Patient and treatment characteristics associated with patient activation in patients undergoing hemodialysis: a cross-sectional study

In this quantitative, observational, cross-sectional, and questionnaire-based study, convenience sampling was used. Participants were recruited in three Flemish dialysis units based on following criteria: (1) diagnosis of ESRD; (2) older than 18 years; (3) Dutch-speaking; (4) cognitively able to understand difficult concepts (at examiner’s discretion: first question contained the word ‘responsibility’ and the examination stopped when the participant did not understand that word correctly); (5) dialysis treatment longer than three months. Systematically, all hemodialysis patients that were treated in the hospital on the day of the investigation and that were eligible, were asked to participate.

To determine the sample size, the number of variables was examined. In order to achieve sufficient statistical power, 50 patients and an additional 8 patients for each variable had to be enrolled [13]. Because of the 15 variables in this study, the sample should thus consist of at least 170 patients.

Data were collected in February and March 2016. The questionnaires were completed independently or together with the interviewer. The same interviewer (LVB) was available on request for all patients. Self-reported questionnaires were used to measure all variables. Hence, data were gathered from the patient’s perspective [see Additional file 1].

Approval for the study was given by the Independent Commission for Medical Ethics of the UZ/KU Leuven, Medical Ethics Committee of hospital Imelda, and Medical Ethics Committee of hospital Sint-Trudo, all located in Belgium. Procedures were in accordance with the declaration of Helsinki [14]. Oral and written information about purpose, duration, and risks of study participation was given to patients before they were asked to participate. All participants provided written informed consent.

The primary outcome was patient activation, which was measured by ‘Patient Activation Measure-13’ (PAM-13, Dutch version) [15]. This instrument is a non-disease specific scale that shows involvement of patients in their health. This 5-item Guttman scale has following possible answers: ‘disagree strongly’ (1), ‘disagree’ (2), ‘agree’ (3), ‘agree strongly’ (4), and ‘not applicable’ (5). Raw scores range from 13 to 52. No score was calculated if no answer or ‘not applicable’ was chosen more than three times. Raw scores were converted to a theoretical score on a scale of 0 to 100. Higher scores indicate a higher level of patient activation. Patients were divided into four levels based on cut-off scores. In level 1 (score: ≤ 47.0), patients do not believe they have an important role regarding their health. In level 2 (score: 47.1–55.1), patients have lack confidence or knowledge to take action. In level 3 (score: 55.2–67.0), patients start to take action. In level 4 (score: ≥ 67.1), patients are maintaining active behavior [16]. The Dutch version of PAM-13 has been shown to be a reliable instrument [15]. Insignia Health provided a license.

Patient-, treatment- and health-related variables were measured though self-reported open and multiple-choice questions [see Additional file 1].

Basic patient-related characteristics included age, sex, highest degree of education, employment status, and living situation. Employment status was questioned as follows: “Do you currently work in paid employment”, with possible answers ‘no’, ‘part-time’, and ‘full-time’. In order to gain information on the level of activation in daily life, leisure-time activities, amount of home care services, presence or absence of children, and perception of social support, were measured. “Do you receive sufficient support from your environment?” with possible answers ‘yes’ or ‘no’, was asked to measure the perception of social support. Leisure-time activities were questioned as follows: “Do you have a hobby, do you do any sport, or are you a member of any organization?”

Treatment- and health-related factors included in this study were time since first dialysis, history of one or more renal transplantations, depressive symptoms, and self-reported health. Depressive symptoms were measured by the “Patient Health Questionnaire-2” (PHQ-2, Dutch version). This instrument is the short version of the PHQ-9 [17]. It questions frequency of a depressed mood and anhedonia (no interest in activities) during the last 2 weeks before the day of the study [17]. Answer possibilities were ‘Not at all’ (0), ‘Several days’ (1), ‘More than half the days’ (2), and ‘Nearly every day’ (3). The maximum score was 6. Our study used a cut-off score of 3, because sensitivity is 87% and specificity is 78% for major depression for this cut-off score [17]. Self-reported health was measured by the EuroQol Visual Analogue Scale (EQ-VAS, Dutch version) [18]. This 20-cm long visual analog scale is a part of the standardized EQ-5D. A scale of 0 to 100 was displayed. A score of 0 represented the worst health, and 100 the best health that could be imagined. A license from EuroQol Research Foundation was obtained.

Analyses were performed using statistical package SPSS (version 23). Patient characteristics are presented as mean ± standard deviation for continuous variables and number and percentage for categorical variables.

In order to detect existence and extent of multicollinearity in the final model, tolerance and variance inflation factor (VIF) were calculated [19]. By using histograms and scatter plots the assumptions of linearity, homoscedasticity and normality were checked and approved. Outliers were identified using Cook’s distance.

A total of 214 patients undergoing hemodialysis were approached in this study of which 197 were willing and able to participate (response rate: 92%) (Fig. 1). Of these patients, 3 patients generated incomplete PAM scores and another 2 patients were excluded because of outlying results that affected the results of the analysis. Ultimately, data from 192 patients were used in the final analysis. Of these patients, 117 (61%) were male and age range was 20–95 years with a mean age of 72 ± 14 (Table 1). In these categories, the sample was representative for the population of patients undergoing dialysis in Flanders (Table 2). Patient characteristics and PAM scores are descripted in Table 1. A total of 138 (72%) participants completed the questionnaire together with the interviewer.

Table 3 shows the results of univariable and multiple analysis with activation as a dependent variable associated with all independent variables.

One of the aims of this study was to investigate the level of activation of patients undergoing hemodialysis. The average activation score was 51 (± 10). Of the 192 patients, 44% patients did not believe they had an important role regarding their health. The high number of 73% of the patients did not take charge of their own health. The difficult combination of diet and fluid restrictions, strict medication regime, intensive dialysis treatment, comorbidities, and the proximity of healthcare workers three times a week, might explain the low activation scores in this population. Although the average activation score of 51 may be overestimated because people with cognitive impairment were excluded in the sample, activation of patients undergoing hemodialysis seems to be low and healthcare workers should be recommended to measure patient activation and intervene upon low levels of activation.

To further increase comprehensibility of the activation scores in our study, a literature search was performed about activation of other chronic patient groups which face similar challenges due to their disease. It appeared that patients with hypertension, depression, asthma, and diabetes have a higher average activation score compared to the patients undergoing hemodialysis in the present study [10, 20‐30]. In the study of Bos-Touwen et al. a comparison was made between the average activation scores of patients with diabetes (55.3), chronic obstructive pulmonary disease (54.7), chronic heart failure (53.6), and chronic kidney disease (51.4) [10]. The kidney patients had the lowest activation score [10]. In the literature only one activation score was found to be lower than the average score measured in our study, namely the average activation score (50) of patients with osteoarthritis in South Korea [31]. Because of the various sample characteristics [22, 26], different countries of origin, cultural backgrounds, access to and cost of health care for patients [15, 22, 31], it can only be assumed and not determined with certainty that patients undergoing hemodialysis in Flanders have lower activation scores compared to other chronic patients.

The second aim was to identify patient- and treatment-related factors associated with activation. Age, self-reported health, hospital, leisure-time activities, and living situation were associated with activation in multiple analysis.

The R² of the model was 0.306. Around 31% of variance of activation could be explained by these five variables. This R² was higher than in previous models [10, 32].

It was demonstrated in our study and in many other studies that higher activation correlates with lower age [12, 15, 21, 22, 26, 30]. Explanations may be that older patients have lower self-efficacy [21], lower health literacy [33], seem to be less compliant [34], and are more accustomed to a paternalistic healthcare system [22] compared to younger people.

As in our study, previous literature showed that patients who reported being in good health, have higher patient activation [10, 15, 21, 26, 35]. This is not surprising, since patients who are more active, report better skills and better knowledge, confidence and behavior needed to manage their health condition [21, 26].

In our study, a significant lower level of activation was found in patients treated in hospital 2. This could be due to differences in the predialysis training, since previous research has shown that predialysis education can lead to higher levels of knowledge [36], and better self-management skills, such as better fluid balance [37]. However, we were unable to investigate this in our study, since no individual data about predialysis education was available. It would be interesting in future studies to investigate the association between organizational features of a hospital, accompaniment of patients before and during dialysis treatment, and patient activation.

When looking at functional disability and level of activation in daily life, previous research showed that patients who were more active, had no difficulties with activities of daily living [12], which could explain why active patients are more likely to have leisure time activities in our study. Although, Fowles et al. did not find a significant association between activation and membership of a health club [32]. Patients who live in a residential care home are less activated compared to patients living with someone. It was previously showed that people who live in their own house or apartment were significantly more activated than those living elsewhere [25].

In our study, age was a confounder in the association between patient activation and educational level, employment status, and history of renal transplant. The average 70-year-old population is less educated compared to a young population, more than half of the patients were on retirement, and older people with comorbidities might have lower chances to get on the transplant list.

To the best of our knowlegde, this study is the first study that has investigated patient activation specifically in the population of patients with hemodialysis.

Strengths of the study are a fairly large sample and a high response rate. Furthermore, the sample is representative for the Flemish hemodialysis population and participants were recruited in three hospitals. International validated questionnaires were used. Finally, because of the clinically meaningful outcome, the study provides information that is useful for practice.

Furthermore, the study has several limitations. First, no information about the directionality of the relationships could be obtained, due to the observational cross-sectional design. Second, data were measured using a self-reported questionnaire, which can induce recall bias or telescoping. However, only two questions had a recall timeframe. Third, certain possibly interesting variables were not included in this study, such as motivation, health literacy, hope, external control, cognitive impairment, genetics, life events, (the amount of) comorbidities, predialysis education on individual level, and organization features. Future research should take these variables into account. Fourth, in our sample, patients that were cognitively unable to understand difficult words were excluded from the study. In addition, no randomization techniques were used. These two factors might reduce the generalizability of our findings and might have created selection bias, which could have affected the results of the study.

Nearly 73% of the patients did not take charge of their own health, which has shown to negatively influence health outcomes and costs. Hence, practitioners and other healthcare workers should be recommended to measure activation of patients and if needed to intervene upon low levels of patient activation.

Because older patients, who reported being in bad health, treated in a particular hospital, without leisure-time activities, and living in a residential care home, had significantly lower patient activation, patients at high risk can be identified using these screening factors. Furthermore, on the long term, provided that longitudinal studies show similar results and provide information on the directionality of the associations, these factors can help healthcare providers to develop personalized interventions to improve patient activation. Hence, improving self-reported health and/or encouraging patients to be more active in daily living could be part of these interventions.

Future research should investigate patient activation and its predictors in longitudinal research and provide information on the usefulness of these predictors in personalized interventions. Furthermore, the consequences of low activation should be further investigated in the particular population of patients undergoing hemodialysis.