Functional assessment of chronic illness therapy—the fatigue scale exhibits stronger associations with clinical parameters in chronic dialysis patients compared to other fatigue-assessing instruments

Ethical considerations

The current study was approved by the institutional review board of National Taiwan University Hospital (NTUH; NO. 201403006RINB), and all participants (or their medical proxies if they were unable to communicate verbally) provided verbal informed consent.

Collection of baseline clinical parameters

Patients with ESRD receiving chronic hemodialysis at the NTUH Jinshan branch were prospectively recruited between 2014 and 2015 (Chao et al., 2015; Chao & Huang, 2015). Those who refused to provide informed consent were excluded from this study. We documented their baseline clinical data on enrollment, including demographic profile (age, sex, education status, and other socioeconomic factors), body mass index (BMI), dialysis duration, and ESRD origin. Comorbidity severity was recorded using Liu’s dialysis comorbidity index (Liu et al., 2009). Blood samples were obtained after enrollment and sent to the reference laboratory of NTUH for analysis of hemoglobin, nutritional and biochemical profiles (albumin, cholesterol, triglyceride, uric acid, pre-dialysis blood urea nitrogen [BUN], creatinine, and electrolyte panels), dialysis clearance, and hormones (ferritin, intact parathyroid hormone) (Chao et al., 2012). Assuming a power of 0.8 and a type 1 error of 5%, at least 32 patients will be required to detect significant differences in serum albumin levels among ESRD patients with and without fatigue.

Assessment of fatigue using different self-report instruments

Trained research assistants administered all of the fatigue assessment tools during the patients’ dialysis sessions nearest to their time of blood collection. We administered these instruments in a random order. A total of seven instruments were used in this study, including the Brief Fatigue Inventory (BFI), Functional Assessment of Chronic Illness Therapy–Fatigue (FACIT-F), Fatigue Severity Scale (FSS), Lee Fatigue Scale (LFS), Fatigue Questionnaire (FQ), Fatigue Symptom Inventory (FSI), and SF36-V. Six out of the seven tools (BFI, FACIT-F, FSS, LFS, FSI, and SF36-V) have been translated into Chinese versions in the literature with established consistency and reliability (Wang et al., 2004; Shun et al., 2006; Tsai et al., 2014; Wang et al., 2016). For those without a Chinese version, at least two staff members performed a forward and backward translation, and any discordant aspects were resolved by group consensus.

Fatigue assessment tools can be stratified according to the dimensions they intend to cover (uni- or multidimensional) (Minton & Stone, 2009). Here we evaluated the utility of both uni- (BFI, FACIT-F, FSS) and multidimensional (LFS, FQ, FSI, SF36-V) instruments for chronic dialysis patients. The BFI, a nine-item questionnaire that was originally developed to evaluate fatigue severity in cancer patients, features the advantages of brevity and avoidance of English-based idioms that tend to confuse Chinese recipients (Mendoza et al., 1999). Scores can be calculated as the average of all the items assessed (range, 0–10). The FACIT-F, part of a comprehensive assessment tool that evaluates HRQoL and the symptomatology of patients with chronic illnesses or cancer, consists of 13 items graded on a four-point Likert scale (http://www.facit.org). The FACIT-F has been used in chronic dialysis patients with good validity (content validity index 0.67–1; good construct validity from an excellent correlation with revised Piper Fatigue Scale results) (Wang et al., 2015). The FSS, a nine-item test graded using a seven-point Likert scale, was initially tested in patients with neurological disorders but has since been expanded for use in patients with other diseases and aims to measure fatigue severity and its influence on one’s daily activity and lifestyle (Krupp et al., 1989). The original LFS consists of 18 items graded on a 10-point Likert scale and is used to assess patient fatigue or energy with different descriptive adjectives (Lee, Hicks & Nino-Murcia, 1991). We previously used a seven-item short-form LFS to increase administration ease along with similarly fair psychometric properties (Tsai et al., 2014). LFS scores are calculated similarly to BFI scores. FQ, or the original Chalder fatigue scale, evaluates two dimensions of patient fatigue, physical and mental, and features the advantage of cultural sensitivity (Chalder et al., 1993). The FSI, which also uses a 10-point Likert scale, measures fatigue intensity at different timings (severity dimension), how fatigue affects patients’ functional statuses (interference dimension), and the persistence of fatigue over time (duration dimension) (Hann et al., 1998). The SF36-V is an integral component of the SF36, which systematically surveys the quality of life of patients with diverse disease forms. The SF36-V is now widely used to gauge fatigue severity and energy levels of ill patients, including those with ESRD (Feroze et al., 2011).

Frailty assessment using an established instrument

Frailty was assessed in these ESRD patients using the Simple FRAIL Scale (SFS). The SFS measures five domains of frailty: fatigue, resistance, ambulation, illnesses, and loss of body weight. The SFS has been applied in ESRD patients with fair validity, the results of which have shown moderate associations with dialysis-related complications (Chao et al., 2014a; Chao et al., 2014b).

Statistical analysis

In this study, continuous and categorical variables were described using mean ± standard deviation and case number with percentage, with group comparisons made using the independent t-test or the Mann–Whitney U-test and chi-square test, respectively. Since diabetes mellitus (DM) and heart failure (HF) might have an influence on fatigue, we also compared the severity of fatigue between ESRD patients with and without DM or HF. Results from different fatigue questionnaires were tested with respect to their association with the clinical parameters (demographic and socioeconomic features), dialysis-related complications (nutritional profiles, azotemia, electrolyte panels, and metabolic results), and frailty severity using Pearson’s correlation coefficients. We then performed multiple regression analyses, incorporating relevant clinical features with dependent variables including serum albumin levels and frailty severity, both important outcome-related factors of ESRD patients. In all analyses, a two-sided p < 0.05 signified statistical significance as determined by SPSS 18.0 software (SPSS Inc., Chicago, IL, USA).

Features of the enrolled chronic dialysis patients

A total of 46 ESRD patients were recruited and participated in the study, higher than the calculated required sample size. Their average age was 67 ± 11.6 years, comparatively higher than that of an average dialysis population (Chao et al., 2014a; Chao et al., 2014b), although the male/female distribution was even (Table 1). Most patients were married, and one-third of our total cohort had no educational history, while another one-third went to elementary school only. The mean patient BMI was 23.2 ± 3.1 kg/m². Diabetes nephropathy was the cause of the ESRD in half of our patients. The dialysis comorbidity index of our patient cohort was 1.54 ± 1.76.

The serum biochemical profiles of these patients indicated that they had fair nutrition status (albumin, 3.9 ± 0.4 g/dL; total cholesterol, 160 ± 45 mg/dL), mild anemia (hemoglobin 10.2 ± 3.6 mg/dL), fair dialytic clearance, and nearly normal electrolyte panels (Table 1).

Among our cohort, half were found to have DM. No significant differences were found between ESRD patients with and without DM regarding demographic profiles, body mass index, and comorbidities including heart failure. Similarly, there were no significant differences regarding their nutritional profiles, azotemic parameters, electrolyte panels, and metabolic profiles) (Table 1). However, ESRD patients with DM had a significantly higher dialysis comorbidity index (p = 0.01) and were more likely to have ESRD originating from diabetic nephropathy (p = 0.001), both of which were expectable judging from the influence of DM. On the other hand, no significant differences were found between ESRD patients with and without HF regarding demographic profiles, body mass index, and all the laboratory data tested.

Fatigue severity among chronic dialysis patients

In our study, FSI and FQ were translated by our staffs, while other questionnaires were translated previously. All patients answered the seven self-report questionnaires, and the fatigue assessment results are shown in Table 2. Using the previously reported diagnostic threshold for each questionnaire (for FACIT-F, score < 30; for FSI, score > 3; for FSS, score > 2.3; for FQ, score > 15; for SF36-V, score < 60; for LFS and BFI, score > 0), we found that the BFI identified fatigue in 65.2% of the patients, while the LFS identified fatigue in 56.6% of the patients. However, the FQ, FSS, and SF36-V identified fatigue in 21.7% of the patients each, with average scores of 14.6 ± 4.1, 1.8 ± 1.2, and 77 ± 15, respectively. On the other hand, the FACIT-F identified fatigue in 15.2% of the patients, with a mean score of 41.1 ± 9.9. According to the dimensions each questionnaire cover, no significant differences were observed between physical and mental components (using the FQ), while energy subscale was more likely to be affected than fatigue subscale (using the SF36-V).

In addition, the presence of DM did not have significant influences on these patients’ fatigue severity, assessed by LFS (p = 0.61), BFI (p = 0.84), FQ (p = 0.89), FACIT-F (p = 0.56), FSI (p = 0.83), FSS (p = 0.88), and SF36-V (p = 0.67). Similarly, the presence of HF did not significantly affect fatigue severity, assessed by LFS (p = 0.61), BFI (p = 0.56), FQ (p = 0.49), FACIT-F (p = 0.57), FSI (p = 0.41), FSS (p = 0.34), and SF36-V (p = 0.37).

Correlation between fatigue severity, clinical features, and frailty severity among chronic dialysis patients

A correlation analysis was performed of the assessment results and the clinical and laboratory parameters. Results from all seven fatigue-assessment tools showed high correlations (Table 3). We further found that the FACIT-F scores exhibited significant associations with age (r = − 0.33, p = 0.03) as well as serum albumin (r = 0.43, p = 0.003), BUN (r = 0.31, p = 0.04), creatinine (r = 0.35, p = 0.02), and serum potassium (r = 0.32, p = 0.03) levels among all ESRD patients. SF36-V scores were also significantly associated with age (r = − 0.34, p = 0.02) as well as BUN (r = 0.29, p = 0.05), creatinine (r = 0.3, p = 0.04), and serum potassium (r = 0.3, p = 0.04) levels but borderline associated with serum albumin (r = 0.28, p = 0.07) level among all ESRD patients. On the contrary, fatigue severity measured by the FSS (r = 0.31, p = 0.04), FSI (r = 0.31, p = 0.04), and BFI (r = 0.36, p = 0.02) was significantly associated with dialysis duration but not biochemical traits. In addition, the FACIT-F, SF36-V, and LFI scores correlated significantly with frailty severity (FACIT-F vs. SFS, r = − 0.52, p = 0.0002; SF36-V vs. SFS, r = − 0.35, p = 0.02; LFI vs. SFS, r = 0.32, p = 0.03), while BFI (r = 0.24, p = 0.12), FQ (r = 0.2, p = 0.19), FSI (r = 0.25, p = 0.09), or FSS (r = 0.24, p = 0.11) did not.

Regression analysis targeting serum albumin levels incorporating fatigue severity

Since serum albumin is an important prognosis determinant for chronic dialysis patients, we next performed a multiple regression analysis with serum albumin as the dependent variable among all ESRD patients (Table 4). Accounting for age, sex, dialysis duration, BMI, and comorbidity index, lower FACIT-F scores and SF36-V scores were both independently associated with lower serum albumin (for the former, p = 0.002; for the later, p = 0.03). However, only the FACIT-F score retained its independent association with serum albumin (p = 0.006). Such an association was not discovered for the BFI (p = 0.13), FQ (p = 0.99), LFI (p = 0.14), FSS (p = 0.29), and FSI (p = 0.12) scores. Similarly, only the FACIT-F scores (p = 0.006) exhibited a significant association with the SFS scores in these patients, while the other scale scores did not.

We further used the sub-cohort with DM for the regression analysis, incorporating the original model components (age, sex, dialysis duration, and BMI). We found that FACIT-F scores still exhibited significant association with serum albumin levels among ESRD patients with DM (t = 2.38, p = 0.03). However, SF36 scores did not have significant association with serum albumin levels among ESRD patients with DM (t = 1.98, p = 0.06). If SF36 and FACIT-F scores were both included in the analysis, only FACIT-F scores retained a significant relationship with serum albumin levels (t = 2.38, p = 0.03). On the other hand, if we used the sub-cohort without DM for regression analysis, no significant association was found between FACIT-F score and serum albumin, or between SF36 scores and serum albumin (for FACIT-F, t = 1.88, p = 0.08; for SF36, t = 0.91, p = 0.38). Using the non-HF sub-cohort, we found that both FACIT-F (t = 3.22, p = 0.003) and SF36 (t = 2.35, p = 0.02) scores exhibited significant associations with serum albumin levels. However, for the HF sub-cohort, the number was too small to permit regression analysis. These findings were largely compatible with results using the entire ESRD cohort.