It takes a village to build a good algorithm – particularly in a field as sensitive as patient preferences

By Nikola Biller-Andorno, Andrea Ferrario, Sophie Gloeckler

Recently, there has been a lot of talk about how artificial intelligence (AI) is going to boost personalized medicine. And, indeed, the field is developing with amazing speed: Digital twins help predict treatment outcomes based on genomic data, AIs can automatically classify lesions from images of the skin or identify early stages of diabetes retinopathy from images of the rear of an eye (fundus).

Patients’ personal values and preferences are typically not part of the algorithmic equation, and they need not be, as long as patients are able to actively engage in treatment choices. But when patients are unable to make decisions for themselves, the lack of knowledge about their preferences becomes a problem in a patient-oriented healthcare setting.

Advance directives (ADs) have tried to approach this issue, and new digital tools such as the Advance Care Compass are becoming increasingly available that can address some of the shortcomings of traditional ADs. Not everyone, however, has an AD and/or surrogate decision-maker who can reliably convey what the patient would have wanted.

This is the use case we make in our feature article “The Ethics of the Algorithmic Prediction of Goal of Care Preferences: From Theory to Practice”. Building on early conceptual work on algorithm-aided prediction of patient preferences and on a qualitative study on health professionals’ attitudes towards a use of AI towards these purposes, we suggest a socio-technical system approach to develop, implement, and evaluate an outcome-specific, real-time prediction of individual patient goal of care preferences in an intensive care setting.

This approach will require a “village” working together: We need clinicians, data scientists, ethicists, computer scientists, designers, patient and people who have experience as surrogate decision-makers or legal representatives. This community is currently coming together in Zurich: The “Mind the Patient Lab” brings together clinicians from two University Hospitals and scholars from ethics, computational linguistics and informatics. The Digital Health Design Living Lab enhances this collaboration by adding different design disciplines, a range of practice partners and providing the possibility to perform extensive user experience testing in a real world setting.

Together, we are striving to build algorithm-based tools that improves patient-oriented, personalized health care while attending closely to the ethical implications of such work. Stay tuned!


Paper title: The Ethics of the Algorithmic Prediction of Goal of Care Preferences: From Theory to Practice          FREE ARTICLE

Author(s): Andrea Ferrario1,2 Sophie Gloeckler3 Nikola Biller-Andorno3

Affiliations: 1) ETH Zurich, 2) Mobiliar Lab for Analytics at ETH, 3) Institute of Biomedical Ethics and History of Medicine (IBME), University of Zurich

Competing interests: None declared.

Social media accounts of post author(s):

(Visited 271 times, 1 visits today)