Construct and Criterion-Based Validity of Brief Pain Coping Scales in Persons with Chronic Knee Osteoarthritis Pain

Abstract

Objectives

A recent trend in clinical practice is to adopt short screening and diagnostic self-report instruments for patients with chronic pain. Brief 2-item pain coping and beliefs measures have recently been developed and have potential to improve decision making in clinical practice. Our study examined the construct and criterion-based validity of the 2-item per scale version of the Coping Strategies Questionnaire (CSQ).

Design

We used data obtained on a community-based sample of 873 persons with chronic knee osteoarthritis pain from the Osteoarthritis Initiative, a large longitudinal cohort study. Persons were administered the two-item per scale version of the CSQ. The International Classification of Functioning framework was used to select a variety of criterion-based measures for comparison to the CSQ. Spearman correlations and hierarchical regression models were used to characterize construct validity and Receiver Operating Characteristic Curves, sensitivity and specificity were used to describe criterion-based validity.

Results

Construct validity of the CSQ scales were generally supported, with the Catastrophizing and Praying or Hoping scales demonstrating the strongest construct validity across criterion measures. Criterion-based validity for the CSQ scales varied depending on the criterion measure. The Catastrophizing and Praying or Hoping scales also had the strongest criterion-based validity with ROC curve areas as high as 0.71 (95%CI= 0.67, 0.75), p<0.001 for identifying persons with substantial physical function deficits.

Conclusions

The findings suggest that several of the 2-item CSQ scales demonstrate a modest level of construct validity along with fair criterion-based validity. The Catastrophizing and Praying or Hoping scales appear to hold the most promise for clinical applications and future longitudinal research.

The use of brief and simple-to-use instruments for quantifying the extent of depression, anxietys, general psychological distress [1,2] and pain-related beliefs and coping [3] has been a recent trend in primary care [4]. Demand has been led by the need for brief screening measures of key pain-related constructs that can be used in clinical settings. Researchers have responded to this need by developing and validating a number of brief measures that can be used as diagnostic and screening instruments. For example, Kroenke and colleagues developed 2-item screening measures for depressive [5] and anxiety disorders [1].

Brief instruments for assessing pain beliefs and coping also have been used extensively to estimate the burden [6] and impact of pain coping strategies in populations [7,8]. Brief instruments have gained popularity, but brevity comes with a price. Reducing item number from a parent scale, for example, eases respondent burden but can increase measurement error [3,9,10]. This increase in measurement error has the potential to result in incorrect diagnostic or treatment-based decisions. When making decisions regarding individual patients, demands for psychometric rigor increase, relative to group level change [11–13]. Thus, before brief instruments can be recommended for routine use in clinical settings, research is needed to confirm their validity.

We sought to determine the extent to which two-item versions of the scales derived from the Coping Strategies Questionnaire (CSQ) [14–16], a commonly reported measure of pain coping, are valid when used on persons in the community with chronic knee OA pain. Persons with chronic pain in the community have not been studied using the two-item version of the CSQ and given the high prevalence [17], economic and health consequences [18] of community-based chronic OA pain, this area is in need of study.

To examine validity, we chose two approaches. Previous work has examined psychometric qualities of the brief versions of CSQ scales only in the context of patient samples [3,9]. To extend and determine the generalizability of the findings from research to community-based individuals, we evaluated construct validity by comparing the 2-item CSQ scale scores to the full spectrum of impairment and function defined in the International Classification of Functioning, Disability and Health (ICF) model, in a large community-based large sample of persons with chronic knee OA pain. Using the ICF conceptual framework to examine construct validity allows for assessment of the association with the CSQ scales to knee-level attributes (e.g., knee pain) as well as person-level generic activities (e.g., sitting and standing) and interactions of the person in life situations (e.g. occupational and recreational activities) [19]. We hypothesized that CSQ scales would demonstrate construct validity in the community-based sample, much like patient samples that have been studied [3,9]. We also examined criterion-based validity of the brief CSQ scales using several approaches. If the 2-item CSQ scales show similar patterns of association to criterion measures as the full scales do, this would provide preliminary support for use of these shorter scales in large sample survey studies and for pain coping screening in busy clinical practices [3,9]. We did not have specific hypotheses regarding the criterion-based validity of the CSQ scales, given that previous researchers have not evaluated this aspect of validity in CSQ scales in either clinical or community samples of individuals with chronic pain. However, given the success of brief versions of other important domains [1,5], as well as the established importance of coping and catastrophizing in particular to chronic pain adjustment [20–22], we thought that this issue was worthy of examination.

Materials and Methods

The Osteoarthritis Initiative (OAI) is a publicly and privately funded prospective longitudinal cohort study of persons in the community who had or were at high risk for knee OA. Subjects received no treatment as part of the study nor were they solicited for treatment by the investigators. The study was approved by the IRB at the OAI Coordinating Center, the University of California at San Francisco.

The study admitted 4,976 persons who were examined over a 4-year period. The sample in the study was community based in that at the 4-year follow-up, the period of recruitment for the current study, 18% of the sample reported currently seeing a health care practitioner for their knee arthritis. Subjects were recruited from: (1) the University of Maryland School of Medicine in Baltimore, Maryland, (2) the Ohio State University in Columbus, Ohio, (3) the University of Pittsburgh in Pittsburgh, Pennsylvania, and (4) Memorial Hospital of Rhode Island, in Pawtucket, Rhode Island. Only data from the 4-year follow-up visits were used in the current study because this was the only visit in which the CSQ was completed by all subjects.

OAI Study Sample

Exclusion criteria were the presence of rheumatoid arthritis, bilateral knee arthroplasty or preexisting plans to undergo bilateral (not unilateral) knee arthroplasty in the next 3 years, bilateral end stage knee OA, positive pregnancy test, inability to provide a blood sample, use of ambulatory aids other than a single straight cane for more than 50% of the time, co-morbid conditions that might interfere with 4 year participation, unlikely to reside in clinic area for at least 3 years.

The inclusion criteria for the current study were that the persons had to have definite knee OA and chronic knee pain. Knee OA was judged to definitely be present when the Kellgren-Lawrence knee OA grade of 2 or higher [23,24] was given for at least one knee by the OAI radiographic readers, as determined by validated standing flexion knee radiographs [25,26]. Chronic knee pain was judged to be present when a person reported knee pain on most days in a month for at least 6 of the past 12 months. A total of 873 persons met these criteria. A total of 421 persons (48.2%) had unilateral chronic arthritic knee pain while 452 (51.84%) had bilateral chronic arthritic knee pain.

The Coping Strategies Questionnaire

The original CSQ is a 50-item self-report measure of eight pain coping strategies [14]. With this measure, respondents rate the frequency of each pain coping response on a 0 to 6 scale. Items for the brief 14-item (two-items per scale) CSQ of seven pain coping strategies (see Table 1) were selected based on their associations with the parent scales [3]. Research in two clinical settings supports the validity of these two-item scales, as evidenced by their strong associations with the parent scales (Pearson r = 0.78 to 0.92), as well as their moderate associations with a variety of criterion variables in samples of patients seeking treatment for chronic pain [3,9]. Four of the scale scores (Diverting Attention, Praying or Hoping, Reinterpreting Pain Sensations and Pain Catastrophizing) were highly skewed to the right in our study. The remaining scales (Ignoring Sensations, Coping Self Statements, Increased Behavioral Activities) demonstrated an approximately normal distribution as evidenced by the skewness statistics reported in Table 2.

Table 1.

The 14-item Coping Strategies Questionnaire

Item	Question
1	I think of things I enjoy doing.
2	I just think of it as some other sensation, such as numbness.
3	It is terrible and I think it is never going to get any better.
4	I don’t pay any attention to it.
5	I pray for the pain to stop.
6	I tell myself I can’t let the pain stand in the way of what I have to do.
7	I do something active, like household chores or projects.
8	I replay in my mind pleasant experiences in the past.
9	I pretend it is not a part of me.
10	I feel I can’t stand it anymore.
11	I ignore it.
12	I try to think years ahead, what everything will be like after I’ve gotten rid of the pain.
13	I see it as a challenge and don’t let it bother me.
14	I do something I enjoy, such as watching TV or listening to music.

All items are scored on a scale from 0 to 6 with 0 = Never do, 3= Sometimes do that, and 6 = Always do that. Items 1 and 8 the CSQ Diverting Attention, items 2 and 9, Reinterpreting Pain Sensations, items 3 and 10, Catastrohpizing, items 4 and 11, Ignoring Sensations, items 5 and 12, Praying or Hoping, items 6 and 13, Coping Self-Statements and items 7 and 14, Increased Behavioral Activities.

Table 2.

Characteristics of the study sample (n=873)

Characteristic	Percentage or Mean (sd)
Age (range = 49 to 83 yr)	65.6 (9.0)
Sex
Women	58.8
Race/Ethnicity
White	78.0
Black or African American	21.0
Asian	1.0
Marital Status
Married	60.1
Body Mass Index (range = 18.2 to 49.9kg/m2)	30.5 (5.1)
Comorbidity (range = 0 to 12)	0.6 (1.2)
Symptom duration
9 years or less	56.5
> 9 years	43.5
7-day Average Pain Intensity (range = 0 to 10)	3.9 (2.4)
WOMAC Physical Function (range = 0 to 68)	18.8 (12.7)
LLFDI Limitation (range = 32.1 to 100)	75.3 (14.7)
LLFDI Frequency (range = 29.2 to 76.3)	53.5 (5.8)
CESD Depression (range = 0 to 47)	8.3 (8.3)
20-meter walk pace (range = 0.33 to 1.95 m/sec)	1.2 (0.2)
CSQ Ignoring Sensations (range = 0 to 6, skewness = −0.05, SE = 0.08)	3.0 (1.7)
CSQ Coping Self Statements (range = 0 to 6, skewness = −0.40, SE = 0.08)	3.5 (1.8)
CSQ Praying or Hoping (range = 0 to 6, skewness = 1.4, SE = 0.08)	1.2 (1.6)
CSQ Reinterpreting Pain Sensations (range = 0 to 6, skewness = 1.3, SE= 0.08)	1.0 (1.3)
CSQ Diverting Attention (range = 0 to 6, skewness = 1.1, SE = 0.08)	1.3 (1.6)
CSQ Increased Behavioral Activities (range = 0 to 6, skewness = 0.06, SE = 0.08)	2.7 (1.8)
CSQ Catastrophizing (range = 0 to 6, skewness = 1.6, SE=0.08)	1.0 (1.4)

WOMAC= Western Ontario and McMasters Universities Arthritis Index, LLFDI Frequency = Late Life Function and Disability Index, Frequency Total Score, LLFDI Limitation = Late Life Function and Disability Limitation Total Score, 20-m Walk= 20-meter walking test, CESD=Center for Epidemiological Studied Depression Scale, CSQ = Coping Strategies Questionnaire, SE = standard error

International Classification of Functioning variables of interest

We used the ICF to guide our choice of variables to examine construct and criterion-based validity [19]. Knee level measures are defined by the ICF as impairments in body system or organ level structure or function and we used a rating of arthritic knee pain, described as pain on average over a 7-day period and measured on a 0 to 10 scale with 0 indicating “no pain” and 10 indicating “pain as bad as you can imagine.” Knee pain ratings were obtained for both knees. When unilateral knee pain was present, we used the rating for the involved knee and when bilateral knee pain was present, we used the more severe of the two ratings.

The activity limitations concept of the ICF reflects person level function of typical generic daily activities such as walking and sitting. It was measured with the WOMAC Physical Function scale which specifically asks the patient to rate the difficulty associated with 17 common activities such as walking, sitting and getting out of bed. There is a great deal of evidence supporting the reliability and validity of the WOMAC Physical Function scale [27–29]. This scale ranges from 0 (normal function) to 68 (severely affected function). We chose the WOMAC Physical Function Scale as the best measure of activity limitation because the scale specifically does not inquire about symptoms associated with the activities but rather only the extent of difficulty in doing each activity. Much like the pain scale, patients were asked to complete a WOMAC scale for each knee. Because the WOMAC Physical Function scale asks about person-level activities, we used the highest (worst) WOMAC score to indicate extent of person level ICF-based activity limitation.

To capture the Participation Restriction concept of the ICF we used the validated Frequency and Limitation summary scales of the Late Life Function and Disability Index (LLFDI) [30,31]. The scale is designed to measure higher order socially driven activities like visiting friends, recreation, travel, and household duties. Each summary scale ranges from 0 to 100 with higher scores indicating higher function. Frequency questions are phrased to ask “how often do you do a particular activity” while Limitation captures capability and asks, “to what extent do you feel limited in doing a particular activity” [30].

We also examined a physical performance measure, the 20-meter walk test, in addition to the full complement of impairment and self-report measures. We added this measure because we were interested in assessing whether pain coping strategies were associated with the actual physical performance of walking, one of the most common activities performed by persons with knee OA. Persons were instructed to walk along a 20 meter walkway at their usual pace. The test result was recorded in 0.10 second increments and the average of 2 tests were recorded in meters/second. Measures of self-paced walking are highly reliable [32].

Finally, because of the frequent co-occurence of chronic pain and depression [33], depressive symptoms were assessed with the validated Center for Epidemiologic Studies Depression scale (CESD) [34]. The scale ranges from 0 (no depression) to 60 (severe depression) with 16 being the validated cut-point for clinically meaningful depressive symptoms [34].

A variety of measures were used to describe the sample and to adjust for potential confounding. In addition to age, sex and the presence of unilateral or bilateral chronic knee pain, comorbidity was measured using a validated scale [35]. Body mass index (BMI) was calculated using calibrated scales. As a proxy for symptom duration we used a question posed to persons at baseline (4 years prior to the data reported in this study). If a person reported that he/she had pain, aching or stiffness in either knee for most days in a month, the person was asked how long ago the symptoms started. The options were 1 year ago or less, 2 years to 5 years ago, and greater than 5 years ago. For this study, we collapsed these data to indicate the person reported pain, aching or stiffness in at least one knee for ≤ 9 years or > 9 years.

Data Analysis

Because of the highly skewed nature of the CSQ Diverting Attention, Praying or Hoping, Reinterpreting Pain Sensations and Pain Catastrophizing subcales of the CSQ, we used Spearman rank order correlations to examine associations among CSQ measures and ICF and performance measures. In addition, we used hierarchical multiple linear regression to more fully examine construct validity of the measures. A separate model was generated for each ICF based variable, depression and the 20-meter walk pace. First, we forced the covariates of age, sex, comorbidity score, unilateral or bilateral chronic knee pain, and symptom duration into the models. Next, we forced all CSQ scale scores into the model in a single step. All variables that were skewed (i.e., Diverting Attention, Praying or Hoping, Reinterpreting Pain Sensations and Pain Catastrophizing) were converted to dichotomous variables to avoid violation of statistical assumptions of normality. The four skewed CSQ scores were coded as 0 (score of 0 or 0.5) or 1 (score of 1 or higher). The rationale for this cut-off was that scores of 0 or 0.5 indicated that the person never or almost never engaged in the coping strategy. A cut-off that differentiated persons that never or almost never engaged in a pain coping strategy from persons that engaged in a coping strategy more often would appear to have potential clinical utility.

To assess criterion-based validity we conducted Receiver Operating Characteristic (ROC) curve analyses to determine the optimal cutpoints for several criterion-based determinations [36]. For the pain ratings and WOMAC Physical Function scales we dichotomized the data to differentiate persons who did (WOMAC Physical Function ≥ 37 and 7-day pain ≥ 6) and did not (WOMAC Physical Function < 37 and 7-day pain < 6) have pain or functional loss that was equivalent to the average scores of persons preparing for knee replacement surgery [37,38]. For the 20-meter walk, we dichotomized scores to differentiate persons who did (≥ 1.22 m/sec) and did not (< 1.22 m/sec) have walking velocity that allowed for safe street crossing [39]. For the LLFDI scores, we dichotomized scores to differentiate persons who did (< 50 for LLFDI Frequency and < 65 for LLFDI Limitation Dimensions) and did not score (≥ 50 for LLFDI Frequency and ≥ 65 for LLFDI Limitation Dimensions in the upper (worst) quartiles. For the area under the curve (AUC), Standard Errors (SE) and 95% Confidence Intervals (CI) were used to describe the ROC results. A test result is no better than random chance when a curve has an area of 0.5 and a test result is perfect when the area under the curve =1. A general recommendation for interpreting AUC values is the following: 0.5 to 0.75 is fair, 0.76 to 0.92 is good, 0.93 to 0.97 is very good and 0.98 to 1.00 is excellent [36,40]. The optimal cut-point for the CSQ scales identified from the hierarchical regressions for each outcome was identified by selecting the measurement with the largest sensitivity + specificity)/2 from each curve.

Results

The characteristics of the sample are provided in Table 2. The Spearman correlations among the CSQ measures and the ICF measures, the 20-meter walk test and the CESD depression measure are presented in Table 3. Correlations among CSQ measures were all low to moderate and ranged from a high of 0.522 between Diverting Attention and Increased Behavioral Activities to a low of −0.05 between Diverting Attention and Ignoring Sensations. Most correlations among CSQ scores (76%) were below 0.40. Correlations among CSQ measures and outcome measures of interest ranged from a high of 0.405 between Praying and Hoping and WOMAC Physical Function and a low of −0.013 for Increased Behavioral Activities and the 20-meter walk test. Most correlations with the outcome measures of interest (71%) were below 0.20 and all significant correlations were in the expected direction, when considering scale coding

Table 3.

Spearman correlations among the CSQ scale measures and the criterion measures

Measure	IGNSENS	CSS	PRAY	RPS	DIVATT	IBA	CAT	7-day Pain	WOMAC	LLDI Freq	LLDI Lim	20-m Walk	CESD
IGNSENS	1.00
CSS	0.37***	1.00
PRAY	−0.14***	0.17***	1.00
RPS	0.08*	0.26***	0.35***	1.00
DIVATT	−0.05	0.25***	0.48***	0.45***	1.00
IBA	0.14***	0.51***	0.27***	0.34***	0.52***	1.00
CAT	−0.16***	0.07*	0.51***	0.28***	0.28***	0.15***	1.00
7-day Pain	−0.06	0.11**	0.39***	0.11**	0.18***	0.03	0.36***	1.00
WOMAC	−.012***	0.05	0.41***	0.18***	0.25***	0.07	0.38***	0.66***	1.00
LLDI Freq	0.09*	0.17***	−0.13**	−0.03	−0.03	0.09*	−0.22***	−0.17***	−0.23***	1.00
LLDI Lim	0.21***	0.07	−0.30***	−0.12**	−0.19***	−0.03	−0.33***	−0.33***	−0.47***	0.45***	1.00
20-m Walk	0.13***	0.04	−0.22***	−0.06	−0.10**	−0.01	−0.17***	−0.27***	−.030***	0.24***	0.37***	1.00
CESD	−0.18***	−0.10**	0.26***	0.10**	−0.10**	0.03	0.33***	0.19***	0.27***	−0.38***	−0.56***	−0.25***	1.00

IGNSENS=Ignoring Sensations, CSS=Coping Self-Statements, PRAY=Praying or Hoping, RPS=Reinterpreting Pain Sensations, DIVATT= Diverting Attention, IBA=Increased Behavioral Activities, CAT= Catastrophizing, WOMAC= Western Ontario and McMasters Universities Arthritis Index, LLDI Freq= Late Life Function and Disability Index, Frequency Total Score, LLDI Lim=Late Life Function and Disability Limitation Total Score, 20-m Walk= 20-meter walking test, CESD=Center for Epidemiological Studied Depression Scale

^***Correlation is significant at the <0.001 level (2-tailed).

^**Correlation is significant at the 0.01 level (2-tailed).

^*Correlation is significant at the 0.05 level (2-tailed).

The hierarchical regression models indicated that the most common CSQ scale scores to enter the models after adjusting for potential confounding were the Praying and Hoping and Catastrophizing scales. The Catastrophizing scale entered all models and in most cases had the strongest association with outcome among all CSQ measures. The Praying and Hoping scale was particularly strongly associated with the pain rating and WOMAC Physical Function. Table 4 summarizes the hierarchical models.

Table 4.

Hierarchical linear regression coefficient models for the outcome measures

Model	Unstandardized β (95%CI)	Standardized β	t	P value	R2	R2 change	Adjusted R2 for model	F (P value)
7-day average pain rating
Constant + covariates^					0.07		0.07	10.74 (<0.001)
PRAY (positive)*	1.2 (0.87, 1.54)	0.25	7.04	<0.001		0.10
CAT (positive)	0.94 (0.61, 1.28)	0.19	5.55	<0.001		0.03
CSS (continuous)+	0.17 (0.07, 0.26)	0.13	3.53	<0.001		0.01
IBA (continuous)	−.16 (−0.25, −0.06)	−0.12	−3.20	0.001		0.01
Complete model					0.22		0.21	22.75 (<0.001)
WOMAC Physical Function
Constant + covariates					0.11		0.10	16.68 (<0.001)
PRAY (positive)	5.75 (3.93, 7.57)	0.22	6.21	<0.001		0.11
CAT (positive)	5.20 (3.45, 6.95)	0.20	5.84	<0.001		0.03
DIVERT (positive)	2.99 (1.17, 4.80)	0.12	3.22	0.001		0.01
IBA (continuous)	−0.58 (−1.06, −0.09)	−0.08	−2.35	0.02		0.01
Complete model					0.27		0.25	28.49 (<0.001)
LLFDI Limitation
Constant + covariates					0.11		0.11	14.47
CAT (positive)	−5.19 (−7.51, −2.87)	−0.17	−4.40	<0.001		0.06
IGNSENS (continuous)	1.09 (0.46, 1.73)	0.13	3.37	0.001		0.03
Pray (positive)	−4.00 (−6.32, −1.68)	−0.13	−3.40	0.001		0.01
CSS (continuous)	0.62 (0.27, 1.21)	0.08	2.06	0.04		0.01
Complete model					0.22		0.20	18.15 (<0.001)
LLFDI Frequency
Constant + covariates					0.10		0.09	12.45 (<0.001)
CSS (continuous)	0.60 (0.38, 0.82)	0.19	5.40	<0.001		0.03
CAT (positive)	−2.32 (−3.16, −1.48)	−0.19	−5.40	<0.001		0.04
Complete model					0.17		0.15	16.68 (<0.001)
20-meter walk test
Constant + covariates					0.30		0.32	57.14 (<0.001)
CAT (positive)	−0.07 (−0.09, −0.04)	−0.15	−4.99	<0.001		0.02
IBA (continuous)	0.008 (0.001, 0.015)	0.069	2.31	0.02		0.01
Complete model					0.33		0.32	47.78 (<0.001)
Depressive Symptoms
Constant + covariates					0.06		0.05	8.72 (<0.001)
CAT (positive)	3.20 (1.96, 4.44)	0.19	5.06	<0.001		0.06
IGNSENS (continuous)	−0.37 (−0.71, −0.35)	−0.08	−2.17	0.03		0.01
Pray (positive)	1.50 (0.22, 2.79)	0.09	2.30	0.02		0.01
CSS (continuous)	−0.49 (−0.82, −0.17)	−0.11	−3.00	0.003		0.01
DIVERT	1.30 (0.14, 2.46)	0.08	2.20	0.03		0.01
Complete model					0.16		0.14	13.46 (<0.001)

PRAY= Praying and Hoping, CAT = Catastrophizing, CSS= Coping Self-statements, IBA= Increased Behavioral Activities, DIVERT= Diverting Attention, IGNSENS= Ignoring Sensations, WOMAC= Western Ontario and McMaster Universities Arthritis Index, LLFDI = Late Life and Function and Disability Instrument

^{^} All models adjusted for age, sex, comorbidity, symptom duration, unilateral or bilateral chronic knee pain and BMI

^*Variables coded as “positive” were converted to dichotomous scales with 0.5 or 0 coded as 0 and 1 or greater coded as 1.

⁺ Variables coded as “continuous” were kept on a continuous scale because they approximated a normal distribution.

Praying and Hoping and Catastrophizing consistently had the largest curve areas in the criterion-based validity analyses with areas ranging from 0.59 for the dichotomized 20-meter walk test to 0.72 for the dichotomized depression outcome. All curve areas for Praying and Hoping and Catastrophizing were significant at p=0.001 or less and were in the “fair” range. Specificity was consistently higher than sensitivity but specificity was never higher than 0.73. Table 5 summarizes the ROC curve characteristics for each outcome measure.

Table 5.

Receiver Operating Characteristic Curve analyese for each outcome measure

Measurement	Curve Area	SE	P value	95% CI	Optimal Cutpoint	Sensitivity	Specificity
7-day verbal pain rating
PRAY	0.66	0.02	<0.001	0.62, 0.70	0.75	0.58	0.69
CAT	0.66	0.02	<0.001	0.62, 0.70	0.75	0.55	0.73
WOMAC Physical Function
PRAY	0.70	0.02	<0.001	0.66, 0.74	0.75	0.66	0.67
CAT	0.71	0.02	<0.001	0.67, 0.75	0.75	0.62	0.71
LLFDI Limitation
CAT	0.68	0.03	<0.001	0.63, 0.73	0.75	0.60	0.71
IGNSENS*	0.61	0.02	<0.001	0.56, 0.66	2.75	0.63	0.53
LLFDI Frequency
CSS*	0.58	0.02	0.001	0.54, 0.63	3.75	0.54	0.62
CAT	0.64	0.03	<0.001	0.59, 0.69	0.75	0.56	0.68
20-meter walk
CAT	0.59	0.02	<0.001	0.55, 0.63	0.75	0.48	0.70
IBA	0.50	0.02	0.82	0.45, 0.55	----	----	----
Depressive symptoms
CAT	0.72	0.03	<0.001	0.67, 0.77	1.25	0.61	0.76

PRAY= Praying and Hoping, CAT = Catastrophizing, CSS= Coping Self-statements, IBA= Increased Behavioral Activities, IGNSENS= Ignoring Sensations, WOMAC= Western Ontario and McMaster Universities Arthritis Index, LLFDI = Late Life and Function and Disability Instrument

^*The IGNSENS and CSS cutpoints predict a positive outcome while the other CSQ measures predict a negative outcome.

Discussion

We hypothesized that persons recruited from the community who also had chronic knee OA pain would demonstrate similar associations to ICF-based criterion measures as compared to persons seeking care for chronic pain. We compared our construct validity findings to previously published research on patients with primarily chronic lower extremity OA pain [41] and a heterogeneous mix of chronic pain patients [3], and veterans treated in a Veteran’s Administration pain management program [9]. Ferreira-Valente and colleagues studied 117 outpatients, most of whom had chronic back, hip or knee pain [41]. They found generally weak associations between 2-item CSQ scores and criterion ratings of pain, physical function and depression, with the exception of Catastrophizing, which was the only CSQ scale that consistently showed significant correlations with a variety criterion measures. The three CSQ scales that showed the strongest significant associations with criterion measures of disability and pain intensity in the 2-item per scale development sample (which consisted of a group of patients presenting with a variety of chronic pain problems) reported by Jensen and colleagues [3] were Ignoring Sensations, Praying or Hoping and Catastrophizing. In a sample of veterans with mixed chronic pain, Tan and colleagues found that Coping Self Statements, Catastrophizing and Increased Behavioral Activities scales consistently demonstrated the strongest associations with depression, pain and disability [9]. Our construct validity findings were similar to some of this previous work, in that the Praying or Hoping and Catastrophizing scales generally had the strongest associations with the criterion measures.

Across all studies, only the Catastrophizing scale was a consistent predictor of outcome, supporting the importance of this response domain (and the 2-item measure of this construct) as a predictor of patient pain and functioning across samples of patients and individuals in the community with chronic pain. Such a finding is not surprising given the consistency with which measures of catastrophizing predict other pain-related domains [42]. However, our results extend this finding to persons in the community, and support the Catastrophizing scale as the 2-item CSQ scale with the most promise as a potential screening measure in individuals in the community with OA-related pain. Given the large volume of evidence supporting the role of pain catastrophizing as an important predictor of poor outcome for a variety of disorders [43–45], our study lends additional support for the use of the CSQ 2-item catastrophizing measure to screen for high pain catastrophizing. Screening for high catastrophizing would be particularly prudent because cognitive behavioral therapies have been shown to reduce catastrophizing and subsequently reduce pain and improve function for a variety of chronic pain conditions [21,46–48].

Our finding that the Praying and Hoping Scale of the CSQ was the second most strongly associated scale after the Catastrophizing scale is consistent with the findings of other researchers who also found that the Praying and Hoping Scale had cross-sectional and longitudinal construct and predictive validity. For example, MacDermid and colleagues reported that baseline scores on the Catastrophizing and the Praying and Hoping Scales of the CSQ showed the strongest associations ranging from Pearson r = 0.24 to 0.50 with baseline, 3- and 6-month follow-up pain and disability measures in 129 persons from the community with shoulder pain [49]. These data in combination with our findings suggest that the Praying and Hoping and Catastrophizing scales of the CSQ show the greatest potential for use in persons with musculoskeletal disorders.

In order to add to our understanding of the construct validity of multiple CSQ scales in explaining pain and functioning in persons with OA, we conducted hierarchical regression models in which we examined the extent to which combinations of CSQ scales enhanced the associations in the models. Examining combinations of CSQ scales has intuitive clinical appeal because it is likely that some patients who have high scores on one CSQ scale may have high scores on others. We found no other studies that examined this approach with the 2-item per scale CSQ. We found that construct validity of combinations of CSQ scales were supported relative to single measures, but that for a number of scales these additional contributions were small, and may not be clinically relevant. For example, Praying or Hoping scores explained an additional 10% of the variance in pain ratings over that of potential confounders, and Catastrophizing contributed an additional 3%. Two other scales (Coping Self Statements and Increased Behavioral Activities) each were statistically significant contributors, but only explained an additional 1% variance each; an amount that likely adds little support to the construct validity for these scales, after adjusting for catastrophizing and praying and hoping.

Catastrophizing explained a potentially important percentage of variance (i.e., greater than 1%) in outcome in all 6 of the regression analyses. Praying or Hoping was particularly important in explaining variance in pain and WOMAC Physical Function, but not for the other criteria. For example, persons scoring a 1 or higher on Praying or Hoping had a WOMAC Physical Function score that was, on average, 5.75 points higher than persons who scored less than 1, a clinically important amount of score variation [50]. Also, both Praying or Hoping and Catastrophizing independently contributed to the models for WOMAC and pain ratings, supporting the construct validity of these CSQ scales. The findings provide construct validity support for combinations of 2-item CSQ scales of Praying or Hoping, Catastrophizing, Coping Self-Statements and Ignoring Sensations..

Peat and Thomas described a sample of patients with chronic knee pain who also had a ≥ 50% worsening of their knee pain relative to 18 months prior [51]. These 57 patients had a mean numerical pain rating, scaled 0 to 10, of 8.77 and a mean score of 34 on the WOMAC Physical Function scale. These scores contrast to our sample in which persons had a mean pain rating of 3.9 and a WOMAC Physical Function score of 18.8. In other words, the sample of 57 patients studied by Peat and Thomas had substantially worse pain and function as compared to our sample. When comparing the dichotomized Catastrophizing scale score in the study of Peat and Thomas to our study, 65% of patients in the Peat and Thomas study were positive for Catastrophizing (i.e., scored a 1 or greater) while 37% of our subjects scored a positive on the Catastrophizing scale. The study by Peat and Thomas provide additional data to aid in interpreting the meaning of Catastrophizing scores and suggest that these scores differentiate among persons with different levels of pain and functional loss.

The criterion-based validity portion of our study was more exploratory, given the fact that no previous studies that have evaluated the CSQ scales using criterion measures. Other self-report screening tools have been shown to have strong diagnostic validity [5,52,53] but this issue has not been explored for the CSQ. We suspect this is the case because, unlike depression or anxiety disorder, which have well defined gold standards for diagnosis, constructs like catastrophizing or coping self-statements have no gold standard. We therefore used surrogate criterion measures to assess criterion-based validity. These surrogate measures had the advantage of strong clinical relevance to persons with chronic osteoarthritic knee pain.

We found the criterion-based validity of the more promising CSQ scales (e.g. Praying or Hoping and Catastrophizing) to be only fair. While the Praying or Hoping and Catastrophizing scales demonstrated clear evidence of statistically meaningful criterion-based validity, the specificity and particularly the sensitivity values suggest substantial error in prediction. While our analyses suggest that 2-item CSQ scales should not be used in isolation to make specific inferences the Catastrophizing and Praying or Hoping scales, in particular, may inform practice, and may be helpful during screening, when scores for these scales are positive (i.e. a score of 0.75 or greater). The specificities for Catastrophizing scales were 0.70 or above for 5 of the 6 “criteria” tested in Table 5. These findings indicate that negative Catastrophizing scores of <0.75 occurs at least 70% of the time, on average, when a person does not score positive on the criterion measure. We view this as a clinically important finding that indicates that the 2-item CSQ scale has a moderate amount of criterion-based power for a variety of clinically important functional and pain measures. However, more research is needed before CSQ scales can be endorsed as screening instruments for routine practice.

Tan and colleagues found Pearson r associations between 0.81 to 0.85 between 2-item CSQ measures and the parent CSQ measures while studying a sample of 563 patients with chronic pain in the VA system [9]. These estimates are in line with original work by Jensen and colleagues [3]. In hierarchical regression models, Tan and colleagues found that while the parent CSQ measures predicted additional variance for some pain and depression measures, the parent scales did not explain additional variance beyond the 2-item scales for disability and pain severity as measured with the Pain Severity scale of West Haven-Yale Multidimensional Pain Inventory. These cross sectional data suggest that parent CSQ scales provide modest additional predictive power for some outcomes but for others, the 2-item scales appear to capture the great majority of predictive power of the parent scales.

Our study has important limitations. The analyses were cross sectional which restricts inferences related to causality. Longitudinal studies that assess the extent of change in CSQ scores and concomitant changes in relevant outcomes and potential covariates are needed to gain insights in the potential causal nature of constructs measured with the CSQ. In addition, the regression analyses may have missed important confounders that may have influenced the associations that were reported. On the other hand, our sample size was much larger than is typical for a study examining the validity of pain coping measures [54,55], which allowed us to generate estimates with a high level of precision. The large sample size also likely resulted in the identification of several statistically important findings that may have limited clinical relevance. Our sample had limited diversity with 21% being African American and only 1% with Asian ethnicity. Studies of more racially and ethnically diverse samples are needed. Finally, content validity of the 2-item CSQ scores is limited given that constructs of interest for these scores are generally broad. With this limitation, the study provides construct and limited criterion-based validity support, particularly for the 2-item CSQ catastrophizing scale as a screening instrument.

In conclusion, the findings identified a group of 2-item per scale CSQ scales that have the greatest relevance to patients which chronic arthritic knee pain. These are the Catastrophizing, Praying or Hoping, Coping Self-Statements and Ignoring Sensations scales. Construct validity of the scales was supported by both the current study and previous work [3,9]. The Catastrophizing and Praying or Hoping scales, in combination, demonstrated the strongest associations with pain and WOMAC Physical Function Scales while the Catastrophizing scale showed the most substantial association with the participation domain of the ICF as well as the 20-meter walk outcomes. Criterion-based validity estimates were in the “fair” range, particularly for the Catastrophizing scale which showed the greatest potential, in combination with the Praying or Hoping scale, for future longitudinal study as screening measures. Clinicians treating persons with knee OA should consider use of the Praying or Hoping and Catastrophizing scales as screening instruments, but should use these scales in combination with other clinical data when making clinical decisions based on the measures.