Intuitive vs Deliberative Approaches to Making Decisions About Life Support

Key Points Question Do seriously ill patients’ decisions about life support interventions differ when made intuitively vs deliberatively? Findings In a randomized clinical trial of 199 hospitalized patients aged 60 years and older with serious oncologic, cardiac, and pulmonary illnesses, encouraging patients to deliberate on end-of-life decisions did not change the content or improve the quality of those decisions. Meaning It is important to evaluate whether commonly advocated decision aids and structured communication interventions improve seriously ill patients’ choices.

Protocol Description: The purpose of this study is to determine whether there are systematic 7 differences between the decisions patients make intuitively versus 8 deliberatively about life-sustaining medical therapies. The targeted 9 population is inpatients at the Hospital of the University of Pennsylvania 10 with serious medical problems. The study will involve facilitated 11 interviews with patients using a survey instrument developed in 12 Qualtrics. 13 14   Research in cognitive psychology has demonstrated convincingly that human cognition involves two 56 different but interrelated modes of processing. The first, which is fast and associative, has been 57 attributed to an experiential or intuitive system (system 1); the second, which is slower, rule-based and 58 analytic, has been attributed to a rational or deliberative system (system 2) 1-3 . It is widely accepted that 59 the intuitive process involves multiple heuristics that can lead to biases, which in turn can lead to serious 60 systemic errors in judgment 4 . The deliberative system is generally thought to help serve as a check 61 against such errors 2 . A minority view, however, holds that decisions made under automatic or 62 unconscious conditions will more faithfully integrate existing values and can lead to normatively optimal 63 judgments, particularly where the decisions in question are complex, requiring the organization and 64 synthesis of several factors 5-6 . The mode of processing used by an individual has been shown to impact 65 such outcomes as the use of deontological versus utilitarian moral reasoning 7-8 , depth of religiosity 9 , and 66 inclination toward social cooperation 10-11 . There has been a strong movement in health care towards the 67 development of decision aids that foster deliberation 12 , but we know little about the use and relative 68 merits of intuitive and deliberative thinking in decisions regarding life-sustaining medical interventions. 69 Patients and surrogates are asked to make judgments about the desirability of initiating or continuing 70 such treatments under widely disparate circumstances: on admission to the hospital, in the context of 71 prolonged discussions over multiple visits with outpatient providers, and in time-sensitive emergency 72 situations.

Study duration
81 Subjects will be accrued over a period of six months starting in June 2015. The total time it will take the 82 research coordinator to explain the study, obtain consent and for a subject to complete the two sessions 83 will, conservatively, take no more than 45 minutes. 84 85 We will be recruiting 200 patients aged 60 or older with chronic respiratory, oncological or 86 cardiovascular diseases. This sample size would provide at least 80% power to detect differences of 20% 87

Target population
above or below baseline rates of 50% in the proportions of patients choosing each form of life support 88 or goal of care with a type I error rate of 0.05. Power to detect the same 20% absolute effects would 89 increase if baseline rates deviate away from 50%. 90 91

Group modifications 92
Subjects enrolled in this trial will be randomized into two groups, an intuitive arm and a deliberative 93 arm. Patients in each arm will be asked all of the same questions. The only difference between the arms 94 will be the instructions regarding how and when to answer the series of hypothetical questions 95 regarding medical interventions. The instructions will be designed to influence patients to think either 96 intuitively or deliberatively about the questions regarding life-sustaining interventions. We will be pilot 97 testing both the deliberative and the intuitive arm prior to starting formal recruitment. The details of the 98 pilot testing are discussed in the Study Procedures section of this proposal. Once we have determined 99 the optimal method, we will submit an IRB modification that reflects those methods 100

Method for assigning subjects to groups
101 Consenting subjects will be randomized with a 50% probability to each trial arm (intuitive versus 102 deliberative procedures) using Qualtrics. The research coordinator/PI will become unblinded to the 103 patients allocation when they open the electronic survey 104

Study instruments
105 Demographics survey: 106 The demographics survey will consist of basic information such as race, education, and religion. We will 107 also ask questions about prior experience with intensive care and whether the patient has a document 108 outlining preferences for care at the end of life. 109 110 Question regarding general treatment preference: 111 We will ask a single general question regarding the patient's treatment priorities in the event of serious 112 illness. The response to this question is modeled on one used in a Study to Understand Prognoses and 113 Preferences for Outcomes and Risks of Treatments (SUPPORT) study 15 . The question acknowledges that 114 in general, most people wish to both live as long as possible and avoid pain and suffering, but that in 115 some situations choosing between these two goals may be necessary. It then asks patients, if they are in 116 a situation where such a choice is needed, if they would they desire a plan of care that focuses on 117 extending life as much as possible, even if it means having more pain and suffering, or a plan of care that 118 focuses on relieving pain and suffering, even if that means not living as long. 119 120 Health state valuation questions: 121 We will ask a series of questions designed to understand how the patient values various health states. 122 We will use three distinct health states: the patient's current state of health and two hypothetical health 123 states that span four domains -cognitive functioning, physical mobility, self-care and pain/discomfort. 124 We will first ask the patient to rate each state of health on a 10-point Likert scale with 0 being much 125 worse than death and 10 being much better than death 16 . We also will ask patients to evaluate a variety 126 of single item health states on a visual analog scale ranging from worst state of health imaginable to 127 best state of health imaginable. 128 129 Rational Experiential Inventory, Short Form 10: 130 The Rational Experiential Inventory is a psychometric instrument designed to evaluate individual 131 differences in intuitive-experiential and analytical-rational thinking, which has been validated in diverse 132 populations. A short, 10-item form has been found to have similar properties as the original version 18 Questions to assess analytical reasoning: 153 During pilot testing of our methods, we will use a variety of questions to assess the analytical processing 154 of the participants, all of which have been used repeatedly by researchers in cognitive psychology as 155 measures of analytical thinking. These include denominator neglect problems 25 , belief bias syllogisms 26-156 27 , and the Cognitive Reflection Test 28 . 157

158
We will recruit patients with respiratory, cardiovascular and oncological diseases from the population of 159 stable inpatients on the general wards at the Hospital of the University of Pennsylvania. Each day, the 160 research coordinator will screen the electronic medical records of patients on the wards (rotating which 161 wards are screened on which day) to determine their eligibility using the eligibility criteria outlined 162 above. The research coordinator will contact each attending physician on the services from which 163 patients will be recruited when the attending physician comes on service to inform the attending of the 164 research project and get the attending physicians' agreement regarding patient recruitment. If the 165 physician does not respond, then the research coordinator will contact the physician once eligible 166 patients have been identified to 1) alert them to their patients eligibility for participation 2) inform them 167 their patients will be recruited for enrollment; and 3) provide them an opportunity to decline or defer 168 any given patients enrollment. Research coordinators will approach potential study participants while 169 they are in their hospital rooms. 170 specify that the questions posed in the study regarding preferences for life-sustaining treatments are 178 hypothetical, but that we would like the patients to approach the questions as if they are answering 179 them in a true clinical context. Following discussion of the study, research coordinators will obtain 180 written consent from patients. The consent forms will contain HIPAA statements of authorization of 181 release of medical records, thus facilitating our collection of data from medical records during the study. 182 The consent form will indicate that patients are being invited to participate in the study because they 183 have one of a series of medical diagnoses. 184

185
The research coordinators will screen electronic medical records of inpatients on the general medicine, 186 pulmonary, cardiology and oncology wards at the Hospital of the University of Pennsylvania for 187 eligibility. Patients' eligibility status will be entered into the eligibility database. We will record ICD9 and 188 ICD 10 codes. Eligible patients will be approached by a research coordinator in the inpatient setting who 189 will seek patients consent to participate in a study exploring patient decision making about life-190 sustaining medical treatments. The research coordinator will specify that the questions are hypothetical 191 and that the answers are not binding, but emphasize that we would like patients to answer the 192 questions as if they were real. Following discussion of the study, research coordinators will obtain 193 written consent from patients. The consent forms will contain HIPAA statements of authorization of 194 release of medical records, thus facilitating our collection of data from medical and billing records during 195 the study. The consent includes clear explanations that different types of decision making will be 196 assigned by chance. The research coordinators will explain who will be enrolled, how many patients are 197 being targeted for enrollment, and what the outcomes of interest are. 198 199 Pilot Testing: During pilot testing, we will test several methods for encouraging patients in the intuitive 200 arm to answer the questions based on intuition including: (a) a time constraint condition, under which 201 patients will be asked to make a decision about the four hypothetical questions based on gut instinct 202 and to answer each question within one minute; and (b) a cognitive loading condition, under which 203 patients will be asked to attend to a second task while also reading the hypothetical scenarios and 204 answering the questions about whether they would or would not want the therapy in question. The 205 second task will either be retention in working memory of a string of numbers or a task that involves 206 pressing a button every time a specific number appears on the bottom of the computer screen. We will 207 ask patients in each condition how difficult the task was to verify that the cognitive load group found the 208 task significantly harder than the other groups. We will also administer a test of analytical reasoning to 209 groups in the various pilot arms to establish that time constraint and cognitive load result in lower ability 210 to engage in analytical reasoning that the group without a time constraint or a cognitive load. The tests 211 we will use to assess analytical reasoning include denominator neglect questions 25 , belief bias 212 syllogisms 26 and the Cognitive Reflection Test 28 . We will also pilot test doing the interviews for both 213 arms in a single survey session and in two separate sessions separated by as much as 24 hours. Once we 214 determine during pilot testing the logistics of how we will organize the intuition and deliberation arms, 215 we will submit a modification describing the final proposed technique. 216 217 The questions will be administered as a Qualtrics survey, facilitated by the research coordinator. At this 218 point, we anticipate that patients in each arm will have two facilitated survey sessions with the research 219 coordinator, as follows (although we may do a single session for one or both arms depending on the 220 results of the pilot testing). 221 222 Intuitive arm: After patient consent is obtained, the research coordinator will administer the 223 hypothetical questions regarding medical interventions, along with the Decision Conflict Scale 224 uncertainty subscale and the short form of the state-trait anxiety scale. At a second visit on the same 225 day, the research coordinator will administer the demographics survey, the questions about health state 226 valuation, and the Rational Experiential Inventory. 227 228 Deliberative arm: After patient consent is obtained, the research coordinator will give the patient 229 instructions and then give the patient a hard copy of the hypothetical questions regarding specific 230 medical interventions. The patient will be given instructions to deliberate on the questions until the 231 following day. At a second visit the following day, the research coordinator will return and collect the 232 patient's answers to the intervention questions. The research coordinator will then administer the 233 Decision Conflict Scale uncertainty subscale, the short form of the state-trait anxiety scale, the 234 demographics survey, the questions about health state valuation, and the Rational Experiential 235 Inventory. 236

Administration of surveys and/or process 237
The research coordinator will screen electronic medical records of inpatients on the general medicine, 238 pulmonary, cardiology and oncology wards at the Hospital of the University of Pennsylvania for 239 eligibility. Patients' eligibility status will be entered into the eligibility database. We will record ICD9 and 240 ICD 10 codes. The research coordinator will contact each attending physician on the services from which 241 patients will be recruited when the attending physician comes on service to inform the attending of the 242 research project and get the attending physician's agreement regarding patient recruitment. If the 243 physician does not respond, then the research coordinator will contact the physician once eligible 244 patients have been identified to 1) alert them to their patients eligibility for participation 2) inform them 245 their patients will be recruited for enrollment; and 3) provide them an opportunity to decline or defer 246 any given patients enrollment. Eligible patients will be approached by a research coordinator in the 247 inpatient setting. The research coordinator will seek patients consent to participate in a study exploring 248 patient decision making about life-sustaining medical treatments. The research coordinator will specify 249 that the questions are hypothetical and that the answers are not binding, but emphasize that we would 250 like patients to take the questions seriously and give their best answer as if they were being asked in a 251 clinical context. Following discussion of the study, the research coordinator will obtain written consent 252 from the patient. The consent forms will contain HIPAA statements of authorization of release of 253 medical records, thus facilitating our collection of data from medical records during the study. The 254 questions will be administered as a Qualtrics survey, facilitated by the research coordinator. Patients in 255 each arm most likely will have two facilitated survey sessions with the research coordinator, as follows 256 (we will be making some adjustments based on results of pilot testing, described in the Study 257 Procedures section; it is possible that subjects in the intuitive arm will have only one survey session). 258 259 Intuitive arm: After patient consent is obtained, the research coordinator will administer the 260 hypothetical questions regarding medical interventions, along with the Decision Conflict Scale 261 uncertainty subscale and the short form of the state-trait anxiety scale. At a second visit on the same 262 day, the research coordinator will administer the demographics survey, the questions about health state 263 valuation, and the Rational Experiential Inventory. 264

265
Deliberative arm: After patient consent is obtained, the research coordinator will give the patient 266 instructions and then give the patient a hard copy of the hypothetical questions regarding specific 267 medical interventions. The patient will be given instructions to deliberate on the questions until the 268 following day. At a second visit the following day, the research coordinator will return and collect the 269 patient's answers to the intervention questions. The research coordinator will then administer the 270 Decision Conflict Scale uncertainty subscale, the short form of the state-trait anxiety scale, the 271 demographics survey, the questions about health state valuation, and the Rational Experiential 272 Inventory. 273 274 We anticipate that the total time for the two sessions spent administering the survey with each patient 275 will not exceed forty minutes. Subject IDs will be assigned at the point of consent. Subject ID numbers, 276 demographic information, medical record number and group assignments will be entered into the 277 analytic database and into Qualtrics. We will maintain the link between the patient name, medical 278 record number and subject ID only until the interviews with that patient are complete and then the 279 names and medical record numbers will be deleted. 280 Prudent steps will be taken to ensure that all information will be kept confidential and secure, including 283 medical and survey data. When a subject consents, he will be assigned a unique randomly generated 284 study identifier. That unique identifier will be entered into every survey/database that is used. A 285 separate password-protected file linking the patient to his unique identifier will be maintained on a 286 secure server only until the final interview is complete. Then the link will be destroyed. Participants 287

Human research protection
answers to questions will be entered in a secured database maintained by Qualtrics. Qualtrics uses 288 Transport Layer Security (TLS) encryption (also known as HTTPS) for all transmitted data. They also 289 protect surveys with passwords and HTTP referrer checking. Their data is hosted by third party data 290 centers that are SSAE-16 SOC II certified. All data at rest are encrypted, and data on deprecated hard 291 drives are destroyed by U.S. DOD methods and delivered to a third-party data destruction service. 292 Qualtrics deploys the general requirements set forth by many Federal Acts including the FISMA Act of 293 2002. They meet or exceed the minimum requirements as outlined in FIPS Publication 200. 294

Subject confidentiality 295
Steps will be taken to ensure that all information will be kept confidential and secure. Unique patient 296 identifiers numbers will be assigned to each subject locally and kept in a secure encrypted file. All paper 297 records will be kept in locked files; all computers will be password protected and kept in locked rooms; 298 all databases will be password protected and maintained on encrypted hard-drives; none will be stored 299 on stand-alone PCs or laptops. The only link between patient name and unique identifier code will be 300 destroyed as soon as the interviews with that patient are complete. 301 regarding whether there is a living will for that patient on file. 332 4. We relaxed the requirement for a second visit to obtain the health state evaluations to facilitate 333 complete data capture. Patients in the deliberative arm were still required to wait for at least 1 minute 334 prior to making health care choices, but in both arms we allowed complete data capture to occur during 335 the same visit. 336 2 Statistical analytic plan 337 The primary outcome will be the choices patients make whether to accept or decline the life-sustaining 338 interventions. The primary exposure will be whether participants were responding in the intuition or the 339 deliberation arm. Chi-square testing will be used to compare these arms. In secondary analyses, we will 340 evaluate congruence between stated valuations of health states and treatment acceptance or refusal, 341 which will be intended as a proxy for decision quality. We will use a logistic regression model to 342 determine how valuations of relevant health states influence the refusal or acceptance of treatments; 343 the resulting odds ratio will represent the incremental odds of refusing treatment with one unit increase 344 on the valuation scale. Because we are interested in the relationship between the scale and treatment 345 refusal across groups defined by randomization, we will additionally explore an interaction between 346 scale and randomization group. Finally, we will analyze the difference between the means of the two 347 randomization groups on the state anxiety scale and the uncertainty subscale of the Decisional Conflict 348 Scale. These data will be described using means and t-tests if normally distributed or medians and 349 Wilcoxon rank sum tests if not normally distributed.