Identifying coping profiles and profile differences in role engagement and subjective well-being

Journal of Basic & Applied Sciences, 2014, 10, 189-204 189 ISSN: 1814-8085 / E-ISSN: 1927-5129/14 © 2014 Lifescience Global Identifying Coping Profiles and Profile Differences in Role Engagement and Subjective Well-Being Saija Mauno1,*, Marika Rantanen2 and Asko Tolvanen2 1University of Tampere, University of Jyväskylä, Finland 2University of Jyväskylä, Finland Abstract: Coping strategies are not necessarily mutually exclusive and can be used simultaneously, a fact which has rarely been examined in coping research. We examined what kinds of coping profiles could be found in data concerning Finnish health care and service employees (n = 2756). We also studied whether role engagement (family-to-workenrichment, work-to-family-enrichment, emotional energy at work, and work engagement) and subjective well-being (life, parental, and marital satisfaction, and psychological distress) differ between coping profiles. The data were analyzed through latent profile (LPA) and covariance analyses (Ancovas). LPA revealed seven distinct coping profiles: two active groups, one passive group, one low and two high copers’ groups and one moderate group. These results indicate that coping strategies are not mutually exclusive and that people might use different strategies simultaneously. The covariance analyses revealed that the most significant differences concerned role engagement: active copers showed higher role engagement (e.g. enrichment, work engagement) than moderate or low copers. The findings imply that the indicators of role engagement deserve more attention in coping research in healthy working adults. Keywords: Coping profiles, coping strategies, latent profile analysis, person-oriented approach, role engagement, well-being. INTRODUCTION Although coping has been actively researched also in the field of work stress since the 1960s, there are still many definitions and theoretical views on coping. “Coping, in sum, is certainly not a unidimensional behavior. It functions at a number of levels and is attained by a plethora of behaviors, cognitions, and perceptions” [1]. The assessment of coping has also been difficult due to inconsistencies in the definition of the construct and to the complexity of the whole phenomenon [2, 3]. Despite these variations, most coping researchers would agree in defining coping as behavioral and cognitive attempts to manage, tolerate, or reduce the stressful demands of a situation [4]. As coping research has progressed coping taxonomies have also been introduced [5]. In these taxonomies, coping has most often been categorized as problem-focused (behavioral coping, e.g. taking direct action), and emotional-focused (cognitive or intrapsychic coping, e.g. positive thinking) even though important limitations have been noticed in such taxonomies [5]. This traditional way of classifying coping strategies into narrow categories is too limited an approach to allow proper understanding of the ultimate nature of coping [5, 6]. Moreover, narrow taxonomies hardly represent the coping strategies that *Address correspondence to this author at the University of Tampere and University of Jyväskylä, Department of Psychology, 40014, University of Jyväskylä, Finland; Tel: + 358 50 3186770; E-mail: [email protected] individuals utilize in real life settings [1, 7]. Coping researchers have therefore recommended investigating coping beyond narrow taxonomies [5, 8, 9]. Accordingly, the starting point in the present study is that coping is a complex construct and that individuals might use different coping strategies simultaneously, so coping strategies are not mutually exclusive. For instance, some individuals might be ‘high copers’ (using a variety of both problemand emotion-focused strategies) whereas others might be ‘low copers’ (not using any strategies very much) [9-11]. Approaching coping from this point of view requires a different methodological approach; that is, a person-oriented analysis instead of a variable-oriented analysis [12]. The basic idea in person-oriented analysis is to cluster or classify individuals into homogeneous sub-groups. In the present study, this meant that we examined how different coping strategies integrated or combined within individuals to form different coping profiles or coping combinations. Specifically, we applied Latent Profile Analysis (LPA) [13] to identify coping profiles and search for homogeneous sub-groups of coping. In addition, in order to externally validate our coping profiles, we examined how well-being and role engagement varied by coping profiles. If differences emerged, this would validate our coping profiles. Moreover, such findings would also have practical value; knowing what sort of coping combinations are most beneficial for well-being and role engagement would help in designing effective coping interventions. 190 Journal of Basic & Applied Sciences, 2014 Volume 10 Mauno et al. To our knowledge, this is the first study of coping strategies to focus on a non-clinical working population (n = 2756) by applying sub-group analysis (LPA) to coping strategies. Earlier studies on coping profiles/typologies have concerned either clinical populations [9, 14] or adolescents [10, 11] but no published study has examined ‘healthy’ workers from this perspective. However, coping is most likely a critical health-promoting resource also for normal, healthy workers [15, 16]. Theoretically, our study relied on the cybernetic coping theory [17, 18], which we introduce briefly below. Coping Strategies in the Cybernetic Coping Theory Edwards [17] developed an integrative theory of stress, coping, and well-being based on the idea of a negative feedback loop. This theory, known as the cybernetic stress theory, was developed in an occupational context and was therefore an appropriate framework for our study, which focused on ’healthy’ working adults. The main idea of this model is that discrepancies between internal needs (desired state) and environmental inputs (perceived state) are crucial in the stressor-strain process. The discrepancies cause stress which, in turn, affects an individual’s well-being and activates coping in order to reduce or prevent the negative impact of stress on well-being [4]. Somewhat similar reasoning is apparent in the previously developed person-environment fit theory on work stress [19]. More specifically, Edwards [17, 18] suggests that stress can activate coping directly, in expectation of possible damage to well-being, or indirectly, after well-being has already been damaged. Coping varies from a conscious careful planning, selection, and implementation to an intuitive coping response and no matter what kind of coping is activated its ultimate task is to prevent or reduce the negative effects of stressors on well-being [1, 17, 18]. Coping efforts can be directed toward the determinants of stress or towards the interpretations that the individual puts on the discrepancy. Specifically, Edwards [17, 18] divided coping behavior into five distinct categories or strategies: changing the situation, accommodation, devaluation, symptom reduction, and avoidance. He argues that these categories are not mutually exclusive and they can be used simultaneously, an idea which offers a good starting point for trying to identify individual-based coping profiles by person-oriented analysis, as we did. Situation changing includes actively solving the problem by modifying the situation and altering the situation to meet one’s desires. Accommodation means that a person’s own desires are matched to the situation, or that the person changes either their personal expectations or the importance given to the stressful situation. Devaluation can be defined as reducing the significance of the discrepancy by devaluing the importance of discrepancies between one’s desires and the situation. The idea of symptom reduction is that one takes direct intra-psychic action to improve one’s well-being, which has been damaged or threatened by the stressful event. Avoidance can be seen as directing one’s attention away from the stressful situation or discrepancies and in this way reducing the impact of the stressors on one’s wellbeing. These five coping strategies, which, in fact, can also be found in other coping models and inventories, e.g., the COPE-scale, [6] can be assessed by the Cybernetic Coping Scale [20], which we used in the present study. Coping Profiles in Previous Person-Oriented Studies As already said, even though the idea that coping strategies are not mutually exclusive and can be used simultaneously is not a new one, only a few previous studies have examined this possibility by trying to establish personal coping profiles (or clusters, typologies). However, none of these earlier studies have sampled the normal working population, as we did. Of these previous studies, two concerned adolescents. Aldridge and Roesch [10] used LPA to examine Hispanic, Asian-Americans, and other minority adolescents. They found three coping profiles: 1) ‘low generic’, 2) ‘active’, and 3) ‘avoidant’ copers. ‘Low generic’ copers used both active and avoidant coping strategies at a low level. The ‘active’ group comprised adolescents who used active and approach strategies (planning, instrumental social support, positive reinterpretation) at a high level whereas ‘avoidant’ copers preferred to use avoidant or passive strategies (such as substance abuse, focusing and venting emotions). Seiffge-Krenke and Klessinger [11] also identified coping profiles in their longitudinal study of German adolescents. They grouped different coping profiles using a factor analytic approach and found four coping profiles: ‘approachers’, ‘avoiders’, ‘high generic copers’, and ‘low generic copers’. ‘Approachers’ used high levels of approach-oriented and low levels of avoidant coping. ‘Avoiders’, on the other hand, preferred avoidant strategies and used approach coping only at Identifying Coping Profiles and Profile Differences Journal of Basic & Applied Sciences, 2014 Volume 10 191 low levels. ‘High generic copers’ used both avoidant and approach strategies at high levels whereas ‘low generic copers’ used both coping styles at low levels. Consequently, it seems that ’avoidant’, ‘active/ approach’, and ‘low generic’ copers were identified in both these studies [10, 11], which suggests that these profiles might be generalizable at least to some extent. We also found two studies which used personoriented analysis in clinical samples when examining coping profiles and their health implications. Walker et al. [14] identified six coping profiles in examining coping with pain by using cluster analysis: ‘infrequent’, ‘self-reliant’, ‘engaged’, ‘inconsistent’, ‘avoidant’, and ‘dependent’. ‘Infrequent’ copers rarely used any of the pain-coping strategies. ‘Self-reliant’ copers used accommodative strategies, for example acceptance, minimizing pain and self-encouragement, at high levels. ‘Engaged’ copers engaged with both personal and interpersonal resources and reported that they often used problem-solving, distraction, selfencouragement, and seeking social support. ‘Inconsistent’ copers used strategies which are inconsistent with each other, for example, in personal coping they used high levels of catastrophizing and self-encouragement and in interpersonal coping preferred high levels of both self-isolation and supportseeking. ‘Avoidant’ copers avoided social contact and kept others from knowing how they felt, and they rarely used self-encouragement or distraction. ‘Dependent’ copers reported high levels of catastrophizing about pain and in addition some support-seeking. Later Luyckx et al. [9] identified four coping profiles in their investigation of coping with illness (type I diabetes) using cluster analysis: ‘active integrated’, ‘passive avoidant’, ‘high generic low integrated’, and ‘low generic high integrated coping’. The ‘active integrated’ group used high levels of what they called ‘tackling spirit’ (for example, having the view that as a result of their own experience they were able to help other people) and diabetes integration (feeling that diabetes is the worst thing that has ever happened, for example) and low levels of passive resignation and avoidance strategies. ‘Passive avoidant’ copers used high levels of passive resignation and avoidance. In addition, they only rarely used tackling spirit and diabetes integration. ‘High generic low integrated’ copers frequently used active coping and moderately high levels of passive resignation and avoidance and moderately low levels of diabetes integration. ‘Low generic high integrated’ copers used high levels of diabetes integration and low levels of all other coping strategies. Only one common coping profile was identified in both these studies: the group of ’passive/avoidant’ copers [9, 14]. Even though our method of analysis (LPA) was data driven, signifying that it is difficult to set precise hypotheses, some hypotheses were posed on the basis of coping theories [4, 17, 18] and the empirical findings presented above on person-oriented coping studies [9, 10, 11, 14]. We expected (Hypothesis 1) to find at least two coping profiles in our data: one group in which active coping strategies (accommodation, symptom reduction, situation changing, and devaluation) are more often used (Group 1) and a second group in which passive (in the present case avoidant) coping strategies are more typical (Group 2). Furthermore, we considered it possible that we would find a group (Hypothesis 2) scoring low in all kinds of coping strategies (Group 3; ‘low generic copers’) as well as a group scoring high in all coping strategies (Group 4: ‘high generic copers’). However, we felt it was equally possible that other kinds of combinations would emerge (various active and passive groups, for instance) especially since we are using a large data set. Previous person-oriented coping studies have used much smaller samples, which also means fewer groups/profiles, because profile computing is based on individual scores, which are likely to show more variation in larger data sets. Differences in Well-Being and Health According to Coping Profiles/Groups The studies that we have already mentioned validated their coping profiles or typologies by examining potential differences in well-being and health shown by each coping profile. This idea is also well in line with coping theory, which argues that different kinds of coping and coping effectiveness, particularly, have implications for well-being and health [1, 4, 15, 16]. In their studies on adolescents’ coping, SeiffgeKrenke and Klessinger [11] found that ‘approach coping’ was related to the lowest symptoms of depression while ‘avoidant coping’ was associated with the highest ones. Another study of adolescents’ coping [10] showed that ‘active copers’ reported less depression and more stress-related growth than ‘low generic copers’. ‘Low generic copers’, for their part, reported less depression than ‘avoidant copers’ and less stress-related growth than ‘active copers’. Person-oriented coping studies which have been based on clinical samples have reported rather similar 192 Journal of Basic & Applied Sciences, 2014 Volume 10 Mauno et al. findings. Walker et al. [14], for example, found that ‘avoidant copers’ reported a higher level of depressive symptoms and lower competence (global, school, and social) than the other coping groups. Furthermore, in the group of ‘engaged copers’, in which problemsolving and self-encouragement and other such strategies were used at high levels, problem-focused coping (for example, perceiving the possibility of doing something to ease the problem) was seen as the most efficient action, and this group reported less depression than the other groups. Luyckx et al. [9], showed that ‘active integrated coping’ was the most effective profile (showing the lowest depressive symptoms and the highest personal control) and ‘passive avoidant’ the least effective profile (showing the highest depressive symptoms, for example). In addition, they showed that the ‘active integrated’ and ‘low generic high integrated’ groups had the highest self-esteem, while the ‘passive avoidant’ and ‘high generic low integrated’ groups had the lowest. Overall, these results, based on both non-clinical adolescent and clinical adult samples, support earlier well-established findings which show that avoidant or passive (often defined as emotion-focused) coping is mostly maladaptive whereas active or engaged (often defined as problem-focused) coping is mostly adaptive in terms of health and well-being outcomes [15, 16, 21]. However, it is worth remembering that these earlier person-oriented studies on coping approached wellbeing quite narrowly, focusing mainly on depression as a major outcome. Our study examines well-being more broadly covering, for example, satisfaction in the family domain and psychological context-free distress. Moreover, our study also covers role engagement, for example, work engagement and work-family enrichment, which has not been looked at before in person-oriented studies on coping. Despite this earlier neglect it could well be argued that role engagement is a relevant outcome because it describes how well a person is functioning psychologically and socially in different, major life domains [22, 23]. On the basis of these previous findings, covering both personand variable-oriented studies on coping strategies, some tentative hypotheses were posed on well-being and role engagement differences by coping profiles. We predicted that those employees who belong to active/approaching coping groups (there might be more than one ‘active group’) will show the highest well-being and role engagement whereas those who belong to passive/avoidant groups (again, there might be more than one ‘passive group’) will show the lowest well-being and role engagement (Hypothesis 3). Moreover, those scoring low in all types of coping (‘low copers’ in active and passive strategies) are expected to show poorer well-being and role engagement than those who score high (‘high copers’ in active and passive strategies) in all types of coping (Hypothesis 4). Finally, it should be remembered that our approach to identifying coping profiles was rather explorative: many different coping profiles might emerge, which means that all our hypotheses should be considered tentative. METHODS Procedure and Participants The data for this study were collected in October 2009 as part of the research project “Work-family coping strategies as promoters of employee wellbeing”. The study was conducted in collaboration with two Finnish trade unions: Tehy and Pam. Members of the former are professional health care workers (including nurses, physiotherapists, social workers, and midwives) and of the latter, service staff (cleaners, waitresses, security staff, and cashiers, for example) employed primarily in the private sector. We used an electronic questionnaire which was distributed by email to each potential participant (N = 7511). Random sampling was carried out by representatives of both unions. A total of 2756 individuals participated in the study, yielding a response rate of 36.7 %. Even though the response rate was rather low, it can be considered acceptable in occupationand organization-based research [24]. In the final data we had altogether 1719 health care professionals and 1037 service employees. 86% of the respondents were women, which corresponds quite well to the real gender distribution in Finnish labor unions: 93 % of Tehy’s and 80 % of Pam’s members are women. The respondents were on average 39.4 (SD = 11.6) years old, again a figure which is comparable with the actual situation in labor unions: the average age among Tehy’s members is 43 and among Pam’s members 40. Thus, in terms of gender and age the respondents corresponded quite well to the target population. Among the respondents the most frequent level of educational achievement was polytechnic or post-secondary, with 58% of respondents in this group, and 33% had intermediate vocational or college education. In terms of family situation, 82% had a spouse or partner and 66% had children. As for their employment, the participants worked on average 36.8 (SD = 9.2) hours per week, Identifying Coping Profiles and Profile Differences Journal of Basic & Applied Sciences, 2014 Volume 10 193 43% of the participants worked in shifts, and 85% of them had a permanent employment contract. Measures Coping strategies were assessed with the Cybernetic Coping Scale [20]. The psychometric properties of the CCS have been validated in previous research [25, 26]. This 15-item scale consists of five sub-scales, each of which was measured in our study by three items: accommodation (e.g. “I try to adjust my expectations”), avoidance (e.g. “I try to avoid thinking about the problem”), devaluation (e.g. “I tell myself the problem is unimportant”), symptom reduction (e.g. “I try to relieve my tension somehow”), and change the situation (e.g. “I try to change the situation to get what I want”). Respondents used a five-point scale ranging from 1 (almost never) to 5 (always). Cronbach’s alpha for accommodation was .62 (M = 3.12, SD = .58), for avoidance .80 (M = 2.68, SD = .75), for devaluation .73 (M = 2.87, SD = .66), for symptom reduction .65 (M = 3.40, SD = .66), and for the situation changing .69. (M = 3.10, SD = .66). Psychological distress was assessed with the Occupational Stress Questionnaire [27]. Six items (concerning for example fatigue, sleeping difficulties, irritation, and depression) which describe context-free or general well-being were assessed on a six-point response scale, ranging from 1 (never) to 6 (almost daily). Cronbach’s alpha for psychological distress was .89 (M = 3.18, SD = 1.11). Marital satisfaction was measured by two items drawn from the Kansas Marital Satisfaction Scale drawn up by Schumm et al. [28] (e.g. “How satisfied are you with your marriage?”). The items correlated highly (r = .79, p <.001). Parental satisfaction was assessed by three items from the Kansas Parental Satisfaction Scale devised by James et al. [29] (e.g. “How satisfied are you with yourself as a parent?”). Items were rated on a seven-point response scale, ranging from 1 (very unsatisfied) to 7 (very satisfied). Cronbach’s alpha for marital satisfaction was .97 (M = 5.79, SD = 1.30) and for parental satisfaction .80 (M = 5.77, SD = .90). Life satisfaction was evaluated by one item, “How satisfied are you with your life?”. This item was rated on a 7-point response scale, ranging from 1 (very unsatisfied) to 7 (very satisfied). Marital satisfaction correlated with life satisfaction (r = .53; p < .001) and with parental satisfaction (r = .28; p < .001). Life satisfaction related to parental satisfaction (r = .43; p < .001). These correlations between the variables were moderate, so they did not measure the same underlying construct. We therefore analyzed them separately. Role engagement was operationalized through four constructs: work engagement, work-to-family enrichment, family-to-work enrichment and emotional energy at work. Work engagement refers to a positive, fulfilling, and fairly persistent affective-cognitive, workrelated state of mind characterized by vigor, dedication, and absorption [30]. In the present study, work engagement was assessed with six items drawn from the short form of the Utrecht Work Engagement Scale (UWES-9) [31]. This 6-item scale consists of the subscales of vigor (e.g. “At my work, I feel that I am bursting with energy”) and dedication (e.g. “I am proud of the work that I do”). Items were assessed on a seven-point response scale, ranging from 1 (never) to 7 (every day). Cronbach’s alpha for work engagement was .93. (M = 5.50, SD = 1.25). Overall, work-family enrichment describes the extent to which experiences in one role (work or family) improve the quality of life (for example, performance or affect), in the other role [32]. In the present study it was measured by eight items of the Work-To-Family Enrichment Scale [33]. Four of these items measured work-to-family enrichment (WFE), describing the extent to which one’s work life facilitated or enriched one’s family life. (e.g. “My involvement in my work makes me satisfied and this helps me be a better family member”). Four items were also used to assess familyto-work enrichment (FWE), illustrating the extent to which one’s family life facilitated or enriched one’s work life (e.g. “My family life puts me in a good mood and this helps me be a better worker”). Each item was assessed on a seven-point response scale, ranging from 1 (totally disagree) to 7 (totally agree). Cronbach’s alpha for WFE was .83 (M = 3.88, SD = 1.29) and for FWE .85 (M = 4.87 SD = 1.17). Finally, emotional energy at work was evaluated by three items (e.g. “I feel capable of being sympathetic to patients/customers”) from the emotional energy sub-scale from the Shirom and Melamed Vigor Scale (SMVM) [34]. Specifically, the sub-scale concerns psychological presence at work in relation to customers and co-workers. We considered this an important aspect of role engagement, especially in the health care and service occupations which we studied. Emotional energy at home was measured by three items similar to those in the emotional energy at work scale except that we replaced “patients/customers” with 194 Journal of Basic & Applied Sciences, 2014 Volume 10 Mauno et al. “family members/significant others” (e.g. “I feel capable of being sympathetic to family members/significant others”). Cronbach’s alpha for emotional energy at work was .89 (M = 5.51, SD = .81) and for emotional energy at home .90 (M = 5.40, SD = .83). The items in these two scales were assessed on a seven-point response scale, ranging from 1 (never) to 7 (always). Because coping has most relevance in stressful circumstances [1, 2, 4], we also wanted to take into account how much stress was reported by the respondents. Stress was operationalized via the constructs of workload and homeload, describing role overload in two life domains. These variables were also used as covariates (together with labour union, gender, age, education) in examining differences in well-being and role engagement by coping profiles. Specifically, to assess homeload we used three items (e.g.” I have lot of responsibilities at home”) from the Family Demand Scale developed by Boyar et al. [35]. The items were assessed on a five-point response scale, ranging from 1 (totally disagree) to 5 (totally agree). Workload was measured by a three-item-based sum-scale (e.g. “Do you have too much to do at work?”) derived from the QPSNordic questionnaire [36]. The response scale ranged from 1 (almost never) to 5 (very often/always). Cronbach’s alpha for homeload was .82 (M = 2.85, SD = .1.02) and for workload .77 (M = 3.24, SD = .82). Of the participants, 70 % reported workload and 37 % homeload at least occasionally (M > 3.0, on a scale 15), implying that workload was more prevalent than homeload. Correlations (Pearson) between the studied variables are presented in Table 1. It is noteworthy that correlation coefficients (bolded) are significant at p <.001 level. Our sample was so large that very small correlations were also significant (p <.05), although these may have little practical value. Statistical Analysis First, we assessed the factor structure of the Cybernetic coping scale by running Confirmatory Factor Analysis (CFA) using the Mplus 5.0 program [37]. The estimation uses the meanand varianceweighted least-square method (WLSMV) and theta parameterization. The goodness of the fit of the CFA models was evaluated by the root mean square error of approximation (RMSEA), Tucker Lewis Index (TLI), Comparative Fit Index (CFI), and Weighted Root Mean Square Residual (WRMR). The statistically nonTable 1: Correlations (Pearson) between the Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 1. Labor union 1 2. Gender .26 1 3. Age -.38 -.10 1 4. Education -.53 -.17 .16 1 5. Workload -.15 -.05 .16 .11 1 6. Homeload .02 -.04 -.22 .00 .16 1 7. Accomodation .01 -.06 .01 .02 .04 .07 1 8. Avoidance .11 -.03 -.10 -.08 -.01 .10 .22 1 9. Devaluation .08 -.02 .00 -.08 -.06 -.02 .32 .54 1 10. Symptom reduction -.02 -.11 -.08 .08 .04 .06 .32 .32 .25 1 11. Situation changing -.00 -.02 -.06 .08 .08 .07 .33 .05 .11 .33 1 12. Distress .13 -.00 -.16 -.07 .34 .34 .06 .13 -.04 .11 .07 1 13. Life satisfaction -.14 -.08 .12 .08 -.12 -.24 -.02 -.08 .06 .02 -.01 -.48 1 14. Marital satisfaction -.01 .01 -.08 -.01 -.07 -.21 -.00 -.06 .01 .02 -.02 -.25 .53 1 15. Parental satisfaction -.07 -.01 .08 .01 -.12 -.22 -.04 -.08 .02 -.01 -.03 -.33 .43 .28 1 16. FWE -.08 -.07 -.03 .09 -.03 -.03 .09 -.04 .02 .09 .11 -.18 .34 .38 .21 1 17. WFE -.22 -.06 .07 .14 -.10 .01 .10 -.06 .02 -.00 .07 -.28 .18 .04 .12 .37 1 18. Work engagement -.20 -.11 .19 .12 -.09 -.11 .08 -.12 .01 -.02 .06 -.40 .32 .09 .20 .21 .45 1 19. Energy at work -.17 -.18 .13 .11 .00 -.03 .08 -.11 .01 .04 .07 -.19 .21 .08 .20 .20 .21 .40 1 20. Energy at home -.08 -.11 .03 .05 -.11 -.18 .04 -.10 .03 .04 .04 -.34 .44 .37 .39 .32 .13 .27 .41 1 Note. For bolded correlation coefficients p .001. Identifying Coping Profiles and Profile Differences Journal of Basic & Applied Sciences, 2014 Volume 10 195 significant -value, the value of RMSEA smaller than 0.06, the values of TLI and CFI greater than 0.95 and WRMR lower than .90 show a good fit of the model [37]. For the next step in analysis the factor scores were saved in the file. After running CFA we continued our analyses with factor scores by identifying coping profiles via mixture modelling, and specifically with Latent Profile Analysis (LPA). LPA is a sub-type of Latent Class Analysis (LCA) but they differ in one respect: LCA is often based on categorical variables whereas LPA is a better alternative for continuous variables, which we used in assessing coping (coping strategies were measured on a 1-5 response scale). Specifically, LPA assumes that the studied constructs, coping strategies in the present case, are independent of one another within their class and consists of latent profiles [37]. LPA uses the categorical latent class variable to group or classify individuals into categories, consisting of individuals who are homogeneous to each other in the same category and heterogeneous between categories. LPA is a useful method when seeking to identify homogeneous sub-groups in a dataset because it also detects small differences between the profiles (or latent groups) and allows statistical testing of the best profile solution (for fit indices, see the following paragraph), which is a clear advantage compared to some of the more traditional group-based methods, for instance, cluster analysis. Furthermore, in applying LPA it is possible to construct measurement error free latent factors, which then can be used as basis for classifying individuals [13, 37, 38]. In the present study, LPA was performed with Mplus 5.0 [37]. There are several statistical and other criteria that can be used to decide the number of latent classes/groups. Here, we used Akaike’s Information Criterion (AIC), the Bayesian Information Criterion (BIC), the sample size-adjusted Bayesian Information Criterion (aBIC), the Bootstrap Likelihood Ratio Test (BLRT), the Vuong-Lo-Mendel-Rubin test (VLMR), and the adjusted Vuong-Lo-Mendel-Rubin test (LMR). The AIC, BIC, and aBIC are relative fit indices: the smaller their values, the better the class solution [37]. Some other fit indices, such as the BLRT, VLMR and LMR tests, compare solutions with different numbers of latent classes: a low p-value (p < .05) indicates that the null hypothesized model with k-1 classes must be rejected in favor of the alternative hypothesized model with k classes [37]. Furthermore, in LCA (also in LPA), the statistical quality of the group classification (i.e., how well the model classifies individuals into subgroups) can be evaluated via Average Latent Class Posterior Probabilities (AvePP) [37]. The values of AvePP vary between 0 and 1: the higher the values, the more distinguishable the latent groups are from each other. Usually, AvePP values greater than .70 is used as a rule of thumb to indicate that the found solution can be interpretable using the mean trajectories [38]. Also, entropy values are often used to assess the goodness of group/class solutions and range from 0 to 1, where high values (> 0.90) indicate that the latent classes are highly discriminative [38]. The criteria that are important in deciding the number of latent classes are the usefulness and clarity of the latent classes. Clarity is evaluated primarily with AvePP and entropy values, while the quality of the classification can be evaluated in terms of the separation of the latent classes. Both AvePP and entropy values were used in this study to evaluate clarity of group/profile solution. Our second aim, after identifying the coping profile groups, was to investigate whether the coping profile groups differ in well-being and role engagement. To examine this, individuals were placed in the class whose posterior probability was highest and saved as an SPSS data file. SPSS version 16 was used to perform the covariance analysis (ANCOVA), which tests whether factors have an effect on the outcome variable after removing the variance which covariates (gender, age, education, labor union, workload and homeload) account for. Bronferoni pairwise comparison was used to determine which groups differed from each other if the general F-test showed significant values. RESULTS Results of CFA for Coping Strategies CFA revealed five distinct coping strategies (accommodation, avoidance, revaluation, symptom reduction, and changing situation), as was hypothesized on the basis of initial scale structure suggested by Edwards and Baglioni [20]. The final factor model, after freeing some of the covariances between residuals, fitted the data sufficiently (  2 (42) =577.93, p<.001 , CFI = .95, TLI = .97, RMSEA = .07, WRMR = 1.82). Detailed psychometrical information (e.g. factor loadings, factor intercorrelations, residual correlations between the observed variables) are provided in Figure 1 and means for normalized factor z-scores according to the coping profiles can be found in Appendix 1. Even though the residual correlations were very small (see 196 Journal of Basic & Applied Sciences, 2014 Volume 10 Mauno et al. Figure 1), the large sample size (N = 2537) in the model decreased the fit of the model to a considerable extent. The standardized factor loadings (ranging between .46 and .85) were all statistically significant. The factors correlated with each other but not very highly: the highest correlation (r = .53) was between situation changing and symptom reduction and the lowest (r = .03) between avoidance and situation changing. Thus, we ended up with a five-factor model, which was identical to the initial model tested by Edwards and Baglioni [20, 25, 26], with the factors accommodation, devaluation, symptom reduction, situation changing, and avoidance. The factor scores were saved and subsequent LPA was based on these factor scores. c1 c6 c11 Situation changing c2 c7 c12 Accommoda -tion c3 c8 c13 Deva lua tion c4 c9 c14 Avoidance c5 c10 c15 Sympton reduction .72 .58 .85 .61 .61 .72 .72 .76 .78 .79 .82 .81 .60 .75 .70 .47 .17 .48 .03 .30 .70 .53 .50 .37 .38 Figure 1: The final factor model of coping scales. Coping Profiles Found in the Data As the model AIC, BIC and aBIC fit indices for the 1 – 8-class solutions presented in Table 2 show, they decreased from 1 to 8, which suggested that the 8class solution was the best fitting model. Also, the BLRT test suggested an 8-class solution. The VLMR and LMR tests, on the other hand, pointed to another conclusion: these tests indicated that the null hypothesized model with the 7-class solution cannot be rejected at p < .05 level when compared to the alternative hypothesized model of the 8-class solution, suggesting that the 7-class solution was the best fitting model. In addition, these 7 classes had the following AvePP values: Class I 0.89, Class II 0.84, Class III 0.95, Class IV 0.84, Class V 0.85, Class VI 0.91, and Class VII 1.0. These values are clearly above 0.70, which is used as the criterion for the clarity of the group solution in AvePP [38]. The analysis was also continued to compute fit indices for the 9-17-class solutions. However, according to the entropy values, when more than 8 class solutions were fitted, the clarity of the class differences decreased (values available from the authors upon request). Consequently, according to these fit indices and a careful contentrelated interpretation, we ended up with the 7-class solution, which was used in the subsequent analysis. The profiles (henceforth for clarity labeled groups) and their sizes are presented in Figure 2. In the figure, being above the median means that this coping strategy is used more than average and being below means that it is used less than average. Group 1 contained employees who used each of the five coping strategies at a moderate level (labeled ‘moderate copers’). This was also the largest group, consisting of nearly half of the participants. Group 2 (labeled ‘low copers’) and Group 3 (labeled ‘passive copers’) consisted of individuals who used every coping strategy less than average, Group 3 even more rarely than Group 2. Group 4 consisted of employees who used changing the situation and symptom reduction more often than average (and devaluation and avoidance less than average). We labeled this group ’blurred copers’ because changing situation as a coping strategy best captures the essence of problemfocused coping whereas symptom reduction describes the essence of emotion-focused coping. Group 5 consisted of employees who used more than average situation changing, accommodation and symptom reduction. They were labeled ‘active copers’ (Group 5) because avoidance and devaluation were not often used in this group. Groups 6 (labeled ‘high copers’) and 7 (labeled ‘the highest copers’) contained employees who used all coping strategies more than average, Group 7 even more than Group 6. Identifying Coping Profiles and Profile Differences Journal of Basic & Applied Sciences, 2014 Volume 10 197 Table 2: Model Fit Indices for the 1-, 2-, 3-, 4-, 5-, 6-, 7-, and 8-Class Solutions Fit index AIC BIC Adj. BIC VLMR LMR BLRT Log-likelihood (df) Entropy 1-class 36018.971 36077.358 36045.585 - - - -17999.485 (10) - 2-class 33621.219 33714.638 33663.802 .000 .0000 .0000 -16794.609 (16) .710 3-class 32385.127 32513.580 32443.680 .0053 .0058 .0000 -16170.564 (22) .814 4-class 31737.364 31900.849 31811.885 .0434 .0456 .0000 -15840.682 (28) .798 5-class 31207.674 31406.191 31298.164 .1470 .1505 .0000 -15569.837 (34) .825 6-class 30818.096 31051.645 30924.554 .0089 .0096 .0000 -15369.048 (40) .825 7-class* 30426.846 30695.428 30549.274 .0000 .0000 .0000 -15167.426 (46) .843 8-class 30167.962 30471.576 30306.359 .2210 .2261 .0000 -15031.981 (52) .844 AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; Adj. BIC = Sample size-adjusted Bayesian Information Criterion; VLMR = Vuong-Lo– Mendall–Rubin likelihood difference test; LMR = Lo-Mendell Rubin; BLRT = Bootstrap likelihood ratio test. *Indicates the selected best profile solution. 1 2 3 4 5 Situation changing Accommodation Devaluation Avoidance Symptom reduction Moderate (N=1251) Low (n=351) Passive (n=34) Blurred (n=300) Active (n=182) High (n=406) The highest (n=13) Figure 2: The level of coping (5 strategies) by seven coping profiles. Table 3 summarizes the differences between coping groups in terms of certain background factors. There were more health care workers than expected amongst the ‘blurred’ and ‘moderate’ copers. ‘High copers’, in contrast, were represented more often than expected among service employees. There were not many gender differences between the coping groups but amongst the ‘low copers’ there were more men than expected. Nor were there many age differences between the coping groups. Only the ‘high copers’ group contained a higher than expected number of individuals who were 28 years or younger. More differences were found in education. The ‘blurred copers’ group contained more individuals than expected with a higher level of education: more employees in this group had polytechnic, Master’s or doctoral degrees than in the group of ‘passive copers’. Also the group of ‘active copers’ contained more individuals with a Master’s degree or a PhD. The ‘high copers’ group had fewer Master’s or doctoral degrees 204 Journal of Basic & Applied Sciences, 2014 Volume 10 Mauno et al. [32] Greenhaus JH and Powell GN. When work and family are allies: A theory of work-family enrichment. Acad Manage Rev 2006; 31: 72-92. http://dx.doi.org/10.5465/AMR.2006.19379625 [33] Carlson DS, Kacmar KM, Wayne JH and Grzywascz JG. Measuring the positive side of the work-family interface. Developing and validation of a work-family enrichment scale. J Vocat Behav 2006; 68: 131-164. http://dx.doi.org/10.1016/j.jvb.2005.02.002 [34] Shirom A. Feeling vigorous at work? The construct of vigor and the study of positive affect in organizations, In Research in organizational stress and well-being, Ed by Ganster D and Perrewe PL. JAI Press 2003; pp. 135-165. [35] Boyar SL, Carr JC, Mosley DC and Carson CM. The development and validation of scores on perceived Work and Family Demand Scales. Educ Psychol Meas 2007; 67: 100115. http://dx.doi.org/10.1177/0013164406288173 [36] Elo A-L, Dallner M and Gamberale F et al., QPSNordic Manual (in Finnish). Helsinki: Finnish Institute of Occupational Health 2006. [37] Muthén LK and Muthén BO (1998-2009). Mplus user’s guide (6th edition). Los Angeles, CA: Muthén & Muthén 1998-2009. [38] Nagin DS. Group-based modelling of development. Harward University Press 2005. [39] Carver CS and Connor-Smith J. Personality and coping. Annu Rev Psychol 2010; 61: 679-704. http://dx.doi.org/10.1146/annurev.psych.093008.100352 [40] Ben-Zur H. Coping styles and affect. Int J Stress Manage 2009; 16: 87-101. http://dx.doi.org/10.1037/a0015731 Received on 14-03-2014 Accepted on 17-04-2014 Published on 23-05-2014 http://dx.doi.org/10.6000/1927-5129.2014.10.27 © 2014 Mauno et al.; Licensee Lifescience Global. This is an open access article licensed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted, non-commercial use, distribution and reproduction in any medium, provided the work is properly cited.