Jane Smith

BACKGROUND Discharge summaries are central to safe transitions of care but are frequently incomplete, delayed, or omitted, particularly in resource-constrained junior doctor workflows. Generative artificial intelligence (AI) tools may support clinical documentation, but their quality, safety, and timeliness in this context require evaluation in Australian settings. OBJECTIVE To assess whether generative AI assistance improved the quality and timeliness of hospital discharge summaries produced by junior doctors compared with human-only drafting. METHODS We conducted a blinded comparative assessment study in which Australian junior doctors produced discharge summaries from de-identified hospital admission records (MIMIC-IV). Three junior doctors used an AI-assisted workflow (Heidi Health) and two drafted summaries without AI assistance. Completed summaries were de-identified and randomly allocated to seven independent specialist reviewers. The primary outcome was overall discharge summary quality assessed using a 20-item rubric (scored 0–100). Secondary outcomes included the Physician Documentation Quality Instrument 9-item (PDQI-9), self-reported completion time, and time to return summaries. Quality outcomes were analysed using linear mixed models with random effects for reviewer and summary. Efficiency outcomes used the Mann–Whitney U test. A post-hoc safety sub-analysis was performed on a randomly selected subset of 25 summaries. RESULTS Seventy-nine summaries were returned (60 AI-assisted, 19 human-only); 230 reviewer assessments were included. AI-assisted summaries scored higher on the primary rubric outcome (adjusted estimated marginal mean 85.56 vs 65.65; F[1,63.320]=137.848; P<.001; Cohen's d=1.38, 95% CI 1.05 to 1.71) and on the PDQI-9 (39.48 vs 31.00; F[1,56.852]=94.639; P<.001; d=1.29). Median completion time was 13.5 vs 25.0 minutes (P=.023). AI-assisted summaries were returned after a median of 20 days compared with 87 days for human-only summaries (P<.001). AI-assisted summaries scored higher across all 20 rubric items and all nine PDQI-9 domains. The safety sub-analysis identified no hallucinations in either group. Material omissions were significantly lower in AI-assisted summaries (median 0 vs 2; P<.001), as were medication-specific errors (median 1 vs 2; P=.002). Overall safety ratings did not differ significantly between groups. CONCLUSIONS In this exploratory blinded assessment study, AI assistance was associated with substantially higher discharge summary quality scores, faster completion, and more timely return than human-only drafting. A post-hoc safety analysis found significantly fewer material omissions and lower medication error counts in AI-assisted summaries, with no hallucinations identified in either group. Findings support further prospective evaluation of supervised AI-assisted discharge summary workflows in Australian hospitals, with explicit safeguards for clinician verification, factual accuracy, privacy, accountability, and equity. CLINICALTRIAL NA

Readiness for practice (RFP) is a critical outcome for medical education, reflecting a graduate’s ability to transition from student to clinician. Objective Structured Clinical Examinations (OSCEs) play a pivotal role in evaluating RFP. This study aimed to investigate the conceptualisation and operationalisation of RFP within the OSCE context and evaluate the contribution of a newly developed RFP-focused rating tool compared to traditional OSCE scoring. Utilising a sequential explanatory mixed-methods approach, this study was conducted across five Australian medical schools. Quantitative data from final-year medical students’ OSCE results were analysed to examine correlations between RFP scores and components of a 5-item Process of Decision-Making (PDM) checklist: safety, reliability, trustworthiness, insight, and an organised/logical approach. In the qualitative phase, semi-structured interviews with OSCE examiners were analysed using inductive thematic analysis to explore their perceptions of RFP and decision-making processes. Quantitative analyses of 550 medical students’ OSCE performance datasets revealed that global rating had the strongest correlation with RFP scores, followed by safety and reliability components of the PDM checklist, whereas trustworthiness showed a weaker association. Failing students consistently scored poorly across all checklist components and RFP, with approximately half or fewer meeting individual checklist requirements. Qualitative analysis of nine examiners’ interviews produced a schema of What, How, and Why: What assessors viewed as RFP was the integration of clinical competence, safety, scope awareness, reasoning, and team readiness; How RFP was judged was through behaviours such as supervision-seeking, recognising unwell patients, attitudinal maturity, and structured consultations; Why RFP judgments were used included: to assess socio-relational skills, address workplace limitations, and make decisions, particularly on borderline performances. Readiness for practice is a complex, multi-dimensional construct that includes application of knowledge, technical skills, reliability, professional identity, safety and trustworthiness. While this study does not directly evaluate examiner training, understanding how examiners apply RFP frameworks may inform future efforts to enhance training consistency, fairness, and assessment utility. The online version contains supplementary material available at 10.1186/s12909-026-09126-6.

The importance of social connection dominated and underpinned many other themes identified as central to achieving safe and effective vocational general practitioner education.

Decision-making in clinical assessment, such as exit-level medical school Objective Structured Clinical Examinations (OSCEs), is complex. This study utilized an empirical phenomenological qualitative approach with thematic analysis to explore OSCE assessors' perceptions of the concept of a “prototypical intern” expressed during focus group discussions. Topics discussed included the concept of a prototypical intern, qualities to be assessed, and approaches to clinical assessment decision-making. The thematic analysis was then applied to a theoretical framework (Cultural Historical Activity Theory—CHAT) that explored the complexity of making assessment decisions amidst potentially contradicting pressures from academic and clinical perspectives. Ten Australasian medical schools were involved with 15 experienced and five less experienced assessors participating. Thematic analysis of the data revealed four major themes in relation to how the prototypical intern concept influences clinical assessors' judgements: (a) Suitability of marking rubric based on assessor characteristics and expectations; (b) Competence as final year student vs. performance as a prototypical intern; (c) Safety, trustworthiness and reliability as constructs requiring assessment and (d) Contradictions in decision making process due to assessor differences. These themes mapped well within the interaction between two proposed activity systems in the CHAT model: academic and clinical. More clinically engaged and more experienced assessors tend to fall back on a heuristic, mental construct of a “prototypical intern,” to calibrate judgements, particularly, in difficult situations. Further research is needed to explore whether consensus on desirable intern qualities and their inclusion into OSCE marksheets decreases the cognitive load and increases the validity of assessor decision making.

Background: Objective structured clinical examinations (OSCEs) are commonly used to assess the clinical skills of health professional students. Examiner judgement is one acknowledged source of variation in candidate marks. This paper reports an exploration of examiner decision making to better characterise the cognitive processes and workload associated with making judgements of clinical performance in exit‐level OSCEs. Methods: Fifty‐five examiners for exit‐level OSCEs at five Australian medical schools completed a NASA Task Load Index (TLX) measure of cognitive load and participated in focus group interviews immediately after the OSCE session. Discussions focused on how decisions were made for borderline and clear pass candidates. Interviews were transcribed, coded and thematically analysed. NASA TLX results were quantitatively analysed. Results: Examiners self‐reported higher cognitive workload levels when assessing a borderline candidate in comparison with a clear pass candidate. Further analysis revealed five major themes considered by examiners when marking candidate performance in an OSCE: (a) use of marking criteria as a source of reassurance; (b) difficulty adhering to the marking sheet under certain conditions; (c) demeanour of candidates; (d) patient safety, and (e) calibration using a mental construct of the ' mythical [prototypical] intern'. Examiners demonstrated particularly higher mental demand when assessing borderline compared to clear pass candidates. Conclusions: Examiners demonstrate that judging candidate performance is a complex, cognitively difficult task, particularly when performance is of borderline or lower standard. At programme exit level, examiners intuitively want to rate candidates against a construct of a prototypical graduate when marking criteria appear not to describe both what and how a passing candidate should demonstrate when completing clinical tasks. This construct should be shared, agreed upon and aligned with marking criteria to best guide examiner training and calibration. Achieving this integration may improve the accuracy and consistency of examiner judgements and reduce cognitive workload.

Participants’ impressions were explored qualitatively, and suggest that simulation can encourage reflection on the non-technical skill of resilience, provided there is careful design and debriefing of the simulation activity.

It was the author's (LWH) observation that pterygium was frequently misdiagnosed by general practitioners that led to this study. The aim was to identify the rate of misdiagnosis of pterygium by op...

A perennial challenge of primary care quality improvement is to establish why interventions work in some circumstances, but not others. This study aimed to identify factors explaining variations in the impact on clinical practice of a facilitation led vascular health intervention in Australian family practice. Our mixed methods study was embedded within a cluster randomised controlled trial of a facilitation intervention designed to increase the uptake of evidence-based prevention of vascular disease in family practices. The study was set in four Australian states using eight of the study’s 16 intervention practices. Facilitators worked with intervention practices to develop and implement improvements in preventive care informed by a vascular risk factor audit. We constructed case studies of each practice’s “intervention narrative” from semi-structured interviews with clinicians, facilitators and other staff, practice observation, and document analysis of facilitator diaries. The intervention narratives were combined with pre- and post-intervention audit data to generate typologies of practice responses to the intervention. We found substantial variability between practices in the changes made to vascular risk recording. Context (i.e. practice size), adaptive reserve (i.e. interpersonal relationships, manager and nurse involvement), and occasional data idiosyncrasies interacted to influence this variability. The findings emphasise the importance of tailoring facilitation interventions to practice size, clinician engagement and, critically, the organisation of, and relationships between, the members of the practice team. The trial was registered with the Australian and New Zealand Clinical Trials Registry (ANZCTR): ACTR N12612000578808 (29/5/2012). This trial registration is retrospective as our first patient returned their consent on the 21/5/2012. Patient recruitment was ongoing until 31/10/2012.

Objective: The objective of this collaborative study was to compare current practices of conducting high-stake, exit-level Objective Structured Clinical Examinations (OSCEs) at all Australian medical schools. We aimed to document similarities and differences between schools, and compare existing practice against available gold standard, evidence-based practice. We also aimed to identify areas where gold standards do not currently exist, and could be developed in the future. Methods: A 72-item semi-structured questionnaire was sent to all 19 Australian medical schools with graduating students. Results: A total of 18/19 schools responded. Of these, 16/18 schools had summative exit-level OSCEs representing content from multiple medical specialties. The total number of OSCE stations varied from 8 to 16, with total OSCE testing time ranging from 70 to 160 min. All schools blueprinted their OSCE to their curriculum, and trained simulated patients and examiners. There was variation in the format of marking rubric used. Conclusions: This study has provided insight into the current OSCE practices of the majority of medical schools in Australia. Whilst the comparative data reveal a wide variation in OSCE practices between schools, many recommended “gold standard” OSCE practices are implemented. The collective awareness of our similarities and differences provides us with a baseline platform, as well as an impetus for iterative quality improvement. Such discourse also serves to develop new gold standards in practice where none have previously existed.