Journal Logo

Invited Foreword

A Call for a Patient Preference Predictor

Wendler, David PhD

Author Information
doi: 10.1097/CCM.0000000000004949
  • Free
  • Editor's Choice


Discussions with families and loved ones regarding the goals of care at the end of life for decisionally incapacitated patients should be guided by two primary considerations: 1) the chances the available treatments might benefit the patient and 2) the patient’s preferences regarding the treatments they receive. Unfortunately, both factors are frequently unknown, often resulting in expensive and excessive care, treatment burdens on patients, stress on families, and moral distress for caregivers.

In recent years, methods involving artificial intelligence have been developed with the goal of making more precise and accurate predictions regarding the chances available treatments will benefit the patient. Although some progress has been made in this regard, accurate prognosis at the end of life remains challenging and, in far too many cases, effectively impossible (1). In this setting, the process of engaging with patients’ surrogates to develop a treatment plan is critical (1).

Ethical and legal standards emphasize the importance of providing patient-centered care, even for patients who have lost the ability to make their own decisions. Current practice thus relies on the substituted judgment standard, which directs surrogates to try to make the treatment decisions the patient would have made for themselves. Unfortunately, in many cases, especially at the end of life, it is unclear what is medically best for the patient, and their surrogate is uncertain as to which treatment course the patient would have chosen for themselves.

Recent proposals suggest that artificial intelligence might offer a means to providing more accurate predictions of patients’ treatment preferences. If they do, this approach would increase the extent to which decisionally incapacitated patients are treated consist with their preferences and might also reduce the burdens on their surrogates. The need for and promise of this approach is illustrated by the now 25-year-old case of a patient with an aggressive brain tumor.


The patient, still in his 20s at the time, was already wildly successful in his field. After dating for years, he and his high school girlfriend had recently married. Soon after, he developed an aggressive brain tumor which left him unable to make his own decisions. To that extent, he was not alone. A significant percentage of hospitalized adults are unable to make their own treatment decisions (2).

Surrogate decision-makers, in this case his new wife, are called on to make decisions for these patients. After discussing the case with her, the clinicians called a bioethics consultation. She was unable to make a decision for her husband, and the clinicians were unsure how to proceed. As the clinicians had explained, there were three options. They could try standard treatment, which offered some chance for a few months extension in life expectancy. They could focus on palliation for her husband’s symptoms and hospice care when needed which, especially at the time, had an unknown impact on life expectancy. Or, she could enroll her husband in a phase 1 trial testing an experimental treatment which, to that point, had been given to four patients.

Surrogates frequently face momentous decisions: the patient’s life is at stake, and it is unclear which course of treatment is clinically best. For example, it is estimated that 70% of treatment decisions near the end of life are made by someone other than the patient (3). As a recent fellow in bioethics, I followed the standard protocol, hoping it might help. I started out by describing advance directives and asking whether her husband had filled one out. Or had they discussed the kinds of treatment her husband would want if he became unable to make decisions? She explained that they had spent their time discussing the beginning of their new life together, not the end. That part of the story is common as well. Most patients do not complete formal advance directives or otherwise discuss their treatment preferences with family or loved ones (4).


When patients do not leave clear instructions regarding their care, and it is unclear what would be medically best for them, ethical norms, as well as clinical and legal standards, endorse substituted judgment. Current practice thus directs surrogates to try to make the treatment decisions the patient would have made for themselves. With this in mind, and in an attempt to relieve some of the burden the patient’s wife was experiencing, I explained that we were not asking what she wanted for her husband. We were asking what decision she thought her husband would make for himself. She looked up, started crying, and explained that this was the problem. She had no idea which of the three options he would choose for himself. This too is not surprising. Surrogates, even those who have known the patient for decades, frequently do not know their treatment preferences (5). Furthermore, studies find that prior discussions between patient and surrogate do not increase surrogates’ ability to predict the patient’s treatment preferences (6).

We then talked about her husband generally, what he was like, what he enjoyed, what he valued, looking for evidence of what he would want us to do. But, I could not see anything and, after another 10 minutes of what seemed increasingly like torture, neither could she. A study in four ICUs, published 20 years later, found that the recommended process I was following, support the surrogate by clarifying the patient’s prognosis and exploring his preferences, can increase the emotional burden on their surrogates (7).

This was essentially all that standard practice offered for assisting surrogates in making treatment decisions for incapacitated patients. Despite numerous attempts to improve it, standard practice has not changed significantly since then. Like countless other cases, it had not helped to clarify what decision the patient’s wife should make. And, by the end of the process, she was experiencing significant distress. I was too. One study found that a third of surrogates charged with making medical decisions for loved ones in ICUs develop symptoms associated with posttraumatic stress disorder, a number that increases to nearly 82% for surrogates who make end-of-life decisions (8).


In light of the significant challenges, some commentators argue that treatment decision-making for decisionally incapacitated patients should focus less on the patient and more on their family and loved ones (9). On this approach, we might have considered which of the three options the patient’s wife preferred, or which option would have been the least stressful for her. Although this makes sense, and the impact on the surrogate and loved ones is critical, making treatment decision-making less patient-centered raises several challenges.

First, this approach seems inconsistent with current ethical and legal standards, which emphasize treating patients based on their own preferences. Second, this practice might undermine clinical care generally. How would patients respond to knowing that their preferences will determine how they are treated only as long as they retain decisional capacity?

Third, although this approach is intended to benefit surrogates, it may be counterproductive. Basing decisions on what surrogates want for the patient, rather than what patients want for themselves, may significantly increase the extent to which surrogates feel responsible for patient outcomes. When the available options all have significant drawbacks, as they often do at the end of life (e.g., continued intubation or withdrawal of aggressive support), the result might be trauma for the surrogate no matter which decision they make.

Fourth, in many cases, this approach is unlikely to make decision-making any easier. Like many surrogates, the wife’s strong preference was to make the decision the patient would have made for himself. What we needed was some way to predict what that was.

Since then, a new method, a patient preference predictor (PPP), has been proposed to do just that (5). The PPP is based on the simple idea that patients’ treatment preferences are frequently correlated with characteristics of them and their lives (10). A computer-based PPP would take these characteristics as inputs and output a prediction of which treatment the patient would want in the circumstances. This prediction might thus increase the extent to which decisionally incapacitated patients receive the treatments they want and avoid the treatments they do not want. For example, when surrogates are unsure what decision they should make, the clinicians could inform them of the PPP’s prediction of which option the patient would choose for themselves. This approach might increase the extent to which surrogates experience the process of making decisions as implementing the patient’s preferences for them rather than making decisions on their behalf.

A very preliminary PPP was found to be as accurate as patient-designated and next-of-kin surrogates (11). Because this version did not include many factors, such as age, which are known to be correlated with patients’ treatment preferences, an improved version, one which incorporates more characteristics of the patient, possibly supplemented by information from their electronic health record and online profile (12), might be more accurate than surrogates alone. If it is, use of a PPP could increase the chances patients are treated consistent with their preferences and values. Furthermore, as in the present case, not knowing which treatment the patient would want can be a significant source of distress. Hence, if the use of a PPP increases surrogates’ predictive accuracy, it might reduce their stress as well.

A survey of almost 1,200 U.S. patients found that, if a PPP increases surrogates’ predictive accuracy, the majority would want it to be included in the process of making treatment decisions for them in the event of decisional incapacity (13). Of course, incorporating a PPP might not increase predictive accuracy. It might just be too hard, or essentially impossible, to predict which treatments decisionally incapacitated patients would choose for themselves (14). Or, for reasons we cannot foresee, incorporating a PPP might not be workable in practice. But, given the importance of respecting patients and minimizing the burdens on their loved ones, it is time to develop a full-fledged PPP and find out. We owe it to our patients and their surrogates.


I thank my collaborators, Sumeeta Varma, David Shalowitz, Annette Rid, Bob Wesley, Dana Howard, and Allan Rivlin and many colleagues who have provided constructive suggestions, especially Scott Kim, Marion Danis, and Christine Grady.


1. Ryon DLS. “Rime” after 25 years: Reflections on medical futility from an ancient intensivist. Crit Care Med. 2021; 49:1–6
2. Torke AM, Sachs GA, Helft PR, et al. Scope and outcomes of surrogate decision making among hospitalized older adults. JAMA Intern Med. 2014; 174:370–377
3. Silveira MJ, Kim SY, Langa KM. Advance directives and outcomes of surrogate decision making before death. N Engl J Med. 2010; 362:1211–1218
4. Coppola KM, Ditto PH, Danks JH, et al. Accuracy of primary care and hospital-based physicians’ predictions of elderly outpatients’ treatment preferences with and without advance directives. Arch Intern Med. 2001; 161:431–440
5. Rid A, Wendler D. Use of a patient preference predictor to help make medical decisions for incapacitated patients. J Med Philos. 2014; 39:104–129
6. Bravo G, Trottier L, Dubois MF, et al. Does promoting research advance planning in a general elderly population enhance completion of a research directive and proxies’ predictive ability? A randomized controlled trial. AJOB Emp Bioethics. 2016; 7:183–192
7. Carson SS, Cox CE, Wallenstein S, et al. Effect of palliative care-led meetings for families of patients with chronic critical illness: A randomized clinical trial. JAMA. 2016; 316:51–62
8. Vig EK, Starks H, Taylor JS, et al. Surviving surrogate decision-making: What helps and hampers the experience of making medical decisions for others. J Gen Intern Med. 2007; 22:1274–1279
9. Berger JT, DeRenzo EG, Schwartz J. Surrogate decision making: Reconciling ethical theory and clinical practice. Ann Intern Med. 2008; 149:48–53
10. Lattie EG, Asvat Y, Shivpuri S, et al. Associations between personality and end-of-life care preferences among men with prostate cancer: A clustering approach. J Pain Symptom Manage. 2016; 51:52–59
11. Shalowitz DI, Garrett-Mayer E, Wendler D. How should treatment decisions be made for incapacitated patients, and why? PLoS Med. 2007; 4:e35
12. Lamanna C, Byrne L. Should artificial intelligence augment medical decision making? The case for an autonomy algorithm. AMA J Ethics. 2018; 20:E902–E910
13. Wendler D, Wesley B, Pavlick M, et al. A new method for making treatment decisions for incapacitated patients: What do patients think about the use of a patient preference predictor? J Med Ethics. 2016; 42:235–241
14. Kim SY. Improving medical decisions for incapacitated persons: Does focusing on “accurate predictions” lead to an inaccurate picture? J Med Philos. 2014; 39:187–195

capacity; surrogate; treatment preferences

Copyright © 2021 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.