Study design and participants
Data were collected in a retrospective cross-sectional survey as part of the ‘Last Year of Life Study-Cologne’ [50]. Sample and data collection has been described in detail elsewhere [51]. Briefly, we included relatives, friends and volunteer workers (all will be referred to as ‘informants’ hereafter) of deceased adult persons in the Cologne area. Accidental and suspicious deaths were not included. Informants had to be aged 18 years and older.
Instrument development
The VOICES-SF survey is a questionnaire about experiences of end-of-life care in the last three months of life, focusing on quality of care and services received. It uses the post-bereavement method to gather information from bereaved relatives, friends or carers acting as proxies. It is a survey instrument rather than a psychometric scale, questions may be added or deleted depending on a survey’s objective [14]. Using the VOICES‑SF as foundation, we developed a culturally adapted German version and called it ‘VOICES-Last Year of Life-Cologne (VOICES-LYOL-Cologne)’. This version evaluates care received in the last year of life, different to the original VOICES‑SF survey, which captures the last three months of life. This modification results from evidence showing that adding palliative care in the last 12 to 24 months of life has proven to be greatly beneficial [52, 53]. The longer period of observation enables the assessment of the introduction of palliative care services.
Figure 1 shows the translation and adaptation process. Procedures were based on the team translation approach ‘TRAPD’ (translation, review, adjudication, pre-testing, documentation)
[54]
. Two native German speakers (GD, NS) produced independent parallel translations, which were discussed with one reviewer. All of them were proficient in English. Three adjudicators (RV, JS, CR) familiar with the research project and the survey design, went through the final review version to decide on further modifications of the translation. Adaptation needs were addressed in different stages. Question adaptations were made to the content, response scales, and visual presentation of parts of the questionnaire. To produce a questionnaire that is culturally appropriate to the German health care system we conducted one group discussion with German healthcare professionals (n = 7) who work in palliative care. The resulting questionnaire was then tested in cognitive pre-tests with think-aloud technique (n = 8) to check whether 1) translated items as well as response categories were clearly understandable, 2) the questionnaire covered all important aspects of healthcare at the end of life, 3) wording and length were considered acceptable for bereaved relatives. Pretesting again resulted in refinement before the adjudicators signed off on the version for final fielding.
The following modifications were made:
- Revised observation period from “last three months of life” to “last year of life”.
- Added questions about the communication of the diagnosis of a life-limiting disease to assess if, by whom, when and how conversations about dying were initiated.
- Added the scale “subjective experiences of process and outcome of care in the last year of life” which comprises four key domains (relief of pain, relief of other symptoms, coordination of care, respect and dignity) in all settings of care.
- Revised terminology to align with German sample.
- Added sections related to healthcare providers additional available in the German healthcare system.
- Added single items about informal carers, place of care in the last two days of life, unsolved problems, financial problems, and demographic and clinical data.
- Harmonised response options and added “do not know” where appropriate.
The VOICES-LYOL-Cologne comprises a total of 106 items and contains skip logic so that informants only answer questions relevant to the care the patient received. In detail, it assesses care provided at home (by a nursing service, specialist palliative home care team, hospice volunteer services, GP, outpatient specialist physicians), care provided in a care home, hospital care (last admission), hospice care (in-patient), care provided in the last two days of life across all care settings, circumstances surrounding the death (communication of a life-limiting disease, carer support, unmet needs, financial needs, preferences and decision-making, communication of imminent death, place of death, bereavement support), and disease specific and sociodemographic data.
For each specified setting/health care provider informants could rate their subjective experiences of process and outcome of care in the last year of life on a four-point scale. Domains assessed are relief of pain, relief of other symptoms, coordination of care (1 = yes, 2 = rather yes, 3 = rather no, 4 = no), respect and dignity (1 = always, 2 = most of the time, 3 = some of the time, 4 = never). The total score is obtained by calculating the average of item scores. In addition, overall satisfaction with care in the last year of life by specified setting as well as taking all services into account is assessed on a four-point scale (1 = good, 2 = rather good, 3 = rather bad, 4 = bad). The questionnaire concludes with one free-text item to make overall comments about the care provided to the patient.
Other instruments administered
The survey also comprised the ‘German Patient Assessment of Chronic Illness Care (PACIC) short form for bereaved persons as proxies (PACIC-S9-proxy)’ [55]. This validated instrument assesses patient-centred care during the last year of life retrospectively and by bereaved relatives.
Data collection
Informants were recruited in cooperation with healthcare practitioners from Cologne between November 2017 and August 2018. Applied strategies (postal distribution, personal distribution, and self-selection) to identify potential participants have been described in detail elsewhere [51]. Study procedures were approved by the Ethics Commission of the Faculty of Medicine of the University of Cologne, Germany (#17-188).
Data analysis
Descriptive statistics were calculated using SPSS Statistics version 26 (IBM Corp., Armonk, NY, USA). Results are presented as mean ± standard deviation (SD) and count (percentage), respectively. The presented p-values are two-sided and considered significant if p < 0.05. To assesses the validity and reliability of VOICES-LYOL-Cologne we followed the procedures applied in the validation of the modified Canadian VOICES survey [46]. This study was informed by the guidelines of the Scientific Advisory Committee of the Medical Outcomes Trust for assessing healthcare surveys [56].
Validity
Construct validity was investigated by generating and testing hypotheses about expected patterns of scores for groups known to differ on relevant variables (nomological network) [56, 57].
- We expected the subjective experiences of process and outcome of care in the last year of life to be different among settings and health care providers. In previous studies informants rated hospice care more positively than homecare or hospital care [16, 18, 46, 58]. Accordingly, we hypothesised that each of the domains within hospice care would be ranked higher than care provided at home or in hospital and tested this using the Wilcoxon signed-rank test for paired samples.
- Satisfaction with care has been found to correlate to place of death. As shown in previous VOICES studies we hypothesised that informants of people who died in hospital rated overall satisfaction with care (taking all services into account) significantly worse than in any other place of death [46, 58]. We tested this hypothesis using the Mann-Whitney U-Test.
- The possibility of providing specialist palliative care at home was introduced in Germany in 2007 [59]. First evaluations showed that this care service enables more patients with complex symptoms and intensive care needs to die in their own home [60, 61]. We therefore expected a higher rate of home deaths in patients who received care by a specialist palliative home care team. This hypothesis was tested using the chi-square test.
- The provision of palliative care has been found to correlate with cancer indication [62, 63]. Accordingly, we hypothesized that cancer patients received care provided by hospice, a hospital palliative care unit or by a specialist palliative home care team more often than patients with non-cancer indications and tested this using the chi-square test.
Criterion (concurrent) validity is the correlation of a scale with some other measure of the trait [64]. The PACIC-S9-proxy served as external criteria to assess criterion validity. We calculated Spearman’s rank correlation between the overall satisfaction rating on the VOICES-LYOL-Cologne and the PACIC-S9-proxy mean score, both taking all services into account. We proceeded analogously with the overall satisfaction rating by specified healthcare provider, i.e., GP, specialist physician and hospital doctor on the VOICES-LYOL-Cologne and the PACIC-S9-proxy (mean score).
Reliability
Cronbach’s alpha was used to assess the internal consistency of the scale “subjective experiences of process and outcome of care in the last year of life”. Cronbach’s alpha indicates whether an item of a scale is appropriate for assessing the underlying concept of its scale [65]. Commonly accepted are values above 0.70 for group comparisons [56].
As a surrogate for test-retest-reliability, we assessed whether overall satisfaction rating of care differed according to the length of time since the patient’s death (reproducibility) and used the Kruskal-Wallis test. This approach was applied by Seow et al. to evaluate the Canadian VOICES survey since retesting bereaved relatives was not feasible [46]. We expected stability of the instrument over time if informants who respond closer to the patient’s death do not differ systematically from those responding longer after a patient’s death. We defined four equally sized groups (≤ 111 days, 112 to 215 days, 216 to 331 days, ≥ 332 days) to perform this test.