Cancer-related fatigue (CRF) is a prevalent condition among patients. Assessing Fatigue in Persons with Cancer

Transcription

1 1685 Assessing Fatigue in Persons with Cancer An Instrument Development and Testing Study Horng-Shiuann Wu, Ph.D., R.N. 1 Maryellen McSweeney, Ph.D. 2 1 Department of Community and Family Medicine, Saint Louis University, St. Louis, Missouri. 2 School of Nursing, Saint Louis University, St. Louis, Missouri. BACKGROUND. Cancer-related fatigue (CRF) is a highly prevalent, subjective experience of patients with and survivors of cancer. Effective assessment of the attributes of CRF from the patient s perspective is essential. The current study developed a measure of CRF from the patient s perspective and determined its psychometric properties for patients with cancer undergoing chemotherapy. METHODS. Chemotherapy outpatients in a free-standing metropolitan area cancer clinic participated in the current multiphase study. In the instrument development phase, 42 items were generated from a qualitative study with 10 semistructured individual interviews and 6 daily fatigue diaries. These items were revised based on content evaluation by 20 cancer patient content experts. In the instrument analysis phase, the 30 content-validated items were pilot tested by 30 patients with breast carcinoma using cognitive interview techniques. The psychometric properties of the resulting 16-item, 5-point CRF rating scale were tested with 82 patients with breast carcinoma. RESULTS. Principal axis factoring gave a one-factor solution accounting for 59% variance. Coefficient alpha reliability was 0.95 for the final 15-item scale. Convergent validity with the Schwartz Cancer Fatigue Scale was 0.84, concurrent validity with the Geriatric Depression Scale was 0.61, and predictive validity with the Cancer-Related Fatigue Distress Scale was CONCLUSIONS. This newly developed instrument to assess the subjective fatigue symptoms of patients with cancer demonstrated effective use of diary and interview methods in instrument development and both cognitive interviewing and traditional psychometric techniques in instrument analysis. The instrument has promising psychometric properties, but confirmatory testing is needed. Cancer 2004;101: American Cancer Society. Supported by a Symptom Assessment and Management Nursing Research Grant of the Oncology Nursing Society Foundation. The St. Louis Cancer and Breast Insititute Chemotherapy Infusion Center offered its support and participation. Horng-Shiuann Wu s current address: School of Nursing, University of California-San Francisco, San Francisco, CA; wuh8@hotmail.com. Address for reprints: Horng-Shiuann Wu, Ph.D., R.N., School of Nursing, University of California- San Francisco, San Francisco, CA 94143; wuh8@hotmail.com Received January 12, 2004; revision received April 19, 2004; accepted June 21, KEYWORDS: cancer-related fatigue, affective feelings, mental dimension, temporal factors. Cancer-related fatigue (CRF) is a prevalent condition among patients with cancer 1 4 and survivors of cancer 1,5,6 that occurs across all ages, genders, cancer diagnoses, stages of disease, and treatment regimens. 7 It is generally recognized as a result of cancer itself and/or various cancer treatments. 5 The development of cancer-related fatigue guidelines by the National Comprehensive Cancer Network (NCCN) 8 and the initial formulation and preliminary testing of a diagnosis of CRF within the framework of ICD-10 diagnoses by the fatigue coalition 9 11 are recent achievements in CRF symptom assessment and management. Their goals are to ensure standardized, comprehensive fatigue assessment in clinical practice and research settings where assessment is the key to recognizing and managing CRF symptoms. 12,13 The currently available self-reported instruments identified in recent CRF literature were evaluated elsewhere 14,15 and are summarized in Table 1. Among these existing CRF instruments, measure American Cancer Society DOI /cncr Published online 13 August 2004 in Wiley InterScience (

2 1686 CANCER October 1, 2004 / Volume 101 / Number 7 TABLE 1 Key Features of Cancer-Related Fatigue Instruments Instruments Dimensions of measure No. of items Reference period Response scale Available languages Single-item measures of fatigue free-standing and incorporated in other instruments Rhoten Fatigue Scale Fatigue severity One item Present (RFS) graphic rating scale with 0, not tired, feelings of energy, and peppy, to 10, total exhaustion Visual-Analog Fatigue Fatigue severity One item Present 10-cm line with 2 end points I do Scale (VAFS) 31 not feel tired at all and I feel totally exhausted M. D. Anderson Symptom Inventory (MDASI) 40 Zung Self-Rating Depression Scale (ZSDS) 41 Fatigue severity Fatigue frequency Total of 19 items (13-item Part I and 6-item Part II); single fatigue item (Part I Item 2) Total of 20 items; single item: Item 10 I get tired for no reason In the last 24 hours 0 10 numeric rating scale with 0, not present, to 10, as bad as you can imagine During the preceding week Multi-item measures of fatigue incorporated in other instruments, typically measuring functional ability or quality of life Profile of Mood States (POMS) 39 EORTC Quality of Life Questionnaire (EORTC QLQ-C30) 42 The MOS 36-Item Short- Form Survey (SF- 36) 44 Functional Assessment of Cancer Therapy- Fatigue (FACT-F) 27 Intensity/presence of trait Vigor Activity (POMS-V) and Fatigue-Inertia (POMS-F) subscales: 8 and 7 items Fatigue intensity Fatigue scale (Items 10, 12, 18): 3 items Fatigue frequency Vitality Scale (Items 9a, 9e, 9g, 9i): 4 items Fatigue subscale: fatigue presence/ intensity, affective aspects, interference with functioning During the past week including today; right now During the past week During the past 4 weeks Fatigue subscale: 13 items During the past 7 days 1 4 four-point Likert scale with 1, none or little of the time, to 4, most or all of the time 0 4 five-point rating scales with 0, not at all, to 4, extremely 4 response levels from 1, not at all, to 4, very much 6 response categories: from all the time to none of the time 0 4 five-point rating scale with 0, not at all, to 4, very much so, Chinese, and Japanese, Iranian, Thai, and Italian, French, Portuguese, German, and Dutch Freestanding unidimensional fatigue measures Brief Fatigue Inventory (BFI) 33 Cancer-Related Fatigue Distress Scale (CRFDS) 25 Severity and impact of fatigue CRF distress with 5 categories: physical, social, psychological, cognitive, and spiritual distress 9 items Now (1 item) and during the past 24 hours (8 items) 20 distress items; with additional 3 items asking fatigue level now and usual and worst fatigue level since having cancer During the past 7 days 0 10 numeric rating scale, German, and Japanese 0 10 numeric rating scale Freestanding multidimensional fatigue measures Multi-dimensional Fatigue Inventory (MFI) 69 Fatigue Symptom Inventory (FSI) 43 5 dimensions: general fatigue, physical and mental fatigue, reduced motivation, and reduced activity Intensity and duration of fatigue and interference with daily life 20 items During the previous days 7-point scale 13 items In the past week 0 10 numeric rating scale, French, Dutch, and Danish Original Piper Fatigue Scale (PFS) includes Piper Fatigue Self- Report Scale-Baseline (PFS-B) and Piper Fatigue Self-Report Scale-Current (PFS- C) 70 Revised Piper Fatigue Scale (Revised PFS) 34 Multidimensional Fatigue Symptom Inventory (MFSI) and the short form (MFSI-SF) 29 Schwartz Cancer Fatigue Scale (SCFS) 23 Revised Schwartz Cancer Fatigue Scale (SCFS- 6) 32 Cancer Fatigue Scale (CFS) 26 Fatigue Management Barriers Questionnaire (FMBQ) 28 7 dimensions: temporal, intensity/ severity, affective, sensory, evaluative, associated symptom, and relief 4 dimensions: behavioral/severity, affective meaning, sensory, and cognitive/mood 5 dimensions: general, emotional, physical, mental fatigue, and vigor 4 dimensions: physical, emotional, cognitive, and temporal 2 dimensions: physical and perceptual 3 dimensions: physical, affective, and cognitive 10 concerns of patients thought to be barriers to fatigue management PFS-B: 42-item and PFS- C: 40 items 22 items and 5 openended questions Rationally derived scales: 83 items; empirically derived scales (MFSI- SF): 30 items Usually 100-mm horizontal visual analog scale (VAS) Now 0 10 numerical scale In the past 7 days 28 items In the prior 2 to 3 days 6 items In the past 2 to 3 days 0 4 five-point scale with 0, not at all, to 5, extremely 1 5 rating scale with 1, not at all, to 5, extremely 1 5 rating scale with 1, not at all, to 5, extremely 15 items Right now 1 5 numeric rating scale with 1, no, to 5, very much 28 items Not specified 1 5 Likert scale with 1, strongly disagree, to 5, strongly agree and Japanese EORTC: European Organization for the Research and Treatment of Cancer.

3 Assessing Fatigue in Persons with Cancer/Wu and McSweeney 1687 ment issues related to subjective features, terminology, dimensionality, symptoms and outcomes, and temporal characteristics of CRF are identified. Many of the instruments need further psychometric testing with cancer populations. 16 Cancer-Related Fatigue as a Subjective Experience One of the major changes in recent years is an increased emphasis on the individual s subjective point of view when assessing fatigue in persons with cancer. 9,16 Persons with cancer describe their fatigue as differing from what they had experienced previously 17,18 and reflecting noticeable loss of energy, exhaustion, increased discomfort, and inability to perform some tasks routinely done before their cancer experience. 2,7,17 20 The literature has shown discordance between patients and health care providers perceptions of cancer-related symptoms in both the level of distress caused by the symptoms and in the perceived attributes of the symptoms Because patients and health care providers often perceive cancer-related symptoms differently, instruments generated by investigators without being validated by persons who have experienced CRF should be viewed cautiously. Few of the instruments to measure CRF have been developed from qualitative studies of the patients experience although some have been content validated by patients with cancer or survivors of cancer. 23,28,29 Cancer-Related Fatigue as a Multidimensional Phenomenon Distinguishing Attributes from Outcomes Four major characteristics of CRF can be identified by focusing on the attributes of CRF revealed in qualitative studies and inferred from the dimensions of CRF in currently available instruments. 30,31 They are 1) the physical sensation of fatigue, including severity and intensity of its indicators; 2) mood, emotion, or affective feelings; 3) the cognitive or mental dimension of fatigue; and 4) temporal components, including the frequency and duration of the fatigue. Although a number of multidimensional measures of CRF are available, few incorporate all three of the physical, emotional, and cognitive components and distinguish these attributes from CRF consequences. 23,25,26,29,32 Many multidimensional instruments mix the effects or consequences of fatigue (such as impairment of functional performance and quality of life) with measures of the manifestations of fatigue (physical, emotional, and cognitive components). As a result, various combinations of high/low symptom scores and impairment scores can yield similar composite scores but very different fatigue experiences, making findings uninterpretable. Measures of fatigue symptoms should be distinguished from measures of fatigue consequences, 33,34 at least through the use of separate subscale scores or separate instruments. Temporal Issues There is evidence that CRF changes over time during the cancer treatments The fluctuating nature of CRF has been described as a roller coaster by patients with cancer who report that CRF fluctuates over time with changes in severity happening on a daily or hourly basis. 38 Multiple, very short reference periods, such as right now, may detect change but be unable to distinguish between random and clinically significant effects. Longer reference periods, such as in the preceding week, may be less sensitive to chance variations but more difficult to retrieve from memory and less accurate. The majority of current measures of CRF examine fatigue at the present time, 26,30,31,33,34,39 during the previous 24 hours, 33,40 in the previous 2 3 days, 23,32 in the past 7 days, 25,27,29,39,41 43 or during the past 4 weeks. 44 Instruments with shorter reference periods may be unable to capture the person s true fatigue experience whereas those with longer reference periods may yield less accurate information. CRF is a highly prevalent, multidimensional, subjective experience affecting all aspects of life of a patient with cancer. Many instruments measure fatigue, but none is a multidimensional assessment of the characteristics of fatigue developed primarily from the patient s perceptions in the United States. The purpose of the current study was to develop a specific measure of CRF from the person s perspective and to determine its psychometric properties for a population of patients with breast carcinoma. MATERIALS AND METHODS Design This multistage instrumentation study had an instrument development phase that consisted of item generation (Stage 1) and content validation (Stage 2) and an instrument analysis phase that consisted of a pilot study (Stage 3) and a main study (Stage 4). The stages are sequential, with each subsequent one building on the analyses of the preceding stage. All four stages took place in the same chemotherapy area of a freestanding outpatient cancer clinic in a metropolitan area. Procedures and Intermediate Analyses Phase 1: instrument development Item generation. Criterion sampling was strategy used in the item generation stage of the study. After informed consent was obtained, each of 10 female che-

4 1688 CANCER October 1, 2004 / Volume 101 / Number 7 motherapy outpatients, with a median age of 52 years (range, years), participated in an audiotaped, semistructured individual interview, lasting minutes, and kept a 2-week daily fatigue diary. Six returned completed diaries. Nine subjects had breast carcinoma. Various cancer stages and chemotherapy regimens were represented. Item generation began with a review of the interview transcripts and the diaries. 45 Content analysis guided the coding process to develop new instrument items. 46 Both the language and expression used by participants during the interviews and in the diaries were retained as much as possible to develop the initial item pool of 73 statements generated from the transcribed interviews and diaries. 45 The statements that were most frequently presented and perceived by the respondents as the most relevant to their subjective fatigue experience were included in the draft instrument. The resulting draft instrument consisted of a question stem, How do you typically feel today? Also, 42 items were related to three perspectives of CRF physical, emotional, and cognitive or mental fatigue with a 5-point rating scale, with 1 strongly disagree, 2 disagree, 3 equally disagree and agree, 4 agree, and 5 strongly agree. Content validation. The 42-item draft instrument was revised and validated based on evaluation from two panels, each of 10 content experts, using the method proposed by Lynn 47 and Grant and Davis 48 combined with nonparametric statistics to determine the retained or deleted items. Content experts were identified as patients with cancer who had experienced CRF. The findings of the first iteration of content validation guided the second iteration. The results of the second iteration were used to content validate the draft instrument. The second panel of content experts consisted of 10 patients with cancer with a median age of 50 years (range, years), who were receiving outpatient chemotherapy primarily for breast carcinoma at various stages. After informed consent was obtained, the content experts were asked to review each of the 42 instrument items and to evaluate the draft instrument as a whole by rating item representativeness/relevance on a 4-point rating scale. Identification of unclear items and items with an inappropriate reading level suggested addition or deletion of items, and additional comments/suggestions were invited at the end of the questionnaire. Based on the general definition of content validity 49 and Lynn s 47 definition of item content validity index (CVI) as the proportion of experts who scored the item as relevant or representative with either a 3 or 4, the 4 response options on representativeness/relevance were arbitrarily classified into 2 categories: representative (including the score of 4, the statement is representative of CRF; or 3, the statement needs minor rewording to be representative of CRF) and not representative (including the score of 2, the statement needs major rewording to be representative of CRF; or 1, the statement is not representative of CRF). The item CVI and item kappa coefficient were calculated for each item to measure the agreement of content experts on item representativeness/relevance. The kappa coefficients of the 42 items ranged from 0.33 to As a result, 25 of the 42 items were retained with or without minor rewording, 2 items were retained with major rewording, and 15 items were deleted. Three additional items I had a short attention span, I got tired all of a sudden, and I had my ups and downs were added based on the comments from content experts and discussion with a consultant who has expertise in both CRF and instrumentation. The resulting 30 items were pilot tested. Phase 2: instrument analysis Pilot study. The content-validated instrument was reformatted based on several considerations of survey design before pilot testing. The pilot instrument consisted of 30 items with a question stem, How much did you experience the symptom yesterday? The respondents rated how much of each fatigue symptom they experienced yesterday on a 5-point rating scale, with 5 very much, 4 quite a bit, 3 somewhat, 2 a little, and 1 not at all. The time reference, yesterday, was chosen based on considerations of fluctuation as one of the characteristics of CRF, 35 37,50 the accuracy of retrieval of information over time, 51,52 and the need to specify a reference period. 52 Because patients with cancer are at risk of mental fatigue, 9,11,53 finer categories of rating 16,54 may not add much to the usefulness of their responses but may decrease the patients capacity for judgment. Consequently, a 5-point rating scale was chosen as the scaling system for the new instrument with verbal labels at all points to enhance scale reliability. 52 From the various response labels measuring quantity in 5 equal intervals in the survey literature, 54,55 the labels 5 very much, 4 quite a bit, 3 somewhat, 2 a little, and 1 not at all were chosen for the new instrument. After informed consent was obtained, the 30 patients with breast carcinoma, with a median age of 51.5 years (range, years), who were undergoing outpatient chemotherapy completed the 30-item in-

5 Assessing Fatigue in Persons with Cancer/Wu and McSweeney 1689 strument and participated in brief, follow-up individual cognitive interviews conducted by the investigator. The retrospective think-aloud method 56 was used to discover whether any items were unclear or difficult to answer. Participants were asked to report how they arrived at their answers. The information was coded and analyzed based on three response components understanding the question, retrieving or arriving at the answer, and formatting and reporting the answer. There were also five problem categories lexical, inclusion/omission, temporal, logical, and computational problems. 57 The cognitive interview after instrument completion provided an insightful understanding of the structure of each item and the dynamics of the whole instrument. Several items were revised based on differing perceptions that surfaced during cognitive interviewing. The use of yesterday as the reference period was raised during the cognitive interview. Most respondents had a consistent interpretation of yesterday as starting from the time of getting up in the morning until the time of going to bed at night. There was greater disagreement about the representativeness of yesterday as a typical day with respect to fatigue experience, although many did identify it as a typical day. The problems revealed in the structured cognitive interview process, added comments made by the participants during the debriefing, and the results of the analysis of item means, standard deviations, interitem correlations, and corrected item-total correlations were used to identify 16 items to retain with or without minor revision. The 16 items were ordered by response sensitivity from less sensitive (e.g., physical fatigue items) to more sensitive questions (e.g., cognitive fatigue items) to enhance response commitment in the main study. 58,59 Main study. The revised instrument for psychometric testing in the main study consisted of 16 items with the question stem, How much did you experience the symptom yesterday? The items were rated on a 5-point scale, with 5 very much, 4 quite a bit, 3 somewhat, 2 a little, and 1 not at all. Scores on the items were summed, with higher scores indicating more fatigue. An additional item Was yesterday a typical day for you? was added at the end to understand if the reference period yesterday was able to reflect the person s fatigue symptoms in general. With approval of the university institutional review board, the participants were instructed to complete four instruments and answer five demographic questions, in the following order: the 16-item newly developed CRF instrument, the Schwartz Cancer Fatigue Scale (SCFS-6), 32 the Geriatric Depression Scale (GDS), 60 and the Cancer Related Fatigue Distress Scale (CRFDS). 25 The sequence of testing was dictated by considerations of instrument length and relevance to CRF. Five demographic questions pertaining to age, highest education level, stage of breast carcinoma, chemotherapy regimen, and the use of epoetin alfa (Procrit) to treat anemia in cancer were asked by the investigator after completion of the instruments. The SCFS-6 32 was developed to measure the physical and perceptual dimensions of CRF. The instrument consists of 6 items measuring CRF with a 5-point rating scale ranging from 1 not at all to 5 extremely in the past 2 to 3 days. Higher scores indicate more CRF. The uniqueness of the SCFS-6 is that each item consists of a single word instead of a sentence, which is the format found in most fatigue instruments. The words were generated by the investigator, but validated by patients with diverse types of cancer and oncology nurse experts. Its reliability was 0.90 for the total scale, 0.88 for the physical subscale, and 0.81 for the perceptual subscale. The two-factor measurement model was established by exploratory factor analysis and confirmed by confirmatory factor analysis. 32 The SCFS-6 was used to assess the convergent validity of the newly created instrument. The GDS Short Form is a 15-item scale with yes or no response options measuring mood over the past week. 60 Scores on the items are summed, with higher scores indicating greater depression. Its psychometric properties were supported by a Cronbach s alpha coefficient of 0.94 and split-half reliability of The GDS was also found to be positively correlated with other tested measures for depression, such as the Zung Self-Rating Depression Scale (r 0.94) and the Hamilton Rating Scale for Depression (r 0.83). 60 As many cancer studies have shown the co-occurrence of depression and CRF, 12,53,61 concurrent validity of the new instrument was evaluated by calculating Pearson correlations between the total fatigue score on the new instrument and the depression score on the GDS Short Form. 60 The CRFDS is a unidimensional measure of distress caused by CRF. It was developed based on qualitative data 18 and content validated by survivors of cancer. The CRFDS has 20 items with a 0 10 rating scale measuring the distress experience of CRF in the past 7 days. Higher scores indicate greater distress from fatigue symptoms. Three additional questions ask the person s current, worst, and usual CRF level. Its internal consistency reliability was 0.98 for the total 20-item scale and the one-factor measurement model was established by exploratory factor analy-

6 1690 CANCER October 1, 2004 / Volume 101 / Number 7 sis. 25 Predictive validity was evaluated by looking at the correlation between CRF and its consequence and perceived future criterion CRF distress. 25 In addition, convergent validity was also assessed by computing the Pearson correlation between the total fatigue score on the new instrument and the single item score, rating from 0 to 10 usual fatigue level since having cancer on the CRFDS 25 for those patients (n 51) who documented that yesterday was a typical day on the new CRF instrument. ANALYSES AND RESULTS OF THE MAIN STUDY After approval was obtained from the institutional review board, data for the main study were collected from 82 female patients with breast carcinoma who were undergoing chemotherapy. The majority were well educated Caucasians with a mean age of 52.4 years (range, years). They were in various stages of breast carcinoma (Stage I, 18.8%; Stage II, 37.5%; Stage III, 12.5%; and Stage IV, 31.3%), received diverse chemotherapy regimens, and were at different points within their chemotherapy cycles. Relatively few (17.3%) either regularly or recently had received epoetin alfa for treating low red blood cell counts. SPSS 10.0 statistical software was used for data management and analysis (SPSS, Chicago, IL). To ensure the quality of data and avoid data entry errors, the data were entered in separate SPSS data files at two different times. The matched files were merged and errors were located and corrected. Among the 82 respondents, 65.9% documented that yesterday was a typical day for them. Yesterday was atypical for 34.1% who either felt more fatigued or had more energy because of various physical conditions such as flu, arthritis, or surgery or because it was 1 day before chemotherapy. Item Analysis Inter-item correlations were evaluated by calculating Pearson correlations. Low correlations existed between Item 8 (I had trouble sleeping) and the remaining 15 items (r ), suggesting removal of the item. A low corrected item-total correlation (r 0.26) also supported the deletion of Item 8. The corrected item-total correlations ranged from 0.65 to 0.86 for the remaining 15 items in the instrument, well above the desired minimum of This suggested that the 15 items in the new instrument consistently measured the same construct. Factorial Construct Validity After removal of Item 8, principal axis factoring with oblimin rotation was used to determine the underlying structure of the remaining 15 items and to TABLE 2 Two-Factor Pattern Matrix with Factor Loadings a Items Factor 1 2 Item 3: I didn t have the energy to get up and do things 1.00 Item 2: I felt like sitting around more than I usually do 0.91 Item 1: I was physically drained 0.89 Item 4: I felt guilty that I was too tired to do the things that I normally do 0.85 Item 7: I got tired easily from activities 0.82 Item 5: I wanted to sleep a lot 0.81 Item 10: I was forcing myself to get up and do things 0.79 Item 11: I was wiped out emotionally 0.65 Item 6: I was more sensitive than usual 0.63 Item 13: I didn t feel rested after sleeping 0.52 Item 15: I could hardly stay awake early in the evening 0.50 Item 12: I got tired all of a sudden 0.49 Item 9: I felt upset because I didn t get enough accomplished 0.49 Item 16: I had trouble remembering things 0.89 Item 14: I had a short attention span 0.70 a Extraction method: principal axis; rotation method: oblimin with Kaiser normalization. test for factorial construct validity of the newly developed CRF instrument. A Kaiser Meyer Olkin criterion equal to 0.91 and a significant chi-square value of the Bartlett test of sphericity indicated a meaningful factor analysis. The exploratory factor analysis yielded a 2-factor solution accounting for 63.4% of the variance with a 15-item model and respective eigenvalues of 9.10 and The results of the factor analysis suggested that a physical and emotional fatigue factor (consisting of Items 1 7, Items 9 13, and Item 15) and a cognitive fatigue factor (consisting of Items 14 and 16) could be the two underlying dimensions of the new 15-item fatigue instrument. Factor loadings of the retained 15 items ranged from 0.49 to 1.00 (Table 2). Although the 2-factor model accounted for a reasonable percentage of the variance (63.4%) in fatigue experiences, the small eigenvalue of Factor 2 ( 1.03) and the moderately high correlation (r 0.65) between the two factors suggested the viability of a single-factor solution. Principal axis factoring, forcing a one-factor solution, examined this possibility. The 1-factor solution accounted for 59.0% of the variance with 15 items with an eigenvalue of Factor loadings of the 15 items ranged from 0.68 to 0.88 (Table 3). The results of the one-factor solution suggested that a single dimension including physical, emotional, and cognitive fatigue underlies the new 15-item fatigue instrument. Both the two-factor and one-factor solutions differ from the three-factor measurement model originally proposed.

7 Assessing Fatigue in Persons with Cancer/Wu and McSweeney 1691 TABLE 3 One-Factor Pattern Matrix with Factor Loadings a Items Factor 1 Item 7: I got tired easily from activities 0.88 Item 3: I didn t have the energy to get up and do things 0.88 Item 5: I wanted to sleep a lot 0.86 Item 1: I was physically drained 0.85 Item 2: I felt like sitting around more than I usually do 0.83 Item 11: I was wiped out emotionally 0.83 Item 10: I was forcing myself to get up and do things 0.79 Item 4: I felt guilty that I was too tired to do the things that I normally do 0.75 Item 14: I had a short attention span 0.70 Item 12: I got tired all of a sudden 0.69 Item 16: I had trouble remembering things 0.69 Item 15: I could hardly stay awake early in the evening 0.69 Item 6: I was more sensitive than usual 0.69 Item 13: I didn t feel rested after sleeping 0.69 Item 9: I felt upset because I didn t get enough accomplished 0.68 a Extraction method: principal axis. Convergent Validity Convergent validity was evaluated by calculating a Pearson correlation between the total fatigue scores on the new 15-item CRF instrument and the total scores on the SCFS The correlation between the two instruments was 0.84 for 82 patients with breast carcinoma. The high correlation between the scores on the new CRF instrument and the scores on the SCFS-6 strongly indicates that the new instrument measures approximately the same construct as the SCFS-6. The correlation of the scores between the new CRF instrument and the single fatigue item on the CRFDS was 0.68 for the 51 patients who documented that yesterday was a typical day. This moderately high correlation provides further evidence of the convergent validity of the new instrument. Criterion-Related Validity A moderate correlation (r 0.61) supported the concurrence of fatigue and depression in these patients with cancer. Predictive validity was evaluated by assessing the correlation between the total fatigue score on the new instrument and the total fatigue distress scores on the CRFDS. 25 A strong correlation (r 0.83) supported the predictive validity of the new instrument. Internal Consistency Reliability Internal consistency reliability of the newly developed 15-item CRF instrument was estimated by Cronbach s alpha. The total scale alpha equal to 0.95 exceeded the usual minimum criterion of internal consistency reliability of 0.70 for a new instrument. 66 Thus, there is evidence that the 15 items consistently measured the same construct. For the 2-factor solution suggested by factor analysis, the value for the 13-item physical and emotional subscale was 0.94 and the value for the 2-item cognitive subscale was Both subscale values were greater than the criterion of DISCUSSION The new CRF instrument is a subjective measure of the characteristics of CRF that was developed from the perspective of patients with cancer. Its validity was established through testing two types of construct validity (factorial construct and convergent validity) and two types of criterion-related validity (concurrent and predictive validity). Either the 1-factor or the 2-factor solution can theoretically and statistically support the factorial validity of the 15-item CRF instrument although neither supports the researcher s a priori 3-factor model. Convergent validity was supported by a strong correlation (r 0.84) between the new CRF instrument and the well developed SCFS-6 and a moderately strong correlation (r 0.68) with the single fatigue item on the CRFDS. Predictive validity was established by a strong correlation (r 0.83) between the fatigue scores on the new CRF instrument and the fatigue distress scores on the CRFDS. The new CRF instrument and the two CRF measures used in its validation are summarized and compared in Table 4. A moderate correlation (r 0.61) between the fatigue scores on the new 15-item instrument and the depression scores on the GDS Short Form was evidence to support the concurrent validity of the new instrument. As shown in the literature, CRF frequently coexists with psychological symptoms such as depression. Stone s 67 suggestion of another factor mediating the relation between depression and CRF is a noteworthy and logical approach to explain the moderate correlation (r 0.61) between fatigue scores and depression scores. A high Cronbach s alpha ( 0.95) indicated that the 15 items of this new instrument were consistently measuring the same construct, which supported its internal consistency reliability. No test retest reliability was tested because CRF fluctuates over time. Its lack of stability would make the results of test retest computation hard to interpret meaningfully. Although this instrument has shown an adequate degree of reliability and moderate to strong construct and criterion-related validity, the three-factor measurement model (physical, emotional, and cognitive fatigue) was not supported by exploratory factor analysis. The two criteria used to determine dimensional-

8 1692 CANCER October 1, 2004 / Volume 101 / Number 7 TABLE 4 Comparisons of Three Cancer-Related Fatigue Instruments Instrument Dimensionality Development Reference period Reliability and validity Feasibility New CRF instrument Revised Schwartz Cancer Fatigue Scale (SCFS-6) 32 Cancer-Related Fatigue Distress Scale (CRFDS) 25 Unidimensional; distinguishes CRF from causes and consequences Multidimensional: physical and perceptual; distinguishes CRF from causes and consequences Unidimensional; distinguishes CRF distress from other attributes Item pool generated from patient interview and diary content; content validated by patients with cancer; cognitive interview technique for pilot testing; psychometric testing with patients with breast carcinoma receiving chemotherapy Item pool generated by the investigator; informally and formally content validated by patients with cancer; rigorous psychometric testing (exploratory and confirmatory) with patients with cancer and long-term survivors Item pool generated from qualitative interview data; content validated by cancer survivors; psychometric testing with diverse patients with cancer Yesterday In the past 2 to 3 days During the past 7 days Reliability: good internal consistency of 0.95 for 15-item scale Construct validity: one-factor solution accounted for 59.0% variance; factor loadings ranged from 0.68 to 0.88; Pearson correlation with SCFS-6 was 0.84; Pearson correlation with the single fatigue item on the CRFDS was 0.68 Criterion-related validity: Pearson correlation with GDS was 0.61; Pearson correlation with total fatigue distress scores on CRFDS was 0.83 Reliability: good internal consistency of 0.90 for 6-item total scale, 0.88 for physical subscale, and 0.81 for perceptual subscale Construct validity: supported by GFI 0.98 and AGFI 0.94; subjects in treatment scored higher on all items, both subscales and total scale than those who had completed treatment; there were significant differences by time since treatment Reliability: good internal consistency: 0.98 for 23-item scale and 0.98 for 20-item scale Content validity: 23 item CVIs ranged from 0.60 to 1.00 with a mean CVI of 0.91 Construct validity: factor analysis confirmed all items loaded on only one factor; 20 of the 23 items had factor loading 0.70 Simple, easily completed 3 minutes; A Flesch Kincaid grade level of 6.5, low literacy level suitable F Brief, simple, easy to use; electronic version (computertouch screen) is available Structure of the questions is not straight forward; some patients have difficulty in answering the questions without assistance CRF: cancer-related fatigue; GDS: Geriatric Depression Scale; GFI: Goodness of Fit Index; AGFI: Adjusted Goodness of Fit Index; CVI: Contant Validity Index. ity in factor analysis (i.e., the Kaiser criterion of eigenvalues 1.00 and the scree criterion of the greatest change in the trend of eigenvalues plotted against the number of factors) both favored a two-factor solution for exploratory factor analysis but with differing degrees of confidence. For the 2-factor model, the eigenvalue of factor 2 ( 1.034) barely met the Kaiser criterion ( 1) whereas there was a marked change in the scree plot between 1 and 2 factors and a relatively flat scree plot after 2 factors. Using the Kaiser criterion as well as interpreting the factor loadings from the one-factor solution suggested that the one factor solution was also plausible. The one-factor solution indicated a single composite score of the 15 CRF symptoms. This numerical score summarizes the individual s overall CRF level. The scores range from 15 to 75, with higher scores indicating more CRF. The one-factor solution is a noteworthy but somewhat unexpected finding of the study. The few emotional (Items 4, 6, 9, and 10) and cognitive items (Item 14 and 16) compared with the 9 physical items (Item 1, 2, 3, 5, 7, 10, 12, 13, and 15) available for the main study may have favored the retention of a single factor solution or at least the inclusion of the emotional and/or the cognitive items together with the physical items in the first factor. A similar pattern has been seen in two other instrumentation studies. For example, in the study of the CRFDS, Holley 25 generated the item pool for the CRFDS based on information gathered in 23 individual interviews with 17 participants who were experiencing CRF. Five themes included cognitive, physical, psychological, social, and spiritual distress. The CRFDS was content validated and tested psychometrically with patients with cancer. However,

9 Assessing Fatigue in Persons with Cancer/Wu and McSweeney 1693 exploratory factor analysis suggested a one-factor solution with all items on the CRFDS reflecting the single dimension of CRF distress. Another example is the development of the SCFS The original instrument was developed based on the conceptualization of CRF as a self-perceived experience and multidimensional phenomenon that contains four dimensions: physical, emotional, cognitive, and temporal aspects of fatigue. After informal content validation by patients with cancer, the four-factor CRF measurement model was supported by exploratory factor analysis but rejected by confirmatory factor analysis. 32 Instead, the best fit of the data was found to be a two-factor model measuring physical and perceptual fatigue. The physical dimension of the SCFS-6 included two physical items and one emotional item as originally proposed and the perceptual dimension of SCFS-6 included two emotional items and one cognitive item as originally proposed in the instrument development stage. Although the multidimensionality of CRF is evidenced in the literature, the findings from the current study as well as others reveal the single dimensionality of CRF in assessing fatigue in persons with cancer. The similarities of the current study and the other two studies 25,32 are that all three of these studies aimed to develop measures specifically for fatigue in persons with cancer; 2) focused on either the consequences of CRF (CRFDS) or on its attributes (this new CRF instrument and SCFS-6) and, therefore, included only items measuring consequent distress or attributes; and 3) placed heavy emphasis on the persons perspective in instrument development. Some of the subsequent problems with dimensionality of the new instrument may be traceable to the content analysis stage. A pronounced problem occurred in asking the first panel of content experts to identify each item with one of the three fatigue dimensions. Three of the 10 subjects expressed verbally or in written format the difficulty of identifying a single fatigue dimension for each item. The patients comments, such as Many questions can be physical, emotional, and cognitive or one can lead to the other: cognitive, emotional, physical (for example, may be reading, feeling refreshed, lose concentration, become frustrated, then give up and go to sleep), showed difficulty in assigning their fatigue experiences according to a defined fatigue dimension and in distinguishing among the dimensions. The well documented discordance between patients and health care professionals perceptions of cancer-related symptoms may be linked to discordance in the perceived dimensionality of CRF between patients and health care professionals. Health care providers or researchers conceptualize CRF as multidimensional, which may reflect the way people are taught to learn and understand certain phenomena in the scientific world. CRF arises from the complex interplay of physical, mental, emotional, environmental, physiologic, and pathologic factors. The difficulty of patients with cancer with the classification of fatigue attributes into physical, emotional, and cognitive dimensions suggests that their view of the attributes of CRF may be more holistic and different from the way health care professionals conceptualize the symptom. There are several advantages of this new instrument over other fatigue scales used in oncology. First, the relatively unique feature of this instrument is its development from the perspective of patients with cancer. Unlike the majority of currently available fatigue instruments for patients with cancer, not only was the original item pool based on the patients own words, the instrument s content validity and psychometric properties were tested on patients with cancer. Second, the instrument was tested extensively by cognitive interview techniques to discover whether any of the items were unclear or difficult to answer. The cognitive interview is a systematic way of understanding the process of answering questions from the respondents point of view. The cognitive interview process is a unique component of the refinement of this instrument in comparison to other instruments used in oncology populations. Third, this instrument aims to measure the characteristics or attributes of CRF. In doing so, it distinguishes between the attributes of fatigue and its causes or outcomes. Unlike many of the fatigue instruments that include outcome measures (such as interference with activity, functional status, quality of life) and attribute items in the same scale, the fatigue score on this new instrument can be more meaningfully and more easily interpreted. Fourth, this instrument is specifically developed to measure CRF. Finally, the new CRF instrument was well perceived by the participants with none refusing to complete this instrument after informed consent was obtained. A Flesch Kincaid grade level of 6.5 was computed by Microsoft Word, indicating a reasonable literacy level required for this new CRF instrument. 68 This instrument can be completed easily in 3 minutes even by elderly patients with cancer. Not burdening people who are already fatigued is a very important consideration when assessing fatigue in patients with cancer. Despite these strengths, there are some limitations. First, the study was limited by a sample that predominantly consisted of female patients with breast carcinoma who were receiving chemotherapy in an outpatient unit. Patients with more advanced illness and possibly with worse fatigue experiences

10 1694 CANCER October 1, 2004 / Volume 101 / Number 7 were not accessible. Although current development supported the use of this new CRF instrument in the population of patients with breast carcinoma, application to other cancer populations without further validation should be done with caution. The new CRF instrument needs to be validated with other types of cancers, other types of cancer treatments, or with patients with more advanced-stage cancer. Second, the study did not explore the temporal concerns of measuring CRF. Using yesterday as the time frame cannot capture the changes of fatigue during a day. Third, the cutoff -points of mild, moderate, or severe for interpreting fatigue scores have not yet been developed for this new CRF instrument. Additional testing is needed to determine the categorization of numerical scores to facilitate its use in clinical assessment. Finally, like other CRF instrumentation studies, the current study faced the dilemma of selecting a good criterion as the external standard for validity testing. Although the three external standards were chosen based on the empirical evidence, no gold standard measure has been established in current CRF literature. This newly developed instrument is best described as a single dimensional measure of CRF that requires further testing of its properties. The potential use of this new CRF instrument in clinical practice and research settings, after more extensive validation of its properties, can be viewed in three ways. It will allow health care providers to acknowledge and better understand what their patients with cancer are experiencing. It will also allow a holistic assessment of CRF from the person s perspective. Another use is providing quantitative information on the symptom status of fatigue to help health care providers to monitor their patient s condition or treatment progress. The information can improve the communication between health care providers and patients, be incorporated into treatment decisions, and help better manage CRF. REFERENCES 1. Curt GA, Breitbart W, Cella D, et al. Impact of cancer-related fatigue on the lives of patients: new findings from the Fatigue Coalition. Oncologist. 2000;5: Curt G, Johnston PG. Cancer fatigue: the way forward. Oncologist. 2003;8(Suppl. 1): Richardson A. Fatigue in cancer patients: a review of the literature. Eur J Cancer Care (Engl). 1995;4: Vogelzang NJ, Breitbart W, Cella D, et al. Patient, caregiver, and oncologist perceptions of cancer-related fatigue: results of a tripart assessment survey. The Fatigue Coalition. Semin Hematol. 1997;34: Flechtner H, Bottomley A. Fatigue and quality of life: lessons from the real world. Oncologist. 2003;8(Suppl. 1): Langeveld NE, Grootenhuis MA, Voute PA, de Haan RJ, van den Bos C. No excess fatigue in young adult survivors of childhood cancer. Eur J Cancer. 2003;39: Winningham ML. Strategies for managing cancer-related fatigue syndrome: a rehabilitation approach. Cancer. 2001; 92: Mock V, Atkinson A, Barsevick A, et al., Panel. NCCN practice guidelines for cancer-related fatigue (Version ). J Natl Comp Cancer Network. 2003;1: Cella D, Davis K, Breitbart W, Curt G, Fatigue Coalition. Cancer-related fatigue: prevalence of proposed diagnostic criteria in a United States sample of cancer survivors. J Clin Oncol. 2001;19: Portenoy RK, Itri LM. Cancer-related fatigue: guidelines for evaluation and management. Oncologist. 1999;4: Sadler IJ, Jacobsen PB, Booth-Jones M, Belanger H, Weitzner MA, Fields KK. Preliminary evaluation of a clinical syndrome approach to assessing cancer-related fatigue. J Pain Symptom Manage. 2002;23: Bender C, Kramer P, Miaskowski C. New direction in the management of cancer-related cognitive impairment, fatigue, and pain. Pittsburgh: Oncology Nursing Society, Patrick DL, Ferketich SL, Frame PS, et al. National Institutes of Health state-of-the-science conference statement: symptom management in cancer: pain, depression, and fatigue, July 15-17, J Natl Cancer Inst. 2003;95: Wu HS, McSweeney M. Measurement of fatigue in people with cancer. Oncol Nurs Forum. 2001;28: Wu HS, McSweeney M. The assessment and measurement of fatigue in people with cancer. In: Armes J, Krishnasamy M, Higginson I, editors. Fatigue in cancer. Oxford, UK: Oxford University Press. 2004: Varricchio C. Measurement and assessment: what are the issues? In: Winningham ML, Barton-Burke M, editors. Fatigue in cancer: a multidimensional approach. Sudbury, MA: Jones & Bartlett Publishers, 2000: Messias DKH, Yeager KA, Dibble SL, Dodd MJ. Patients perspectives of fatigue while undergoing chemotherapy. Oncol Nurs Forum. 1997;24: Holley S. Cancer-related fatigue. Suffering a different fatigue. Cancer Pract. 2000;8: Glaus A, Crow R, Hammond S. A qualitative study to explore the concept of fatigue/tiredness in cancer patients and in healthy individuals. Eur J Cancer Care (Engl). 1996;5: Ream E, Richardson A. Fatigue in patients with cancer and chronic obstruction airways disease: a phenomenological enquiry. Int J Nurs Stud. 1997;34: Knowles G, Borthwick D, McNamara S, Miller M, Leggot L. Survey of nurses assessment of cancer-related fatigue. Eur J Cancer Care (Engl). 2000;9: Newell S, Sanson-Fisher RW, Girgis A, Bonaventura A. How well do medical oncologists perceptions reflect their patients reported physical and psychosocial problems? Data from a survey of five oncologists. Cancer. 1998;83: Schwartz AL. The Schwartz Cancer Fatigue Scale: testing reliability and validity. Oncol Nurs Forum. 1998;25: Velikova G, Wright P, Smith AB, et al. Self-reported quality of life of individual cancer patients: concordance of results with disease course and medical records. J Clin Oncol. 2001; 19: Holley SK. Evaluating patient distress from cancer-related fatigue: an instrument development study. Oncol Nurs Forum. 2000;27: