Context Hyperglycemia in hospital inpatients without pre-existing diabetes is associated with increased mortality. However, the independent contribution of hyperglycemia to healthcare-associated infection (HAI), acute kidney injury (AKI), and stroke is unclear. Objective To investigate the relationship between hyperglycemia and adverse clinical outcomes in hospital for patients with and without diabetes. Design Diabetes IN-hospital: Glucose and Outcomes (DINGO) was a 26-week (October 2019 – March 2020) prospective cohort study. Clinical and glucose data were collected up to the 14^th day of admission. Primary stratification was by hyperglycemia, defined as ≥2 random capillary blood glucose (BG) measurements ≥11.1 mmol/L (≥200 mg/dL). Propensity weighting for nine clinical characteristics, was performed to allow interrogation of causality. To maintain the positivity assumption, patients with HbA1c > 12.0% were excluded and pre-hospital treatment not adjusted for. Setting The Royal Melbourne Hospital, a quaternary referral hospital in Melbourne, Australia. Patients Admissions with at least two capillary glucose values and length of stay >24 hours were eligible, with half randomly sampled. Outcome measures HAI, AKI, stroke, and mortality. Results Of 2,558 included admissions, 1,147 (45%) experienced hyperglycemia in hospital. Following propensity-weighting and adjustment, hyperglycemia in hospital was found to, independently of nine covariables, contribute an increased risk of in-hospital HAI (130 [11.3%] vs.100 [7.1%], adjusted odds ratio [aOR] 1.03, 95% confidence interval [95%CI] 1.01-1.05, p = 0.003), AKI (120 [10.5%] vs. 59 [4.2%], aOR 1.07, 95%CI 1.05-1.09, p < 0.001), and stroke (10 [0.9%] vs. 1 [0.1%], aOR 1.05, 95%CI 1.04-1.06, p < 0.001). Conclusions In hospital inpatients (HbA1c ≤ 12.0%), irrespective of diabetes status and pre-hospital glycaemia, hyperglycemia increases the risk of in-hospital HAI, AKI, and stroke compared with those not experiencing hyperglycemia.

The number of inherited metabolic diseases (IMDs) in newborn screening (NBS) programs has increased significantly in the past decades. For some of the IMDs included in NBS (e.g., tyrosinemia type I), there are clear and substantial health benefits of NBS, while for others (e.g., very long chain acyl CoA dehydrogenase deficiency and 3‐methylcrotonyl CoA carboxylase 1 deficiency), this is less clear as NBS identifies individuals who are asymptomatic or have milder forms of the disease. Therefore, knowledge of the full disease spectrum (including later onset forms) is needed when setting diagnostic metabolite cut‐offs for NBS. Insights into the clinical, genetic and biochemical characteristics of different patient subsets can be used to redefine NBS protocols to identify patients with more severe forms of the disease who are most likely to benefit from identification in the newborn period. These insights require life‐long monitoring of individuals identified based on symptoms versus those identified by NBS to determine long‐term health outcomes and quantify the benefits of NBS. Adult metabolic specialists should be included in the development of NBS programs to provide data from this long‐term monitoring and to contribute specific knowledge about later onset phenotypes of the IMDs included in NBS programs. The goal should be to develop NBS programs that identify newborns that benefit from early disease detection and treatment, without increasing psychological, social and management burden for individuals who may develop disease in adulthood with milder phenotype or potentially even not at all.

Objective: Machine learning (ML) models have emerged as tools to predict length of stay (LOS) and disposition decision (DD) in emergency departments (EDs) to combat overcrowding. However, site-specific ML models are not transferable to different sites. Our objective was to develop interpretable ML models to predict LOS and DD at specific time points, all while establishing a transparent data analysis framework. This framework was designed to be easily adapted by other institutions for the development of their own ML models.Methods: We analysed data from 297 392 ED visits of patients aged 18 and above at a quaternary hospital between 30 June 2019 and 31 December 2022. Eight ML algorithms were evaluated, and ultimately, twelve lasso models built from 21 features were trained to predict four outcomes of LOS and DD at three time points post-triage. Hold-out testing and cross-validation were conducted for these models.Results: The area under the curve values were 0.862/0.868/0.878 for binary LOS predictions at 10, 60 and 120-minute time points and 0.839/0.851/0.863 for binary DD predictions. The accuracies were 60.2%/60.7%/61.9% for ternary LOS predictions and 61.5%/62.3%/63.4% for ternary DD predictions.Conclusions: Interpretable ML models demonstrated outstanding performances in predicting both LOS and DD. The transparent data analysis framework can be easily adapted by other institutions.

Background: Research exploring perspectives on inpatient portals reports that patients desire the information affordances of inpatient portals, and clinicians recognize their value for improving patient experience but also express caution regarding sharing aspects of the medical record. This study contributed to the existing literature on inpatient portals by considering the psychosocial dimension of clinician resistance to information sharing with inpatients and the power dynamic associated with clinician-patient information asymmetry. Along with the information affordances commonly discussed in this area, this study explored perspectives on the novel option to audio record consultations via an inpatient portal. Objective: This study aims to understand patient, proxy, and clinician perspectives on the value and impact of an inpatient portal within the Australian context. It explores clinician resistance and receptivity to sharing aspects of the medical record with patients and the power dynamic that characterizes the relationship between clinician and patient. It considers how an inpatient portal might assist in the transformation of this relationship such that this relationship could be characterized by greater information symmetry. Methods: Interviews were conducted with patients (n=20), proxies (n=4), and clinicians (n=21) recruited from 3 areas within the Royal Melbourne Hospital, where the portal would later be implemented. A largely inductive reflexive thematic analysis was conducted. Results: Patient and proxy participants reported that they wanted to understand what is happening in their care for peace of mind and that an inpatient portal could support this understanding. Clinician participants reflected on how they might transform their information-sharing practice to provide greater transparency in their relationship with patients. Participants considered the types of information that could be shared and how this information could be shared via an inpatient portal. Four key themes were generated: (1) affording the patient and proxy awareness, control, and reassurance through sharing accessible and meaningful information; (2) protecting the clinician and safeguarding quality health care in information sharing; (3) flexibly deploying the functions depending upon clinician, patient, proxy, and context; and (4) moving toward person-centered care: empowerment and equity via an inpatient portal. Conclusions: An inpatient portal provides an opportunity to reconceptualize the medical record and how this information might be shared with patients while they are admitted to the hospital, such that they have more understanding as to what is happening in their care, which ultimately supports their well-being. The transition to a more transparent information-sharing culture in the Australian hospital context will take time. An inpatient portal is a critical step in facilitating this transition and creating more informational symmetry in the clinician-patient relationship.

Background: A multidisciplinary approach is required for the management of long COVID. The aim of this study was to determine the feasibility (demand, implementation, practicality, acceptability, and limited efficacy) of an allied-health-led multidisciplinary symptom management service (ReCOV) for long COVID. Methods: A single-group observational cohort feasibility study was conducted to determine demand (referrals), acceptability (survey), implementation (waitlist times, health professions seen), practicality (adverse events), and limited efficacy (admission and discharge scores from the World Health Organization Disability Assessment Scale, Brief Illness Perception Questionnaire (BIPQ), Patient Health Questionnaire, and EuroQol 5D-5L). Data are presented as median [interquartile range] or count (percentage). Results: During the study, 143 participants (aged 42.00 [32.00–51.00] years, 68% women) participated in ReCOV. Participants were waitlisted for 3.86 [2.14–9.86] weeks and engaged with 5.00 [3.00–6.00] different health professionals. No adverse events occurred. The thematic analysis revealed that ReCOV was helpful but did not fully meet the needs of all participants. Limited efficacy testing indicated that participants had improved understanding and control (p < 0.001) of symptoms (BIPQ) and a small improvement in EQ VAS score (median difference 5.50 points [0.00–25.00], p = 0.004]). Conclusions: A multidisciplinary service was safe and mostly acceptable to participants for the management of long COVID. Further research should investigate the clinical and cost effectiveness of such a service, including optimal service duration and patient outcomes.

Information extraction (IE) from specialized clinical texts such as brain MRI reports is important for various clinical and population health contexts. However, this topic is under-explored due to privacy concerns limiting data availability and the inherent complexity and domain-specificity of clinical language. Common methods relying on substantial amounts of training data fail. The recent advances in large language model (LLM) research provide a promising solution to bridge the data scarcity gap, with improved ability to adapt to novel tasks with little supervision. We introduce a new, challenging dataset of 100 expert-annotated brain MRI reports, featuring 152 fine-grained entity types and 4 relation types, characterised by low inter-annotator agreement. This task reflects the inherent complexity and real-world ambiguity of medical text. We evaluate a small, open-weight LLM across span detection, named entity recognition, and relation extraction tasks. We compare few-shot prompting and parameter-efficient fine-tuning against specialized off-the-shelf biomedical IE systems. Our results demonstrate that both few-shot and fine-tuned LLM approaches substantially outperform off-the-shelf baselines. While LLMs show superiority, absolute performance, particularly for complex relations and fine-grained entities, remains modest, correlating with the dataset’s inherent difficulty and the extreme low-resource setting.