Feature: Human versus machine: Investigating the impact of AI technologies on compassionate care

By Alexandra Burza, MMJC'19

As a Clinical Fellow in Hematology and a Postdoctoral Associate at the Roman Institute of Philosophy, Dr. Ben Chin Yee is uniquely positioned to tackle a burgeoning question in medical ethics: How will the increasing integration of artificial intelligence (AI) technologies impact compassionate and equitable patient care?

Since his AMS Fellowship in Compassion and Artificial Intelligence began in January of this year, he has been working to find an answer to this question.  

Within medicine, AI can be employed to create algorithms that perform diagnostic testing, interpret medical imaging results, or analyze blood films. Currently, the majority of these applications of AI are in domains that are less patient facing; however, there is widespread interest in developing more applications that are integrated into clinical care, that clinicians can use in conjunction with traditional tools in order to diagnose or make therapeutic decisions.

“If a tool, an algorithm, is being used in clinical decision making, there are certain variables that are being considered and there are certain things that are being left out,” Chin-Yee said.

“My concern, which others share, is that is that sometimes what is being left out can be significant and can be considerable, such as ethnic background or socioeconomic status, which might not be captured by the datasets the algorithm employs,” he explained.

His research will explore the impact of new genomic AI technology on the patient-physician relationship, as well as investigate potential implicit biases associated with the use of genomics and AI, within the context of the oncology patient population at London Health Sciences Centre.

“Datasets often systematically exclude particular groups, and are not representative of our diverse patient populations,” he explained.

“For instance, there's some evidence that particular patient populations, especially racialized groups and non-white patients, may be underrepresented in genomic databases. Particular patient populations might have genomic variants that are less characterized or of uncertain significance. That can impact the ability to prognosticate on the basis of those biomarkers.”

Using machine learning models as well as traditional statistical techniques, this research will develop predictive algorithms to measure outcomes among the local cancer patient population.

By comparing the accuracy of algorithms informed by different data sets: hospital admissions, complications related to treatment, socio-economic status or other socio-demographic factors, determinations can be made as to the efficacy of using genomic biomarkers to predict outcomes across a diverse patient population. Ultimately, these findings can help inform how AI tools can be developed and implemented in order to support equitable patient care in the future.

Genetic biomarkers are readily used for prediction of clinical outcomes in oncology already. There are also considerable disparities in cancer outcomes between patients from different socio-economic backgrounds. Chin-Yee says the intersection between strong technological innovation and the need for contextualized and individualized patient care makes this type of research especially important in oncology.

“We need to be really reflective about how we integrate these technologies into clinical practice and research, which includes attending to questions like: who benefits from the application of these technologies? Who is being left out?”

Chin-Yee says he recognizes that the excitement around AI innovation is well deserved, as new technology allows for more precise diagnostic tools and may enable personalized treatments for patients. However, he emphasizes the importance of the humanistic aspect of health care delivery for gaining the context and nuance necessary for effective individualized and compassionate care.

“We all want to have more accurate predictive tools; better means of determining what the best treatment is going to be for a given individual. Though as clinicians and researchers, we need to be aware that these new technologies are not going to be a cure-all for all the problems that currently exist in our health care system.”

This research is part of a collaborative project being undertaken with Dr. Alejandro Lazo-Langner, Associate Professor of Medicine, Oncology, and Epidemiology and Biostatistics, and Dr. Bekim Sadikovic, Professor of Pathology and Laboratory Medicine and Director of the Verspeeten Clinical Genome Centre.