We designed a cross-sectional survey to be conducted at 5 tertiary care teaching hospitals across Canada. From west to east, they were St. Paul's Hospital, Vancouver, BC; Royal Alexandra Hospital, Edmonton, Alta.; Toronto General Hospital, Toronto, and Kingston General Hospital, Kingston, Ont.; and Queen Elizabeth II Health Sciences Centre, Halifax, NS. Each site admitted seriously ill patients to the acute care wards under the care of a primary service and had regional or hospital palliative care consultation services available upon request. The research ethics board at each participating institution approved the study design. All study subjects provided written informed consent.
Eligibility criteria for the study were chosen to define a patient population having advanced disease, with a 50% probability of survival at 6 months.25
Although our focus was on elderly patients, some patients with advanced cancer or end-stage medical disease would be younger than 65 years; to enhance the feasibility and generalizability of the study, we lowered the age limit to 55. Patients were therefore eligible if they were aged 55 years or older; were expected to stay in hospital for at least 72 hours; and had an advanced stage of 1 or more of these 4 diseases:
• Chronic obstructive lung disease, with at least 2 of these 4 conditions: baseline Paco2 of at least 45 mm Hg; cor pulmonale; an episode of respiratory failure during the past year; forced expiratory volume in 1 second of 0.75 L or less
• Congestive heart failure, with New York Heart Association class IV symptoms or a left-ventricular ejection fraction measured at 25% or less
• Cirrhosis, confirmed by imaging studies or documentation of esophageal varices, and any of hepatic coma, Child's class C liver disease or Child's class B liver disease with gastrointestinal bleeding
• Cancer, diagnosed as metastatic cancer or stage IV lymphoma
On the basis of a chart review of a patient's progress notes, conversation with staff or, in some circumstances, direct evaluation of the patient, we excluded participants likely to have language or cognitive difficulties. Consenting patients identified a family member or other close person who provided some form of care in the home setting, if such a relationship existed. There was no age restriction for family members or social caregivers to participate in the study.
To develop the questionnaire, we first reviewed published taxonomies of quality end-of-life care to generate an initial list of key care elements.12–15
Additional elements were then generated from further literature review and discussion with the multidisciplinary End of Life Research Working Group at Queen's University, Kingston, Ont. Next, to determine whether any elements had been overlooked or were ambiguously phrased, we conducted 12 semistructured interviews with eligible seriously ill patients in hospital.
Finally, we developed a comprehensive list of 28 elements of care, organized into 5 domains: medical and nursing care; communication and decision-making; social relationships and support; meaningful existence; and advance planning of care. We used response options to assess differing degrees of importance using a 5-point ordinal scale: 1 = not at all important; 2 = somewhat important; 3 = important; 4 = very important; and 5 = extremely important. The draft questionnaire was then tested extensively to assess comprehensiveness, readability, sensibility and clarity in focus groups with health care providers and in individual interviews with patients and family members.
To collect data, a research nurse was hired at each hospital and trained by project staff to conduct the interview and perform study procedures. The research nurses screened medical records of current inpatients to identify potential participants and confirmed their suitability with the attending physician or assigned bedside nurse. Consecutive patients who seemed emotionally, physically and cognitively able to participate were approached for informed consent, as were the designated family members of enrolled patients. Both patients and family members were then administered the questionnaire in separate, face-to-face interviews.
The first item on the questionnaire was an open-ended question; the research nurse asked the interviewee to talk about the illness and treatments. Next, the nurse presented the respondent with a list of the 28 elements and asked them to rate each element for its level of importance, in their own eyes, to end-of-life care. Except for pronouns (e.g., “his or her disease” instead of “your disease”), the family-member survey was identical to the patient survey; respondents were instructed to report their own views, not to offer proxy responses on behalf of the patient. The nurse also collected demographic data including age, sex, marital status, ethnicity, religion, education, admission diagnosis and functional status.26
Survival status at 6 months was determined by contacting the participating patients or their family or family doctor.
Because the main purpose of this study was descriptive, we set out to obtain a consecutive sample of 100 eligible patients in each of the 5 participating hospitals. A sample size of 100 was a convenience sample thought to be large enough to provide a representative sample of responses for each locale. For each survey question, we noted the frequency for each of the response options from 1 to 5, separately for both patients and family members. In tables, we ranked the elements according to the proportion of responding patients and family members who rated each element as 5 (extremely important).
Using Pearson's χ2
we explored key differences in ratings of importance (the percentage of ratings of “extremely important” v. all other ratings) between groups of patients with different diagnosis categories (cancer v. other end-stage medical) and levels of family support (patients with a family caregiver v. those without), and different relationships to the patient of the family members responding (spouse, child or “other”). Finally we analyzed differences in ratings between patients and their corresponding caregivers, using McNemar's test for matched-pair data.28