Beyond Medical Malpractice: AI, Clinical Decision-Making, and the Transformation of Health Law

Abstract

Artificial intelligence (AI) is transforming clinical decision making, and has seriously undermined the traditional medical-malpractice regimes that place the lion's share of blame on doctors. This study explores the implications of clinical decisions with AI: who is liable, informed consent, standards of care, explainability, patient rights, algorithmic fairness and institutional accountability. The Pile of Law data set is filtered to include only domain-specific health law documents, and then augmented by relevant regulatory and policy documents to create a domain-specific Health Law and AI corpus. A hybrid approach of legal-NLP combines TF-IDF to define influential legal terms, named entity recognition to extract legal and institutional actors, semantic relationship analysis using Sentence-BERT and analysis of the underlying legal themes using BERTopic, which is subsequently validated in a doctrinal way. Analysis reveals eight key themes, one of which is liability and accountability, which is the most predominant (24.62%), followed by transparency and explainability (18.74%), patient rights and informed consent (15.48%), algorithmic fairness (12.83%) and institutional governance (11.67%). The results for coherence score of BERTopic is 0.71 and topic diversity is 0.86, whereas for NER, they are 90.58% F1. Themes from human validation agree with 89.30% of those from NLP. The results show that there is a lack of a cohesive or holistic approach to responsibility as it cuts across different clinicians, healthcare institutions and technology service providers. The study adds an NLP-based layered accountability system for the governance of an AI medical decision support system, in addition to the existing malpractice law.