We conducted a prospective, cross-sectional study with outpatients in the HD programme at the Fresenius Medical Care clinics in the province of Cordoba (San Rafael, Pintor García Guijo, Cabra and Palma del Río), Spain. From the potential sample from the four clinics (n=316), we selected patients (n=186) who met the inclusion criteria: 1) ability to understand and speak Spanish; 2) aged 18–90; 3) have provided a signed informed consent form for the study; 4) more than six months on HD; and 5) not suffer from cognitive impairment or a serious mental disorder. A total of 130 patients were therefore excluded, which resulted in a 58% response rate to the study. After that, we requested authorisation from the Hospital Universitario Reina Sofía Independent Ethics Committee (IEC), which we received in June 2017.

The following instruments were applied:

• •

  COOP-WONCA quality of life charts using the Arenas et al version.26 This is a general questionnaire for assessing health-related quality of life consisting of nine items that rate the following dimensions: 1) physical fitness; 2) feelings; 3) daily activities; 4) social activities; 5) changes in health; 6) overall health; 7) pain; 8) social support; and 9) general quality of life. Each of the dimensions refers to what happened in the last two weeks and is answered with a Likert scale that scores from 1 to 5, with the highest scores reflecting the worst perceived health. In the end, an overall score is obtained corresponding to the sum of all dimensions, except for item 5 which, having a bipolar structure, has a different reading from the rest of the dimensions.27 For the present study it showed a good Cronbach's Alpha (0.77).

• •

  Acceptance and Action Questionnaire-ii for patients on haemodialysis (AAQHD-II) This instrument (Appendix Table) assesses the level of PI in the context of certain thoughts (items 2 and 3), fears (item 6), treatment regimens (items 4 and 5), situations related to the HD therapy (item 1) and how the treatment regimens affect the patient's life (item 7). It is a seven-item questionnaire with a five-point Likert scale (1= never true, 2= very seldom true, 3= sometimes true, 4= frequently true, 5= always true). The instrument showed a good Cronbach's Alpha (0.72), grouping the items into one single factor, just like the original questionnaire.

• •

  Hospital Anxiety and Depression Scale (HADS), Spanish version by Terol et al.28 This instrument assesses the emotional responses of anxiety and depression in patients with physical and mental disorders and in general population without taking into account somatic symptoms, so limiting the chance of the symptoms of the disease contributing to the scores.29,30 It contains 14 items with a four-point Likert response format ranging from 0 to 3: the higher the score the worse the anxiety or depression. It is made up of two sub-scales: one for anxiety and the other for depression. In this study, a good Cronbach's Alpha (0.83) was obtained, both for the anxiety and the depression sub-scales.

Sociodemographic variables included the dialysis centre, gender, age, marital status, educational level and employment status. Clinical variables included active prescription of psychiatric drugs, time on HD, transplant waiting list, Charlson comorbidity index score, presence of diabetes and type of vascular access. In addition, the average dialysis dose for the month was estimated, measured as Kt/V calculated using the On-line Clearance Monitor (Fresenius Medical Care), Kt On-line Clearance Monitor or eKt/V, the mean pre-HD systolic blood pressure for the month and interdialysis weight gain. We also collected the lab analysis results for that month for potassium, sodium, haemoglobin, transferrin saturation, C-reactive protein, albumin, phosphorus (P) and calcium. Finally, we measured the patients' relative hydration status by bioimpedance spectroscopy and their body composition using the BCM® device (Body Composition Monitor; Fresenius Medical Care), recording the relative overhydration, lean tissue mass, adipose tissue mass, fat tissue mass, lean tissue index, fat tissue index, relative lean tissue mass and relative fat tissue mass. The information for the clinical variables was obtained from the Euclid database31 belonging to Fresenius Medical Care. The data were collected in June and July 2017.

The AAQHD-II was created after a review of items in the Spanish version of the AAQ-II15 and other versions of said instrument used for the study of diabetes,19 cancer20 and obesity.21 The whole AAQHD-II development process was reviewed and agreed on by an interdisciplinary team (nephrologist, psychologist, social worker and nurse) with extensive experience in the treatment of patients on HD. A battery of ten items was then put together and reviewed by another psychologist, author of the Spanish version of the AAQ-II. That battery was administered to a group of six patients on HD to assess their degree of understanding of the items. After evaluating the patients' opinions on completing the questionnaire, three items which seemed ambiguous were removed, while the term "HD" was replaced by the word "dialysis", as the patients are more familiar with that. By consensus from the start, the specialists involved in adapting the questionnaire considered the use of a five-point Likert scale, rather than the seven points in the original scale, to make it easier to understand the response format in view of the advanced age of the target patients.

The evaluation protocol was carried out in the treatment rooms during the patients' haemodialysis sessions by three assessors previously trained in completing the questionnaires. The aims of the study were explained to the patients prior to the questionnaire evaluation, and the questionnaires along with the study instructions were administered to those who agreed to take part by signing the informed consent form. Completion time ranged from 20 to 30min.

Cronbach's alpha and its 95% confidence interval (CI) were calculated to analyse the internal consistency of the AAQHD-II. The item-total correlations were calculated to identify items that should be removed due to a low discrimination index (values below 0.20). The suitability of the data for a factor analysis was measured with the Kaiser-Meyer-Olkin test. After that, an exploratory factor analysis was carried out using the robust unweighted least squares extraction method using polychroric correlations with the program Factor 10.8.03.32 The number of dimensions was determined by means of a parallel analysis based on the Minimum Rank Factor Analysis.33

In order to verify the convergent validity of the instrument, Pearson's correlations were made between PI and the psychological symptoms (anxiety and depression), and health-related quality of life dimensions.

The data are presented as mean±standard deviation for continuous variables with normal distribution, median and interquartile range (25th percentile-75th percentile) for non-normal continuous variables or percentages for categorical variables.

Those who scored more than 20 points on the AAQHD-II test were defined as "inflexible" patients. This value was chosen because it is a standard deviation above the mean of the AAQHD-II variable, and also for being the cut-off value of the upper quartile of the distribution.

To verify the possible existence of significant differences between the flexible and inflexible patients for the different variables recorded in the study, the Chi-square statistical tests were applied for the categorical variables, Student's t for the normal continuous variables and the Mann-Whitney U test for continuous non-normal variables.

To identify variables associated with PI, a multiple linear regression analysis was performed in which clinical or psychological variables that were significant in the bivariate analysis were included as independent variables.

All statistical analyses were performed using the statistical package SPSS version 23.0. The level of statistical significance was set at P<0.05.

The characteristics of the sample used are detailed in Table 1. A total of 186 patients were assessed, 113 (60.8%) male and 73 (39.25%) female. The mean age was 70.17±1.01 years, 97.85% were pensioners or retired, 33.87% had diabetes and the Charlson comorbidity index was 4.

Table 2 shows a comparison of patient scores based on a high or low PI score. Among the sociodemographic variables we found a significant association between PI and gender (P=0.005). In terms of health-related quality of life, the most inflexible patients obtained the worst scores (24.67±5.91; P<0.000). In the emotional symptoms, the most inflexible patients had the highest scores for both anxiety (P<0.000) and depression (P<0.000). Among the medical variables, the most inflexible patients were more likely to be taking psychoactive drugs (P=0.034). The fat tissue index (18.99±8.91; P=0.033), phosphorus levels (4.66±1.38; P=0.001) and interdialytic weight gain (1.89±0.93; P=0.016) were also increased in patients with a higher PI score.

As already mentioned in the section on instruments, the AAQHD-II showed acceptable internal consistency in this study. Table 3 shows the content of the items and corrected item-total correlations. The corrected item-total correlations varied from 0.28 (item 4) to 0.61 (item 3). As all scores were greater than 0.20, no items were eliminated because of a low level of discrimination.

The results showed that the data were appropriate for performing an exploratory factor analysis in view of the fact that Bartlett's test was statistically significant (290.8 [21], P<0.001) and the results of the Kaiser-Meyer-Olkin test were acceptable (0.71). The parallel analysis recommended extracting a single factor that explained 44.31% of the variance (self-value of 3.10). The factor loadings of the items varied from 0.35 (item 4) to 0.81 (item 3) (Table 3)

Table 4 shows the correlations between the AAQHD-II and the other instruments. The total AAQHD-II score was significantly correlated with both the anxiety (r=0.54, P=0.000) and the depression (r=0.51, P=0.000) sub-dimensions, assessed by the Hospital Anxiety and Depression Scale, and with the total score for health-related quality of life (r=0.45, P=0.000), assessed with the COOP-WONCA charts, and the following sub-dimensions: social activity (r=0.42, P=0.000), daily activities (r=0.40, P=0.000), feelings (r=0.36, P=0.000), overall health (0.32, P=0.000), quality of life in general (r=0.31, P=0.000), physical fitness (r=0.23, P=0.001) and change in health (r=0.14, P=0.004).

The results of the multivariate regression analysis are presented (Table 5) with the variables that showed statistical significance in the prediction of PI in the univariate analysis. This analysis shows how the variables being female, health-related quality of life, anxiety, depression, time on HD, albumin, fat tissue index, calcium levels, phosphorus levels and inter-HD weight gain explain a large portion of the PI (R2=46%, P=0.000). However, the only ones to show statistical significance were: anxiety (P=0.000), depression (P=0.009) and phosphorus levels (P=0.013).

The results of the multivariate regression analysis are presented (Table 6) with the variables that showed statistical significance in the prediction of phosphorus levels in the univariate analysis. This analysis shows how the variables being a woman, age, being single, married, separated/divorced, widowed, PI, anxiety, systolic blood pressure, potassium levels and inter-HD weight gain partly explain the variability of phosphorus levels (R2=23%, P=0.000). However, the only ones to show statistical significance were being female (P=0.014), age (P=0.000), PI (P=0.026), systolic blood pressure (P=0.016) and potassium levels (P=0.004).