The pragmatic derivation and validation of measures of adaptive and maladaptive coping styles

Abstract There remain challenges involving the theoretical and empirical variability of the operationalization of coping styles. The current analyses sought to identify a parsimonious number of coping styles, from the commonly used Ways of Coping Questionnaire presented in dispositional format, and present internal consistency, construct validity and effectiveness information for the obtained scales. Data were collected from 317 adult community participants. Measures assessed coping, personality, stress, and demographic characteristics such as age and gender. Factor analyses identified that only two unique constructs, and therefore subscales, could be supported by the available data. However, three reliable parallel versions for each coping style, comprising 20, 10 and 5 items, respectively, were able to be derived. Face validity led to these uncorrelated scales being termed adaptive and maladaptive coping. Personality, stress and demographic details provided initial construct validity for these scales. This research proposes empirically sound measures for adaptive and maladaptive coping styles in dispositional format derived from a popular existing instrument. The independence of the scales demonstrates the fundamental distinction between styles, while the availability of scales of differing length with equal reliability offers particular value. These new options will provide clarity and consistency in the quantification of two commonly cited coping styles.


PUBLIC INTEREST STATEMENT
Stress in modern life is inevitable. However, it is how we cope with stress and stressors that determines our capacity to continue to thrive in our day-to-day activities. Agreed terminology and associated measures are therefore critical for clinicians and researchers to be able to reliably study people's different styles of coping. Our broader research program concerns the role played by coping styles in affecting positive or negative outcomes following stressful life events. During this research we became concerned by the inconsistency with which coping activities were described by researchers. Therefore, the analyses reported in this paper illustrate a set of brief, practical, yet robust measures that characterize coping as either adaptive or maladaptive. The broad adoption of our proposed measures will result in the consistent assessment and description of coping styles, allowing comparisons across different samples, settings, and circumstances.

Introduction
It is generally accepted that coping reflects attempts to manage the internal and/or external demands perceived to follow from challenging or threatening stimuli (Lazarus & Folkman, 1984). While this is largely a consensus view, there remain multiple options to operationalize coping. Available measurement models demonstrate the complexity of the construct, with coping variously construed as an outcome, process, behavior, and a set of either trait-based or state-based strategies or styles (Greenaway et al., 2015;Schwarzer & Schwarzer, 1996;Skinner, Edge, Altman, & Sherwood, 2003).
Coping strategies are practical behavioral or cognitive techniques of which examples are humor, self-blame, positive reappraisal, distancing, planning, self-pity, and information searching (Greenaway et al., 2015;Schwarzer & Schwarzer, 1996;Skinner et al., 2003). It is common practice for scales to be developed by nominating individual items that characterize specific coping strategies such as these. However, the attempt to provide breadth of coverage may result in an unwieldy profile of coping dimensions. For example, while the Coping Checklist (Billings & Moos, 1981) surveys only five coping strategies, the Ways of Coping Questionnaire (WCQ; Folkman & Lazarus, 1988) measures eight, and the Coping Orientation to Problems Experienced inventory (COPE; Carver, Scheier, & Weintraub, 1989), Brief COPE (Carver, 1997), and the Coping Strategies Inventory (CSI; Tobin, Holroyd, Reynolds, & Wigul, 1989) assess 14 strategies.
As it is rare for a research study to be able to accommodate such a large number of items or variables to evaluate the single construct of coping, perhaps inevitably researchers seek to aggregate coping strategies into a more manageable number of higher-order coping styles (e.g., emotionfocused versus problem-focused coping; adaptive versus maladaptive coping). For example, the CSI describes a hierarchical model of coping which includes the higher-order coping styles of engagement and disengagement. Similarly, the Coping Styles Questionnaire (CSQ; Roger, Jarvis, & Najarian, 1993) directly quantifies an adaptive style (rational and detached), and a maladaptive style (emotional and avoidant), as does the Maladaptive and Adaptive Coping Styles Questionnaire (MAX; Moritz et al., 2016) which comprises the dimensions of maladaptive, adaptive, and avoidance coping. An additional confusion is that the conceptual labels applied to these styles often refer differentially to function (e.g., appraisal) or the potential for effectiveness (e.g., adaptive/maladaptive).
Beyond these examples, however, researchers commonly engage in ad hoc dimension reduction, the cost of which is often theoretical and/or empirical variability in the definition and operationalization of coping styles depending on sample size and composition, method of data reduction, and the scale used (Schwarzer & Schwarzer, 1996;Skinner et al., 2003). This has led to long-standing confusion surrounding the best way to characterize coping styles. Using the Brief COPE as an illustrative example, its 14 strategies have been used to represent both problem-and emotionfocused styles (Tuncay, Musabak, Engin Gok, & Kutlu, 2008), and adaptive and maladaptive styles (Rohland, 2000). Further, while Meyer (2001) aggregated eight Brief COPE strategies and termed the sum "adaptive", Badura, Reiter, Altmaier, Rhomberg, and Elas (1997) used only four in their "adaptive" scale. In summary, the benefits of aggregated coping style measures include a reduced load for participants, providing a more intense focus on fewer coping practices than if multiple scales are employed. However, a lack of reliable and repeatable scales has often led to inconsistent definitions and item content for the same named styles.
A further contextual issue is that coping may reflect either situational (unique to a stressor) or dispositional (habitual methods) modes of responding to stress (Ayers, Sandler, West, & Roosa, 1996;Bouchard, Guillemette, & Landry-Léger, 2004;Livneh & Martz, 2007). Situational (state) theories emphasize the actual coping of an individual and/or the outcome of the application of coping methods or strategies. Such an approach leads to the microanalytic study of a wide variety of specific and concrete coping strategies. Conversely, dispositional (trait) theories focus on the more general recognition of a person's resources and tendencies related to coping. That is, a macroanalytic approach concentrating on fundamental and abstract coping methodologies (styles). The former (situational) theory appears to be dominant, comprising the examination of strategies chosen to suit the specifics of the threat experienced (i.e., coping responses relevant to specific environmental demands).
Other potential associations include that dispositional coping emotions predict situational coping emotions when participants experience high distress, but not in low distress situations (Carver & Scheier, 1994). The connection between dispositional coping and health outcomes has also been noted to be mediated by situational coping (Bouchard et al., 2004;Catanzaro et al., 2000;Sasaki & Yamasaki, 2005). These latter examples suggest an interaction between situational and dispositional styles that is complex, but that both nevertheless have a role to play in describing coping.
Notwithstanding these observations, relatively few instruments measure coping in dispositional format, or permit the comparison of dispositional and situational coping. Those that do include the COPE and Brief COPE (Carver et al., 1989), Children's Coping Strategies Checklist (CCSC), How I Coped under Pressure Scale (HICUPS) (Ayers et al., 1996), and Emotional Approach Coping Scales (Stanton et al., 2000). However, despite being one of the oldest and most frequently used measures of coping strategies, the WCQ has not yet been changed into a dispositional format. While Folkman and Lazarus (1988) did not rule out dispositional measurement using the WCQ, they noted that the psychometric properties of the WCQ may not hold for a dispositional version.

Current study
The data to be reported are sourced from a yet to be published parent study concerned with the mediating role of coping style in the association between life event stress and stress appraisal. Using the WCQ in dispositional format, poor internal reliability data were noted for the original eight scales. This is perhaps not surprising as there remains uncertainty concerning the structure of these scales even in situational format (e.g., Lundqvist & Ahlström, 2006;Senol-Durak, Durak, & Elagöz, 2011). The replication of the original eight factors has been noted as challenging (Cook & Heppner, 1997;Edwards & O'Neill, 1998;Rózsa et al., 2008), with the suggestion that researchers conduct independent factor analyses to construct their own scales (Parker, Endler, & Bagby, 1993).
The initial research objective was therefore to propose a parsimonious number of coping styles in dispositional format using the WCQ. Exploratory factor analysis was used as a heuristic tool to examine the clustering patterns within the item pool, with the results serving as a pragmatic guide to scale membership (Kim & Mueller, 1985). The internal consistency of the derived scales was also assessed. The second research objective was to present valuable first steps toward the construct validity of the derived scales which comprised the exploration of variables previously reported to be associated with coping style. These were personality (Bouchard et al., 2004;Carver & Connor-Smith, 2010;Panayiotou, Kokkinos, & Kapsou, 2014;Watson & Hubbard, 1996), stress (Holahan & Moos, 1987;Holton, Barry, & Chaney, 2015;Mahmoud, Staten, Hall, & Lennie, 2012;Meng & D'Arcy, 2016;Parasuraman & Hansen, 1987), age (Ben-Zur, 2002;Diehl et al., 2014), education level (Holahan & Moos, 1987), marital status, and gender (Kaiseler et al., 2012;Tamres, Janicki, & Helgeson, 2002).

Participants and procedure
A convenience sample was sourced from adult educational institutions, leisure organizations, and occupational groups in Australia. A brief précis of the study was presented (at group meetings where possible), with those providing informed consent then given a written summary, questionnaire booklet and envelope. Questionnaires comprising basic demographic questions and the scales described below were either completed at the time or returned later via reply-paid post. The sample consisted of 317 participants with a mean age of 39.8 years (SD = 14.0, range 18-65 years). The majority were female (51%), living with a partner (61% married), and had achieved post-secondary education (51%). The study was conducted with the approval of the authors' institutional ethics committee.

Personality
The Mini-IPIP is a 20-item version of the 50-item International Personality Item Pool (Goldberg, 1999). It assesses Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism (4 items per trait). Participants indicate their agreement with each item using a 5-point scale ("very inaccurate" to "very accurate"), with responses summed for each trait. Comparable convergent, discriminant and criterion-related validity to that of longer Big Five measures has been reported (Donnellan, Oswald, Baird, & Lucas, 2006), with internal consistency (α) ranging from 0.65 to 0.82. The current sample provided similar coefficients (0.67-0.81).

Stress
A revised Social Readjustment Rating Scale (Hobson & Delunas, 2001) comprising 51 life events was used. Participants indicate which they have experienced within the last year. Allocated stress points range from 22 (minor law violation) to 87 (death of spouse/mate). Life event stress was determined by summing the points associated with the events experienced. Participants were also asked to use a 4-point scale to provide an appraisal of the stress currently felt in relation to the events experienced (0 = "not at all stressful" to 3 = "extremely stressful"). Stress appraisal was the sum of the products of the stress points for each life event experienced and the relevant appraisal rating.

Results
An exploratory factor analysis was conducted with maximum likelihood extraction as it allows generalization from a sample to a population, and gives more weight to correlations with more unique variance and less error variance (Kim & Mueller, 1985). Parallel Analysis criteria (Lautenschlager, 1989) identified five factors for retention (35% variance accounted for) with satisfactory goodness-of-fit (χ 2 (1825) = 2751.65, p < .001). Varimax rotation was then used to maximize uniqueness. This model lacked clarity. Scales derived from Factors 1 and 4 were highly correlated (r = −0.67, p < .001), and Factor 3 was also correlated with Factor 1 (r = −0.35, p < .001) and Factor 4 (r = 0.33, p < .001). Factors 2 and 5 were also highly correlated (r = 0.53, p < .001). These observations suggested that as few as two unique constructs may be reflected in the data.
A two-factor solution (goodness-of-fit: χ 2 (2014) = 4028.05, p < .001) accounted for 23% of variance. With loadings of 0.25 or above on their secondary factor, two Factor 1 items and four Factor 2 items were not considered further for scale membership. Inspection of the remaining top 20 items of each factor led to the labels of Adaptive Coping (Factor 1) and Maladaptive Coping (Factor 2) being applied. Table 1 shows original item numbers and scale membership, and factor loadings. Three parallel scales were tested from each factor (20, 10 and 5 items). In all cases these were uncorrelated, signifying a non-linear association between adaptive and maladaptive coping (Table 2). Conversely, all versions of both Adaptive (r = 0.90-0.95) and Maladaptive Coping (r = 0.88-0.95) were highly correlated. High α coefficients (Table 2) were also achieved for all versions suggesting that the shorter scales may be used with little loss of precision.
Tables 3 and 4 present preliminary construct validity data. In Table 3, correlations with age, personality and stress measures are shown. Consistently, Adaptive Coping was unrelated to age, while maladaptive strategies were more likely from younger participants. In order of strength, Extraversion, Openness, Agreeableness and Conscientiousness were consistently positively associated with adaptive strategies. Neuroticism showed modest negative associations with Adaptive Coping but was strongly positively related to Maladaptive Coping. Conscientiousness was negatively related to Maladaptive Coping while Extraversion and Agreeableness showed no such association. The coefficients for Openness were small and inconsistent. Higher life event stress was indicative of greater endorsement of both adaptive and maladaptive styles, while current stress appraisal was positively associated with Maladaptive Coping, but shared no association with Adaptive Coping. Further, the re-analysis of current stress appraisal with coping controlling for life event stress produced equivalent results, albeit with slightly more modest associations with Maladaptive Coping. Table 4 presents t test results that explore demographic differences in coping styles. Adaptive Coping varied only in terms of education, being consistently higher among participants with postsecondary education. Variations in Maladaptive Coping were present for all variables, dependent on the scale version. Females reported greater use of maladaptive strategies (10 items), participants with no partner used more maladaptive strategies (20 and 10 items), and those with postsecondary education recorded a lower use of maladaptive strategies (20 items). Note that for all significant analyses effect sizes were modest.

Discussion
These analyses sought to determine the number and nature of coping style dimensions inherent in the WCQ when measured in dispositional format. Only two unique factors could be extracted, proposed to reflect adaptive and maladaptive coping, respectively. Preliminary construct validity data for three reliable versions of each of these coping styles, comprising 20, 10 and 5 items, respectively, was then provided. The independence of the two styles was maintained regardless of the length of the scales, further reinforcing the fundamental distinction between adaptive and maladaptive coping, and providing scale options that are rarely available.
It is perhaps surprising that, from among 66 items, only two factors could be realistically supported. An important observation is that the five factors initially extracted were far from distinct. Plausible labels were (1) Planful Problem Solving, (2) Escape-Avoidance (with Accepting Responsibility), (3) (inverse of) Social Support (with Self-Controlling), (4) (inverse of) Positive Reappraisal, and (5) Escape-Avoidance, Distancing and Confrontive Coping. Yet the decision to rotate only two factors resulted in a much clearer interpretation. The lack of correlation between the scales derived from the two factors is distinctive and speaks to the independence of these coping styles. This provides confidence in the ability of the proposed scales to discriminate between participants reporting, for example, a clear reliance on one style over the other, or the tendency to use both styles. Current alternative measures of adaptive and maladaptive coping, such as in the MAX (Moritz et al., 2016), share a significant negative correlation. Yet as noted by Moritz et al. (2016) themselves, there is no reason to suggest that adaptive and maladaptive styles should be reciprocal (opposing poles of a continuum). They further note that the study of concurrent coping styles, as reflected in the current research, has emerged only relatively recently.      Note: * p < .05; ** p < .01; *** p < .001 (two-tailed).

Item
Primary support that the identified constructs may be labelled "adaptive" and "maladaptive" was evident from item content (i.e., face validity), and also from previous similar scales in the literature (Badura et al., 1997;Meyer, 2001;Moritz et al., 2016;Roger et al., 1993;Rohland, 2000;Tuncay et al., 2008). The putative adaptive scales were heavily represented by items previously reported as members of the Positive Reappraisal and Planful Problem Solving scales, with modest representation from Seeking Social Support. Three adaptive coping items had not previously loaded on a WCQ scale (e.g., "I try to analyze the problem in order to understand it better"). Overall, the adaptive scale characterizes a person who deals with stressors through personal growth, optimism, solution-focused actions, creativity, and flexibility.
The maladaptive scales predominantly tapped former Escape-Avoidance and Accepting Responsibility scales and, again, items not previously represented in WCQ scales also contributed (e.g., "I wish that I could change what had happened or how I felt"). A person scoring highly on maladaptive coping could be characterized as engaging in denial, self-criticism, and passivity. While this content and label largely accord with a consensus view of the nature of "maladaptive" coping, it could perhaps be argued that emotional distancing at a time of significant crisis is in fact adaptive for those who find the stress experience overwhelming. The need for further validity testing against external criteria such as successful/unsuccessful adaptation to stressors is therefore encouraged.
Preliminary construct validity was also supportive of the naming of the derived scales. For example, personality is a key predictor of how individuals choose to deal with stressors (Carver & Connor-Smith, 2010), and has previously been linked to adaptive and maladaptive styles. The current results largely replicated previous observations. Specifically, associations involving extraversion, openness and conscientiousness with adaptive coping, and neuroticism with maladaptive coping, are common (Bouchard et al., 2004;Carver & Connor-Smith, 2010;Panayiotou et al., 2014;Watson & Hubbard, 1996). The current data additionally linked agreeableness with adaptive coping. Further, while conscientiousness has primarily been positively related to an adaptive style, the current data demonstrated a negative association with maladaptive coping. Importantly, all associations with personality were consistent across parallel adaptive and maladaptive scales, respectively.
Stress and coping relationships were also relatively consistent across the three scale versions. The observation that higher maladaptive scores were associated with a higher current stress appraisal reflects previous research illustrating that maladaptive rather than adaptive coping is more directly related to stress outcomes (Mahmoud et al., 2012;Meng & D'Arcy, 2016). The finding that both adaptive and maladaptive coping styles were higher as more stress was experienced is also consistent with previous research (Holahan & Moos, 1987;Holton et al., 2015;Parasuraman & Hansen, 1987). Further, the significant partial correlations between current stress appraisal and coping styles, controlling for the amount of life stress experienced, suggest that the increased use of maladaptive strategies occurs when and if the stress experienced is appraised more negatively, while the use of adaptive coping strategies appears to be independent of the level of appraisal. Beyond this observation, it is also clear that the greater experience of stress mandates the engagement of more coping responses per se, demonstrated by the endorsement of higher levels of both coping styles, perhaps individually chosen to deal with specific stressors. It may be that such a pattern would not be observed if situational coping responses were sought, forcing respondents to nominate strategies for a single stress experience.
Demographic characteristics provided less consistent relationships, with the exception of age. While adaptive coping was not related to age, the use of a maladaptive style decreased with maturity. Although there is limited research on age-related coping styles (Rubio, Dumitrache, Cordon-Pozo, & Rubio-Herrera, 2016), this association has been noted previously (Ben-Zur, 2002;Diehl et al., 2014). Beyond age there were indications that maladaptive coping was more likely among women, those without a current life partner, and those with less education. However, these effects were variable across the three scale versions. Conversely, adaptive coping was strongly and consistently used by those with at least post-secondary education, replicating the findings of Holahan and Moos (1987). No other demographic effects were evident for adaptive coping.
There are alternate theories (e.g., sex role, socialization) to support either similarities or differences in the use of coping styles by men and women (Gattino, Rollero, & Piccoli, 2014;Kaiseler et al., 2012;Sigmon, Stanton, & Snyder, 1995). Kaiseler et al. (2012) suggest both biological and social reasons why men and women use different coping styles in similar stressful conditions (dispositional), and also note that differences in situational coping may reflect differential appraisals of stressors rather than an overall preference for a coping style. Generally, there is limited information on gender for the WCQ. Commonly, no gender differences are reported in situational format (Porter et al., 2000). Similarly, few gender differences were found using a dispositional format in the current study. However, the one noted effect was in line with Hampel and Petermann's (2005) research with adolescents, illustrating that females report higher maladaptive scores. More common is that women simply report more coping efforts than men overall, regardless of style (Kaiseler et al., 2012;Tamres et al., 2002). This was not supported by the current data.

Future research
While the adaptive and maladaptive measures reported are extremely encouraging for those who desire brief, pragmatic scales, caveats are in order. Clearly, replication is necessary (in both situational and dispositional form), as is further validation in general. Consideration of criterion validity, for example, and convergent validity against existing coping measures are clear next steps. Such studies should involve large, representative samples and apply confirmatory factor analysis as a more stringent test of the factor structures nominated. This has not been attempted in the current paper due to cautions against using both exploratory and confirmatory procedures with the same data (Kline, 2016). Further, given the size of the WCQ (66 items), splitting the current sample into development and validation samples (approximately 150 participants in each) would particularly compromise the development analyses. The availability of a larger sample would also allow more complex statistical techniques such as multiple group structural equation modeling, latent profile analysis or multidimensional scaling to be considered which may be of value in testing the measurement invariance of the proposed scales. Finally, context may influence item responses. Therefore, an important test will be whether the proposed scales have the same properties, such as internal reliability, when items are not presented as components of the 66-item WCQ.
Beyond psychometric issues, fruitful inquiry will involve the description of the patterns of use of adaptive and maladaptive coping styles. Given the uncorrelated nature of the scales, high scores on one do not automatically equate to low scores on the other. That is, adaptive and maladaptive do not represent two ends of a single continuum. Rather, their association is likely to be complex, perhaps varying according to a range of individual differences. The current analyses have not addressed whether one particular style predominates for individuals. Cluster analysis or latent class analysis may prove informative in characterizing specific patterns of use of the two styles. Finally, as coping styles may best be considered to be sets of behaviors that mediate the relationship between stress (or stressors) and outcomes, future research should also seek to determine whether the proposed maladaptive and adaptive scales effectively predict differential successful/ unsuccessful outcomes as might be suggested by the labels "maladaptive" and "adaptive".

Conclusion
Providing operationalization of the higher order adaptive and maladaptive coping style constructs, these analyses have proposed three empirically sound measures for each in dispositional format derived from an existing instrument. While the terms are frequently used descriptively, it is rare that specific adaptive and maladaptive measures are available. The choice of scales of differing length with equal reliability is even less common. The longer scales may be employed when number of items is not an issue and slightly greater measurement precision is perhaps desired, while the shorter scales may prove ideal when longer scales are not practical. It is suggested that these scales would allow both researchers and practitioners to assess adaptive and maladaptive coping styles with greater confidence than offered by existing options.