Dr Mark Howard1
1Monash University
Current evidence for the application of machine learning-based clinical decision support (MLCDS) for treatment planning in newly-diagnosed epilepsy is insufficient to establish the downstream benefits to patients. While MLCDS has demonstrated good analytic accuracy and could provide crucial outputs for predicting which initial antiseizure medications (ASM) are most likely to result in seizure control, stakeholders’ acceptance of MLCDS has been poorly studied: the latter is critical as beyond accuracy, securing downstream therapeutic benefits from MLCDS is dependent on how the outputs of the model are accepted by physicians and patients. In response to this problem, we conducted a qualitative study utilising semi-structured interviews to examine neurologists’ and patients’ attitudes, values, beliefs, and perceptions regarding the use of MLCDS tools in the newly-diagnosed epilepsy clinical care paradigm.
Results: Neurologists and people with epilepsy (PWE) saw the potential usefulness of MLCDS for guiding initial ASM choice, but willingness to accept the recommendations depended on key factors such as: quality and completeness of, and access to, information input into the MLCDS; interpretability and justification of the clinical decision; and system status (locked or dynamic). PWE reported they would trust their neurologists’ use of MLCDS, but the “tool” should not compromise shared clinician-patient decision-making. Finally, responsibility and legal ramifications were important considerations.
Conclusions: To secure therapeutic benefits for PWE, MLCDS must be analysed as a socio-technical system and, minimally, onboarding should respond to the interruption of patient-clinician relationships, and account for factors such as MLCDS-clinician disagreement, interpretability, moral and legal responsibility, and trust.
Biography:
Dr Mark Howard is an Early Career Researcher and Research Fellow with the Philosophy Department at Monash University. He writes on technology and society, and his current research program, encompassing both theoretical and empirical ethics, focuses on the social impacts of automating technology in healthcare diagnostics and treatment planning.