Title: Fatigue is the predominant patient-reported outcome measure in hemodialysis patients: Results of a multicenter cross-sectional ePROMs Study

Abdallah. Guerraoui (  abdallah.guerraoui@calydial.org ) calydial Mathilde. Prezelin-Reydit AURAD-Aquitaine Anne. Kolko Association pour l'Utilisation du Rein Arti ciel en région Parisienne (AURA) Paris Marie. Lino-Daniel calydial Charlotte. Dumas Roque AURAD-Aquitaine Pablo. Urena Association pour l'Utilisation du Rein Arti ciel en région Parisienne (AURA) Paris Philippe. Chauveau AURAD-Aquitaine Catherine. Lasseur AURAD-Aquitaine Julie. Haesebaert Université Lyon, Université Claude Bernard Lyon 1, RESHAPE INSERM U1290 Agnes. Caillette-Beaudoin calydial

The collection of electronic patient-reported outcome measures (ePROMs) is an innovative method to better take these symptoms into consideration [8]. Moreover, ePROMs allow patients to express their symptoms on a regular basis and allow the medical team to adapt treatment plans accordingly [8].
However, the use of ePROMs for CKD patients treated with HD in routine care remains limited [9,10]. The objective of the study was to describe the prevalence of symptoms reported by patients treated with hemodialysis using ePROMs. The secondary objective was to explore predictive factors of the presence of the patient reported symptoms.

Study design
A multicenter cross-sectional observational study was conducted in three hemodialysis centers in France (Vienne, Bordeaux and Paris) between January and March 2020. Patients were included if they were aged at least 18 years old and treated with hemodialysis for at least 3 months in one of the participating centers. Patients were excluded if they refused to participate in the study, or could not read or understand French well enough to complete the questionnaires. The study was declared to the data protection authority in France, known as the Commission Nationale de l'Informatique et des Libertés (CNIL), and was approved by the CPP Ile-de-France VII on 26 December 2019 in accordance with French regulations. Choice of symptoms to collect We chose to report the symptoms which were the most frequently reported in a preliminary qualitative study. This qualitative study was conducted to identify relevant symptoms reported by patients and affecting quality of life between dialysis sessions 11 . A total of 20 patients were interviewed by a nephrologist trained to conduct semi-structured interviews. Patients were encouraged to provide examples and expand their answers to collect further details. Interviews were then transcribed and coded with thematic analysis to identify theme and subthemes from the data. Interviews were ceased when no new codes were identi ed and data saturation was reached. We explored inter-dialysis symptom and we selected the three most frequent symptoms that were fatigue, mental and sleep disorder (Table 1). Data were collected once by the patient, or with the help of a caregiver if necessary, with a selfadministered ePROMs through an electronic tablet during a HD session or consultation. The collection of data through a tablet was chosen due to the ease of use and simple interface for the patients. Data collected included the presence of post-dialysis fatigue with a binary question (yes/no), its perceived intensity with a visual analog scale ranged from 0 to 10 (0 was no perceived intensity and 10 was the most possible perceived intensity) [12], and the recovery time of this symptom after a session as a Likert scale question [13] (Table 2). Perceived stress scale (PSS 10) was questioned using a scale adapted from Cohen and Williamson [14,15]. Sleep quality the day before the dialysis session was questioned in the form of a Likert scale format, the current state of health of the patient using a visual analog scale [16], and the one-year change using a Likert scale were also collected. Patients characteristics were described from the patient's medical records (demographics, dialysis situation, BMI, comorbidities, hemoglobin).

Statistical analysis
We included all patient with eligibility criteria during the study period, our aim was to include all patient during a 3 months period to be representative of patients.
We described continuous variables using medians and interquartile ranges. Numbers and percentages were used for qualitative variables. We rst described the prevalence of each ePROM was estimated with 95% con dence interval. We compared patient's characteristics and ePROMS between the 3 participating centers. Association between ePROMs (recovery time of fatigue (≤ 6-hour versus > 6-hour), PSS stress level (A versus BC), sleep quality (disrupted versus not disrupted) and patients' characteristics and clinical variables was explored using chi square comparisons or Student's t and Wilcoxon tests according to the nature and distribution of the variables.
Multivariate logistic regression models were conducted to identify if patients characteristics or clinical variables were associated with recovery time of fatigue (model 1), perceived stress using PSS (model 2) and sleep quality (model 3). Variables included in the model were chosen based on available literature, expert discussion and results of the bivariate analysis. A multivariate regression model was built for each e-PROMS, with the ePROMs as dependent variable, and clinical characteristics (age, gender, dialysis duration, cardiovascular history), denutrition, hemoglobin, duration of haemodialysis session and number of sessions per-week, and the two other e-PROMs as independent variables. Denutrition was de ned by at least two of three following criteria: Serum Albumin < 35 g/l, Serum Prealbumin < 300 mg/l, nPCR < 1.2 g/Kg/d. Cardiovascular history was de ned by at least one of the following: diabetes, Coronary artery disease, Heart failure, Stroke. A bilateral threshold of 5% was considered to de ne the statistical signi cance. The analysis was performed with SAS 9.1 software.

Results
In total, 173 patients were included during the study period. The mean age of patient was 66.2 ± 14.4 years and majority of patients were males (67.6%). Furthermore, patients were treated with hemodialysis for a mean total duration of 48.9 ± 58.02 months (median 31 months) and were mainly treated in selfdialysis unit (67%). The mean length of dialysis per week was 11.46 ± 1.41 corresponding to 3.09 ± 0.56 sessions per week (shown in Table 3).

Discussion
In this study, we identi ed a high prevalence of self-reported fatigue at 72.1% and important stress at 39.3% for CKD patients treated with hemodialysis. To our knowledge, this study was the rst of its kind conducted in France with questionnaires lled at the healthcare facility via a tablet on describing the use of ePROMs.
Fatigue was the most prevalent symptom identi ed in our study, in comparison to the other symptoms assessed. The prevalence of 72.1% of patients was consistent with the range of previous published literature, which presented a gure from 60-97%. These results were additionally similar to the weighted mean prevalence of 71% estimated in a systematic review [17,18]. These results did not differ between centers even though patient characteristics and comorbidities differed.
Items collected from a patient were with a simple binary question and visual analog scale instead of a dedicated measure and thus may not have re ected the speci city of fatigue from patients under hemodialysis. Such speci c questionnaire was not available at the time of protocol de nition and therefore a generic questionnaire was used. In future studies, items may be collected via the recently published measure SONG-HD questionnaire speci cally designed for patients treated with hemodialysis [19]. This innovative tool designed through an international study included several components of fatigue including tiredness, lack of energy and inability to participate in social situations [20]. This tool, however, did not distinguish between interdialytic fatigue and post-dialysis fatigue [21].
In our study, the post-dialysis fatigue through the after-session delay recovery time in hours was chosen to be assessed as expressed by the patient. The recovery time inferior to 6 hours found in the study for 75.1% of patients were similar compared to an international study where 73% of patients declared the same timing, as well as a recovery time of more than 12 hours declared by 11.6% of patients from our study compared to 10% in the international DOPPS study [22]. Increased fatigue and higher levels of perceived stress were associated in the multivariate analysis indicating the potential interrelation between these two symptoms.
The stress assessment through the perceived stress score seemed to be more informative than the visual analog scale (VAS) to assess its intensity as it allowed to better discriminate patient groups with different stress levels. The two scales are different as the VAS describe the stress at the time of the questionnaire while the PSS includes the stress felt in the past week. Additionally, the analog scale may re ect the stress level more at the time of the questionnaire while the PSS ndings may re ect the stress tendency over the past weeks more and thus may be a better estimation of the patient stress level at home.
On the other hand, results on sleep quality were different to those reported in the literature. While we found only 14.5% of patients with reported altered sleep, sleep disturbance was reported at weighted mean prevalence of 44% with a range of 20-83% [12,23,24] in various countries. The observed difference may be due to the questionnaire used to assess the sleep quality in our study. The questionnaire in our study focused on the sleep quality the night before the dialysis session in comparison to the previous nights, to assess the potential impact of pre-dialysis anxiety on sleep quality. Additionally, these results may be in favor of a limited impact of the dialysis on sleep quality. The difference in results compared with literature is thus explained as those results concerning the disturbance in overall sleep quality for patients treated by hemodialysis in comparison to their situation before the dialysis initiation [25].
The main strengths of this study relied on the cross-sectional and multicenter design from various HD center settings and the use of simple questionnaires to collect data from patients through a tablet directly at patient side. The collection of data from patients during a consultation or HD session through a tablet device also allowed to not have missing data that could have weaken the interpretation of the results. On the other hand, the answers provided by patients may as well have been in uenced by the settings in which they were to reply.
Despite signi cant differences in patient characteristics from the three centers including age, comorbidities or type of dialysis, no differences were found on the prevalence of the various PROMs, in favor of internally coherent results. The study population was not matching with the population pro le of the French Renal Epidemiology and Information Network (REIN) [26, 27,26] and consequently, in terms of comorbidities, coronary artery disease, congestive heart failure and cancer comorbidities the prevalence were higher in our study population. Contrarily, prevalence of diabetes and cerebrovascular disease were lower compared to REIN. This difference may lead to a selection bias and therefore our results may not be applicable to all hemodialysis patients in France. The 3 centers in our study have different characteristics, and no statistical differences were observed between centres for the three e-PROMS.
The main limitations of this study included the observational design, limited number of patients included, and the absence of linkage of PROMs with clinical outcomes such as cardiovascular events, hospitalizations or mortality. However, the objective was to describe the symptoms of patients and not to explain these symptoms with their clinical situation or outcomes.
Nonetheless, meta-analysis of oncology trials identi ed baseline fatigue as an independent prognostic factor for overall survival above performance status and quality of life in oncology patients, recommending collecting this information in routine oncology care for patient strati cation [29]. Due to the clinical impact of fatigue on daily QoL of patients undergoing hemodialysis, it may be however relevant to consider the presence of reported fatigue in such patients to be a clinically relevant item to consider as itself, despite the need for further research in this area [30]. Additionally, recent studies identi ed an association between fatigue and all-cause mortality in those patients as well as between frailty and worse health related QoL [22,31,32].
To improve daily routine care of CKD patients treated with HD, the collection and integration of ePROMs into the care plan could be promoted in a standardized approach. The clinical use of PROMs has been most extensively studied in the oncology literature, in which randomized trials have reported improved outcomes in several areas, including patient-provider communication, quality of life, care satisfaction, and even survival. However, to derive such bene ts from the routine use of PROMs, they must be incorporated into clinical care. A renal ePROM system can play a supportive role in the routine clinical management of ESRD patients and improve the patient centred care.
Electronic PROMs records as well as various other electronic methods of communication between the clinician and patient may serve to accelerate the trajectory toward patient-centered care using patient- All methods were carried out in accordance with the relevant guidelines and regulations.
Informed consent was obtained from all subjects, no subjects were under 18 years of age Consent for publication: Not applicable Availability of data and materials: The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.