Development of the Cogniphobia Scale for Headache Disorders (CS-HD): A Pilot Study

Abstract

Cogniphobia refers to the specific fear and avoidance of cognitive exertion, which is believed to precipitate or exacerbate headache. This parallels a well-documented phenomenon, kinesiophobia (fear of movement), in chronic pain. The existing measure of cogniphobia (C-Scale) was developed in persons with post-traumatic headache, and item content may not generalize to the broader headache population. This study aimed to develop and begin the initial examination of the psychometrics of a novel measure of cogniphobia for headache disorders: the Cogniphobia Scale for Headache Disorders (CS-HD). Candidate CS-HD items were developed through content analysis, patient evaluation and peer-debriefing. Eighty adults with migraine recruited from a headache center completed 23 candidate items for the CS-HD and surveys of demographics, anxiety, and headache locus of control. A series of dimensionality analyses identified a single component, composed of 15-items, which accounted for 54.05% of the variance in the CS-HD items. The CS-HD demonstrated high internal consistency in this sample (α = 0.94). Preliminary convergent validity analyses found CS-HD total scores were positively associated with anxiety (ρ = .37, p = .001) and locus of control (internal r = .45, p < .001, chance r = .30, p = .009). This study provides initial evidence supporting the use of the CS-HD to evaluate cogniphobia in people with headache.

Public Significance Statement

This study describes the development of a self-report measure to assess cogniphobia, a phenomenon where people with headache begin to fear (and even avoid) thinking too hard because of the belief that it causes headaches. This pilot study suggests that a 15-item version of the scale is reliable and demonstrates preliminary validity in people with migraine.

Migraine is prevalent, impacting 1 in 7 people in the United States (Smitherman, Burch, Sheikh, & Loder, 2013). Globally, migraine is the leading specific cause of disability among neurologic conditions (Leonardi & Raggi, 2013). Persistent cognitive difficulties, such as poor attention or concentration, could contribute to the high burden of migraine, particularly in work and school settings (Blumenfeld et al., 2011).

Research suggests fear of pain leads to fear of behaviors perceived to increase pain, avoidance of these behaviors, and increases in disability (French, France, Vigneau, French, & Evans, 2007; Black, Fulwiler, & Smitherman, 2015). The episodic, unpredictable, and incredibly aversive experience of migraine can lead to complex belief systems regarding potential triggering factors (Wober, 2010), which can lead to disengagement from daily functioning and increased physiologic reactivity when inevitably confronted with perceived trigger factors (Martin, 2010). Cogniphobia is the specific fear and avoidance of cognitive exertion, which is believed to precipitate or exacerbate headache (Martelli, Grayson & Zasler, 1999). Originally described in post-traumatic headache (Todd, Martelli, & Grayson, 1998), cogniphobia could have important ramifications for headache disorders broadly. Cognitive exertion is a critical component of everyday life; avoidance of cognitive exertion could result in diminished engagement in cognitively demanding tasks including work, school, and high stakes testing such as standardized tests and neuropsychological tests.

The chronic pain literature has produced several scales relevant to cogniphobia. The Pain Anxiety Symptom Scale includes a “cognitive” subscale, which describes difficult thinking when currently in pain (McCracken, Zayfert, & Gross, 1992; α = 0.87). Although certain items are tangentially relevant to cogniphobia, items all pertain to current pain, not interictal periods between episodes of pain. A measure of cogniphobia should pertain to fear and avoidance of cognition in between headache episodes (not merely during a headache episode, when cognitive exertion would be expected to be associated with increased pain). The Tampa Kinesiophobia Scale (TSK; Miller, Kori & Todd, 1991) is a 17 item scale which measures kinesiophobia, or fear of movement because it is believed to exacerbate chronic pain; internal consistency is good (α = 0.84) and the TSK is associated with pain symptoms, anxiety, and disability (French, et al., 2007). Researchers recommend using the total TSK score as literature regarding the TSK factor structure is mixed (Lundberg, Styf, & Jansson, 2009; French et al, 2007). However, the TSK is specifically a measure assessing fear and avoidance of movement; the content of a measure of cogniphobia should address fear and avoidance of cognitive exertion or thinking.

No consensus measure of cogniphobia exists. The first measure of cogniphobia directly adapted the TSK (Miller, Kori & Todd, 1991) to assess cogniphobia in people with post-traumatic headaches (C-Scale; Todd, Martelli & Grayson, 1998; Martelli, MacMillan & Grayson, 1999). However, this direct adaptation resulted in content primarily focused around injury and re-injury. This is not particularly relevant for most primary headache disorders which are not caused by a specific injury. Measures of cogniphobia should emphasize content that addresses fear of thinking due to the threat of head pain itself, not solely the threat of re-injury. Further, no psychometric information was provided in the published abstract for this scale (Martelli, MacMillan & Grayson, 1999). Authors of a more recent study of 74 undergraduates who reported frequent headaches (M = 6.67 headaches/month) due to stress addressed some of these issues by adding three items addressing fear of thinking due to the threat of head pain (Suhr & Spickard, 2012). Content analysis identified two subscales: Headache Dangerousness (α = .86) and Escape/Avoidance (α = .83). However, the authors did not perform statistical analysis to evaluate these subscales. The only other relevant available study adapted the TSK to evaluate avoidance of cognitive exertion due to stress (“mental kinesiophobia”) and found this measure had a higher mean score among 20 patients with chronic stress compared to 34 workers in one department of a mental health service organization (Schmidt, 2003).

The current pilot study aimed to further develop and initially evaluate the properties of the Cogniphobia Scale for Headache Disorders (CS-HD) to be appropriate for a broad sample of people with primary and secondary headache disorders. Specific aims included: 1) Develop a set of content valid candidate items acceptable to patients with primary headache disorders and providers; 2) Evaluate scale structure and select a set of CS-HD items that load onto scale component(s) and demonstrate a breadth of content; 3) Evaluate the internal consistency of the CS-HD; 4) Provide preliminary evidence for construct validity of the CS-HD by evaluating relationships with related symptoms and headache-related beliefs. As the fear-avoidance cycle is also implicated in anxiety (Olatunji, Moretz, & Zlomke, 2010), we expected strong positive relationships between the CS-HD and a measure of anxiety symptoms. Headache-specific locus of control describes the belief that individuals headaches are primarily determined by a) the individual (Internal), chance or fate (Chance), or one’s doctor or healthcare provider (Medical Professional). We expected positive relationships with internal and chance locus of control as mental exertion is both personally controllable, and also an unavoidable component of everyday life. We did not expect a significant relationship between CS-HD and Medical Professional locus of control.

Methods

Item Development

Items from existing measures (Todd, Martelli, & Grayson, 1998; Suhr & Spickard, 2012) were refined and new items developed with the goal of producing a final item bank in which each item was consistent with the construct of cogniphobia: fear and avoidance of cognitive exertion due to the belief that it precipitates or exacerbates headache. Development occurred iteratively through: 1) Cognitive interviewing with patients with primary and secondary headache disorders, 2) Peer-debriefing with clinicians specializing in headache, asking them to evaluate the content validity, wording, and clinical utility of the items, and 3) content analysis of existing measures of pain-related fear and avoidance. For example, all items referring to “pain” were modified to refer to “headache” or “headache pain.” Further, items which posited, “I have an injury,” were modified to, “I am hurting myself,” which was more relevant to people with primary headache disorders.

Participants

Consecutive patients were recruited from the waiting room of a headache clinic in the Bronx, NY. Participants were included if they had a current physician diagnosis of migraine, reported current criteria for migraine (Headache Classification Committee of the International Headache Society, 2013), were aged 18 and over, and had the capacity to read English and consent. For the purpose of scale development, participants were only included in the present study if they completed all candidate items for the CS-HD.

One hundred and thirty participants were approached for the study. Of the 110 participants consented, 95 participants returned the surveys. Of these, 15 participants were excluded from analysis (10 did not meet migraine criteria, 5 did not complete CS-HD items). Therefore, 80 participants were included in the CS-HD scale development analyses.

Procedure

Recruitment and enrollment occurred from June – August, 2014. Psychology graduate students performed informed consent and administered a series of surveys which took approximately 20 minutes to complete. The majority of participants completed all surveys in the clinic (n = 75); five began the surveys in clinic and mailed the completed surveys back to the clinic. Participants received a $10 gift card. This study was in compliance with the Institutional Review Board of the Albert Einstein College of Medicine (IRB # 2013–2859).

Measures

Cogniphobia

Participants completed the 23 candidate items for the CS-HD. Participants responded on a 4-point scale ranging from 1–4 (1 “strongly disagree,” 2 “disagree,” 3 “agree,” 4 “strongly agree”). Higher scores indicate higher levels of cogniphobia.

Migraine Diagnosis and Symptoms

Chart review obtained physician diagnosis of migraine. To confirm current diagnosis, all participants completed the American Migraine Survey/American Migraine Prevalence and Prevention diagnostic screener (Lipton et al., 2001), which evaluates migraine diagnostic criteria using the 2004 criteria (Headache Classification Subcommittee of the International Headache Society, 2004), which are essentially unchanged in the 2013 criteria (Headache Classification Committee of the International Headache Society, 2013). In a large epidemiologic study which sampled the US population, the screener demonstrated a sensitivity of 100% and specificity of 82% for migraine (Lipton et al., 2001). Migraine diagnosis was categorized as either episodic (<15 headache days/month) or chronic (≥15 headache days/month).

Participant Characteristics

Demographic questions assessed age, gender (“female” or “male”), race/ethnicity (“White, Non-Hispanic”, “Hispanic”, “Black/African American”, “Asian”, or “other”), and education (“8 grades or less”, “some high school”, “high school graduate or GED”, “some college or technical school”, “college graduate”, or “graduate degree”).

Anxiety Symptoms

The Generalized Anxiety Disorder – 7 (GAD-7)(Spitzer, Kroenke, Williams, & Lowe, 2006) is 7-item survey which evaluates Diagnostic and Statistical Manual (DSM) – IV diagnostic criteria for Generalized Anxiety Disorder (American Psychiatric Association, 2000), which are essentially unchanged in the updated DSM-V (American Psychiatric Association, 2013). Development studies demonstrate good internal consistency, criterion and construct validity (Spitzer et al., 2006). In this sample, the GAD-7 demonstrated excellent internal consistency (α = 0.91).

Headache Specific Locus of Control

The Headache Specific Locus of Control (HSLC) (Martin, Holroyd, & Penzien, 1990) is a 33-item survey which evaluates beliefs about the primary determinants of headache in people with recurrent headache disorders, such as migraine. Three 11-item subscales comprise the HSLC; subscales assess the extent to which individuals believe headache onset and course is primarily determined by their own behaviors (Internal), chance or fate (Chance), or the actions of their medical provider (Medical Professionals). Items are coded on a 5-point scale ranging from “strongly disagree” to “strongly agree.” In the development literature, each subscale demonstrated good internal consistency (αs = 0.80–0.89), adequate 3-week test-retest reliability (rs = 0.72–0.78) and good construct validity (Martin et al., 1990). In this sample, Internal and Chance HSLC demonstrated good internal consistency (αs = 0.86 and 0.80, respectively), but Medical Professionals HSLC demonstrated questionable internal consistency (α = 0.68). Results with Medical Professionals HSLC are therefore interpreted with caution.

Data Analysis

Dimensionality of the CS-HD was evaluated using 1) Horn’s Parallel Analysis using 95^th percentile of distribution of random data eigenvalues, 2) the Minimum Average Partials test, and 3) principal components analysis (PCA) with a promax rotation. Spearman’s rho evaluated inter-item correlations. The CS-HD total score distribution was described and evaluated for skewness and kurtosis. Cronbach’s alpha evaluated internal consistency. Correlations between the CS-HD and anxiety symptoms (GAD-7) and locus of control beliefs (HSLC) provided preliminary evidence for construct validity. Validity analyses were calculated both with complete case analysis (ns = 76–79), and with missing values imputed at the item-level using multiple imputation (n = 89). All analyses were conducted using SPSS version 22 (SPSS IBM, New York, U.S.A). Alpha was set at .05, two-tailed, for all inferential analyses; the Bonferroni-Holm method corrected for multiple comparisons, a sequential process which sets the threshold for p-value significance at alpha divided by the number of remaining comparisons (Holm, 1979).

Results

Demographics

Participants were majority female (87.5%) and White, Non-Hispanic (78.9%) with an average age of 44.95 years (SD = 12.65). Approximately half of the sample had a graduate degree (49.37%). Participants most commonly experienced episodic migraine (65.79%).

Structure and Item Reduction

Horn’s Parallel Analysis suggested extraction of a single component (Component 1 eigenvalue = 9.53, 95^th percentile eigenvalue cut-off = 4.20), as the eigenvalue for the second component (Component 2 eigenvalue = 2.03) did not exceed the cut-off of the 95^th percentile of distribution of random data eigenvalues (2.07) (Figure 1). However, the minimum average partials test recommended extraction of 3 components. Therefore, we explored using extracting 3 components using PCA with promax rotation. Components 1 and 2 were most highly correlated (.61), whereas Component 3 demonstrated modest correlations with both Components 1 and 2 (.36 for each). Together, all three components accounted for 57.16% of the variance in the CS-HD items; however, the majority of this variance (41.46%) was accounted for by the first component, with 8.84% accounted for by the second component, and 6.85% accounted for by the third component. Evaluation of the pattern matrix did not result in a theoretically consistent division of items (e.g., there was no clear distinction in content between each component; rather, avoidance and dangerousness were present across all three components); further, two items loaded at approximately the same magnitude across the first and second components. A PCA extracting only two components also did not result in a theoretically consistent distinction between the components. Together, the two components (correlated at 0.55) accounted for 50.31% of the variance in the CS-HD items; however, the majority of this variance (41.57%) was accounted for by the first component. Evaluation of the pattern matrix suggested only four items loaded onto the second component; however, these items did not differ in a consistent way from the content of items which loaded onto the first component. Therefore, consistent with the results from the Horn’s Parallel Analysis, the single component model was chosen.

Figure 1.

Scree plot from Horn’s Parallel Analysis depicting actual eigenvalues by components for all 23 candidate items for the CS-HD, and the cut-off of the 95^th percentile of distribution of random data eigenvalues.

Items were chosen for inclusion in the final CS-HD scale if they had high component loading in the single-component model, while maintaining a breadth of item content across the scale and moderate to low inter-item correlations. Items with loadings ≥ 0.60 were retained. A final PCA was conducted on the final 15-item CS-HD. A single component with an eigenvalue of 8.11 was extracted, which accounted for 54.05% of the variance in CS-HD items. Item component loadings ranged from 0.55–0.85 (Table 1). The majority of correlations among the final 15 items fell below ρ = 0.50, with a range of 0.26 – 0.74.

Table 1.

CS-HD Component Matrix

Item	Component Matrix Loadings
I’m afraid that I might make the cause of my headache pain worse if I concentrate too much	.74
I worry that when I have to think or concentrate too hard, that I will bring on a headache	.77
My headaches put my head and brain at risk for the rest of my life	.56
Headaches always mean I’ve hurt myself or have done something to make it worse	.66
I’m afraid that I might make my headaches worse by concentrating too much or being too mentally active	.79
Simply being careful not to concentrate too hard or too long is the safest thing I can do to prevent my headache pain from worsening	.77
I wouldn’t have this much headache pain if there weren’t something potentially dangerous going on in my head	.67
Headache pain lets me know when to stop concentrating so that I don’t hurt myself	.79
Performing a difficult mental task frequently brings on my headache pain	.85
I can’t do all the things normal people do because it’s too easy for me to cause harm to my headache condition	.67
It’s really not safe for a person with a headache condition like mine to engage in too much thinking and concentrating	.79
No one should ever concentrate on difficult mental tasks when he/she is experiencing headache pain	.67
I will stop concentrating as soon as I sense headache pain coming on	.69
I avoid thinking/concentrating too hard because it causes headaches	.81
When I have a headache, I’m afraid that thinking/concentrating too hard will make the headache pain worse	.75

Reliability and Distribution in this Sample

The 15 CS-HD items demonstrated excellent internal consistency in this sample (α = 0.94), and large item-total correlations (0.61–0.80). Total scores ranged from 15–48 (out of a possible 15–60). In this sample, the mean CS-HD total score was 27.26 (SD = 8.28) and distribution of scores did not vary substantially from a normal distribution (kurtosis = −.93, SE = .53; skewness = .13, SE= .27).

Initial Validity Examination

Content validity of the CS-HD was achieved through retaining broad item content during the item selection procedure. The set of 15 final CS-HD items had a Flesch readability index of 70.5, indicating that the final subset of items chosen were fairly easy to read (Flesch, 1948).

Relationships between the CS-HD and measures of anxiety and headache-related beliefs provided initial evidence regarding construct validity. Correlations using complete case analysis did not differ in direction or significance compared to correlations using multiple imputation and pooled estimates; complete case analysis results are presented below.

The GAD-7 was positively skewed; the majority of participants (67.5%) fell in the “no anxiety” range (scores = 0–4), 21.6% fell in the “mild anxiety” range (scores = 6–10), and 10.9% fell in the “moderate-severe anxiety” range (scores > 10). Higher scores on the CS-HD were associated with higher levels of anxiety symptoms (GAD-7) ρ = .37, p = .001. When evaluated by clinical category, participants whose GAD-7 scores fell in the “mild anxiety” or “moderate-severe anxiety” range had higher average CS-HD scores (M = 31.0, SD = 7.6) than participants whose GAD-7 scores fell in the “no anxiety” range (M = 25.4, SD = 7.6), t (75) = 3.02, p = .003.

HSLC scales were normally distributed. Higher scores on the CS-HD were associated with higher internal HSLC, r = 0.45, p < 0.001, and higher chance HSLC, r = .30, p = .009 (Table 5). The CS-HD was not significantly associated with Medical Professional HSLC, r = .08, p = .504.

Discussion

This study described the initial development of the Cogniphobia Scale for Headache Disorders (CS-HD). The current study suggests the CS-HD is a single-component scale which demonstrated excellent internal consistency in this sample (α = 0.94). This single component accounted for over half of the variance in item scores in this sample. It is possible that the structure of the CS-HD may differ based on patient population; for example, headache dangerousness may comprise a more distinct component of cogniphobia for people with post-traumatic headache compared to people with primary headache disorders. Further data in a variety of headache patient populations is needed to support the single-factor structure of the CS-HD.

Initial content validity was addressed through obtaining input from multiple stakeholders and retaining broad item content in the final scale. Preliminary convergent validity analyses found that higher CS-HD scores were associated with clinically significant generalized anxiety disorder symptoms; generalized anxiety disorder is characterized by heightened worry about a variety of factors (Spitzer et al., 2006) which, when comorbid with migraine, we expect might include potential migraine trigger factors (such as cognitive exertion). As expected, higher Internal and Chance HSLC were associated with higher scores on the CS-HD. Medical Professionals HSLC was not associated with the CS-HD, providing initial evidence for discriminant validity.

Future studies should evaluate construct validity of the CS-HD by evaluating its relationship with fear of headache pain and observed avoidance of cognitive exertion, such as failure of neuropsychological measures of effort during the interictal period. Given the expected transient cognitive changes expected during and immediately following a migraine attack (Meyer, Thornby, Crawford, & Rauch, 2000), researchers should ensure that people with migraine are in the interictal period when evaluating neuropsychological measures for validation of the CS-HD. Further, validation of the CS-HD using neuropsychological criterion measures may be complicated in certain subsets of people with migraine, such as those with chronic migraine (≥15 headache days/month).

Final item selection and evaluation of the psychometric properties of the CS-HD was accomplished using data from people with migraine. Migraine is a relevant population for the development of the CS-HD because it is the most common diagnosis for which people seek medical treatment for headache, and is associated with more disability worldwide than any other neurologic condition (Leonardi & Raggi, 2013). Migraine is also the headache disorder in which subsets of patients have demonstrated diminished performance on measures of cognitive functioning (Suhr & Seng, 2012). The TSK-adapted cogniphobia scale items were originally developed for a secondary headache population; further developing the measure in a primary headache disorder population, such as migraine, increased the clinical utility of the measure. Continued evaluation of the reliability and validity of the CS-HD in samples with a variety of both primary and secondary headache disorders is essential to further establish the psychometric properties and utility of the measure.

This was a small pilot study with a sample derived from a single population in a tertiary care headache center. Study participants were disproportionately white, female, and educated, and represented a more severe headache sample than those observed in general neurology and primary care settings. Thus, we cannot generalize the psychometric properties described in this study to other samples. However, even among this sample, scores on the CS-HD predominantly fell in the lower and middle range (15–58) of the possible scale range (15–60). Future research should continue to evaluate the dimensionality, internal consistency, and both convergent and discriminant validity of the CS-HD in a larger sample derived from a more diverse clinical setting, and which includes headache patients specifically seeking assessment of cognitive concerns.