CARE-RAG: Clinically Aligned Risk-Calibrated Evidence-Aware Retrieval-Augmented Generation for Safe Medical Question Answering

The first future direction is empirical implementation. The framework should be implemented with open models and public datasets, then compared against strong baselines. The initial target should be claim-level support and risk-coverage performance rather than raw accuracy alone. The second direction is clinical entailment modeling. General natural language inference models often miss domain-specific qualifiers. A dedicated medical support verifier should learn population, intervention, comparator, outcome, time, and jurisdiction distinctions. This verifier could become a reusable component for medical RAG research. The third direction is multimodal extension. Medical AI increasingly uses images, waveforms, and structured records. CARE-RAG can be extended by adding image-derived evidence nodes, structured lab nodes, and multimodal claim support. The same answerability principle applies, but the evidence graph becomes richer. The fourth direction is prospective workflow study. After benchmark validation, CARE-RAG should be tested in retrospective clinical workflows and then carefully in prospective settings with oversight. Outcome measures should include safety, efficiency, user trust, understanding of limitations, and effects on documentation quality. The fifth direction is policy alignment. Different institutions and jurisdictions may set different risk thresholds. CARE-RAG should support configurable policy profiles while preserving transparent reporting. A system used in an emergency department may require stricter thresholds than a system used for general education. The final direction is continuous monitoring. Medical knowledge changes, model behavior changes, and user behavior changes. The long-term research challenge is not only building a safe model once, but maintaining calibrated safety over time. This requires monitoring, incident review, recalibration, and accountable change control.

[1] P. Lewis, E. Perez, A. Piktus, F. Petroni, V. Karpukhin, N. Goyal, H. Kuettler, M. Lewis, W. Yih, T. Rocktaeschel, S. Riedel, and D. Kiela, 'Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks,' Advances in Neural Information Processing Systems, 2020. [2] A. Asai, Z. Wu, Y. Wang, A. Sil, and H. Hajishirzi, 'Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection,' International Conference on Learning Representations, 2024. [3] S. Li, S. Park, I. Lee, and O. Bastani, 'TRAQ: Trustworthy Retrieval Augmented Question Answering via Conformal Prediction,' Proceedings of NAACL-HLT, pp. 3799-3821, 2024. [4] R. Yang, Y. Ning, E. Keppo, M. Liu, C. Hong, D. S. Bitterman, J. C. L. Ong, D. S. W. Ting, and N. Liu, 'Retrieval-augmented generation for generative artificial intelligence in health care,' npj Health Systems, vol. 2, article 2, 2025. [5] M. Abo El-Enen, S. Saad, and T. Nazmy, 'A survey on retrieval-augmentation generation models for healthcare applications,' Neural Computing and Applications, vol. 37, pp. 28191-28267, 2025. [6] K. Wu, E. Wu, K. Wei, A. Zhang, A. Casasola, T. Nguyen, S. Riantawan, P. Shi, D. Ho, and J. Zou, 'An automated framework for assessing how well LLMs cite relevant medical references,' Nature Communications, vol. 16, article 3615, 2025. [7] World Health Organization, Ethics and governance of artificial intelligence for health: guidance on large multi-modal models, Geneva: WHO, 2024. [8] C. Autio, R. Schwartz, J. Dunietz, S. Jain, M. Stanley, E. Tabassi, P. Hall, and K. Roberts, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile, NIST AI 600-1, National Institute of Standards and Technology, 2024. [9] Q. Jin, B. Dhingra, Z. Liu, W. W. Cohen, and X. Lu, 'PubMedQA: A Dataset for Biomedical Research Question Answering,' Proceedings of EMNLP-IJCNLP, pp. 2567-2577, 2019. [10] G. Tsatsaronis, G. Balikas, P. Malakasiotis, I. Partalas, M. Zschunke, M. R. Alvers, D. Weissenborn, A. Krithara, S. Petridis, D. Polychronopoulos, et al., 'An overview of the BioASQ large-scale biomedical semantic indexing and question answering competition,' BMC Bioinformatics, vol. 16, article 138, 2015.