Development of the PROMIS-based Research Assessment and Clinical Tool-Fatigue (ReACT-F)

Purpose Evidence has shown that cancer-related fatigue (CRF) may be a treatment-limiting symptom and often impairs health-related quality of life. Accurate assessment of the multidimensional nature of CRF could help drive interventions to mitigate this debilitating symptom. Currently, there are no clinical tools to effectively and efficiently assess the multidimensionality of CRF. The purpose of this paper is to introduce a CRF-specific short form that can assess the multidimensional nature of CRF for use in the clinical setting. Methods The CRF-specific short form was developed using the 95-item PROMIS® fatigue bank. Bi-factor analysis was used to evaluate dimensionality of the alternative model using fatigue for the general factor and physical, cognitive, affective, global, and motivational for the local factors. After unidimensionality was confirmed (loading factor > 0.3), one item from each local factor was selected using discrimination power for inclusion in the CRF-specific short form. Results The Research Assessment and Clinical Tool-Fatigue (ReACT-F) was created from the 95-item PROMIS fatigue bank using established item parameters. The ReACT-F assesses five common dimensions of CRF as well as perceived burden of the fatigue dimensions. Conclusions The ReACT-F is a CRF-specific self-report short form that addresses the need for a brief, clinically useful tool to quickly assess the multidimensional nature of CRF. We anticipate that the ReACT-F can be completed in the clinical setting in approximately 3 minutes, providing clinicians with meaningful data to drive personalized interventions. Further validation of the ReACT-F is highly encouraged.


Introduction
Cancer-related fatigue (CRF) is a highly prevalent, complex, multidimensional symptom that can greatly impair the healthrelated quality of life of cancer patients [1,2]. Clinical guidelines have adopted a single-item, 0 to 10 numeric rating scale to initially screen for CRF due to its easily administered nature [3][4][5][6]. Several guidelines recommend a more comprehensive evaluation when patients rate their fatigue ≥ 4 (i.e., moderate fatigue) using a 0-10 numerical rating scale. This evaluation includes a focused history, assessment of treatable contributing factors (anemia, nutrition deficits, pathologic/physiologic abnormalities, etc.), and concurrent symptoms (pain, depression, sleep disturbance, etc.) and conditions (cardiac, renal, pulmonary, etc.) [3][4][5][6]. Moreover, inclusion of a measure of the multidimensional nature of CRF would be advantageous to understand the full fatigue experience of cancer patients.
The fatigue experienced by cancer patients is often reported to be multidimensional in nature including physical, emotional, and cognitive dimensions, although the exact terminology for each dimension can vary (i.e., affective, motivational, behavioral, functional, etc.) [7,8]. Therefore, when conducting an in-depth evaluation of CRF, clinicians should consider the multidimensional nature of CRF to fully capture the CRF experience and optimize management. Consistent with the Precision Medicine Initiative of the National Institutes of Health (NIH), understanding the specific dimension of CRF that most affects the patient can help guide the clinician to develop a more tailored and personalized management strategy.
Though comprehensive multidimensional fatigue assessments are available (e.g., revised Piper Fatigue Scale, Multidimensional Fatigue Inventory, and the Fatigue Questionnaire), most of them were developed using classical test theory resulting in measures that may not be best suited for a clinical environment [9]. Applications from the PROMIS® (Patient-Reported Outcomes Measurement Information System, http:// www.healthmeasures.net) fatigue item bank are expected to overcome these limitations [10]. As part of the NIH's roadmap project, the PROMIS® was developed to offer a set of person-centered measures to evaluate symptoms of individuals with or without chronic conditions [11]. One of these measures is the PROMIS instrument that assesses fatigue and the impact of fatigue on daily living [10]. The PROMIS fatigue bank consists of 95 items generated from a comprehensive literature review, focus groups, and individual interviews which were then calibrated using item response theory (IRT) models [10,12], allowing for briefyet-precise fatigue estimation via tailored, individualized computer adaptive test (CAT), or short forms with fixed numbers of items. For the latter, multiple short forms can be created to meet users' needs and scores from these short forms are comparable as long as scores are generated using item parameters established in the original calibrated item banks. Yet precision levels may vary as demonstrated in Lai et al. (2011) in which three short forms were developed targeting patients with mild fatigue, severe fatigue, and for fatigue across the whole severity continuum. Several short forms derived from the PROMIS fatigue item bank are available [13][14][15], yet none of them target fatigue content areas that are important to cancer patients. Therefore, to fill this void, a content-specific CRF short-form was developed that can be used in the clinical setting.

Methods
Fatigue dimensions To determine the fatigue dimensions of interest, current multidimensional fatigue assessments were reviewed (Table 1) [16][17][18][19][20][21]. The most commonly assessed fatigue dimension was physical (20/20) followed by cognitive (16/20), affective (7/20), global (6/20), and motivational (5/20). Thus, these five dimensions were selected moving forward. The physical dimension of CRF was conceptualized as fatigue related to energy level. The cognitive dimension of CRF was conceptualized as fatigue related to thought processes, memory, and executive function. The affective dimension of CRF was conceptualized as fatigue related to emotions or feelings. The global dimension of CRF was conceptualized as encompassing the subjective experience of fatigue. Lastly, the motivational dimension of CRF was conceptualized as fatigue related to actions that maintain a meaningful or purposeful existence.
Assigning PROMIS fatigue items All 95 items in the PROMIS fatigue item bank were reviewed by the primary author (KD), who then assigned them to one of the five dimensions. This classification was then reviewed by the second author (DLK) for consensus. If there was disagreement with any classification, a third reviewer (LS) was included to achieve consensus.

Statistical analysis
The current PROMIS fatigue item bank was modeled to have one general fatigue factor with two local factors (experiences and impacts), which was psychometrically proven to be sufficiently unidimensional [10]. For this paper, in order to develop a content-specific CRF short form that produces scores comparable to the PROMIS fatigue item bank and its short forms, we first evaluated the sufficient dimensionality of the alternative model as discussed above using bi-factor analysis.
Bi-factor analysis includes two classes of factors: a general factor, defined by loadings from all of the items in the scale, and local factors, defined by loadings from pre-specified groups of items related to that sub-domain [22][23][24][25]. Items are considered sufficiently unidimensional when standardized loadings are > 0.3 for all the items on the general factor. Similarly, if the loadings of all the items on a local factor are salient, this would indicate that the local factor is well defined even in the presence of the general factor, and it is more appropriate to report scores of local factors separately [22,24,26].
In the model used by the current study, the general factor was Bfatigue^and the 5 local factors were physical, cognitive, affective, global, and motivational. Once sufficient unidimensionality was supported, we then created a content-specific CRF short form by selecting items from each local factor by reviewing item content, as well as using item parameter threshold values obtained from item response theory (IRT) estimation, particularly the discrimination parameter.
Discrimination power describes the strength of an item's discrimination between people at different fatigue levels below and above the threshold, indicating the degree of association between item responses and the fatigue latent trait. Items with the highest discrimination parameters typically produce the highest information function (i.e., lowest measurement errors) were considered the best candidates to be included in the short form.

Results
The 95 items from the PROMIS fatigue bank were organized into the five fatigue dimensions (physical, cognitive, affective, global, and motivational) as listed in Table 2. After consensus was achieved, there were 12 items from the PROMIS fatigue bank that fit into the physical dimension, 13 in the cognitive dimension, 4 in the affective dimension, 32 for the global dimension, and 34 for the motivational dimension. Essential dimensionality of these items was supported with acceptable fit indices: RMSEA = 0.04, CFI = 0.985, TLI = 0.985. All items showed higher loading to the general factor than to their  Table 3 for a summary of the PROMIS item selection information.
The PROMIS-based Research Assessment and Clinical Tool-Fatigue (ReACT-F) CRF-specific short form was created using established item parameters (Fig. 1). Each item was selected based upon the discriminative value and is considered representative to each fatigue dimension. We added a numeric rating scale at the beginning of the questionnaire per the current fatigue assessment guidelines and we added an additional item, BWhich aspect of fatigue is most bothersome to you^to assess the overall burden of the CRF dimensions and inform treatment decisions to optimize CRF management.

Scoring
This short form can be scored using similar approaches as used by other PROMIS fatigue short forms [27]. A 5-item questionnaire can only be scored when at least 4 of the items are completed. However, patients should be encouraged to complete all items to minimize measurement errors. Each question has a Likert scale with values ranging from one to five. A total raw score is calculated by summing the five items on the questionnaire and a prorated value will be used to replace missing value; therefore, the total score on the instrument ranges from 5 to 25. Higher scores indicate worse fatigue. A raw score can then be translated into a PROMIS based T-score to create the final score for a respondent (details are shown in www.healthmeasures.net/score-and-interpret/ calculate-scores).

Discussion
The purpose of this paper was to develop a brief tool to capture the multidimensional nature of CRF. This was carried out by examining items included in the 95-item PROMIS fatigue bank to determine if specific items could be selected to measure different fatigue dimensions. The final CRF assessment tool, the ReACT-F, consists of five PROMIS items, where each item is expected to screen a specific dimension of CRF: physical, cognitive, affective, global, and motivational.
The ReACT-F is a self-report short form that addresses the previously identified gap in the literature, which is the lack of a brief, clinically useful tool to quickly assess the multidimensional nature of fatigue in the cancer population. This new content-specific short form requires further validation to determine its clinical and scientific relevance. In the clinic, it is expected that the ReACT-F can aid clinicians to quickly assess the specific fatigue experience of their patients to allow for a more focused evaluation and tailored management. For example, patients reporting physical fatigue may be further evaluated for deconditioning, cardiopulmonary status, or musculoskeletal impairment, so physical rehabilitative strategies can be planned. Individuals who report affective fatigue may be referred for comprehensive psychological evaluation, while those who report affective fatigue or cognitive fatigue may benefit from occupational psychotherapy for behavioral adaptive coaching and a neuropsychology consult for comprehensive cognitive function evaluation, respectively.
Scientifically, this evaluation tool will be useful to determine the phenotypic characteristics of each fatigue dimension     To obtain permission to use, please contact the corresponding author © K. Dickinson, D. Lynch Kelly, J-S. Lai, L. Saligan within the global fatigue construct. The ReACT-F tool can assist in identifying clinical and demographic attributes, as well as the biologic profile of the specific fatigue experience, to advance our understanding of the etiology of CRF. Understanding the etiology of CRF is important for treatment development and generation of algorithms to identify individuals at risk to develop clinically meaningful fatigue related to the progression of their disease or as a side effect of their treatment.

Limitations
The five items have high discriminative value demonstrating the ability to allow for the assessment of multidimensions of fatigue; however, they did not have sufficient power when factor loading to be independent from the construct of fatigue, as assessed through comparison of factor loadings between the general factor (fatigue) and the local (subdomain) factors. Thus, the dimensions are not independent constructs, but components of a general fatigue construct.

Conclusions
In conclusion, the ReACT-F is a CRF-specific self-report short form that addresses the need for a brief, clinically useful tool to quickly assess the multidimensional nature of CRF. The ReACT-F assesses five common dimensions of CRF as well as perceived burden of the fatigue dimensions. This tool has clinical and scientific promise, to advance our understanding and management of CRF. We anticipate that the ReACT-F can be completed in the clinical setting in approximately 3 minutes, providing clinicians with meaningful data to drive personalized interventions. Further validation of the ReACT-F is highly encouraged to assess its psychometric properties and determine its clinical utility.