- Email: [email protected]
Rasch Analysis of the Coping Inventory for Stressful Situations in Individuals With Moderate to Severe Traumatic Brain Injury
Archives of Physical Medicine and Rehabilitation journal homepage: www.archives-pmr.org Archives of Physical Medicine and Rehabilitation 2015;96:659-66
ORIGINAL RESEARCH
Rasch Analysis of the Coping Inventory for Stressful Situations in Individuals With Moderate to Severe Traumatic Brain Injury Hillary A. Greene, MA,a Lisa J. Rapport, PhD,a Scott R. Millis, PhD,b,c Robin A. Hanks, PhD,b,c Michael W. Williams, MAa From the aDepartment of Psychology and bSchool of Medicine, Department of Physical Medicine and Rehabilitation, Wayne State University, Detroit, MI; and cRehabilitation Institute of Michigan, Detroit, MI.
Abstract Objective: To evaluate psychometric properties of the Coping Inventory for Stressful Situations (CISS) in individuals with traumatic brain injury (TBI). Design: Archival study using Rasch analysis. Setting: Postacute rehabilitation hospital. Participants: Adults (NZ331) 1 to 15 years after moderate to severe TBI, recruited consecutively. Interventions: Not applicable. Main Outcome Measure: CISS. Results: Indices of unidimensionality and model fit supported the scale’s proposed multidimensional structure consisting of Task, Emotion, and Avoidant coping style; 3 unidimensional scales showed better fit than a single combined scale. The 3 scales met Rasch expectations of reliability and separation for persons and items, as well as adequate response category functioning. The scales were generally well targeted but showed some evidence of ceiling effect for Task, and floor effects for Emotion and Avoidant coping; item difficulties did not fully capture extreme ranges demonstrated by some participants, suggesting that measurement of coping after TBI on the CISS would be improved with additional items at low and high ranges of difficulty. Results were generally equivalent for cross-sectional groups representing short-term (1y), intermediate (2y), and long-term (5e15y). Conclusions: The CISS showed good psychometric properties as a measure of coping style among persons with moderate to severe TBI in acute and chronic phases of recovery, and showed evidence of multidimensionality as predicted by theory, consistent with 3 unidimensional scales. Added items tapping broader (or more accessible, less cognitively complex) ranges of coping responses would likely benefit the scale overall and improve correspondence with the response needs of people with TBI. Archives of Physical Medicine and Rehabilitation 2015;96:659-66 ª 2015 by the American Congress of Rehabilitation Medicine
Coping style is an aspect of psychological functioning that may influence recovery from traumatic brain injury (TBI).1-3 Prevailing theories suggest a 3-factor structure of coping style: problem-, emotion-, and avoidance-focused.4,5 Avoidant and emotion coping styles are associated with relatively poorer functioning, physical
Presented in part as a poster to the American Psychological Association, February 6e9, 2013. Supported by the National Institute on Disability and Rehabilitation Research (grant no. H133A020515-03) and Wayne State University Thesis Research Support. Disclosures: none.
health, and subjective well-being than task coping.5-7 Given the high prevalence of cognitive and emotional deficits, this 3-factor structure of coping style might not well characterize coping among people with TBI.8 Cognitive deficits may alter how people think about and engage in managing stressors.9,10 Cognitive impairments can undermine attention, memory, and executive functions needed for identifying, planning, and implementing complex coping responses necessary for task coping.11-14 Emotional disturbances, low frustration tolerance, or inertia post injury may favor an
0003-9993/14/$36 - see front matter ª 2015 by the American Congress of Rehabilitation Medicine http://dx.doi.org/10.1016/j.apmr.2014.11.006
660 emotion-focused response to stressors over active, task-focused coping.15-17 Identifying the need to invoke adaptive action may be disrupted by TBI via impaired awareness of deficits associated with neurologic dysfunction, impaired appreciation of deficits, or denial,18,19 which may increase the use of avoidant coping strategies.19 Most studies of coping style in TBI have falsely assumed that reliability generalizes from psychometric support established for non-TBI populations.20 Although multiple scales are available to assess coping in general populations (eg, Ways of Coping Questionnaire,21 COPE),22 they do not appear to be the strongest options psychometrically20,23,24 or clinically.22,25 The Coping Inventory for Stressful Situations (CISS) was developed by Endler and Parker4 to address psychometric weaknesses of earlier scales. The CISS includes 3 scales to assess task-, emotion-, and avoidance-focused coping, and is shorter than other scales, which is desirable for assessments of people with TBI. Research using traditional psychometric approaches (ie, classical test theory) indicates that the CISS is a reliable, multidimensional, valid measure of coping style among healthy and clinical populations, making it potentially a good choice for examining coping in TBI.4,7,26,27 Rasch modeling28 is well suited to examine the CISS. Rasch models are developed with the goal of establishing an ideal scale with interval-level measurement and examining how well the data fit the ideal model, whereas classical test theory approaches strive to develop a model to describe the data.29 Rasch analyses also enable examination of scale functioning at item and overall levels, scale dimensionality, and category response functioning. The primary aim of this study was to use Rasch modeling to improve understanding of how the CISS functions in TBI and whether it functions best in its original form or if alterations might improve its measurement properties. A secondary aim was to evaluate whether time since injury meaningfully affects CISS reliability by using Rasch analyses to assess the psychometric properties separately for respondents in acute versus chronic stages of recovery from TBI. This aim was particularly well suited for Rasch analysis, which affords the opportunity to calculate separate reliabilities for persons and items. Lastly, CISS dimensionality was of interest. Although it was not a primary purpose of the approach, Rasch analysis can provide some information regarding whether the proposed 3-dimensional structure of the CISS holds among respondents with TBI versus a single dimension of coping response, as might be driven by global cognitive impairment.
Methods Participants A total of 331 adults were recruited consecutively from a large longitudinal study of TBI. All participants had medically documented TBI, received treatment at an affiliated level-1 trauma center within 24 hours of injury, and required inpatient rehabilitation. Persons whose injuries resulted from anoxic encephalopathy, who were non-English speaking, or whose injury severities precluded valid evaluation were excluded.
List of abbreviations: CISS Coping Inventory for Stressful Situations TBI traumatic brain injury
H.A. Greene et al Table 1
Characteristics of the 331 participants
Characteristics
Values
Age at follow-up (y) Sex Women Men Education (y) Ethnicity African American White Latino(a) Native American Asian or Pacific Islander Time since injury (y) GCS (total at emergency department admission) Days to follow commands* Disability Rating Scale
44.013.5 61 (18.4) 270 (81.6) 11.92.1 234 (70.7) 87 (26.3) 5 (1.5) 3 (0.9) 2 (0.6) 6.14.9 9.3 (3e15) 7.212.1 6.52.9
NOTE. Values are mean SD, n (%), or median (range). Abbreviation: GCS, Glasgow Coma Scale. * Days to follow commands is a motor subscale of GCS; days from injury to follow commands.
Participants completed the CISS between 1 and 15 years after TBI (mean SD: 6.14.9y). The median Glasgow Coma Scale total at admission was 9 (range, 3e15), which corresponds to moderate TBI. Persons with complicated mild injuries with intracranial hemorrhage were included because their neuropsychological profiles are similar to those of individuals with moderate TBI.30,31 Demographic and injury severity characteristics are shown in table 1.
Measures: CISS The CISS4 is a 48-item self-report measure that uses a 5-point scale (1 [not at all] to 5 [very much]). It provides three 16-item scales (Task, Emotion, Avoidance) to assess specific coping styles, which are scored independently of one another. The Task scale focuses on problem solving, planning, and efforts to alter difficult situations; the Emotion scale focuses on self-oriented responses, fantasizing, or emotions, including efforts to deal with stress; and the Avoidance scale focuses on efforts to relieve stress by distraction with unrelated activities or social interactions.4 The CISS has well-demonstrated reliability and validity when used in healthy4 and clinical27 populations.
Procedures The study was approved by the institutional review board. These archival data were collected at scheduled follow-ups of 1, 2, 5, 10, and 15 years post injury. Only 1 CISS per participant contributed to the current dataset. Three cross-sectional time points post injury were examined: 1 year, 2 years, and chronic (5y post injury).
Data analyses Descriptive statistics and normative scores for the CISS scales were examined; univariate analyses of variance were used to compare the scales across each follow-up time. Rasch analyses were conducted for each of the scales at each follow-up time. Examining the scales at multiple time points post injury affords the opportunity to examine how well the CISS functions psychometrically among www.archives-pmr.org
Rasch analysis of Coping Inventory for Stressful Situations in TBI
661
individuals with TBI at different stages of recovery and adjustment to TBI that potentially could influence coping responses and reliability of responding. Rasch analyses also were conducted at each time point using all 48 items to test unidimensionality. The Rasch rating scale model32 was used to evaluate the CISS rather than the partial credit model because the CISS items all share the same rating scale structure. The partial credit model is used when each item has a unique rating scale structure. Measurement properties assessed included estimates of reliability, dimensionality, item and person targeting, and response category functioning. Models of interest for comparison included the widely accepted 3-factor model, a unifactorial model, and a null model (ie, a lack of dimensionality in the coping data). Winsteps 3.74 softwarea was used to conduct Rasch analyses. SPSS (version 20) softwareb was used for the remaining analyses. Unidimensionality was examined using Rasch principal component analysis of residuals, which is supported by unexplained variance in the first contrast of eigenvalues <3 and variance in data explained by the measure 60%.33 Unidimensionality also was evaluated with tests of local independence of the items, or response dependency34; a nonsignificant test indicates that after considering the Rasch factor, no associations remain among items other than random associations. Raw score residual correlations were computed, which correspond with Yen’s Q3 statistic. Test of fit to the model was evaluated with infit and outfit statistics. Infit is a weighted statistic of overall performance of an item with the model, whereas outfit is an unweighted statistic that reflects the extent to which outliers are present in the model.29 The metrics of infit and outfit statistics are weighted and unweighted normalized mean-squares, respectively.29 When data are predicted well by the Rasch model, infit and outfit values approximate the expected value of 1.0, with extreme values ranging from 0 to infinity.29 When data poorly fit the model (unpredictability), fit statistics are >1.0, and when data overfit the model (redundancy), fit statistics are <1.0.29 Because the measurement of coping has implications for assessment and prediction of real-world functions for persons with TBI, relatively stringent mean-square values from 0.7 to 1.3 were considered “excellent” fit statistics in this study, whereas fit statistics between 0.5 and 2.0 are considered “good” generally.33,35 Infit and outfit statistics for the total CISS scale and individual Task, Emotion, and Avoidance scales were each examined at the 3 time points, corresponding with the aim to compare the psychometric functioning of the CISS across early and late stages of recovery from TBI. Person reliability (analogous to Cronbach a) and item reliability were interpreted using Fisher’s recommendations,33 with “good” reliability >.80. Person (or item) separation indices >2.0 were interpreted as support for internal consistency. Targeting, which refers to matching of item difficulty to participant ability for a specific measure used with a specific sample, was assessed by examining person-item maps. The Rasch analysis itemperson map, also called the Wright map, facilitates assessment of targeting. On the map, items are plotted by difficulty on a logit scale, which are the raw scores converted to a scale with interval properties; persons are plotted by symbols that represent either 4 persons (#) or 1 to 3 persons (.). Symbols on the map denote the mean (M), 1 SD (S), and 2 SDs (T) from the mean for persons (left side of the midline) and items (right side of the midline). “Ability” is reflected from most to least: cases with the highest scores (at the top of the map) most easily endorsed (agreed with or performed correctly) the items. Similarly, “difficulty” is presented from most to least: items with the highest scores (top of the map) were most difficult to endorse. Good targeting of difficulty and ability ideally results in a symmetrical spread of
items and persons along the vertical axis of the item-person map.29 To obtain “uniformly precise measurement,” the items should cover a wide difficulty range, such that the item difficulties are spaced consistently, there are few gaps between items, and item difficulty levels correspond to person ability levels.36 Additionally, scores should span the range possible on the scale, with the sample mean approximating the midpoint of the scale and less than 15% obtaining the highest or lowest possible scores; exceeding that percentage provides evidence of ceiling and floor effects, respectively.37 Adequacy of the response categories at scale and item levels was assessed via consistency with Rasch category ordering principles, which expect that average category endorsement should increase monotonically.38 Andrich thresholds were examined for step disordering, which also expects monotonic increases.29,38
www.archives-pmr.org
Results For raw scores, the sample used more Task (mean SD: 57.313.1) than Emotion (mean SD: 44.113.7) and Avoidant (mean SD: 41.912.2) coping, which is consistent with the pattern observed for the normative sample. Univariate analyses of variance indicated that mean scores did not differ significantly across the follow-up time points on the Task scale (F2,328Z.40, PZ.67, h2Z.00), Emotion scale (F2,328Z1.75, PZ.18, h2Z.01), or the Avoidance scale (F2,328Z2.09, PZ.13, h2Z.01). However, normative scores indicated that the total sample used relatively less Task (zZ.15) and more Emotion (zZ.36) and Avoidant (zZ.27) coping compared with healthy community-dwelling individuals.39
Unidimensionality and model fit Across all time points, infit and outfit mean-square statistics for the CISS as a total scale and the individual Task, Emotion, and Avoidance scales were generally acceptable (0.5e2.0), suggesting that items fit the expectations of the rating scale models. No items had “poor” (<.33 or >3.0) fit with the model.33 Rather, across all analyses, most items had “excellent” fit (0.7e1.3) with the model, and all items with the exception of 2 (items 6, 48) had “good” fit (0.5e2.0) with the model.33 Item 6 had good to excellent fit at year 2 and chronic times, and had slightly less than good fit at year 1 (eg, infit 2.03, outfit 1.92). Item 48 had excellent fit at year 1 and chronic times, and had slightly less than good fit at year 2 (eg, infit 1.57, outfit 2.1). In the context of this generally good performance, most occasions in which items fell outside the “excellent” range occurred in the year-1 group, followed by the year-2 group. For all Rasch principal component analyses of residuals, the observed raw variance explained by the measure closely approximated or maximized the Rasch model value, which is the best value attainable if the data exactly fit the Rash model, with all differences between empirically observed and modeled variances being less than or equivalent to 1% (table 2). Although these values were below the ideal 60% threshold,33 models that examined the CISS scales separately performed uniformly better on variance explained than did the models that examined all items as a single scale. Furthermore, all models testing a single scale (all items) had eigenvalues >3 for unexplained variance in the first contrast, which is evidence against unidimensionality.33 In contrast, all models testing scales separately had eigenvalues <3 for the unexplained variance in the first contrast, indicating that the CISS showed superior fit as 3 separate scales (Task, Emotion, Avoidance), consistent with the structure proposed by the test authors. Raw score residual correlations also were computed for all analyses, which correspond with Yen’s Q3 statistic, and no response dependency was found.
662
Table 2
Select diagnostic data for Rasch analysis with CISS*
Scale Year
y
Infit statistic Outfit statistic Person separation ratioz Person reliabilityx Item separation ratioz Item reliabilityx Raw variance in data explained by measures in principal component analysis of residuals Empirical (%) Modeled (%) Unexplained variance in 1st contrast in principal component analysis of residuals Raw %
Task
Emotion
Avoidance
All Items
1
2
Chronic
1
2
Chronic
1
2
Chronic
1
2
Chronic
1.04.32 1.05.32 2.42 0.85 3.14 0.91
1.02.22 1.04.28 2.65 0.88 2.63 0.87
1.01.17 1.04.22 2.24 0.83 3.96 0.94
1.02.23 1.07.31 2.15 0.82 3.94 0.94
1.02.19 0.981.08 2.11 0.82 3.45 0.92
1.01.15 1.02.20 2.13 0.82 5.13 0.96
1.02.24 1.07.32 2.25 0.83 4.73 0.96
1.02.22 1.03.36 2.1 0.81 4.59 0.95
1.02.13 1.04.14 1.81 0.77 6.74 0.98
1.02.24 1.04.27 2.89 0.89 4.53 0.95
1.01.22 1.03.24 2.59 0.87 4.38 0.95
1.00.14 1.03.14 2.83 0.89 6.82 0.98
43.2 44.1
48.9 49.9
41.3 42.2
41.9 42.4
39.0 39.8
41.2 41.8
43.4 43.9
44.0 44.8
37.6 38.1
31.9 32.4
32.3 32.3
32.8 32.9
2.2 7.9
2.0 6.3
1.9 7.1
2.2 7.9
2.6 10.1
2.1 7.8
2.7 9.4
2.7 9.3
2.5 9.7
8.3 11.8
9.4 13.3
8.1 11.3
NOTE. Values are mean SD or as otherwise indicated. * All values reflect inclusion of all relevant items (no dropped items). y Year 1 (nZ104), year 2 (nZ90), chronic (nZ205). z Separation ratios: poor (<2), fair (2e3), good (3e4), very good (4e5), excellent (>5).33 x Reliability estimates: poor (<.67), fair (.67e.80), good (.81e.90), very good (.91e.94), excellent (>.94).33
H.A. Greene et al
www.archives-pmr.org
Rasch analysis of Coping Inventory for Stressful Situations in TBI
663
Fig 1 Item-person map of the CISS, Task scale, chronic group. Abbreviations: M, mean; S, 1 SD from the mean; T, 2 SDs from the mean.
Reliability Table 2 shows person and item separation ratios and reliability estimates. Person reliability estimates ranged from .77 to .89; with 1 exception, all estimates met or exceeded the criterion for “good” (.80) reliability.33 Item reliability estimates ranged from .87 to .98; all estimates met or exceeded “good” reliability criterion, and nearly all corresponded to “very good” (.91e.94) to “excellent” (>.94) reliability.33 Person and item reliabilities for the Task, Emotion, and Avoidant scales were generally equivalent across the 3 follow-up time groups, falling within the same classification range. Person separation ratios were generally “fair” (2e3), whereas item separation ratios were generally “good” (3e4) to “very good” (4e5). Person separation ratios appear equivalent for the 3 time-point groups; item separation ratios appear higher for the chronic group (“very good” to “excellent”) as compared with the 1- and 2-year postinjury groups (generally “good”).
Targeting Examination of person distributions based on the percentage of participants who obtained the highest or lowest possible scores on the CISS scales across all analyses indicated that there were no significant problems with ceiling or floor effects (ie, none www.archives-pmr.org
Fig 2 Item-person map of the CISS, Emotion scale, chronic group. Abbreviations: M, mean; S, 1 SD from the mean; T, 2 SDs from the mean.
exceeded the 15% criterion).37 Additionally, the means for item difficulties and person abilities (shown in item-person maps) across the CISS scales for all time points generally were matched well.37 The item-person maps for Task, Emotion, and Avoidant scales were very similar for the 3 postinjury times; maps for the chronic group are presented as examples. As shown in figure 1, for the Task scale, there was some evidence of extreme scores at the high end of the logit scale on the person map, extending beyond the cluster of scores on the item map, with much less evidence of extreme scores at the low end of the map; this pattern was evident across all time points. For the Emotion scale (fig 2), extreme scores on the person map extended at the high and low ends, though more extreme at the low versus high end of the scale, beyond the spread of item scores for all 3 time groups. For the Avoidance scale (fig 3), extreme person scores appeared generally
664
H.A. Greene et al disordering, count data and fit statistics were considered to evaluate whether items needed to be dropped from the scale. The Andrich thresholds (step-calibration estimates) showed disordered thresholds across all scales and time points assessed. However, there were no problems with fit statistics or category disordering overall, and no further action was warranted to address threshold disordering.
Discussion
Fig 3 Item-person map of the CISS, Avoidance scale, chronic group. Abbreviations: M, mean; S, 1 SD from the mean; T, 2 SDs from the mean.
more at the low ends than the high ends of the scales across all time points. Although participants reported higher and lower amounts of coping ability than individual items were able to assess fully, the CISS scales did not show overall marked problems with targeting through ceiling or floor effects.
Response categories The CISS showed adequate category functioning, with no major problems with category disordering across any of the scales or time points. Each sequential response option was associated on average with increasing item difficulty (ie, intensity of agreement), which indicated that the categories increased monotonically. Category ordering was also examined at an item level across scales and time points. For items that showed category
Coping style among individuals with moderate to severe TBI can be reliably assessed using the CISS. The CISS performed as a multidimensional measure of coping in TBI and reliably captured task-, emotion-, and avoidance-focused coping styles across shortand long-term adjustment to injury. Establishing the reliability of the CISS among persons recovering from TBI provides a necessary foundation for examining a broad range of questions about stress and coping within this population. These findings support the ability of persons with moderate to severe TBI to complete measures such as the CISS reliably, given the array of cognitive impairments often experienced.40 The CISS demonstrated good reliability even at 1 year after injury, which is a relatively early stage in recovery often associated with significant global cognitive impairments that might otherwise undermine reliable self-estimates.41 Similarly, it might have been the case that global cognitive deficits associated with TBI would yield an undifferentiated (global) coping response, in which specific types of coping are not discernible and instead give rise to a generalized mass of positive or negative reactions to stress. Therefore, the findings regarding clear dimensionality of the CISS in this population at 1 year post injury are particularly striking. These findings are consistent with prior work demonstrating the reliability of the CISS among healthy populations,7,39,42-44 persons with physical health problems,26 and persons with mental health problems.27 The CISS functions much better psychometrically as a multidimensional rather than a unidimensional measure of coping style, which is consistent with the structure and purpose designed by Endler and Parker5 and prior research that conceptualizes coping as a multidimensional construct characterized by task, emotion, and avoidant responses.4,22,45,46 Therefore, it is likely that adults with TBI engage in similar types of coping responses as none brain-injured adults; however, the relative reliance on each strategy appears to differ. Normative comparisons indicated that persons with TBI more heavily relied on emotion-focused and avoidant coping than do healthy adults, which highlights the importance of assessing coping after TBI to inform rehabilitation planning. Rasch analyses enhanced understanding regarding the psychometric characteristics of the CISS as used in TBI beyond confirming reliability and multidimensionality. Consistent with the intended function of the CISS,39 the findings indicated that all 48 items of the CISS and all 5 response options contributed meaningfully to the assessment of coping in TBI, with no evidence to support dropping any items or altering the response options. Some evidence indicates that items for the CISS scales could improve in precision of measurement by assessing a broader and more evenly distributed range of coping.36 Findings based on examination of item-person targeting showed that scale items characteristically underrepresented the range of coping reported by some individuals with TBI, such that scores underestimated
www.archives-pmr.org
Rasch analysis of Coping Inventory for Stressful Situations in TBI
665
their task coping or overestimated their emotion or avoidance coping, and thus limited differentiation of persons with differences at the very high or low end of these constructs, respectively. The CISS might be improved for TBI by adding items that better assess the lower and upper ranges of specific coping styles, although alterations should also balance practical considerations, such as administration time. For example, with task coping, clear patterns identified those items that were most challenging to endorse (eg, item 1: I schedule my time better) and items that were least challenging to endorse (eg, item 6: I do what I think is best). Conceptually, the most difficult task items appear to reflect a stronger approach orientation for dealing with stressors than the easier task items. Perhaps more difficult task coping items might intensify commitment and immediacy to take action, such as “I make dealing with the stressor my top priority.” Additionally, it seems noteworthy that actions at the upper end of item difficulty, such as outlining priorities and scheduling, require complex higher-order thinking as well as sustained action; in contrast, items at the low end are generally global statements of intention that do not require complex thinking or action. Consequently, it may be that items needed at the upper end of task coping for this population with cognitive limitations reflect “active” coping with activity that is relatively less cognitively complex, such as “I work hard at getting better.” Discriminating coping items are needed at the low ends of the emotion and avoidance coping scales. For Emotion coping, even the lowest items currently focus on intense emotional responses to stress (guilt, preoccupation, denial, regret) and intense experience of negative emotion (eg, very upset, angry). Perhaps lower-level emotion coping items might reflect milder forms of distress or a decreased emphasis on changing responses in the future, such as “I feel down more than usual.” Given the unique cognitive challenges associated with TBI and the increased risk for depression, it might be worthwhile to test items that tap less active emotions (eg, feelings of hopelessness, being overwhelmed, helplessness, frustration). For Avoidant coping, easier items might assess distraction that does not involve costly or out-of-home activities, or those that require high levels of functional independence such as watching television, listening to music, or playing video games, or general attempts to get one’s mind off the problem. Essentially, the findings indicate that the items need to spread more to capture the needs of the people endorsing them; it likely reflects a need for different ways to express the coping style at high and low levels of intensity rather than only a need for milder and stronger descriptive language.
for improving item discrimination when used for people with TBI, which future studies should address through adding items at the extreme levels of coping styles, the measure generally performed well without changes to the structure or content. Despite physical, psychological, and cognitive deficits common to this population, the CISS appears to be a good candidate for research and clinical assessment of coping and coping change during TBI treatment and recovery.
Study limitations These findings may not generalize to all persons with TBI given the exclusion of mild or very severe injuries, although individuals with mild injuries would not be expected to have long-term difficulties.47 Future research should use prospective designs and replicate with independent samples; although the cross-sectional groups in this study did not show meaningful differences related to time post injury, within-subjects longitudinal designs would best address this issue.
Conclusions These findings build on prior research3,48-50 regarding the assessment of coping after TBI. The CISS appears reliable across the short- and long-term adjustment to injury and taps distinct aspects of coping style. Although Rasch analyses identified areas www.archives-pmr.org
Suppliers a. Winsteps 3.74 software. Available at: http://www.winsteps. com. b. SPSS version 20 software, IBM Corp.
Keywords Brain injuries, traumatic; Coping behavior; Psychometrics; Rehabilitation
Corresponding author Lisa J. Rapport, PhD, 5057 Woodward Ave, 7th Fl, Detroit, MI 48202. E-mail address: [email protected].
References 1. Dawson DR, Schwartz ML, Winocur G, Stuss DT. Return to productivity following traumatic brain injury: cognitive, psychological, physical, spiritual, and environmental correlates. Disabil Rehabil 2007;29:301-13. 2. Anson K, Ponsford J. Evaluation of a coping skills group following traumatic brain injury. Brain Inj 2006;20:167-78. 3. Curran CA, Ponsford JL, Crowe S. Coping strategies and emotional outcome following traumatic brain injury: a comparison with orthopedic patients. J Head Trauma Rehabil 2000;15:1256-74. 4. Endler NS, Parker JD. Multidimensional assessment of coping: a critical evaluation. J Pers Soc Psychol 1990;58:844-54. 5. Endler NS, Parker JD. Assessment of multidimensional coping: task, emotion, and avoidance strategies. Psychol Assess 1994;6:50-60. 6. Endler NS, Parker JD, Summerfeldt LJ. Coping with health problems: conceptual and methodlogical issues. Can J Behav Sci 1993; 25:384-99. 7. Cosway R, Endler NS, Sadler AJ, Deary IJ. The Coping Inventory for Stressful Situations: factorial structure and associations with personality traits and psychological health. J Appl Biobehav Res 2000;5:121-43. 8. Moore AD, Stambrook M. Cognitive moderators of outcome following traumatic brain injury: a conceptual model and implications for rehabilitation. Brain Inj 1995;9:109-30. 9. Lazarus RS, Folkman S. Stress, appraisal, and coping. New York: Springer; 1984. 10. Godfrey HP, Knight RG, Partridge FM. Emotional adjustment following traumatic brain injury: a stress-appraisal-coping formulation. J Head Trauma Rehabil 1996;11:29-40. 11. Krpan KM, Levine B, Stuss DT, Dawson DR. Executive function and coping at one-year post traumatic brain injury. J Clin Exp Neuropsychol 2007;29:36-46. 12. Rabinowitz AR, Arnett PA. A longitudinal analysis of cognitive dysfunction, coping, and depression in multiple sclerosis. Neuropsychology 2009;23:581-91.
666 13. Wilder-Willis KE, Shear PK, Steffen JJ, Borkin J. The relationship between cognitive dysfunction and coping abilities in schizophrenia. Schizophr Res 2002;55:259-67. 14. Hurt CS, Landau S, Burn DJ, et al. Cognition, coping, and outcome in Parkinson’s disease. Int Psychogeriatr 2012;24:1656-63. 15. Hesdorffer DC, Rauch SL, Tamminga CA. Long-term psychiatric outcomes following traumatic brain injury: a review of the literature. J Head Trauma Rehabil 2009;24:452-9. 16. Rogers JM, Read CA. Psychiatric comorbidity following traumatic brain injury. Brain Inj 2007;21:1321-33. 17. Hanks RA, Temkin N, Machamer J, Dikmen SS. Emotional and behavioral adjustment after traumatic brain injury. Arch Phys Med Rehabil 1999;80:991-7. 18. Prigatano GP. Disturbances of self-awareness and rehabilitation of patients with traumatic brain injury: a 20-year perspective. J Head Trauma Rehabil 2005;20:19-29. 19. Kortte KB, Wegener ST, Chwalisz K. Anosognosia and denial: their relationship to coping and depression in acquired brain injury. Rehabil Psychol 2003;48:131-6. 20. Kieffer KM, MacDonald G. Exploring factors that affect score reliability and variability in the ways of coping questionnaire reliability coefficients: a meta-analytic reliability generalization study. J Individ Differ 2011;32:26-38. 21. Folkman S, Lazarus RS. Ways of Coping Questionnaire. Palo Alto: Consulting Psychologists Pr; 1988. 22. Carver CS, Scheier MF, Weintraub JK. Assessing coping strategies: a theoretically based approach. J Pers Soc Psychol 1989;56:267-83. 23. Edwards JR, O’Neill RM. The construct validity of scores on the Ways of Coping Questionnaire: confirmatory analysis of alternative factor structures. Educ Psychol Meas 1998;58:955-83. 24. Lundqvist LO, Ahlstro¨m G. Psychometric evaluation of the Ways of Coping Questionnaire as applied to clinical and nonclinical groups. J Psychosom Res 2006;60:485-93. 25. Carver CS. You want to measure coping but your protocol’s too long: consider the Brief Cope. Int J Behav Med 1997;4:92-100. 26. Hurt CS, Thomas BA, Burn DJ, et al. Coping in Parkison’s disease: an examination of the Coping Inventory for Stressful Situations. Int J Geriatr Psychiatry 2011;26:1030-7. 27. McWilliams LA, Cox BJ, Enns MW. Use of the Coping Inventory for Stressful Situations in a clinically depressed sample: factor structure, personality correlates, and prediction of distress. J Clin Psychol 2003;59:1371-85. 28. Rasch G. Probabilistic models for some intelligence and attainment tests. Copenhagen: Danmarks Paedagogiske Institut; 1960. 29. Bond TG, Fox CM. Applying the Rasch model: fundamental measurement in the human sciences. 2nd ed. Mahwah: Lawrence Erlbaum; 2007. 30. Levin HS, Mattis S, Ruff RM, et al. Neurobehavioral outcome following minor head injury: a three-center study. J Neurosurg 1987; 66:234-43. 31. Kashluba S, Hanks RA, Casey JE, Millis SR. Neuropsychologic and functional outcome after complicated mild traumatic brain injury. Arch Phys Med Rehabil 2008;89:904-11.
H.A. Greene et al 32. Andrich D. A rating formulation for ordered response categories. Psychometrika 1978;43:561-73. 33. Fisher WP Jr. Rating scale instrument quality criteria. Rasch Meas Trans 2007;21:1095. 34. Tennant A, Conaghan PG. The Rasch measurement model in rheumatology: what is it and why use it? When should it be applied, and what should one look for in a Rasch paper? Arthritis Rheum 2007;57: 1358-62. 35. Wright BD, Linacre JM. Reasonable mean-square fit values. Rasch Meas Trans 1994;8:370. 36. Baghaei P. The Rasch model as a construct validation tool. Rasch Meas Trans 2008;22:1145-6. 37. McHorney CA, Tarlov AR. Individual-patient monitoring in clinical practice: are available health status surveys adequate? Qual Life Res 1995;4:293-307. 38. Linacre JM. Category disordering (disordered categories) vs. threshold disordering (disordered thresholds). Rasch Meas Trans 1999;13:675. 39. Endler NS, Parker JDA. Coping Inventory for Stressful Situations (CISS) manual. 2nd ed. North Tonawanda: Multi-Health Systems; 1999. 40. Dikmen SS, Corrigan JD, Levin HS, Machamer J, Stiers W, Weisskopf MG. Cognitive outcome following traumatic brain injury. J Head Trauma Rehabil 2009;24:430-8. 41. Williams MW, Rapport LJ, Hanks RA, Millis SR, Greene HA. Incremental validity of neuropsychological evaluations to computed tomography in predicting long-term outcomes after traumatic brain injury. Clin Neuropsychol 2013;27:356-75. 42. Rafnsson FD, Smari J, Windle M, Mears SA, Endler NS. Factor structure and psychometric characteristics of the Icelandic version of the Coping Inventory for Stressful Situations (CISS). Pers Individ Diff 2006;40:1247-58. 43. Furukawa T, Suzuki-Moor A, Saito Y, Hamanaka T. Reliability and validity of the Japanese version of the Coping Inventory for Stressful Situations (CISS): a contribution to the cross-cultural studies of coping. Seishin Shinkeigaku Zasshi 1993;95:602-20. 44. Boysan M. Psychometric properties of the Turkish version of the Coping Inventory for Stressful Situations. Arch Neuropsychiatry 2012;49:196-202. 45. Pearlin L, Schooler C. The structure of coping. J Health Soc Behav 1978;19:2-21. 46. Folkman S, Moskowitz JT. Coping: pitfalls and promise. Annu Rev Psychol 2004;55:745-74. 47. Belanger HG, Curtiss G, Demery JA, Lebowitz BK, Vanderploeg RD. Factors moderating neuropsychological outcomes following mild traumatic brain injury: a meta-analysis. J Int Neuropsychol Soc 2005;11:215-27. 48. Anson K, Ponsford J. Coping and emotional adjustment following traumatic brain injury. J Head Trauma Rehabil 2006;21:248-59. 49. Malia K, Powell G, Torode S. Coping and psychosocial function after brain injury. Brain Inj 1995;9:607-18. 50. Bryant RA, Marosszeky JE, Crooks J, Baguley I, Gurka J. Coping style and posttraumatic stress disorder following severe traumatic brain injury. Brain Inj 2000;14:175-80.
www.archives-pmr.org
ORIGINAL RESEARCH
Rasch Analysis of the Coping Inventory for Stressful Situations in Individuals With Moderate to Severe Traumatic Brain Injury Hillary A. Greene, MA,a Lisa J. Rapport, PhD,a Scott R. Millis, PhD,b,c Robin A. Hanks, PhD,b,c Michael W. Williams, MAa From the aDepartment of Psychology and bSchool of Medicine, Department of Physical Medicine and Rehabilitation, Wayne State University, Detroit, MI; and cRehabilitation Institute of Michigan, Detroit, MI.
Abstract Objective: To evaluate psychometric properties of the Coping Inventory for Stressful Situations (CISS) in individuals with traumatic brain injury (TBI). Design: Archival study using Rasch analysis. Setting: Postacute rehabilitation hospital. Participants: Adults (NZ331) 1 to 15 years after moderate to severe TBI, recruited consecutively. Interventions: Not applicable. Main Outcome Measure: CISS. Results: Indices of unidimensionality and model fit supported the scale’s proposed multidimensional structure consisting of Task, Emotion, and Avoidant coping style; 3 unidimensional scales showed better fit than a single combined scale. The 3 scales met Rasch expectations of reliability and separation for persons and items, as well as adequate response category functioning. The scales were generally well targeted but showed some evidence of ceiling effect for Task, and floor effects for Emotion and Avoidant coping; item difficulties did not fully capture extreme ranges demonstrated by some participants, suggesting that measurement of coping after TBI on the CISS would be improved with additional items at low and high ranges of difficulty. Results were generally equivalent for cross-sectional groups representing short-term (1y), intermediate (2y), and long-term (5e15y). Conclusions: The CISS showed good psychometric properties as a measure of coping style among persons with moderate to severe TBI in acute and chronic phases of recovery, and showed evidence of multidimensionality as predicted by theory, consistent with 3 unidimensional scales. Added items tapping broader (or more accessible, less cognitively complex) ranges of coping responses would likely benefit the scale overall and improve correspondence with the response needs of people with TBI. Archives of Physical Medicine and Rehabilitation 2015;96:659-66 ª 2015 by the American Congress of Rehabilitation Medicine
Coping style is an aspect of psychological functioning that may influence recovery from traumatic brain injury (TBI).1-3 Prevailing theories suggest a 3-factor structure of coping style: problem-, emotion-, and avoidance-focused.4,5 Avoidant and emotion coping styles are associated with relatively poorer functioning, physical
Presented in part as a poster to the American Psychological Association, February 6e9, 2013. Supported by the National Institute on Disability and Rehabilitation Research (grant no. H133A020515-03) and Wayne State University Thesis Research Support. Disclosures: none.
health, and subjective well-being than task coping.5-7 Given the high prevalence of cognitive and emotional deficits, this 3-factor structure of coping style might not well characterize coping among people with TBI.8 Cognitive deficits may alter how people think about and engage in managing stressors.9,10 Cognitive impairments can undermine attention, memory, and executive functions needed for identifying, planning, and implementing complex coping responses necessary for task coping.11-14 Emotional disturbances, low frustration tolerance, or inertia post injury may favor an
0003-9993/14/$36 - see front matter ª 2015 by the American Congress of Rehabilitation Medicine http://dx.doi.org/10.1016/j.apmr.2014.11.006
660 emotion-focused response to stressors over active, task-focused coping.15-17 Identifying the need to invoke adaptive action may be disrupted by TBI via impaired awareness of deficits associated with neurologic dysfunction, impaired appreciation of deficits, or denial,18,19 which may increase the use of avoidant coping strategies.19 Most studies of coping style in TBI have falsely assumed that reliability generalizes from psychometric support established for non-TBI populations.20 Although multiple scales are available to assess coping in general populations (eg, Ways of Coping Questionnaire,21 COPE),22 they do not appear to be the strongest options psychometrically20,23,24 or clinically.22,25 The Coping Inventory for Stressful Situations (CISS) was developed by Endler and Parker4 to address psychometric weaknesses of earlier scales. The CISS includes 3 scales to assess task-, emotion-, and avoidance-focused coping, and is shorter than other scales, which is desirable for assessments of people with TBI. Research using traditional psychometric approaches (ie, classical test theory) indicates that the CISS is a reliable, multidimensional, valid measure of coping style among healthy and clinical populations, making it potentially a good choice for examining coping in TBI.4,7,26,27 Rasch modeling28 is well suited to examine the CISS. Rasch models are developed with the goal of establishing an ideal scale with interval-level measurement and examining how well the data fit the ideal model, whereas classical test theory approaches strive to develop a model to describe the data.29 Rasch analyses also enable examination of scale functioning at item and overall levels, scale dimensionality, and category response functioning. The primary aim of this study was to use Rasch modeling to improve understanding of how the CISS functions in TBI and whether it functions best in its original form or if alterations might improve its measurement properties. A secondary aim was to evaluate whether time since injury meaningfully affects CISS reliability by using Rasch analyses to assess the psychometric properties separately for respondents in acute versus chronic stages of recovery from TBI. This aim was particularly well suited for Rasch analysis, which affords the opportunity to calculate separate reliabilities for persons and items. Lastly, CISS dimensionality was of interest. Although it was not a primary purpose of the approach, Rasch analysis can provide some information regarding whether the proposed 3-dimensional structure of the CISS holds among respondents with TBI versus a single dimension of coping response, as might be driven by global cognitive impairment.
Methods Participants A total of 331 adults were recruited consecutively from a large longitudinal study of TBI. All participants had medically documented TBI, received treatment at an affiliated level-1 trauma center within 24 hours of injury, and required inpatient rehabilitation. Persons whose injuries resulted from anoxic encephalopathy, who were non-English speaking, or whose injury severities precluded valid evaluation were excluded.
List of abbreviations: CISS Coping Inventory for Stressful Situations TBI traumatic brain injury
H.A. Greene et al Table 1
Characteristics of the 331 participants
Characteristics
Values
Age at follow-up (y) Sex Women Men Education (y) Ethnicity African American White Latino(a) Native American Asian or Pacific Islander Time since injury (y) GCS (total at emergency department admission) Days to follow commands* Disability Rating Scale
44.013.5 61 (18.4) 270 (81.6) 11.92.1 234 (70.7) 87 (26.3) 5 (1.5) 3 (0.9) 2 (0.6) 6.14.9 9.3 (3e15) 7.212.1 6.52.9
NOTE. Values are mean SD, n (%), or median (range). Abbreviation: GCS, Glasgow Coma Scale. * Days to follow commands is a motor subscale of GCS; days from injury to follow commands.
Participants completed the CISS between 1 and 15 years after TBI (mean SD: 6.14.9y). The median Glasgow Coma Scale total at admission was 9 (range, 3e15), which corresponds to moderate TBI. Persons with complicated mild injuries with intracranial hemorrhage were included because their neuropsychological profiles are similar to those of individuals with moderate TBI.30,31 Demographic and injury severity characteristics are shown in table 1.
Measures: CISS The CISS4 is a 48-item self-report measure that uses a 5-point scale (1 [not at all] to 5 [very much]). It provides three 16-item scales (Task, Emotion, Avoidance) to assess specific coping styles, which are scored independently of one another. The Task scale focuses on problem solving, planning, and efforts to alter difficult situations; the Emotion scale focuses on self-oriented responses, fantasizing, or emotions, including efforts to deal with stress; and the Avoidance scale focuses on efforts to relieve stress by distraction with unrelated activities or social interactions.4 The CISS has well-demonstrated reliability and validity when used in healthy4 and clinical27 populations.
Procedures The study was approved by the institutional review board. These archival data were collected at scheduled follow-ups of 1, 2, 5, 10, and 15 years post injury. Only 1 CISS per participant contributed to the current dataset. Three cross-sectional time points post injury were examined: 1 year, 2 years, and chronic (5y post injury).
Data analyses Descriptive statistics and normative scores for the CISS scales were examined; univariate analyses of variance were used to compare the scales across each follow-up time. Rasch analyses were conducted for each of the scales at each follow-up time. Examining the scales at multiple time points post injury affords the opportunity to examine how well the CISS functions psychometrically among www.archives-pmr.org
Rasch analysis of Coping Inventory for Stressful Situations in TBI
661
individuals with TBI at different stages of recovery and adjustment to TBI that potentially could influence coping responses and reliability of responding. Rasch analyses also were conducted at each time point using all 48 items to test unidimensionality. The Rasch rating scale model32 was used to evaluate the CISS rather than the partial credit model because the CISS items all share the same rating scale structure. The partial credit model is used when each item has a unique rating scale structure. Measurement properties assessed included estimates of reliability, dimensionality, item and person targeting, and response category functioning. Models of interest for comparison included the widely accepted 3-factor model, a unifactorial model, and a null model (ie, a lack of dimensionality in the coping data). Winsteps 3.74 softwarea was used to conduct Rasch analyses. SPSS (version 20) softwareb was used for the remaining analyses. Unidimensionality was examined using Rasch principal component analysis of residuals, which is supported by unexplained variance in the first contrast of eigenvalues <3 and variance in data explained by the measure 60%.33 Unidimensionality also was evaluated with tests of local independence of the items, or response dependency34; a nonsignificant test indicates that after considering the Rasch factor, no associations remain among items other than random associations. Raw score residual correlations were computed, which correspond with Yen’s Q3 statistic. Test of fit to the model was evaluated with infit and outfit statistics. Infit is a weighted statistic of overall performance of an item with the model, whereas outfit is an unweighted statistic that reflects the extent to which outliers are present in the model.29 The metrics of infit and outfit statistics are weighted and unweighted normalized mean-squares, respectively.29 When data are predicted well by the Rasch model, infit and outfit values approximate the expected value of 1.0, with extreme values ranging from 0 to infinity.29 When data poorly fit the model (unpredictability), fit statistics are >1.0, and when data overfit the model (redundancy), fit statistics are <1.0.29 Because the measurement of coping has implications for assessment and prediction of real-world functions for persons with TBI, relatively stringent mean-square values from 0.7 to 1.3 were considered “excellent” fit statistics in this study, whereas fit statistics between 0.5 and 2.0 are considered “good” generally.33,35 Infit and outfit statistics for the total CISS scale and individual Task, Emotion, and Avoidance scales were each examined at the 3 time points, corresponding with the aim to compare the psychometric functioning of the CISS across early and late stages of recovery from TBI. Person reliability (analogous to Cronbach a) and item reliability were interpreted using Fisher’s recommendations,33 with “good” reliability >.80. Person (or item) separation indices >2.0 were interpreted as support for internal consistency. Targeting, which refers to matching of item difficulty to participant ability for a specific measure used with a specific sample, was assessed by examining person-item maps. The Rasch analysis itemperson map, also called the Wright map, facilitates assessment of targeting. On the map, items are plotted by difficulty on a logit scale, which are the raw scores converted to a scale with interval properties; persons are plotted by symbols that represent either 4 persons (#) or 1 to 3 persons (.). Symbols on the map denote the mean (M), 1 SD (S), and 2 SDs (T) from the mean for persons (left side of the midline) and items (right side of the midline). “Ability” is reflected from most to least: cases with the highest scores (at the top of the map) most easily endorsed (agreed with or performed correctly) the items. Similarly, “difficulty” is presented from most to least: items with the highest scores (top of the map) were most difficult to endorse. Good targeting of difficulty and ability ideally results in a symmetrical spread of
items and persons along the vertical axis of the item-person map.29 To obtain “uniformly precise measurement,” the items should cover a wide difficulty range, such that the item difficulties are spaced consistently, there are few gaps between items, and item difficulty levels correspond to person ability levels.36 Additionally, scores should span the range possible on the scale, with the sample mean approximating the midpoint of the scale and less than 15% obtaining the highest or lowest possible scores; exceeding that percentage provides evidence of ceiling and floor effects, respectively.37 Adequacy of the response categories at scale and item levels was assessed via consistency with Rasch category ordering principles, which expect that average category endorsement should increase monotonically.38 Andrich thresholds were examined for step disordering, which also expects monotonic increases.29,38
www.archives-pmr.org
Results For raw scores, the sample used more Task (mean SD: 57.313.1) than Emotion (mean SD: 44.113.7) and Avoidant (mean SD: 41.912.2) coping, which is consistent with the pattern observed for the normative sample. Univariate analyses of variance indicated that mean scores did not differ significantly across the follow-up time points on the Task scale (F2,328Z.40, PZ.67, h2Z.00), Emotion scale (F2,328Z1.75, PZ.18, h2Z.01), or the Avoidance scale (F2,328Z2.09, PZ.13, h2Z.01). However, normative scores indicated that the total sample used relatively less Task (zZ.15) and more Emotion (zZ.36) and Avoidant (zZ.27) coping compared with healthy community-dwelling individuals.39
Unidimensionality and model fit Across all time points, infit and outfit mean-square statistics for the CISS as a total scale and the individual Task, Emotion, and Avoidance scales were generally acceptable (0.5e2.0), suggesting that items fit the expectations of the rating scale models. No items had “poor” (<.33 or >3.0) fit with the model.33 Rather, across all analyses, most items had “excellent” fit (0.7e1.3) with the model, and all items with the exception of 2 (items 6, 48) had “good” fit (0.5e2.0) with the model.33 Item 6 had good to excellent fit at year 2 and chronic times, and had slightly less than good fit at year 1 (eg, infit 2.03, outfit 1.92). Item 48 had excellent fit at year 1 and chronic times, and had slightly less than good fit at year 2 (eg, infit 1.57, outfit 2.1). In the context of this generally good performance, most occasions in which items fell outside the “excellent” range occurred in the year-1 group, followed by the year-2 group. For all Rasch principal component analyses of residuals, the observed raw variance explained by the measure closely approximated or maximized the Rasch model value, which is the best value attainable if the data exactly fit the Rash model, with all differences between empirically observed and modeled variances being less than or equivalent to 1% (table 2). Although these values were below the ideal 60% threshold,33 models that examined the CISS scales separately performed uniformly better on variance explained than did the models that examined all items as a single scale. Furthermore, all models testing a single scale (all items) had eigenvalues >3 for unexplained variance in the first contrast, which is evidence against unidimensionality.33 In contrast, all models testing scales separately had eigenvalues <3 for the unexplained variance in the first contrast, indicating that the CISS showed superior fit as 3 separate scales (Task, Emotion, Avoidance), consistent with the structure proposed by the test authors. Raw score residual correlations also were computed for all analyses, which correspond with Yen’s Q3 statistic, and no response dependency was found.
662
Table 2
Select diagnostic data for Rasch analysis with CISS*
Scale Year
y
Infit statistic Outfit statistic Person separation ratioz Person reliabilityx Item separation ratioz Item reliabilityx Raw variance in data explained by measures in principal component analysis of residuals Empirical (%) Modeled (%) Unexplained variance in 1st contrast in principal component analysis of residuals Raw %
Task
Emotion
Avoidance
All Items
1
2
Chronic
1
2
Chronic
1
2
Chronic
1
2
Chronic
1.04.32 1.05.32 2.42 0.85 3.14 0.91
1.02.22 1.04.28 2.65 0.88 2.63 0.87
1.01.17 1.04.22 2.24 0.83 3.96 0.94
1.02.23 1.07.31 2.15 0.82 3.94 0.94
1.02.19 0.981.08 2.11 0.82 3.45 0.92
1.01.15 1.02.20 2.13 0.82 5.13 0.96
1.02.24 1.07.32 2.25 0.83 4.73 0.96
1.02.22 1.03.36 2.1 0.81 4.59 0.95
1.02.13 1.04.14 1.81 0.77 6.74 0.98
1.02.24 1.04.27 2.89 0.89 4.53 0.95
1.01.22 1.03.24 2.59 0.87 4.38 0.95
1.00.14 1.03.14 2.83 0.89 6.82 0.98
43.2 44.1
48.9 49.9
41.3 42.2
41.9 42.4
39.0 39.8
41.2 41.8
43.4 43.9
44.0 44.8
37.6 38.1
31.9 32.4
32.3 32.3
32.8 32.9
2.2 7.9
2.0 6.3
1.9 7.1
2.2 7.9
2.6 10.1
2.1 7.8
2.7 9.4
2.7 9.3
2.5 9.7
8.3 11.8
9.4 13.3
8.1 11.3
NOTE. Values are mean SD or as otherwise indicated. * All values reflect inclusion of all relevant items (no dropped items). y Year 1 (nZ104), year 2 (nZ90), chronic (nZ205). z Separation ratios: poor (<2), fair (2e3), good (3e4), very good (4e5), excellent (>5).33 x Reliability estimates: poor (<.67), fair (.67e.80), good (.81e.90), very good (.91e.94), excellent (>.94).33
H.A. Greene et al
www.archives-pmr.org
Rasch analysis of Coping Inventory for Stressful Situations in TBI
663
Fig 1 Item-person map of the CISS, Task scale, chronic group. Abbreviations: M, mean; S, 1 SD from the mean; T, 2 SDs from the mean.
Reliability Table 2 shows person and item separation ratios and reliability estimates. Person reliability estimates ranged from .77 to .89; with 1 exception, all estimates met or exceeded the criterion for “good” (.80) reliability.33 Item reliability estimates ranged from .87 to .98; all estimates met or exceeded “good” reliability criterion, and nearly all corresponded to “very good” (.91e.94) to “excellent” (>.94) reliability.33 Person and item reliabilities for the Task, Emotion, and Avoidant scales were generally equivalent across the 3 follow-up time groups, falling within the same classification range. Person separation ratios were generally “fair” (2e3), whereas item separation ratios were generally “good” (3e4) to “very good” (4e5). Person separation ratios appear equivalent for the 3 time-point groups; item separation ratios appear higher for the chronic group (“very good” to “excellent”) as compared with the 1- and 2-year postinjury groups (generally “good”).
Targeting Examination of person distributions based on the percentage of participants who obtained the highest or lowest possible scores on the CISS scales across all analyses indicated that there were no significant problems with ceiling or floor effects (ie, none www.archives-pmr.org
Fig 2 Item-person map of the CISS, Emotion scale, chronic group. Abbreviations: M, mean; S, 1 SD from the mean; T, 2 SDs from the mean.
exceeded the 15% criterion).37 Additionally, the means for item difficulties and person abilities (shown in item-person maps) across the CISS scales for all time points generally were matched well.37 The item-person maps for Task, Emotion, and Avoidant scales were very similar for the 3 postinjury times; maps for the chronic group are presented as examples. As shown in figure 1, for the Task scale, there was some evidence of extreme scores at the high end of the logit scale on the person map, extending beyond the cluster of scores on the item map, with much less evidence of extreme scores at the low end of the map; this pattern was evident across all time points. For the Emotion scale (fig 2), extreme scores on the person map extended at the high and low ends, though more extreme at the low versus high end of the scale, beyond the spread of item scores for all 3 time groups. For the Avoidance scale (fig 3), extreme person scores appeared generally
664
H.A. Greene et al disordering, count data and fit statistics were considered to evaluate whether items needed to be dropped from the scale. The Andrich thresholds (step-calibration estimates) showed disordered thresholds across all scales and time points assessed. However, there were no problems with fit statistics or category disordering overall, and no further action was warranted to address threshold disordering.
Discussion
Fig 3 Item-person map of the CISS, Avoidance scale, chronic group. Abbreviations: M, mean; S, 1 SD from the mean; T, 2 SDs from the mean.
more at the low ends than the high ends of the scales across all time points. Although participants reported higher and lower amounts of coping ability than individual items were able to assess fully, the CISS scales did not show overall marked problems with targeting through ceiling or floor effects.
Response categories The CISS showed adequate category functioning, with no major problems with category disordering across any of the scales or time points. Each sequential response option was associated on average with increasing item difficulty (ie, intensity of agreement), which indicated that the categories increased monotonically. Category ordering was also examined at an item level across scales and time points. For items that showed category
Coping style among individuals with moderate to severe TBI can be reliably assessed using the CISS. The CISS performed as a multidimensional measure of coping in TBI and reliably captured task-, emotion-, and avoidance-focused coping styles across shortand long-term adjustment to injury. Establishing the reliability of the CISS among persons recovering from TBI provides a necessary foundation for examining a broad range of questions about stress and coping within this population. These findings support the ability of persons with moderate to severe TBI to complete measures such as the CISS reliably, given the array of cognitive impairments often experienced.40 The CISS demonstrated good reliability even at 1 year after injury, which is a relatively early stage in recovery often associated with significant global cognitive impairments that might otherwise undermine reliable self-estimates.41 Similarly, it might have been the case that global cognitive deficits associated with TBI would yield an undifferentiated (global) coping response, in which specific types of coping are not discernible and instead give rise to a generalized mass of positive or negative reactions to stress. Therefore, the findings regarding clear dimensionality of the CISS in this population at 1 year post injury are particularly striking. These findings are consistent with prior work demonstrating the reliability of the CISS among healthy populations,7,39,42-44 persons with physical health problems,26 and persons with mental health problems.27 The CISS functions much better psychometrically as a multidimensional rather than a unidimensional measure of coping style, which is consistent with the structure and purpose designed by Endler and Parker5 and prior research that conceptualizes coping as a multidimensional construct characterized by task, emotion, and avoidant responses.4,22,45,46 Therefore, it is likely that adults with TBI engage in similar types of coping responses as none brain-injured adults; however, the relative reliance on each strategy appears to differ. Normative comparisons indicated that persons with TBI more heavily relied on emotion-focused and avoidant coping than do healthy adults, which highlights the importance of assessing coping after TBI to inform rehabilitation planning. Rasch analyses enhanced understanding regarding the psychometric characteristics of the CISS as used in TBI beyond confirming reliability and multidimensionality. Consistent with the intended function of the CISS,39 the findings indicated that all 48 items of the CISS and all 5 response options contributed meaningfully to the assessment of coping in TBI, with no evidence to support dropping any items or altering the response options. Some evidence indicates that items for the CISS scales could improve in precision of measurement by assessing a broader and more evenly distributed range of coping.36 Findings based on examination of item-person targeting showed that scale items characteristically underrepresented the range of coping reported by some individuals with TBI, such that scores underestimated
www.archives-pmr.org
Rasch analysis of Coping Inventory for Stressful Situations in TBI
665
their task coping or overestimated their emotion or avoidance coping, and thus limited differentiation of persons with differences at the very high or low end of these constructs, respectively. The CISS might be improved for TBI by adding items that better assess the lower and upper ranges of specific coping styles, although alterations should also balance practical considerations, such as administration time. For example, with task coping, clear patterns identified those items that were most challenging to endorse (eg, item 1: I schedule my time better) and items that were least challenging to endorse (eg, item 6: I do what I think is best). Conceptually, the most difficult task items appear to reflect a stronger approach orientation for dealing with stressors than the easier task items. Perhaps more difficult task coping items might intensify commitment and immediacy to take action, such as “I make dealing with the stressor my top priority.” Additionally, it seems noteworthy that actions at the upper end of item difficulty, such as outlining priorities and scheduling, require complex higher-order thinking as well as sustained action; in contrast, items at the low end are generally global statements of intention that do not require complex thinking or action. Consequently, it may be that items needed at the upper end of task coping for this population with cognitive limitations reflect “active” coping with activity that is relatively less cognitively complex, such as “I work hard at getting better.” Discriminating coping items are needed at the low ends of the emotion and avoidance coping scales. For Emotion coping, even the lowest items currently focus on intense emotional responses to stress (guilt, preoccupation, denial, regret) and intense experience of negative emotion (eg, very upset, angry). Perhaps lower-level emotion coping items might reflect milder forms of distress or a decreased emphasis on changing responses in the future, such as “I feel down more than usual.” Given the unique cognitive challenges associated with TBI and the increased risk for depression, it might be worthwhile to test items that tap less active emotions (eg, feelings of hopelessness, being overwhelmed, helplessness, frustration). For Avoidant coping, easier items might assess distraction that does not involve costly or out-of-home activities, or those that require high levels of functional independence such as watching television, listening to music, or playing video games, or general attempts to get one’s mind off the problem. Essentially, the findings indicate that the items need to spread more to capture the needs of the people endorsing them; it likely reflects a need for different ways to express the coping style at high and low levels of intensity rather than only a need for milder and stronger descriptive language.
for improving item discrimination when used for people with TBI, which future studies should address through adding items at the extreme levels of coping styles, the measure generally performed well without changes to the structure or content. Despite physical, psychological, and cognitive deficits common to this population, the CISS appears to be a good candidate for research and clinical assessment of coping and coping change during TBI treatment and recovery.
Study limitations These findings may not generalize to all persons with TBI given the exclusion of mild or very severe injuries, although individuals with mild injuries would not be expected to have long-term difficulties.47 Future research should use prospective designs and replicate with independent samples; although the cross-sectional groups in this study did not show meaningful differences related to time post injury, within-subjects longitudinal designs would best address this issue.
Conclusions These findings build on prior research3,48-50 regarding the assessment of coping after TBI. The CISS appears reliable across the short- and long-term adjustment to injury and taps distinct aspects of coping style. Although Rasch analyses identified areas www.archives-pmr.org
Suppliers a. Winsteps 3.74 software. Available at: http://www.winsteps. com. b. SPSS version 20 software, IBM Corp.
Keywords Brain injuries, traumatic; Coping behavior; Psychometrics; Rehabilitation
Corresponding author Lisa J. Rapport, PhD, 5057 Woodward Ave, 7th Fl, Detroit, MI 48202. E-mail address: [email protected].
References 1. Dawson DR, Schwartz ML, Winocur G, Stuss DT. Return to productivity following traumatic brain injury: cognitive, psychological, physical, spiritual, and environmental correlates. Disabil Rehabil 2007;29:301-13. 2. Anson K, Ponsford J. Evaluation of a coping skills group following traumatic brain injury. Brain Inj 2006;20:167-78. 3. Curran CA, Ponsford JL, Crowe S. Coping strategies and emotional outcome following traumatic brain injury: a comparison with orthopedic patients. J Head Trauma Rehabil 2000;15:1256-74. 4. Endler NS, Parker JD. Multidimensional assessment of coping: a critical evaluation. J Pers Soc Psychol 1990;58:844-54. 5. Endler NS, Parker JD. Assessment of multidimensional coping: task, emotion, and avoidance strategies. Psychol Assess 1994;6:50-60. 6. Endler NS, Parker JD, Summerfeldt LJ. Coping with health problems: conceptual and methodlogical issues. Can J Behav Sci 1993; 25:384-99. 7. Cosway R, Endler NS, Sadler AJ, Deary IJ. The Coping Inventory for Stressful Situations: factorial structure and associations with personality traits and psychological health. J Appl Biobehav Res 2000;5:121-43. 8. Moore AD, Stambrook M. Cognitive moderators of outcome following traumatic brain injury: a conceptual model and implications for rehabilitation. Brain Inj 1995;9:109-30. 9. Lazarus RS, Folkman S. Stress, appraisal, and coping. New York: Springer; 1984. 10. Godfrey HP, Knight RG, Partridge FM. Emotional adjustment following traumatic brain injury: a stress-appraisal-coping formulation. J Head Trauma Rehabil 1996;11:29-40. 11. Krpan KM, Levine B, Stuss DT, Dawson DR. Executive function and coping at one-year post traumatic brain injury. J Clin Exp Neuropsychol 2007;29:36-46. 12. Rabinowitz AR, Arnett PA. A longitudinal analysis of cognitive dysfunction, coping, and depression in multiple sclerosis. Neuropsychology 2009;23:581-91.
666 13. Wilder-Willis KE, Shear PK, Steffen JJ, Borkin J. The relationship between cognitive dysfunction and coping abilities in schizophrenia. Schizophr Res 2002;55:259-67. 14. Hurt CS, Landau S, Burn DJ, et al. Cognition, coping, and outcome in Parkinson’s disease. Int Psychogeriatr 2012;24:1656-63. 15. Hesdorffer DC, Rauch SL, Tamminga CA. Long-term psychiatric outcomes following traumatic brain injury: a review of the literature. J Head Trauma Rehabil 2009;24:452-9. 16. Rogers JM, Read CA. Psychiatric comorbidity following traumatic brain injury. Brain Inj 2007;21:1321-33. 17. Hanks RA, Temkin N, Machamer J, Dikmen SS. Emotional and behavioral adjustment after traumatic brain injury. Arch Phys Med Rehabil 1999;80:991-7. 18. Prigatano GP. Disturbances of self-awareness and rehabilitation of patients with traumatic brain injury: a 20-year perspective. J Head Trauma Rehabil 2005;20:19-29. 19. Kortte KB, Wegener ST, Chwalisz K. Anosognosia and denial: their relationship to coping and depression in acquired brain injury. Rehabil Psychol 2003;48:131-6. 20. Kieffer KM, MacDonald G. Exploring factors that affect score reliability and variability in the ways of coping questionnaire reliability coefficients: a meta-analytic reliability generalization study. J Individ Differ 2011;32:26-38. 21. Folkman S, Lazarus RS. Ways of Coping Questionnaire. Palo Alto: Consulting Psychologists Pr; 1988. 22. Carver CS, Scheier MF, Weintraub JK. Assessing coping strategies: a theoretically based approach. J Pers Soc Psychol 1989;56:267-83. 23. Edwards JR, O’Neill RM. The construct validity of scores on the Ways of Coping Questionnaire: confirmatory analysis of alternative factor structures. Educ Psychol Meas 1998;58:955-83. 24. Lundqvist LO, Ahlstro¨m G. Psychometric evaluation of the Ways of Coping Questionnaire as applied to clinical and nonclinical groups. J Psychosom Res 2006;60:485-93. 25. Carver CS. You want to measure coping but your protocol’s too long: consider the Brief Cope. Int J Behav Med 1997;4:92-100. 26. Hurt CS, Thomas BA, Burn DJ, et al. Coping in Parkison’s disease: an examination of the Coping Inventory for Stressful Situations. Int J Geriatr Psychiatry 2011;26:1030-7. 27. McWilliams LA, Cox BJ, Enns MW. Use of the Coping Inventory for Stressful Situations in a clinically depressed sample: factor structure, personality correlates, and prediction of distress. J Clin Psychol 2003;59:1371-85. 28. Rasch G. Probabilistic models for some intelligence and attainment tests. Copenhagen: Danmarks Paedagogiske Institut; 1960. 29. Bond TG, Fox CM. Applying the Rasch model: fundamental measurement in the human sciences. 2nd ed. Mahwah: Lawrence Erlbaum; 2007. 30. Levin HS, Mattis S, Ruff RM, et al. Neurobehavioral outcome following minor head injury: a three-center study. J Neurosurg 1987; 66:234-43. 31. Kashluba S, Hanks RA, Casey JE, Millis SR. Neuropsychologic and functional outcome after complicated mild traumatic brain injury. Arch Phys Med Rehabil 2008;89:904-11.
H.A. Greene et al 32. Andrich D. A rating formulation for ordered response categories. Psychometrika 1978;43:561-73. 33. Fisher WP Jr. Rating scale instrument quality criteria. Rasch Meas Trans 2007;21:1095. 34. Tennant A, Conaghan PG. The Rasch measurement model in rheumatology: what is it and why use it? When should it be applied, and what should one look for in a Rasch paper? Arthritis Rheum 2007;57: 1358-62. 35. Wright BD, Linacre JM. Reasonable mean-square fit values. Rasch Meas Trans 1994;8:370. 36. Baghaei P. The Rasch model as a construct validation tool. Rasch Meas Trans 2008;22:1145-6. 37. McHorney CA, Tarlov AR. Individual-patient monitoring in clinical practice: are available health status surveys adequate? Qual Life Res 1995;4:293-307. 38. Linacre JM. Category disordering (disordered categories) vs. threshold disordering (disordered thresholds). Rasch Meas Trans 1999;13:675. 39. Endler NS, Parker JDA. Coping Inventory for Stressful Situations (CISS) manual. 2nd ed. North Tonawanda: Multi-Health Systems; 1999. 40. Dikmen SS, Corrigan JD, Levin HS, Machamer J, Stiers W, Weisskopf MG. Cognitive outcome following traumatic brain injury. J Head Trauma Rehabil 2009;24:430-8. 41. Williams MW, Rapport LJ, Hanks RA, Millis SR, Greene HA. Incremental validity of neuropsychological evaluations to computed tomography in predicting long-term outcomes after traumatic brain injury. Clin Neuropsychol 2013;27:356-75. 42. Rafnsson FD, Smari J, Windle M, Mears SA, Endler NS. Factor structure and psychometric characteristics of the Icelandic version of the Coping Inventory for Stressful Situations (CISS). Pers Individ Diff 2006;40:1247-58. 43. Furukawa T, Suzuki-Moor A, Saito Y, Hamanaka T. Reliability and validity of the Japanese version of the Coping Inventory for Stressful Situations (CISS): a contribution to the cross-cultural studies of coping. Seishin Shinkeigaku Zasshi 1993;95:602-20. 44. Boysan M. Psychometric properties of the Turkish version of the Coping Inventory for Stressful Situations. Arch Neuropsychiatry 2012;49:196-202. 45. Pearlin L, Schooler C. The structure of coping. J Health Soc Behav 1978;19:2-21. 46. Folkman S, Moskowitz JT. Coping: pitfalls and promise. Annu Rev Psychol 2004;55:745-74. 47. Belanger HG, Curtiss G, Demery JA, Lebowitz BK, Vanderploeg RD. Factors moderating neuropsychological outcomes following mild traumatic brain injury: a meta-analysis. J Int Neuropsychol Soc 2005;11:215-27. 48. Anson K, Ponsford J. Coping and emotional adjustment following traumatic brain injury. J Head Trauma Rehabil 2006;21:248-59. 49. Malia K, Powell G, Torode S. Coping and psychosocial function after brain injury. Brain Inj 1995;9:607-18. 50. Bryant RA, Marosszeky JE, Crooks J, Baguley I, Gurka J. Coping style and posttraumatic stress disorder following severe traumatic brain injury. Brain Inj 2000;14:175-80.
www.archives-pmr.org