Setting
This cross-sectional study was conducted between 1 May, 2018, and 30 April, 2019, in the internal medicine ward of Lausanne University Hospital, Switzerland. The internal medicine ward has a 166-bed capacity and admits 5550 patients per year, two-thirds of who are ≥65 years old.
Sample size
Sample size was based on data reported by Rurup et al. [2] and calculations were performed using PS Power and Sample Size version 3.0 [14]. Considering a WTD prevalence of 15% [2], an alpha value of 5%, and a power of 80%, a sample size of 250 was considered adequate. Considering a 50% exclusion rate, a 40% refusal rate, and 10% of the records to have missing data, it was necessary to screen 900 patients.
Selection procedure
Before being considered eligible for the study, patients aged ≥65 years were screened for language skills and cognitive impairment. For pragmatic reasons, screening was conducted on the second day of hospitalization; the first day was too busy owing to all the clinical procedures performed during the admission process. As a large number of patients were admitted to the ward, it was not possible to screen them all. The study nurse visited a different sector each day according to a previously established schedule. All patients within the selected sector were screened. Patients were excluded from this screening if one of the following conditions was present: 1) previous participation in the study; 2) refusal of cognitive screening; and 3) cognitive screening not feasible (e.g. patient transferred to another unit or too ill to cooperate). Cognitive evaluation was performed using the 6-Item Cognitive Screener [15] and cognitive impairment was defined as a score < 3. Patients unable to speak French or who had cognitive impairment were considered ineligible to participate. Patients eligible for participation were invited to participate and given written information about the study. In accordance with the relevant Swiss legislation and ethics committee guidelines, eligible patients had a 24-h period to make their decision about participating; after this 24-h period, patients willing to participate signed a consent form and were included in the study.
Data collection
Data were collected in face-to-face interviews with a research nurse trained in palliative care. Participants completed several standardised study questionnaires, which have been previously published elsewhere and are free to use for research purposes. These instruments are described below. The interviews lasted approximately 1 h and were conducted in a closed room, where the research nurse was alone with the participant. Special attention was paid to study participants: the nurse listened patiently and carefully to each participant.
Wish to die
As WTD instruments vary [3], we assessed WTD using two validated instruments to ensure that the construct was adequately measured in this large sample. We considered patients to show a WTD if responses on either or both instruments were positive.
The French version of the Schedule of Attitudes toward Hastened Death-senior (SAHD-senior) assesses the intensity of the WTD in elderly persons by evaluating attitudes to death, the wish to die/live, and the fear of physical/psychological suffering [3]. It is a modified version of the SAHD, which was originally validated for young terminal stage patients [16]. The SAHD-senior consists of 20 true–false statements; the total score ranges between 0 and 20. Scores ≥10 indicate a WTD.
The French version of the Categories of Attitudes toward Death Occurrence (CADO) measures quality of the WTD [3]. It is based on a qualitative study conducted by Schroepfer [17] and contains six categories (1: neither ready nor accepting; 2: not ready but accepting; 3: ready and accepting; 4: ready, accepting, and wishing death would come; 5: considering a hastened death but having no specific plan; and 6: considering a hastened death with a specific plan). Categories 4 to 6 express an active WTD.
Other covariates
QoL was assessed using two instruments. The first was the French version of the Quality of life – Control, Autonomy, Self-realization and Pleasure (CASP-12) [18], which was specifically developed for elderly respondents [19, 20]. The CASP-12 consists of 12 questions; responses are on a Likert scale of 0 to 3. The total score ranges between 0 and 36; higher scores indicate better QoL. The second instrument was a single question scored on a numerical scale between 0 (‘worst QoL’) and 10 (‘best QoL’). This scale has been shown to be reliable compared with multi-item scales [21, 22].
Social and family support were evaluated using the French version of the Medical Outcomes Study Social Support Survey (MOS-SSS) [23, 24]. This validated tool contains 19 questions that rate four dimensions (tangible support, affective support, positive social interaction, and emotional/informational support). Responses are on a Likert scale ranging from 1 to 5. The pooled responses are converted to a score ranging between 0 and 100; high scores indicate excellent social support.
Depressive symptoms were assessed using the 20-item Center for Epidemiologic Studies-Depression (CES-D) scale [25]. The CES-D has been validated for elderly respondents and translated into French [26]. It comprises 20 questions, and responses are on a 4-point Likert scale. The final score ranges between 0 and 60; scores ≥16 indicate clinical depression.
Spirituality was assessed using three questions (spiritual person, place in life, importance during illness) from the semi-structured clinical interview SPIR (Spiritual needs and preferences) [27]. Responses to each question are on a numerical scale ranging between 0 and 10; higher values indicate that the respondent places great importance on spirituality. The adapted SPIR score was the average of all responses.
The degree of stress generated by an interview querying the WTD attitude was assessed at the end of the interview on a numerical scale between 0 and 10. Higher scores indicate greater stress.
Do-not-resuscitate (DNR) orders are requested from all patients hospitalized at Lausanne University Hospital [13]. This information was collected at admission and recorded in the patient’s electronic record.
Information about age, gender, case severity (Charlson score) [28], number of medications (at admission), and functional status (Combined Braden activity and mobility subscale) [29] were obtained from medical records.
Ethical statement
The ethics commission of Canton Vaud (CER-VD, www.cer-vd.ch) approved the study (reference 2017–01875, decision of 6 December, 2017). The study was performed in accordance with the Helsinki declaration and its former amendments, and in accordance with the relevant Swiss legislation.
As a safety issue, we systematically monitored if participants needed to begin psychiatric treatment related to the WTD during the first 2 weeks of hospitalization. If over 10% had experienced distress, the study would have been stopped. In the end, no participants presented a worsening of their psychological status following the interview.
Statistical analysis
Statistical analyses were performed using Stata version 15.1 (Stata Corp, College Station, Texas, USA). Missing data (which were infrequent) were not replaced. WTD prevalence was expressed as percentages and 95% confidence intervals (CI) calculated using the Poisson method. Agreement between the two WTD instruments (SAHD-senior and CADO) was assessed using Cohen’s kappa. Following previous studies [4], data for all patients expressing a WTD on at least one of the two instruments were pooled. Descriptive results for the WTD (yes/no) were expressed as averages ± standard deviations (SD) or as medians [interquartile ranges] for continuous variables, and as number of participants (percentages) for categorical variables.
Bivariate analyses were performed using Student’s t-test or the Mann–Whitney test for continuous variables and the chi-square test for categorical variables. Variables significantly associated with the WTD in a bivariate analysis were then tested using a multivariable analysis, as follows: first, each variable was tested after adjusting for age; second, all variables were included in a single model; third, the results of the second step were confirmed using stepwise backward and forward procedures, with a P-value for entry of .05 and a P-value for removal of .10. Multivariable analysis was performed using logistic regression and the results were expressed as odds ratios (OR) and 95% CIs. For multivariable analysis, the effects of a 5-year increase in age, a 5-point increase in the CASP-12 score, and a 2-point increase in the CES-D score were used. Statistical significance was assessed at P < .05.