See acknowledgments for members of the LUMINA Study Group
Systemic lupus erythematosus in three ethnic groups. X. Measuring cognitive impairment with the cognitive symptoms inventory
Article first published online: 5 JUN 2002
Copyright © 2002 by the American College of Rheumatology
Arthritis Care & Research
Volume 47, Issue 3, pages 310–319, 15 June 2002
How to Cite
Alarcón, G. S., Cianfrini, L., Bradley, L. A., Sanchez, M. L., Brooks, K., Friedman, A. W., Baethge, B. A., Fessler, B. J., Bastian, H. M., Roseman, J. M., McGwin, G., Reveille, J. D. and for the Lumina Study Group (2002), Systemic lupus erythematosus in three ethnic groups. X. Measuring cognitive impairment with the cognitive symptoms inventory. Arthritis & Rheumatism, 47: 310–319. doi: 10.1002/art.10457
- Issue published online: 5 JUN 2002
- Article first published online: 5 JUN 2002
- Manuscript Accepted: 24 OCT 2001
- Manuscript Received: 5 APR 2001
- NIH, National Institute of Arthritis and Musculoskeletal and Skin Diseases. Grant Number: R01-AR-42503
- General Clinical Center Research Grants, University of Texas-Houston Health Science Center. Grant Number: M01-RR-02558
- General Clinical Center Research Grants, University of Texas Medical Branch. Grant Number: M01-RR-00073
- General Clinical Center Research Grants, University of Alabama at Birmingham. Grant Number: M01-RR-00032
- Systemic lupus erythematosus;
- Cognitive impairment;
- Cognitive assessment
To determine the factor structure of the Cognitive Symptoms Inventory (CSI) in patients with systemic lupus erythematosus (SLE) participating in a multiethnic longitudinal study of outcome, the Lupus in Minority populations, Nature versus nurture (LUMINA) study.
LUMINA patients of Hispanic (n = 48), African American (n = 64), and Caucasian (n = 44) ethnicity who had a study visit (enrollment or followup) between January 1 and September 30, 2000 were included. Patients completed the CSI, a 21-item self-report measure of cognitive function. Sociodemographic, clinical, immunologic, psychosocial, and behavioral variables were ascertained per protocol and as previously described. Data were analyzed with SPSS. The factor structure of the CSI was determined using the principal axis method with oblique rotation as decided by Gorsuch. All factors having an Eigenvalue greater than 1 were considered. A 4-factor solution was derived that accounted for 42.6% of the common variance. The correlations between patient factor scale scores and variables from the demographic, clinical, psychosocial, and behavioral domains were then examined.
The four factors and their respective variance are, Attention/Concentration (28.8%), Pattern Recognition/Activity Management (5.7%), Intermediate Memory (4.7%), and Initiation of Executive Functions (3.4%); each factor correlated with the total CSI score. Overall, patients' factor scale scores were positively and significantly correlated with other measures of cognitive dysfunction such as the Systemic Lupus Activity Measure (neuromotor domain) or the Systemic Lupus International Collaborating Clinics Damage Index (neurocognitive impairment), as well as with measures of fatigue, maladaptive coping skills, poor mental functioning, poor social support, and helplessness. They were, however, not correlated with sociodemographic or clinical variables.
In addition to demonstrating that the CSI can be used to measure cognitive impairment in patients with SLE in the research setting, we have determined a 4-factor solution for the CSI that appears to have adequate metric properties. At present, the CSI may best be used as a screen for difficulties in daily activities involving intermediate memory, concentration, attention, and executive function. Nevertheless, further work with the CSI items and factor scales is necessary to establish internal and test–retest reliability of the factor scales; and provide additional evidence of the convergent and predictive validity of these scales in larger samples of patients from each ethnic subgroup.
Patients with systemic lupus erythematosus (SLE) often report difficulties in concentrating, remembering, and performing other cognitive-dependent activities of daily living (ADLs). Indeed, investigators have identified neurocognitive impairment, with or without preceding or concomitant evidence of central nervous system involvement, in 20% to 60% of these patients (1–8). The wide variation in frequency of neurocognitive impairment across studies may be due to several factors, such as differences in patient sample characteristics, cognitive assessment methods, and criteria for defining impairment (3). Little is known regarding the natural history of neurocognitive impairment in patients with SLE or the extent to which impairment may adversely affect patients' quality of life over long-term followup (3, 6). Nevertheless, neurocognitive impairment must be properly evaluated in order to identify patients whose cognitive dysfunction may increase their risk of physical injury or reduce their abilities to properly adhere to treatment regimens as well as to function effectively in their home or work environments.
It is generally agreed that physicians should use well-validated, brief screening measures to identify patients with SLE who should be referred for formal neuropsychological evaluation (2–4, 7). Several interview-based cognitive screening measures are available, such as the Mini-Mental State and the Neurobehavioral Cognitive Status examination (9, 10). However, many physicians find it difficult to use even brief interviews for cognitive assessment in busy practice settings. These physicians would benefit from the development of a brief, self-administered, paper-and-pencil screening instrument that may accurately identify patients who require additional interview-based screening or perhaps formal neuropsychological evaluation. There is only one report of the initial development of such an instrument. Pincus and colleagues (11) administered a 21-item Cognitive Symptoms Inventory (CSI) to 974 patients with various rheumatic diseases. No reliability data were reported, but it was found that patients' scores on the CSI were most strongly correlated with their scores on measures of fatigue and helplessness. It is not known, however, to what extent CSI scores correlate with scores on other measures of cognitive dysfunction. It also is not known whether patients' responses to the CSI may be aggregated into two or more relatively homogeneous scales that measure different aspects of cognitive dysfunction.
We report the first attempt to identify the underlying factor structure of the CSI using patients from the LUMINA (Lupus in Minority populations, Nature versus nurture) cohort (12, 13). This cohort allowed us to evaluate the CSI using a large sample of well-characterized patients with SLE from diverse ethnic backgrounds. We determined the factor structure of the CSI based on the responses of our cohort and produced 4 factor scales. We then examined the correlations among the scores on these scales with patients' responses to 3 self-report measures of cognitive dysfunction, as well as to self-report measures of fatigue, helplessness, self-efficacy, pain, social support and use of maladaptive coping strategies. This procedure allowed us to determine whether the CSI scales are more strongly associated with measures of cognitive dysfunction (i.e., convergent validity) than with demographic or psychosocial factors (divergent validity).
SUBJECTS AND METHODS
Participants in this study were part of the LUMINA study, a multiethnic (Hispanic, African American, and Caucasian) cohort of SLE patients from 2 states (Alabama and Texas) and 3 institutions (The University of Alabama at Birmingham, The University of Texas–Houston Health Science Center, and The University of Texas Medical Branch, Galveston) and their collaborating clinics and medical centers. The constitution of this cohort, as well as its baseline sociodemographic, clinical, immunogenetic, behavioral, and psychological features have been described in detail previously (12, 14, 15). Likewise, predictors of disease activity, damage caused by the disease or secondary effects of its treatments, correlates of fatigue and functioning, and predictors of early mortality have been described (15–17). One hundred and fifty-six LUMINA patients (48 Hispanic, 64 African American, and 44 Caucasian) who had a study visit (enrollment or followup) between January 1 and September 30, 2000 were eligible to participate in this study. All instruments used in the LUMINA study were first translated into Spanish by 2 bilingual investigators; subsequently, they were back-translated into English by 2 other bilingual researchers, so that the Spanish translation could be refined for meaning as well as cultural equivalence and competence (18). The Spanish versions were applied to those Hispanic patients who indicated that Spanish was their primary and/or preferred language (27%). Hispanics were, however, examined as a group regardless of the language in which the CSI and other instruments were administered. All of the African American and Caucasian patients were administered the English language versions of the instruments.
As previously noted, the CSI is a paper-and-pencil, self-administered inventory that consists of 21 questions aimed at determining the subject's ability to perform several cognitive functions and ADLs (11). The large majority of literate subjects required only minimal assistance to adequately complete the CSI.
The Systemic Lupus Activity Measure (SLAM)
Disease activity during the month preceding the study visit was measured with the Systemic Lupus Activity Measure. The SLAM is a validated instrument consisting of various clinical manifestations characteristic of SLE. These manifestations are grouped by organ systems (19, 20) and include the following: constitutional, integument, eyes, reticuloendothelial, cardiovascular, gastrointestinal, neuromotor, and joints. In addition, the values of several laboratory tests, including urinary sediment, hematocrit, erythrocyte sedimentation rate, serum creatinine, as well as white blood cell, lymphocyte, and platelet counts, are recorded in the SLAM. Manifestations in the neuromotor system include, stroke syndrome, seizures, cortical dysfunction, headaches, and myalgias. All items in the SLAM are graded on a 0–3 scale (absent to present and severe). The total SLAM score is the sum of the scores for all organ systems (range 0–69) and laboratory values (range 0–21). Thus the total SLAM score ranges from 0–90.
The SLICC (Systemic Lupus International Collaborative Clinics) Damage Index (SDI)
This validated instrument measures the extent of irreversible organ damage caused either by the disease or by the treatments used for it (21) from the time of diagnosis. Manifestations need to be present for at least 6 months to be scored in the SDI. Because the SDI is a cumulative construct, a manifestation is scored only once (1 point) and it is carried forward subsequently. A few manifestations can be scored twice (2 points) if a new distinctive event such as a stroke, myocardial infarction, loss of a limb, bowel resection, osteonecrosis or a second malignancy occur. Organ systems included in the SDI are ocular, neuropsychiatric, renal, pulmonary, cardiovascular, peripheral vascular, gastrointestinal, musculoskeletal, skin, premature gonadal failure, diabetes, and malignancy. The maximum possible SDI score is 50. Manifestations included in the neuropsychiatric domain include cognitive impairment, major psychosis, stroke, cranial or peripheral neuropathy, and transverse myelitis.
The Medical Outcome Study Short Form-36 (SF-36)
Health status was assessed by patients' responses to the SF-36, a validated instrument for self-reported health status in healthy populations as well as populations with disease (22, 23). This instrument consists of 36 items that reflect individuals' perceptions of their physical and mental functioning. The 36 items are aggregated into 8 scales: physical functioning, role-physical, bodily pain, general health, vitality, social functioning, role-emotional, and mental health. These 8 scales are used to produce 2 summary measures: physical health (physical component summary [PCS]) and mental health (mental component summary [MCS]). The results are normalized, with 0 representing lack of function and 100 representing full function. The data can then be compared with normative data for the general US population (which includes Caucasians and African Americans) or data from populations with various chronic medical disorders.
The level of pain for the week preceding the visit was measured with a 10-cm visual analog scale where 0 represents no pain and 10 most pain possible.
Arthritis Self-Efficacy Scale
This validated, self-report instrument, developed by Lorig et al, measures the extent to which individuals may successfully perform specific actions to attain health-related goals (24). This instrument consists of 3 sections (function, pain, and other symptoms). For this study, we only administered the “other symptoms” section and substituted the word “Lupus” for “Arthritis” in the instructions as well as in all the items. Scores on the self-efficacy scales range from 10 to 100, with higher scores indicating greater confidence in one's ability to perform behaviors necessary to better control symptoms of SLE. The maximum score for this section of the instrument is 700.
Arthritis Helplessness Index (AHI)
This 15-item, validated self-report instrument measures the extent to which individuals believe they are unable to control the symptoms of arthritis as well as the effects of arthritis on their health status (25). Responses are graded on a 5-point Likert scale (strongly agree to strongly disagree). The scores are reversed for a few items for consistency, so that higher scores represent higher degree of helplessness (range 0 to 75).
Fatigue Severity Scale (FSS)
This validated, 9-item, self-report scale is a measure of fatigue severity experienced by individuals while performing activities of daily living (26). Each FSS item is scored from 0 representing no fatigue to 7 representing most fatigue possible. The total score, which also ranges from 0 to 7, represents the mean of the scores across the 9 items.
We used a modified version of the Pain Catastrophizing Scale (27) to measure the extent to which our participants use maladaptive strategies for coping with their illnesses. This catastrophizing scale is a validated, 13-item, self-report measure of the belief that one is unable to cope effectively with persistent pain. It has been shown that high scores on this measure are associated with high levels of pain, psychological distress, and functional disability among a large number of patient groups, including those with musculoskeletal disorders. We substituted the word “Lupus” for “Pain” in the instructions. Scores on the catastrophizing scale ranged from 0 to 13, with higher scores indicating greater usage of maladaptive coping.
Social support was ascertained using the Interpersonal Support Evaluation list or ISEL (28), a 40-item instrument composed of 4 scales (appraisal, belonging, tangible, and self-esteem) and a summary measure. Each scale ranges from 0 to 10 and the summary measure from 0 to 40. Higher scores indicate better social support.
One week prior to the study visit, participants were mailed a package with all questionnaires described above except for the CSI. Upon arrival at the General Clinical Research Center, a Research Associate verified that all questionnaires had been properly completed. Immediately thereafter, the patients were given the CSI and asked to complete all items. The Research Associates aided patients who requested assistance in responding to the CSI; however, this assistance was limited (e.g., clarifying the wording of an item) in order to minimize extraneous influences on the patients' responses.
The SPSS statistical software (SPSS, Chicago, IL) was used to perform all analyses described. The baseline features of the LUMINA patients who participated in this study were examined using analysis of variance for continuous variables and chi-square for categorical variables. We followed procedures recommended by Gorsuch in performing the factor analysis of patients' responses to the CSI (29). Specifically, a correlation matrix of the 21 CSI variables was formed with communality estimates consisting of squared multiple correlations placed in the diagonal cells. The communality estimates were derived using 34 computer iterations in order to achieve factor stability. The principal axis factoring method was used to extract all factors having eigenvalues greater than 1, therefore, explaining a significant amount of the common variance. This method of factor extraction was used since it minimizes distortion of the factor solution (30). Scree tests suggested a 4-factor solution which accounted for 42.6% of the total variance; each of the factors also had an eigenvalue greater than 1. Because there was no indication that the 4 factors should be independent of one another, the 4 extracted factors were first rotated by a Promax oblique rotation procedure with kappa set to a power of 2. This analysis produced the simplest structure with the least correlation among factors. An attempt was also made to extract higher order factors. As suggested by Gorsuch (29), the original 4 factors were then rotated using the varimax solution in order to determine if an orthogonal rotation was more appropriate than an oblique rotation. We also performed a higher order factor analysis on the factor intercorrelation matrix using the principal factor method in order to determine if it was possible to form a linear combination of the 4 first-order factors.
Finally, in order to determine the convergent and divergent validity of the CSI and its factors, we used multiple regression procedures to estimate the patients' factor scale scores. We examined the associations among the estimated factor scores and the variables from the socioeconomic-demographic, clinical, psychosocial, and behavioral domains using Pearson's correlation coefficients.
The baseline socioeconomic-demographic, clinical, psychological, and behavioral features of patients in this study are shown in Table 1. They are comparable to the features of the entire LUMINA cohort, as previously reported.
|Variable||Hispanic n = 48||African American n = 64||Caucasian n = 44||Total n = 156||P*|
|Socioeconomic and demographic features|
|Age, years, mean ± SD||33.8 ± 44.6||31.9 ± 11.7||43.6 ± 14.1||35.8 ± 13.3||<0.001|
|Sex, % female||98||91||75||89||0.002|
|Education, years, mean ± SD||10.0 ± 3.9||12.8 ± 2.6||13.7 ± 3.1||12.2 ± 3.5||<0.001|
|Marital status, % married||40||27||71||43||<0.001|
|Below poverty line, %||37||37||9||29||0.003|
|Health insurance, %||55||81||91||76||0.001|
|House ownership, %||64||63||75||66|
|Disease duration, months, mean ± SD||17.1 ± 14.2||16.5 ± 13.9||19.0 ± 16.3||17.4 ± 16.4|
|Disease onset type, % acute||57||34||25||38||0.006|
|Number of ACR criteria, mean ± SD||5.7 ± 1.2||5.5 ± 1.3||5.5 ± 0.9||5.5 ± 1.2|
|Cortical dysfunction,† %||29.2||34.4||34.1||32.7|
|Neurocognitive impairment,‡ %||15.8||3.9||7.9||8.7|
|SLAM score at enrollment or baseline, mean ± SD||9.7 ± 6||10.3 ± 5.5||7.9 ± 3.6||9.4 ± 5.3|
|SDI score, mean ± SD|
|Enrollment or baseline||0.42 ± 0.76||0.55 ± 1.08||0.47 ± 0.80||0.49 ± 0.91|
|Last visit||1.15 ± 1.82||0.94 ± 1.74||0.57 ± 0.97||0.90 ± 1.60|
|Behavioral and psychological features|
|Helplessness, mean ± SD||39.4 ± 6.4||39.6 ± 6.8||40.1 ± 6.1||39.7 ± 6.4|
|Fatigue,§ mean ± SD||4.8 ± 1.6||4.9 ± 1.4||5.1 ± 1.4||4.9 ± 1.5|
|Coping with Illness, mean ± SD||15.7 ± 14.2||19.4 ± 15.2||14.3 ± 10.7||17.0 ± 13.7|
|Self-efficacy, mean ± SD||64.7 ± 19.3||56.3 ± 20.8||56.0 ± 23.8||58.7 ± 21.5|
|Mental functioning,¶ mean ± SD||43.9 ± 11.9||44.1 ± 12.0||46.3 ± 10.3||44.7 ± 11.5|
Table 2 shows the loading of each CSI item on the 4 first-order rotated factors. A factor loading of ≥ 0.50 was considered to be sufficient for the purpose of factor interpretation.
|Item number and description||Attention/Concentration||Pattern Recognition/Activity Management||Intermediate Memory||Initiation of Executive Functions|
|1. Dial telephone?||0.124||0.025||−0.068||0.573|
|2. Organize meals?||0.297||0.317||0.288||0.810|
|3. Recognize people?||0.329||0.538||0.237||0.260|
|4. Learn new things?||0.380||0.195||0.085||0.175|
|5. See colors as black and white?||0.023||0.699||0.238||0.088|
|6. Remember details of recent experiences?||0.659||0.411||0.430||0.255|
|7. Remember important past experiences?||0.596||0.441||0.324||0.197|
|8. Shop (food, others) without list?||0.362||0.403||0.593||0.203|
|9. Manage money and pay bills?||0.326||0.594||0.220||0.145|
|10. Remember to take medicines?||0.301||0.636||0.255||0.091|
|11. Remember details at home or work?||0.582||0.264||0.449||0.275|
|12. Stay awake during movie or TV show?||0.341||0.420||0.342||0.107|
|13. Concentrate on reading book/newspaper?||0.378||0.461||0.438||0.079|
|14. Concentrate on one task?||0.581||0.233||0.286||0.269|
|15. Concentrate on more than one task?||0.689||0.236||0.319||0.370|
|16. Find correct word during conversation?||0.539||0.177||0.429||−0.012|
|17. Remember why you came into a room?||0.773||0.178||0.266||0.160|
|18. Remember where you keep your glasses?||0.600||0.191||0.296||0.143|
|19. Remember things from 2 days ago?||0.530||0.288||0.493||0.109|
|20. Find your way while driving?||0.204||0.270||0.204||0.080|
|21. Keep track of things to do without a list?||0.348||0.284||0.919||0.108|
Factor 1: Attention/Concentration
This factor accounted for 28.8% of the common variance and was defined by 9 of the 21 items (item 6, 7, 11, 14–19). The item content reflected the ability to concentrate on multiple tasks and to attend to everyday events over moderate time periods. For example, item 15 asked respondents to indicate how much of a problem, over the last month, it has been to “concentrate on more than one task you need to do.” Items 18 and 19 evaluated attention to everyday events (e.g., “How much of a problem has it been to remember what you ate or wore two days ago?”).
Factor 2: Pattern Recognition/Activity Management
This factor accounted for 5.7% of the common variance and was defined by 4 items (item 3, 5, 9, 10). This factor was relatively heterogeneous in content. Items 3 and 5 involve respondents' abilities to recognize faces of familiar people and to correctly identify colors (e.g., “How much of a problem has it been to see different colors only as black and white?”). Items 9 and 10 ask respondents to indicate to what extent they can successfully manage their finances and to remember to take their medicines. These two sets of items may load on the same factor, as pattern recognition may be essential for individuals to successfully work with numbers and to identify medication containers.
Factor 3: Intermediate Memory
This factor accounted for 4.7% of the common variance and was defined by 2 items (item 8, 21). These reflected the ability to perform tasks without the organizational aid of a list. For example, one item asked, “How much of a problem has it been to shop for food or things you need without a list?”
Factor 4: Initiation of Executive Functions
This factor accounted for 3.4% of the common variance and was defined by items 1 and 2. The items assess the ability to perform 2 basic activities of daily living that require concentration, and performing a sequence of simple or complex behaviors, such as dialing a telephone and organizing meals.
Table 3 shows the correlation matrix among the 4 factors described above. The correlations, (0.13–0.35) were moderately large, and all but one correlation were statistically significant (P ≤ 0.001). Intermediate Memory, however, was not significantly correlated with Initiation of Executive Functions. Each of the 4 factors was also significantly correlated with the total 21-item CSI score, with r = 0.45–0.92 (P ≤ 0.001 in all cases).
|Factor||Number of items||Attention/ Concentration||Pattern Recognition/Activity Management||Intermediate Memory||Initiation of Executive Functions|
|2. Pattern Recognition/Activity Management||4||–||–||0.310*||0.184*|
|3. Intermediate Memory||2||–||–||–||0.126|
|4. Initiation of Executive Functions||2||–||–||–||–|
|Total CSI score||0.917†||0.672†||0.708†||0.449†|
The higher order factor analysis using Promax rotation produced 2 factors. The first second-order factor consisted of a linear combination of Attention/Concentration, Pattern Recognition/Activity Management, and Intermediate Memory. Each of the 3 first-order factors had loadings >0.40 on the second-order factor. However, only Initiation of Executive Functions comprised the remaining second-order factor.
Finally, factor rotation using the varimax solution revealed that the same items which loaded on the original 4 factors following oblique rotation continued to define the same 4 factors after orthogonal rotation (Table 4). Given the magnitudes of the correlations among the factors rotated by direct oblique procedures and the negligible changes produced by orthogonal rotation upon the relationships among the factors and the items, the oblique solution was retained.
|Item number and description||Attention/ Concentration||Pattern Recognition/Activity Management||Intermediate Memory||Initiation of Executive Functions|
|1. Dial telephone?||0.095||−0.060||−0.088||0.582|
|2. Organize meals?||0.177||0.239||0.201||0.760|
|3. Recognize people?||0.237||0.504||0.089||0.169|
|4. Learn new things?||0.358||0.160||−0.024||0.110|
|5. See colors as black and white?||−0.095||0.702||0.123||0.017|
|6. Remember details of recent experiences?||0.583||0.330||0.252||0.123|
|7. Remember important past experiences?||0.530||0.379||0.144||0.073|
|8. Shop (food, others) without list?||0.254||0.329||0.495||0.101|
|9. Manage money and pay bills?||0.239||0.570||0.059||0.046|
|10. Remember to take medicines?||0.208||0.615||0.095||−0.012|
|11. Remember details at home or work?||0.516||0.178||0.318||0.166|
|12. Stay awake during movie or TV show?||0.266||0.377||0.221||0.014|
|13. Concentrate on reading book/newspaper?||0.291||0.410||0.310||−0.027|
|14. Concentrate on one task?||0.537||0.161||0.146||0.170|
|15. Concentrate on more than one task?||0.640||0.149||0.160||0.259|
|16. Find correct word during conversation?||0.506||0.108||0.318||−0.117|
|17. Remember why you came into a room?||0.761||0.010||0.092||0.037|
|18. Remember where you keep your glasses?||0.573||0.122||0.159||0.039|
|19. Remember things from 2 days ago?||0.467||0.212||0.370||−0.004|
|20. Find your way while driving?||0.155||0.245||0.128||0.024|
|21. Keep track of things to do without a list?||0.224||0.180||0.874||0.001|
Table 5 shows that all 4 of the first-order CSI factors were associated with another standard self-report measure of cognitive dysfunction, lending support to the use of the CSI as a measure of this construct. Specifically, these factors were significantly positively correlated with a measure of cortical dysfunction from the SLAM (r = 0.18–0.36). It should be noted that Attention/Concentration was more highly correlated with the SLAM cortical dysfunction measure compared to the total CSI score. Attention/Concentration also was the only CSI factor scale that was significantly associated with cognitive impairment scores on the SDI, a measure of irreversible damage as a result of lupus or its treatments (r = 0.20). However, Attention/Concentration, Intermediate Memory, and Initiation of Executive Functions were all negatively correlated with scores on the Mental Component Scale of the SF-36 (r = −0.33, −0.21, and −0.24, respectively). As self-reports of cognitive impairment on the CSI factors increased, SF-36 mental functioning scores decreased.
|Factor||Cortical dysfunction†||Cognitive impairment‡||Mental functioning§|
|2. Pattern Recognition/ Activity Management||0.21||NS||NS|
|3. Intermediate Memory||0.22||NS||−0.21|
|4. Initiation of Executive Functions||0.18||NS||−0.24|
|Total CSI score||0.16||NS||−0.32|
Table 6 shows the relationship between the 4 first-order CSI factors and measures of social, psychological, and behavioral distress. All 4 factors were positively associated with the FSS (r = 0.21–0.31), the Catastrophizing Scale (r = 0.34–0.46), the AHI measure of helplessness (r = 0.16–0.29), and negatively associated with self-esteem, a measure of social support (r = −0.19, −0.37), and with the self-efficacy scale (r = −0.22, −0.35). Finally, all of the CSI factors, with the exception of Intermediate Memory, were significantly associated with pain. Thus, higher levels of cognitive dysfunction measured by the CSI factor scales were associated with higher levels of fatigue, pain, psychological distress, maladaptive coping skills, and lower levels of social support and of self-efficacy.
|2. Pattern Recognition/Activity Management||0.21||0.40||0.16||−0.19||−0.22||0.22|
|3. Intermediate Memory||0.29||0.41||0.26||−0.24||−0.35||NS|
|4. Initiation of Executive Functions||0.23||0.34||0.21||−0.24||−0.25||0.27|
|Total CSI score||0.31||0.52||0.27||−0.34||−0.36||0.26|
In contrast to the results described above, Table 7 reveals that none of the 4 CSI factors was associated with socioeconomic-demographic variables such as age, sex, ethnicity, education, poverty level, or housing ownership status. Furthermore, no factor correlated with clinical features selected such as number of American College of Rheumatology criteria and disease duration.
|Factor||Age||Sex||Ethnicity||Education||ACR criteria #||Disease duration|
|2. Pattern Recognition/Activity Management||−0.03||−0.15||0.05||−0.03||−0.13||−0.07|
|3. Short-term Memory||−0.03||−0.04||0.09||0.09||0.10||−0.03|
|4. Initiation of Executive Functions||0.04||0.06||−0.01||−0.14||0.02||−0.04|
|Total CSI score||0.04||−0.04||0.07||−0.00||−0.00||0.02|
Although patients with SLE may report variable degrees of impaired cognition, they often do not appear to be impaired during scheduled clinical encounters with the health system. This incongruence between patients' self-reports and physicians' observations of patient behavior is probably due to several factors, such as the difficulty in distinguishing the effects of fatigue from those of cognitive dysfunction as well as the presence of subtle cognitive impairments that are not readily elicited by standard mental status questions. These factors may be particularly important in clinical settings in which the ethnic backgrounds of patients and physicians differ from one another. As a consequence, physicians and other health care providers may not fully appreciate the presence or magnitude of impaired cognition as well as its potential impact on patients' abilities to comply with medical regimens, independently care for themselves, and avoid injuries.
Formal neuropsychological evaluations can be very helpful in identifying patients' cognitive impairments and possible strategies for coping with those impairments. Indeed, there are now computerized neuropsychological tests that do not require strong English reading skills or verbal interactions between patient and assessor (31). However, it usually is not feasible to perform formal neuropsychological evaluations of large numbers of patients with SLE due to several factors. These include limited access to well-trained and experienced neuropsychologists, especially in rural areas, as well as the economic barriers to thorough assessments of cognitive function that require large amounts of professional time and well-equipped laboratories. Moreover, these tests are generally expensive, and they are not covered by third-party payors. It is clear that it would be very useful to have a reliable and valid screening instrument that would accurately identify patients whose reports of cognitive impairments might be validated and accurately measured by neuropsychological assessment.
Our results suggest that the CSI is a promising self-report instrument for initial assessment of cognitive impairment per se in patients with SLE from diverse backgrounds. We found that all of our patients completed the CSI with only minimal assistance from the staff in approximately 10 minutes, regardless of the differences in ethnic background or the language in which it was administered. Only a handful of patients voiced concerns over the additional time required for their study visit due to the administration of the CSI. Moreover, there were no differences on the CSI factor scales among patients from different ethnic backgrounds. Thus, we believe that it is feasible to request large numbers of patients of various backgrounds to complete the CSI in busy clinical settings.
In addition to establishing the feasibility of using the CSI, the results of our principal factor procedure indicated that 4 dimensions underlie patients' responses to the instrument. These were: Attention/Concentration, Pattern Recognition/Activity Management, Intermediate Memory, and Initiation of Executive Functions. There were significant, albeit modest, associations among each pair of these 4 dimensions with the exception of Intermediate Memory and Initiation of Executive Functions. This indicates that, although the 4 dimensions of cognitive impairment evaluated by the CSI share a moderate amount of common variance, the factor scales are not so highly correlated that they duplicate one another. Indeed, the higher order factor analysis produced 2 factors. One higher order factor was comprised of a linear combination of dimensions involving concentration and memory abilities required to perform many ADLs such as keeping track of finances and remembering to take medications, i.e., Attention/Concentration, Pattern Recognition/Activity Management, and Intermediate Memory. The second higher order factor was comprised solely of Initiation of Executive Functions. This dimension is required to successfully perform sequences of behaviors required in daily life such as dialing a telephone and organizing meals.
Our correlational analyses produced initial evidence of the convergent and divergent validity of the 4 CSI factors. With regard to convergent validity, we found significant correlations among patients' scores on each of the CSI cognitive factor scales and their scores on the SLAM measure of Cortical Dysfunction. Each of these factor scales, with the exception of Pattern Recognition/Activity Management, was significantly associated with the SF-36 measure of Mental Functioning. Attention/Concentration showed the highest correlations with these validity criterion measures as well as with the SDI measure of Cognitive Impairment. This is probably due primarily to the fact that Attention/Concentration was comprised of a larger number of CSI items than the other three factors. However, this pattern of correlations also suggests that the item content of SLAM, SDI, and SF-36 instruments may largely relate to cognitive functions that are dependent on attention and concentration.
Our analysis of convergent validity also revealed that high scores on the CSI factor scales generally were significantly associated with measures of fatigue, psychological distress, social support, maladaptive coping skills, self-efficacy, and pain. These findings are similar to those of Pincus and colleagues who reported that total CSI scores were most highly correlated with measures of fatigue and helplessness (11). Nevertheless, all of the significant correlation coefficients produced by our convergent validity analyses were modest in magnitude, i.e., ranging from 0.18 to 0.46. These findings suggest that fatigue and pain make moderate contributions to patients' reports of cognitive dysfunction on the CSI. The correlational analyses also suggest that cognitive dysfunction may contribute to patients' use of maladaptive coping skills as well as to perceptions of high levels of helplessness and low self-efficacy levels.
With respect to divergent validity, our correlational analyses indicated that patients' scores on the 4 CSI factor scales were not significantly associated with their responses to any of the sociodemographic or clinical variables explored (e.g., age, sex, education, ethnicity, disease duration, number of ACR criteria). Thus, the patients' responses to the CSI factors were not confounded by these variables.
Despite the positive findings discussed above, 4 items did not load on any of the CSI factors. Effort should be devoted to revising these items and evaluating the effect of these revisions on the factor structure of the instrument. For example, items 12 and 13 ask subjects to report how much of a problem it has been to stay awake during a movie or television show and to concentrate on reading a newspaper or book. It may be that falling asleep during a movie or having difficulty in reading a newspaper produces relatively minor consequences for persons with SLE compared to problems in paying bills or organizing meals for family members. Changing the content of items 12 and 13 may reduce the variability in persons' CSI responses and thus allow both items to load highly on 1 of the 4 factors. In addition, a few CSI items may benefit from revision despite the fact that they already load highly on a CSI factor. One is the item: “How much of a problem has it been to shop for food or things you need without a list?” Most persons who successfully cope with cognitive dysfunction by using shopping lists will correctly produce a high rating for this item. However, some of these individuals may incorrectly perceive that the item asks them whether they have difficulty remembering to use shopping lists as a memory aid. These persons are likely to produce a low rating for this item and thereby increase the variability of their CSI responses.
In summary, the results of our study suggest that the 4 CSI factors that were empirically derived by principal factor analysis are promising measures of cognitive dysfunction in patients with SLE. Patients of diverse backgrounds rapidly complete this instrument with little difficulty. In addition, their scores on these scales are moderately correlated with measures of cognitive impairment that are independent of background and demographic variables. We wish to emphasize, however, that a great deal of additional work must be devoted to revisions of the CSI and additional validation studies of these revised instruments. First, it is desirable to develop additional items for the 4 factor scales that may increase the internal reliability of these scales. Increasing internal reliability may strengthen the associations between the CSI factor scales and the criterion measures of cognitive impairment (e.g. SLAM, SDI). It also would be desirable to develop additional items that may be used to evaluate other dimensions of cognitive function.
Second, it will be necessary to cross-validate the factor structure of the CSI on independent samples of patients with SLE. Although the present results suggest that 4 dimensions underlie the CSI responses of a sample of patients from 3 ethnic backgrounds, it will be very important to determine the extent to which this 4-factor structure replicates across samples of patients from each ethnic subgroup. Moreover, given that the patients in the present study are relatively young, it will be important to determine the extent to which older age influences the CSI responses of individuals in these ethnic subgroups. If these studies produce positive results, there would be strong evidence that patients with SLE produce similar responses to the CSI items regardless of age and ethnic background.
Finally, it will be necessary to establish the psychometric properties of the CSI. For example, future studies must examine the reliability and validity coefficients of the factor scales that are derived within each of the ethnic subgroups that are studied. Once adequate internal and test–retest reliability have been documented for these scales, it will be necessary to further assess convergent validity by determining the associations between patients' CSI scores and their performance on established neuropsychological measures, especially those that assess memory (e.g., Wechsler Memory Scale-Third Edition [WMS-III], Verbal Paired Associates Immediate and Delayed Recall subtests), attention/concentration (e.g., WMS-III Digit Span subtest), and executive function (e.g., Wisconsin Card Sorting Test). It also will be important to determine the predictive validity of the revised CSI factor scales. That is, it will be necessary to determine within each ethnic subgroup whether patients' scores on these scales may be used to predict future clinical outcomes such as changes in functional ability, both cognitive and otherwise, over time.
Although much work remains to be performed with the CSI, we believe that this instrument currently can be used as a screening tool for patients with SLE who report neurocognitive impairment that cannot be observed in the regular clinical encounter. It is probably best used as a screen for difficulties in daily activities involving intermediate memory, concentration, attention, and executive function. Moreover, we advocate that investigators devote effort to revising and improving this instrument in research settings so that it may eventually be used with confidence in both clinical and research settings in patients of diverse backgrounds. However, it should be emphasized that the CSI cannot be used, either at present or in the future, as a substitute for formal neuropsychological evaluation of patients with SLE.
Current LUMINA Investigators and staff:
At the University of Alabama at Birmingham: Graciela S. Alarcón, MD, MPH, Holly M. Bastian, MD, MSPH, Alfred A. Bartolucci, PhD, Barri J. Fessler, MD, Jeffrey M. Roseman, MD, PhD, MPH, Gerald McGwin, Jr., MS, PhD, Martha L. Sanchez, MD, MPH, and Ellen Sowell, Assistant Coordinator.
At the University of Texas–Houston Health Science Center: John D. Reveille, MD, Alan W. Friedman, MD, Chul Ahn, PhD, Jo McLain, BSN, RN, Robert Sandoval, BA, and Rudyard Lanete, BA.
At the University of Texas Medical Branch at Galveston: Bruce A. Baethge, MD, Shrilekha Sairam, MD, and Teresa Solis.
The authors express their gratitude to current and past LUMINA investigators, staff, and patients without whom this study could not have taken place, to Dr. Renato Alarcón for his most helpful comments, and to Ella Henderson, OA, for her technical assistance in the preparation of this manuscript.
- 3Cognitive dysfunction can occur without a neuropsychiatric event: identifying cognitive deficits in systemic lupus erythematosus. J Musculoskel Med 1999; 16: 356–63., , .
- 8Use of a new computerized test of cognitive function in systemic lupus erythematosus (SLE): a preliminary analysis [abstract]. Arthritis Rheum 1993; 36 Suppl 9: S87., , , , , .
- 11A self-report cognitive symptoms inventory to assess patients with rheumatic diseases: results in eosinophilia-myalgia syndrome (EMS), fibromyalgia, rheumatoid arthritis (RA), and other rheumatic diseases [abstract]. Arthritis Rheum 1996; 39 Suppl 9: S261., , .
- 18LUMINA Study Group. Systemic lupus erythematosus in three ethnic groups. V. Acculturation, health-related attitudes and behaviors and disease activity in Hispanic patients from the LUMINA cohort. Arthritis Care Res 1999; 12: 267–76., , , , , , for the
- 23SF-36 physical and mental health summary scales: a user's manual. Boston: The Health Institute. New England Medical Center; 1994., , .
- 28Measuring the functional components of social support. In: SarasonI, SarasonBR, editors. Social support: theory, research, applications. Boston: Martinus Nijhoff; 1985. p. 73–94., , , .
- 29Factor analysis. Philadelphia: W.B. Saunders; 1974..
- 31Assessment of cognitive functioning. In: The psychological corporation. ed. MicroCog. Texas: Harcourt, Brace & Company; 1996., , , , , .