How Can We Ensure Use of Artificial Intelligence in Pediatric Medicine Is Ethical?

How Can We Ensure Use of Artificial Intelligence in Pediatric Medicine Is Ethical? 915 341 Jessica Nye, PhD

Team diversity, data contextual factors, and transparency are key factors for the ethical use of artificial intelligence (AI) in pediatric medicine.

AI has been increasingly used in medicine to help clinicians predict patient risk and make treatment decisions. A discussion about the ethical translation of AI into pediatric medicine was published as a comment in The Lancet Child & Adolescent Health.

Emre Sezgin, PhD, principal investigator at the Center of Biobehavioral Health in the Abigail Wexner Research Institute at Nationwide Children’s Hospital and assistant professor of Pediatrics at The Ohio State University College of Medicine and his colleagues explained that the first important consideration is the AI implementation team. An ethnically diverse and trained team of pediatric health care providers, AI developers, researchers and community stakeholders is essential such that accurate interpretation of an algorithm output is made.

In addition, input data into any AI algorithm should be sourced from a diverse population that reflects real-world patient populations to ensure output accuracy and equitable care.

“We have made huge progress with AI in recent years. Now, we are looking at the data and seeing that some tools supported by AI generate biased results based on patient race or ethnicity. It is unfortunate, but as with all of medical research, we don’t have the data that reflects real-world populations. So, it is not a problem with AI solely, but reflects a lack of inclusivity within the health system,” says Dr. Sezgin.

One way to begin to address the lack of diversity in datasets used to train AI algorithms is to ensure the team developing these tools are themselves inclusive, as inherent biases may be propagated by the AI developers.

“One of the unique factors about collecting data for use in AI algorithms in the field of pediatrics is the nature of the data. Medical records for pediatrics are different from adult medical records since there are more actors or informants in these records (parents, legal guardians, siblings and relatives) than you may see in an adult medical record. This makes pediatric medical records unique and challenging for AI solutions because there are multiple sources for data combined into a single note. In addition, there could be multiple providers (for example, the care team for children with medical complexities),” says Dr. Sezgin.

However, multiple aspects of input data should also be considered. For instance, as Dr. Sezgin and his colleagues discussed in their comment, data collected in juvenile detention facilities not being integrated with larger datasets may decrease the representativeness of at-risk populations. Similarly, data collected during infancy may not be relevant as the child is getting older (e.g., infancy to the adolescence). It is for these reasons that context of the data and transparency about collection and reporting methods is imperative for AI in pediatric medicine.

Clinicians should also have a realistic outlook about what AI models can and cannot do in practice. AI models will not replace clinicians, but these tools could be viewed as a second opinion.

“AI will not replace any job in the medical domain. My idea is it will promote and boost the decision-making process. Eventually, I think AI tools have high potential to improve efficiency and effectiveness of clinical decision-making process,” concludes Dr. Sezgin.

 

Reference:

Boch S, Sezgin E, Linwood SL. Ethical artificial intelligence in paediatrics. Lancet Child & Adolescent Health. 2022;S2352-4642(22)00243-7.

 

 

About the author

Jessica Nye, PhD, is a freelance science and medical writer based in Barcelona, Spain. She completed her BS in biology and chemistry and MS in evolutionary biology at Florida State University. Dr. Nye studied population genetics for her doctorate in biomedicine at University of Pompeu Fabra. She conducted her postdoctoral research on the inheritance of complex traits at the Autonomous University of Barcelona.