This was a parallel arm, open-label randomized controlled trial with 1:1 allocation to intervention and control groups. The study was approved by the Ottawa Health Sciences Network Research Ethics Board. The trial is registered at clinicaltrials.gov (NCT03287895). The approved protocol can be found in Supplement 1.
Setting and Participants
Participants were recruited at the General Hospital and the Civic Hospital in Ottawa, Canada, two affiliated academic tertiary-care teaching hospitals. All patients admitted to internal medicine or neurology wards had a 1-year risk of death calculated by an automated, validated risk prediction model running on hospital databases.11 Patients were eligible if they were older than 55 years old, had a 1-year risk of death greater than 10%, and had an order in their medical record for CPR in the event of cardiac arrest, or no documented order regarding CPR in their medical record. Research assistants recruited patients in their hospital rooms. Internal medicine and neurology wards were selected because of high risk of death and longer median length of stay in these populations making CPR discussions more relevant. In our jurisdiction, patients must verbally consent to any order regarding CPR during the hospital admission. In cases where the eligible patient did not have capacity to make medical decisions for themselves, their healthcare proxy (HCP) was eligible for participation.
Participants were randomized using a centralized automated computer-generated sequence to intervention or control groups using randomly permuted block sizes of 2, 4, or 6 for allocation concealment.
The DSI was administered by trained research assistants in the patient’s hospital room. The first component consisted of two values clarification exercises that helped patients prioritize health goals (Supplement 2). The first exercise presents the patient with 2 values at opposite ends of a linear scale and asks them to select the point on the scale that represents the relative importance of the values. For example, the first scale is labeled with “how well you live” at one end of the scale and “how long you live” at the other. The second values clarification exercise uses a methodology called “best-worst scaling” to establish the relative importance of four values. To prioritize the values, the user is asked sequentially to select which value is most important and least important from sets of 3 of the 4 values. The four values are as follows: maintain independence, quality of life/well-being, live as long as possible, avoid machines to keep me alive.
The second component of the DSI was a video decision aid about CPR viewed immediately after the values clarification exercises.12 The video is 7 min long, fulfills 7/7 criteria of International Patient Decision Aids Standards to be considered a decision aid, and has been shown to be acceptable and increase patient knowledge.13,14,15 The decision aid presents the harms and benefits of CPR for people who have serious chronic disease, similar to those in this trial.
Participants then received a “Dear Doctor Letter” that summarized their responses to each part of the intervention and their response to the question about CPR. Patients were instructed to give the letter to their doctor in hospital to discuss what is important to them and their wishes regarding CPR.
Patients randomized to usual care were told to ask their physician in hospital if they had questions about CPR. Usual care at the participating hospital includes a hospital policy that patients should have an order to administer or withhold CPR within 48 h of admission. There is no structure to enforce the policy and all aspects of the communication are at the discretion of the physicians.
The primary outcome was the difference in no-CPR orders between groups at the time of death, discharge, or 14 days after enrollment, whichever came first. Secondary outcomes were as follows: patient stated preferences for CPR immediately after the intervention, decisional conflict, satisfaction with decision-making, health care proxy’s decision-making self-efficacy, quality of documentation about goals for future medical care, and concordance between values and stated preference for CPR.
Orders for CPR were abstracted from patient charts by a blinded outcome assessor. To assess the quality of patient’s decisions, we used the decisional conflict scale (DCS) a measure of uncertainty in decision-making and factors that contribute to uncertainty.16 The DCS is scaled 0–100 with lower scores meaning less decisional conflict. For satisfaction in decision-making, we used the decision-making domain of the CANHELP survey.17 The CANHELP survey, scaled from 0 to 100, measures satisfaction with end-of-life care and has been shown to be valid and internally consistent. We used the HCP Self Efficacy scale (range from 1 to 4) to measure HCP self-efficacy to make decisions on behalf of the patient.18 All surveys were completed immediately after the intervention for the intervention group and at enrollment for the control group. After completing the intervention, patients were read the following script: “Whether you choose to have CPR or not, treatment is still focused on helping you live as well as you can for as long as you can. You will always receive treatment to help you with symptoms and care for your needs. Given your current health condition, at this point in time, what would you want if your heart were to stop beating?
Use machines and all possible measures including Cardiopulmonary Resuscitation (CPR) with a focus on keeping me alive at all costs.
Allow a natural death with no artificial prolongation of life and no CPR.
I am unsure.”
We report patient responses to the values clarification question about the trade-off between “I want doctors to do everything possible, including use of machines, to keep me alive for as long as possible” and “I would want a peaceful death without being attached to machines” because responses to this question are expected to correlate with responses to the question about CPR preferences.
We assessed the frequency and quality of documented conversations about CPR and goals for future medical care using a standardized chart abstraction tool that contained 17 components in 4 domains.19 Abstraction of the primary outcome and quality of documented conversations was done by a blinded outcome assessor who reviewed the chart.
Patient and Admission Characteristics
Marital status, place of residence, education, race, information needs, language, and quality of life were collected by self-report before randomization. Participants self-assessed their frailty using a validated tool.20 We collected age, sex, risk of death within 1 year, comorbidities, and information on prior healthcare encounters from our institutional data warehouse.
The sample size calculation was based on a previous CPR video decision aid study.21 A sample size of 200 patients had 80% power to detect a 15% absolute difference in orders for CPR (primary outcome) between groups with an alpha of 0.05. The primary analysis was the risk difference between groups. As a secondary analysis, we performed logistic regression to determine the odds ratio (OR) of the primary outcome in intervention group compared to the control group adjusted for the CPR order at the time of enrollment (either order for CPR or no order regarding CPR). We used a chi-square test to test consistency with the null hypothesis for the primary outcome, and for other categorical variables. We used Fisher’s Exact test for categorical variables with cells containing fewer than five observations. For ordinal scales (DCS, CANHELP, and HCP self-efficacy), we used the Wilcoxon signed-rank test. Analysis was done by a statistician who was blinded to allocation. We used complete case analysis for missing outcome data.