI90: .031–.037), SRMR = .050, CFI = .938].

The analyses of metric invariance revealed an essentially tau-equivalent measure (version 4) in comparing study 1 and 2 (see Table 3). Stepwise invariance tests of the essentially tau-equivalent model compared to the congeneric model revealed non-significant for the subscales Skill Adequacy, Cognitive Demands, Autonomy, Participation, Organizational Stressors, Work Overload and Information Problems as well as for the whole instrument. The subscale Lecturer Feedback resulted partial essentially tau-equivalent as two of three items revealed invariant. As (at least partial) metric invariance has to be established for a test to be meaningful [64] – which was confirmed – another important indicator for psychometric validity of the instrument is given.

Table 3

Invariance analyses | study 1 vs. study 2 | female vs. male | 3 subject groups

	Study 1 vs. study 2			Female vs. male			Different subjectsa
Essentially tau-equivalent model	df	CMIN	p	df	CMIN	p	df	CMIN	p
Skill Adequacy	3	2.488	.48	3	2.894	.41	6	6.420	.38
Cognitive Demands	2	2.473	.29	2	.473	.79	4	2.111	.71
Lecturer Feedback	2	7.857	.02*	2	2.075	.35	4	23.832	.00**
Autonomy	3	3.332	.34	3	0.748	.86	6	8.808	.18
Participation	1	2.004	.16	1	0.168	.68	2	0.196	.91
Organizational Stressors	4	1.395	.84	4	1.567	.81	8	2.254	.97
Work Overload	3	3.344	.34	3	4.246	.24	6	3.82	.70
Information Problems	2	.643	.72	2	0.333	.85	4	1.522	.82
WA-S Screening
(version 4)	20	23.645	.26	20	12.62	.89	40	44.932	.27

Note.  < .05; ** p < .001; a psychology - vs. medicine - vs. other students.

Due to the satisfying invariance results, the item parameters (Table 4) of the WA-S Screening in study 1 and 2 were analyzed at once (N = 755).

Table 4

Item parameters | study 1 & 2 | version 4

Item	M	SD	Kurtosis	Skewness	r it
Skill Adequacy
1	3.65	1.07	–.33	–.54	.43
2	4.20	0.85	.95	–1.04	.48
3	4.07	0.92	.38	–.89	.46
4	3.74	1.02	–.14	–.61	.45
Cognitive Demands
5	3.78	0.98	–.32	–.56	.57
6	3.69	1.07	–.78	–.38	.72
7	3.59	1.04	–.58	–.39	.69
Lecturer Feedback
8	2.76	1.08	–.55	.22	.71
9	2.26	1.00	–.16	.53	.77
10	2.29	1.06	–.39	.52	.72
Autonomy
11	3.61	1.06	–.25	–.55	.69
12	3.77	1.02	.13	–.72	.72
13	3.54	1.10	–.36	–.50	.63
14	2.92	1.12	–.69	.13	.62
Participation
15	2.16	0.98	–.39	.52	.74
16	2.01	0.93	–.10	.67	.74
Organizational Stressors
17	3.05	1.24	–1.00	–.06	.58
18	2.59	1.17	–.63	.45	.58
19	2.76	1.13	–.61	.39	.60
20	2.92	1.28	–1.07	.14	.57
21	2.83	1.15	–.80	.23	.66
Work Overload
22	2.56	1.12	–.52	.46	.64
23	3.43	1.15	–.68	–.37	.74
24	3.30	1.17	–.82	–.18	.72
25	2.99	1.18	–.83	.17	.73
Information Problems
26	2.62	1.12	–.58	.41	.77
27	2.54	1.15	–.64	.43	.81
28	2.66	1.12	–.60	.34	.77

Note. N = 755; Min-max for all items: 1–5; r corrected item-total correlation.

Accordingly, also additional invariance analyses regarding gender and the three largest groups of studies (see above) were computed in one total set of available data (study 1 and 2). These analyses confirmed an essentially tau-equivalent measure with partial essentially tau-equivalent Lecturer Feedback in comparing the three groups and full metric invariance comparing females and males (Table 3).

Criterion validity

The correlation matrix revealed expected relations between the eight subscales and the criteria (Table 5). Nearly all of the 32 correlations were significant (p < .05) and appeared in the expected directions. The strongest relations were found between Skill Adequacy and two criteria of Burnout (EE: r =–.41; IE: r =–.59) as well as Work Engagement (r = .34) and between the stressors and the Burnout dimensions (e.g., Organizational Stressors and EE/CY: r = .31, Work Overload and EE: r = .59, Work Overload and IE: r = .35, Information Problems and EE: r = .34).

For a multivariate perspective, four structural equation model (SEM) analyses were performed with the eight subscales of the WA-S Screening as predictors and Work Engagement as well as the three Burnout dimensions as criteria (Table 6). Model fits were acceptable including the outcome Work Engagement [χ2(593) = 1392.8, p < .000; χ2/df = 2.35, RM SEA = .064 (CI90: .059–.068), CFI = .876].. The fit indices including the Burnout dimensions revealed good results [χ2 (428–491) = 847.2–955.0, p < .000; χ2/df = 1.94–1.98, RMSEA = .053–.056 (CI90: .048 –.061), CFI = .895 –.916].

Table 6

Standardized path coefficients from the WA-S Screening subscales on Work Engagement and Burnout dimensions tested in four structural equation models | study 2 (N = 333)

	Work engagement	Emotional exhaustion	Cynicism	Inefficacy
Variable	β	β	β	β
Skill Adequacy	.42***	–.25***	–.31***	–.84***
Cognitive Demands	.39***	.08	–.26***	–.21***
Lecturer Feedback	.16*	–.11	–.13	–.08
Autonomy	.10	–.16**	–.17**	–.05
Participation	–.01	.06	.16*	–.09
Organizational Stressors	–.24*	.09	.44***	.05
Work Overload	.04	.47***	–.07	.10
Information Problems	.03	–.04	–.11	–.03

Note. *p < .05; **p < .01 ***p < .001.

The results of the SEM analyses showed that Skill Adequacy was the work characteristic, that was significantly associated with all four outcomes in the expected ways (β= |.25|–|84.|). Further high path coefficients (in expected directions) were found for Cognitive Demands on Work Engagement (+), CY (-) and IE (-), for Work Overload on EE (+) and Organizational Stressors on CY (+). For both Burnout and Work Engagement, it is critical that the stressor Information Problems (in each case) showed no significant effect on the outcomes when tested in the models (see Table 6).

Summary

Study 2 confirmed the factor structure and good psychometric scale properties of the WA-S Screening. Invariance analyses between study 1 and 2, different groups of studies and gender confirmed (partial) metric invariance. Like in study 1 the eight-factor model (Model 2) as well as the four-factor model including two second order factors for resources and stressors (Model 3) showed good model fits.

Although both models were confirmed it is recommended to score the WA-S Screening on the level of the eight subscales as their distinction reveals more relevant information and the eight-factor model showed a slightly better model fit. Hypothesis 1 is therefore confirmed.

Criterion validity was tested for the work-related outcomes Burnout and Work Engagement. From a bivariate perspective, the correlations with the resources (Hypothesis 2) revealed small but without exception significant coefficients of Lecturer Feedback and Autonomy with all measured criteria. In SEM analyses, these relations were confirmed for Work Engagement, although not always for all Bur-nout dimensions (see Table 6). Participation was only and inconsistently related to Burnout. While correlations showed very small, but significant relations with EE (r =–.12, p < .05) and IE (r =–.15, p < .001) in expected ways, SEM analyses revealed a small, but unexpected positive impact on CY (β= 16, p < .05). Hypothesis 2 therefore is partly confirmed for Lecturer Feedback and Autonomy and is not confirmed for Participation. The subscale Skill Adequacy (Hypothesis 3) was significantly correlated with all dimensions of Burnout and Work Engagement. These results were verified in the SEM analyses. Hypothesis 3 is therefore confirmed. The subscale Cognitive Demands (Hypothesis 4) revealed an expected significant relation with Work Engagement (r = .28, p <.001) and low CY (r =–.20, p < .001), but also an unexpected positive correlation with EE (r = .23; p < .001). The fact that Cognitive Demands are also positi-vely related to the stressor Work Overload (r = .39; p < .001) could be a reasonable explanation. Indeed, the relation between Cognitive Demands and EE disappeared in the SEM when Work Overload was statistically controlled. In the SEM analyses, Cognitive Demands displayed the expected relations to Work Engagement, IE and CY. Hypothesis 4 therefore is partly confirmed. Finally, each of the stressors Organizational Stressors, Work Overload and Information Problems (Hypothesis 5) was significantly re-lated to all of the measured criterion scales (p < .05) as well as to each other (r = .42–.57, p < .001). The sub-scale Information Problems did not show significant paths in SEM analyses. A probable reason is the middle to high intra-correlation of the stressor-scales, which could have subsequently led to an indistinct overlap of variance explanation. Nevertheless, each of the eight subscales could at least once prove expected relationships to the examined criteria (in correlations and/or SEM analyses). Hypothesis 5 thus is partly confirmed for Organizational Stressors as well as Work Overload and is not accepted for the subscale Information Problems.

Discussion

This article aimed to outline the development of a theory-driven, short measure of student work characteristics and the investigation of the psychometric properties as well as the predictive criterion validity. Study 1 revealed a) a measure model of eight subscales with good reliability (internal consistency) and b) a notable alternative model with four subscales. Study 2 confirmed the factor structure, metric invariance and additionally displayed the relations of the subscales with the criteria burnout and work engagement. In sum, the two studies found substantial evidence for the reliability and validity of the developed WA-S Screening. They fill the shortage of theoretically grounded measurements in the study-work context of university students based on the latest developments in the Model of Learning Demands, Work-related Resources, and Stressors. In the following, we discuss the detailed findings also within the context of university students and suggest research and practical implications.

Measure development and psychometric properties

The Screening TAA indicated an excellent base for a measure-adaptation into other particular work settings (study work). The focus of the TAA au-thors, taking the quality of working-structure and -processes into account [24], is also a crucial aspect in terms of learning and studying at university. Nevertheless, it took many steps (adaptation in cooperation with one of the original authors, qualitative and quantitative pre-tests) to identify the central aspects of work characteristics for students to create a short scr-eening instrument with stable and satisfactory psychometric (item) properties. Only the subscale Skill Adequacy still shows moderate internal consistencies (ωt) between .6 and .7, which is however justified by its formative character (section 3.2.). Additionally, the subscale Participation persisted with only two items in the final version (the former third item in version 1 was removed due to a too small factor loading), which does not fulfil the formal criterions of a test scale. However, the scale covers a conceptually important aspect of work characteristics and therefore we decided to keep it in the instrument. Altogether, the instrument shows stable and satisfactory psychometric properties. As of now, it should be considered in future surveys in the context of university students as it is based on the latest theoretical models regarding the distinction of work characteristics.

The WA-S Screening and mental health-related outcomes

Concerning the indicated prediction of criteria by each of the eight WA-S Screening subscales the results showed that especially the resource Lecturer Feedback is positively associated with Work Engagement and Autonomy negatively at least with two of the burnout dimensions. Participation is associated with Cynicism concerning study work and shows no positive impact on Work Engagement. A possible exp-lanation for the unexpected impact of Participation on Cynicism in SEM analyses could be the general perception of low Participation (M = 1.96; SD = .89) in the sample. Only a few participants were probably aware of such possibilities of e.g. co-designing curricula and taking part in organizational decisions, due to being involved in such activities themselves. Consequently, the involvement and often voluntary additional work could be a separate factor leading to cynicism and not directly be related to the work characteristic of participation itself. Another explanation could be that the perceived participation opportunities offered by some universities appear as “pseudo opportunities” not including real impact and thus understandably leading to higher cynicism. However, this connection should be kept in mind for further analyses. For practical implications of low Participation, see 5.4.

Taking all analyses together, Skill Adequacy app-eared to be a very important prerequisite in the con-text of university students being associated with mo-tivation as well as work-related well-being (work engagement) and indicators of mental health (low burnout). As mentioned above it concerns the effort taken by organizations (e.g. universities) to take care of the fit between their demands and the students’ skills, for example on the one hand in being mind-ful with the recruitment of students, inherent req-uirements or application processes to courses and concurrently on the other hand supporting skill-pro-moting tutorials for students.

Regarding the indicated prediction of the typical burnout dimension Emotional Exhaustion, the characteristics Work Overload, low Autonomy and low Skill Adequacy appeared to be key variables. For the association with Work Engagement, the resource Lecturer Feedback, and especially Skill Adequacy as well as the challenging Cognitive Demands came to the fore. As described before, the latter was also correlated to Emotional Exhaustion. This is not surprising as also in terms of action regulation demands “it is the amount that makes the poison”. Learning Demands hereby clearly appear as work characteristics to be seen in a differentiated view. Whereas resources and stressors are more explicit supporting or impairing factors. These results fit to the Model of Learning Demands, Work-related Resources, and Stressors [29] we proposed in this study.

In sum, the instrument should be considered in future surveys when it comes to identify health- and personality-relating conditions in the context of university students.

Limitations

Several limitations should be kept in mind concerning the results of the two studies. First, the design of our studies was cross-sectional, which in particular restricts the validity of the results from regression analyses in study 2. Future studies should apply longitudinal study designs in order to examine criterion validity and ensure the causal predictability of work engagement and burnout by the WA-S Screening scales over time. Second, the samples were limited to occasionally acquired samples in German speaking countries, mostly representing medical and psychology students, including an imbalance between the number of females and males. As the results are potentially not universal, they should be verified in different student samples (e.g., technical fields of study) in the future. Nevertheless, the metric invariance between the two studies and therefore between the countries and explicitly also between the fields of study as well as between females and males was already confirmed. Compared to the other existing student measure we involved some “special” kinds of studies like medical students. Possible research sho-uld further aim to verify the predictability of the WA-S Screening towards relevant criteria, including more and different criteria of well-being and health (e.g. life satisfaction and/or physical complaints). Additionally, more objective (performance) outcomes like study dropouts and grades could be taken into account to complete the picture of work characteristics and their potential consequences.

As we decided to develop the WA-S Screening measure for students’ working conditions based on the comprehensive Screening TAA [47], we did not consider other possible work characteristics beyond the Screening TAA that are associated with work-related well-being and health, for example social support from colleagues or supervisors [29, 40]. We had to eliminate the Screening TAA subscales Social Climate and Social Stressors due to a high employment focus and the aim to create a short measure (Table A.1). In retrospect, it would still have made sense to include new items regarding social support by colleagues, friends, and/or lecturers. Other burnout-relevant working conditions in study contexts may be the organizational structure of studies, their curricula or number of exams. It is of course possible to extend the WA-S Screening with established and easy-to-adapt instruments measuring curricular conditions or social support [65].

Practical implications and recommendations

Apart from the essential extension possibilities, the application of the WA-S Screening in practice also “allows” the researcher to reduce the assessment via selected subscales and even dimensions (resources and stressors). As the models of our analyses showed, it should nevertheless be the first choice to assess work characteristics at university on subscale level.

Potential practical implication examples for universities and other organizations in the sector of higher education can be drawn from the content of the subscales and are proposed in the following. We hereby point out that we did not examine those implications in our studies. Resulting subscale mean values above/below the response scale mean (= 3) can be considered as criteria for high/low levels.

(Low) Skill Adequacy: Focusing even more on the fit between the students‘ skills and their range of courses in the process of application, inscription and the first years of study in being mindful with the recruiting and also providing supporting tutorials.

(Too low or high) Cognitive Demands: In order to ensure e.g. an appropriate level of cognitive demands, enough autonomy and to avoid work overload for students, persons responsible should take a look at existing curricula verifying if those conditions realistically are provided.

(Low) Lecturer Feedback: Additionally, each lecturer could be trained or at least be informed concerning the importance of lecturer feedback in class and its positive effect on students‘ work engagement. Many lecturers are obviously not implementing this behavior yet, due to several reasons. It could also be helpful to implement in the curriculum explicitly the type of feedback that has to be given (e.g., verbally, in writing and/or according to certain rules to be perceived as constructive). In addition, the introduction of periodical peer feedback could be considered.

(Low) Autonomy: In order to avoid burnout symptoms, the curricula should not determine too many preconditions regarding order and attendance of modules and seminars. High Autonomy should go hand in hand with secured transparent information, goals and instructions (see also high Organizational Stressors and Information Problems). (Low) Participation: Serious structures of (democratic) participation opportunities need to be provided and communicated both in courses and in the university in general.

(High) Organizational Stressors: University departments, curricula and study management should be clear and transparent. All members need to communicate and collaborate in a constructive way to enhance also the organizational work climate.

(High) Work Overload: Training lecturers to change perspectives into the students’ view and to collaborate more with colleagues, to avoid concurrent exams etc. (see also improvement of curricular management). In equivalent addition, offers for training students’ communication and planning skills should be institutionalized.

(High) Information Problems: Clear and transparent information needs to be accessible for students by coordinators and lecturers. Competent contact persons (staff) and tutors are recommended.

In general and in accordance with the model of Glaser et al. [29]: Resources should always complement high challenge demands. Moderate to high stressors combined with high demands should be avoided or at least be complemented by more or higher levels of resources.

As the WA-S Screening is a condition-related instrument, we suggested potential condition-related practical implications. Nevertheless, it is also recommended to address the responsibility of the students in the interplay of study conditions at university and individual health-related outcomes in practice. This responsibility can also be addressed mindfully and be encouraged to a certain degree.

Conclusion

In these studies, we found indicators for the validity and reliability of the Work Analysis Measure for Students (WA-S Screening). The instrument was dev-eloped to measure students’ working conditions in an economic and simultaneously theoretically grou-nded way, addressing the shortage of theory-based measures in this context. The WA-S Screening is associated with relevant university-related aspects of well-being and mental health. It is ready to be applied, further examined and validated in different student contexts such as other forms of tertiary educational institutions, academic subjects or German-speaking regions. The instrument can form the base to appropriate actions that might subsequently help young adults at university to stay engaged and to prevent them from being affected by the burnout syndrome, before even having entered the job markets.

Conflict of interest

None to report.

Table A.1

Adaptation process from the original Screening TAA to the first WA-S Screening version (1) and the final version (4)

Original Screening TAA	Adaptation	WA-S Screening version for pre-tests (version 1)	Adaptation	WA-S Screening final (version 4)
	Items			Items			Items
Cognitive Demands	4	B	Cognitive Demands	4	D	Cognitive Demands	3
Learning Requirements	3	A
Skill Acquisition	3	–>	Skill Acquisition	3	A, B’, D
Skill Adequacy	4	–>	Skill Adequacy	4	B’, D	Skill Adequacy	4
Skill Applicability	3	A
Task Transparency	3	B	Task Transparency	3	C, D
Task Predictability	3	A
Supervisor Feedback	3	B	Lecturer Feedback	3	–>	Lecturer Feedback	3
Autonomy/Job Control	9	–>	Autonomy	9	D	Autonomy	4
Participation	4	A’, B	Participation	3	D	Participation	2
Social Climate	5	A, C
Organizational Stressors	4	B, B’	Organizational Stressors	4	B’	Organizational Stressors	5
Social Stressors	4	A, C
Work Overload	3	B, B’	Work Overload	4	Work Overload	4
Goal Conflicts	4	A
Information Problems	4	A’	Information Problems	3	B	Information Problems	3
Work Interruptions	4	A
Quality Impairments	3	A
Additional Effort	3	A, C
Unfavorable Work Environment	3	A, C
Physical Workload	4	A
21 subscales	80		10 subscales	40		8 subscales	28

Notes: A: Exclusion of subscale | A’: Exclusion of item | Due to no/too little study fit (low content validity, high employment focus) | B: Important subscale, but larger textual adaptation on item level | B’: Supplement of new item(s) | C: No central aspect on work and task characteristic level (aiming for a short measure) | D: Exclusion of item(s) due to psychometric properties.