As Automatic Speech Recognition (ASR) is increasingly deployed in clinical dialogue, standard evaluations still rely heavily on Word Error Rate (WER). This paper challenges that standard, investigating whether WER or other common metrics correlate with the clinical impact of transcription errors. We establish a gold-standard benchmark by having expert clinicians compare ground-truth utterances to their ASR-generated counterparts, labeling the clinical impact of any discrepancies found in two distinct doctor-patient dialogue datasets. Our analysis reveals that WER and a comprehensive suite of existing metrics correlate poorly with the clinician-assigned risk labels (No, Minimal, or Significant Impact). To bridge this evaluation gap, we introduce an LLM-as-a-Judge, programmatically optimized using GEPA to replicate expert clinical assessment. The optimized judge (Gemini-2.5-Pro) achieves human-comparable performance, obtaining 90% accuracy and a strong Cohen’s kappa of 0.816. This work provides a validated, automated framework for moving ASR evaluation beyond simple textual fidelity to a necessary, scalable assessment of safety in clinical dialogue.

As healthcare services deploy AI to automate patient-facing communication, concerns persist about the interactional work through which empathy is made relevant. We examine empathy not as an internal state but as an interactional accomplishment, asking how patients display orientations to an LLM-powered voice assistant’s turns as (non-)empathic in real clinical telephone calls. Using Conversation Analysis (CA) to analyse post–cataract surgery follow-up calls conducted by AI-powered voice assistant Dora (Ufonia), we compare patient responses across earlier and later system versions.Earlier calls show minimal, delayed, prosodically closed responses to wellbeing enquiries, consistent with treating Dora as a transactional information-gathering device. Later calls more often feature socially rich formats, for example colloquial upgrades, gratitude tokens, occasional return enquiries, and increased turn-final rising intonation, suggesting patients hear Dora’s talk as socially implicative and thus opening space for affiliative/empathetic uptake. We discuss implications for CA-informed conversation design and for evaluating “empathy” via participant orientations in situ rather than post-hoc self-report.

Clinical voice agents are now deployed in routine care, where real patients do not wait their turn: they interrupt. These systems typically use a cascaded architecture (speech-to-text → LLM → text-to-speech), so when a patient cuts the agent off mid utterance, clinically required content can be lost even when the model handles cooperative transcripts well. Yet clinical conversational-AI benchmarks almost universally assume patients wait for the agent to finish, missing interruption-induced loss of required content. We present a transcript-based evaluation of interruption recovery, adapting conversation-analytic overlap categories into three operational types (recognitional, competitive, transitional sub-unit) and testing four deployment-oriented, non-reasoning LLM configurations across four cells spanning history-taking (information gathering) and FAQ (information provision), scored on whether the agent preserves the clinically required content. In the gathering cells, target-question failure varied across models; in the provision cells, where arms are directly comparable, failure rose for every model. Rankings differ across cells, and competitive FAQ interruption produced 30/30 provision-coverage failures for all four models (Wilson 95% CI: 88.6–100.0%; baseline 0/30 for three, 4/30 for Llama). A brief apology marker (“sorry to interrupt”) shifts recovery by tens of percentage points, inconsistently across models, and for one it reduces recovery. Interruption robustness therefore cannot be a single score: evaluation must be content-grounded, reported per cell, and matched to the deployment’s interruption profile.

Large Language Models (LLMs) have shown impressive potential in clinical question answering (QA), with Retrieval Augmented Generation (RAG) emerging as a leading approach for ensuring the factual accuracy of model responses. However, current automated RAG metrics perform poorly in clinical and conversational use cases. Using clinical human evaluations of responses is expensive, unscalable, and not conducive to the continuous iterative development of RAG systems. To address these challenges, we introduce ASTRID - an Automated and Scalable TRIaD for evaluating clinical QA systems leveraging RAG - consisting of three metrics: Context Relevance (CR), Refusal Accuracy (RA), and Conversational Faithfulness (CF). Our novel evaluation metric, CF, is designed to better capture the faithfulness of a model’s response to the knowledge base without penalising conversational elements. Additionally, our metric RA captures the refusal to address questions outside of the system’s scope of practice. To validate our triad, we curate a dataset of over 200 real-world patient questions posed to an LLM-based QA agent during surgical follow-up for cataract surgery - the highest volume operation in the world - augmented with clinician-selected questions for emergency, and clinical and non-clinical out-of-domain scenarios. We demonstrate that CF predicts human ratings of faithfulness more accurately than existing definitions in conversational settings. Furthermore, using eight different LLMs, we demonstrate that the three metrics can closely agree with human evaluations, highlighting the potential of these metrics for use in LLM-driven automated evaluation pipelines. Finally, we show that evaluation using our triad of CF, RA, and CR exhibits alignment with clinician assessment for inappropriate, harmful, or unhelpful responses. We also publish the prompts and datasets for these experiments, providing valuable resources for further research and development.

Recent advances in large language models (LLMs) have generated significant interest in their application across various domains including healthcare. However, there is limited data on their safety and performance in real-world scenarios. This study uses data collected using an autonomous telemedicine clinical assistant. The assistant asks symptom-based questions to elicit patient concerns and allows patients to ask questions about their post-operative recovery. We utilise real-world postoperative questions posed to the assistant by a cohort of 120 patients to examine the safety and appropriateness of responses generated by a recent popular LLM by OpenAI, ChatGPT. We demonstrate that LLMs have the potential to helpfully address routine patient queries following routine surgery. However, important limitations around the safety of today’s models exist which must be considered.