The EI was tested by way of a cluster randomised controlled trial with repeated measures [22].
Data collection
Data were collected in the HD units of ten hospitals in French-speaking Switzerland: Geneva (University Hospital, n=18 nurses), Jura (sites of Delémont and Porrentruy, n=16 nurses), Lausanne (University Hospital, n=11 nurses), Martigny (n=7 nurses), Monthey (n=7 nurses), Morges (n=5 nurses), Payerne (n=7 nurses), Sierre (n=7 nurses), Sion (n=14 nurses), and Yverdon (n=9 nurses). To participate in the study, nurses had to meet the following inclusion criteria: (1) at least six months’ HD nursing experience and (2) willingness to participate. There were no exclusion criteria. Data were collected at four time points: immediately prior to the EI (T0), immediately after the EI (T1), six months later (T2), and one year after the EI (T3). There were 101 nurses at T0, 96 at T1, 86 at T2, and 74 at T3. We first met with management and head nurses of each HD unit to explore their interest in participating in our study. If interested, we next met with nurses to present the project and distribute informed consent forms. Then, we met with the nurses again two weeks later to collect the forms. Nurses had the option of mailing them in. A coordinator was appointed for each HD unit to facilitate data collection and communication with researchers. Recruitment took place from April 2016 to January 2017. Our data were naturally structured into 10 clusters (i.e., the ten HD units), so cluster randomisation was used to define the experimental and control groups. In 2017, prior to data collection, the size of the target population was estimated at 126 HD nurses. We required that both the experimental and the control groups comprise at least 50 nurses. One of the possible cluster combinations that satisfied these conditions was randomly selected using R [23]. The procedure yielded an experimental group (EG) of 54 nurses and a control group (CG) of 72 nurses. The inclusion process was completed in March 2018 but, owing to the partial defection of one HD unit, the CG shrank considerably. The overall final sample thus totalled 105 HD nurses. Nonetheless, our criterion of at least 50 nurses each in the CG and EG was respected. The final EG comprised 51 nurses and the final CG, 50 (96.2% participation rate in both cases). At each of the four data collection time points, two meetings with nurses were organised: the first to distribute the hardcopy questionnaires and the second to retrieve them. Data thus collected were then transcribed onto digital support. To reduce the risks of introducing transcription errors, data entry was performed twice by different members of the research group and the resulting databases were compared. Up to 1.9% of the values had to be corrected across time points. Because of organisational constraints, researchers were not completely blind to participant group assignment. However, the EI, group assignment, data collection and data analysis were tasks performed for the most part by different members of the research group. Biases were thus kept to a minimum.
Intervention
Our EI ([MASKED FOR REVIEW]) consisted of 3.5-hour training sessions once a week for four weeks. Nurses were organised into groups of no more than five to facilitate interaction and feedback. Theoretical lectures alternated with practical exercises, such as simulations, to ensure the EI fit the reality of each HD unit. At the beginning of each session, focus exercises were proposed to help participants enter a more receptive mind-set and clearly separate the training from previous activities.
In the first session, the core concepts of Watson’s [17,18] Theory of Human Caring were introduced and a first clinical situation was presented to the participants for discussion. In the second session, Watson’s ten carative factors were introduced and discussed. Another clinical situation was presented to show a real-life application of the concepts covered. The third session was dedicated entirely to the concept of “hope” and how to help patients developing it. The concept was presented theoretically and through exercises. Finally, in the fourth session, a simulation activity was organised to revise and practise the concepts covered in the previous sessions. This activity was bookended by a briefing that introduced the situation and helped participants prepare their actions, and a debriefing that served to collect feedback from both participants and instructors. A final evaluation by all the participants wrapped up the training ([MASKED FOR REVIEW]). For ethical reasons and out of fairness, the EI was offered to the HD units in the CG at study completion.
Instruments
Data were collected using a three-part questionnaire. The first served to collect sociodemographic and work-related data. Questions covered gender, age, marital status, and presence of children, prior training in counselling, as well as years of nursing and HD nursing experience. This part of the questionnaire was previously used in related studies ([MASKED FOR REVIEW]).
The second part consisted of the French version of the Caring Nurse-Patient Interaction Scale (EIIP-70; [24,25,26]). The instrument covers 10 dimensions, corresponding to Watson’s 10 carative factors [17]: humanism (6 items), hope (7 items), sensibility (6 items), helping relationship (7 items), expression of emotions (6 items), problem solving (6 items), teaching (11 items), environment (7 items), assistance for basic needs (10 items), and spirituality (6 items). The items are rated on a five-point Likert scale ranging from “almost never” (1) to “always” (5) to indicate frequency of a given behaviour. This questionnaire had been used before (see the recent review by Cossette [26]) and demonstrated satisfactory psychometric characteristics, obtaining Cronbach’s alphas for the 10 dimensions ranging from 0.73 to 0.91 [25]. A pilot study [MASKED FOR REVIEW] had shown that respondents had no difficulty understanding the questionnaire.
The third part of the questionnaire consisted of the Quality of Work Life Questionnaire by Elizur and Shye [27] translated and validated in French by Delmas, Escobar, and Duquette [4]. This instrument yields a total score that can be divided into four dimensions: psychological (items 1–4), physical (items 5–8), social (items 9–12) and cultural (items 13–16). Respondents must choose from six answers on a Likert scale ranging from 1 “very little” to 6 “a very large part”. Internal consistency was estimated at α = 0.90 for the overall English version [27] and ranged from 0.87 to 0.89 for the dimensions of the French version [4].
Ethical considerations
Ethical approval was obtained from the Vaud (Switzerland) Ethics Committees on Research Involving Humans (N 2017–00946). Nurses were given two weeks’ time to decide whether to participate and to sign the informed consent form. They were free to refuse to participate and to withdraw at any time without consequence. Informed consent was obtained from all participants who participated to the study.
Because of the strong emotions that the EI could elicit, psychological support from a certified psychologist was available for participants at no charge, if needed.
Our research was registered on clinicaltrials.gov (NCT03283891, first registration 14/09/2017).
Blinding
Given the type of intervention, no blinding was possible for participating nurses. As for the research team, data allocation, intervention, data collection and analyses were done by different members of the research team. Nevertheless, no complete blinding was possible for organisational reasons.
Data analysis
First, means, standard deviations, and frequencies were calculated for sociodemographic and work-related data, as well as data on caring attitudes and behaviours and nurse QWL. Second, EG and CG characteristics were compared at baseline. We used chi-squared tests for categorical variables and linear regressions for numeric variables to test for independence. The Bonferroni correction for p-values was applied, given the large number of comparisons made and the fact that these were not part of the core analysis [28]. Third, the EI’s effects on both nurse caring attitudes and nurse QWL were evaluated using random-intercept regression models. This is a particular regression technique that allows the intercept to vary depending on predictors referring to groups of units of analysis instead of the units themselves. This technique is generally used when data are structured in clusters and/or are longitudinal. Our models comprised three levels: observations (regardless of time point), nurses (seen as clusters of observations), and hospitals (seen as clusters of nurses). This model was used to address the problem of lack of independence across observations by the same nurse but at different time points and by nurses working in the same hospital. Statistical significance for all tests was set at p < 0.05, following a frequentist approach. All analyses were run on R [23].