East and North Hertfordshire NHS Trust

Patients with heart failure with reduced ejection fraction require rapid initiation and uptitration of guideline-directed medical therapy (GDMT), which is resource-intensive. In a prospective, open-label pilot trial, we assessed the feasibility, acceptability, and safety of a generative artificial intelligence-powered virtual assistant (VA), with retrieval-augmented generation and expert prompt engineering, to optimize GDMT. Patients with new heart failure with reduced ejection fraction (n = 60) were randomized to VA-guided care, delivered by nonmedical staff at 2-weekly intervals or standard-of-care treatment delivered by doctors or nurses. At 12 weeks, patients in the VA arm had superior GDMT optimization across all medication classes, lower N-terminal pro-B-type natriuretic peptide, and fewer hospitalizations. Patient-reported acceptability, appropriateness and feasibility scores were high, with no safety disagreements between VA and clinician recommendations. Treatment by an artificial intelligence-powered VA, run by nonmedical staff, with minimal remote medical supervision, is acceptable to patients, and can safely and effectively optimize GDMT, representing a scalable strategy to optimize treatment and health care resource utilization.

Although ICUs generate rich multimodal data, most studies examine only narrow modality combinations, limiting progress toward clinically useful decision-support systems. We propose a systematic framework for evaluating multimodal fusion strategies in the ICU, enabling structured comparisons of early, intermediate and late fusion across combinations of imaging, tabular, text and time-series data. To assess generalisability across diagnoses and patient groups, we apply the framework to six cardiovascular conditions covering a broad diagnostic spectrum and report performance broken down by age, sex and insurance. Rather than asking whether multimodal fusion helps, we ask under which conditions it helps. We find that multimodal models frequently outperform unimodal baselines, with gains of up to five percentage points in AUC, particularly when modalities provide complementary information or when prediction uncertainty is high. Across the architectures we explored, late fusion is the most robust and best-performing strategy. Using uncertainty-based ranking, we show that multimodal integration is especially effective for correcting high-uncertainty cases, substantially reducing both false-negative and false-positive rates relative to single-modality models. The most informative data are typically available within the first 24 hours of ICU admission. These findings offer guidance for prioritising modalities and deploying multimodal models in real-world clinical settings.

Background.Implantable brain-computer interfaces (iBCIs) decode neural signals to generate command signals for effector devices to restore lost functions, such as movement or speech. However, maintaining device performance over time requires recalibration of decoding algorithms due to inherent instability in neural signals.Objective.To systematically review recalibration procedures in iBCIs for patients with motor impairments, focusing on the clinical implications of recalibration requirements and strategies which can enable long-term, independent use.Approach.A systematic search was conducted across EMBASE, MEDLINE, and CINAHL databases to identify studies involving recalibration of iBCIs. Data on recalibration frequency, duration, staff requirements, and location were extracted and analyzed.Main results.Recalibration practices varied widely amongst studies and were typically performed according to predetermined study protocols, rather than practical need following deteriorating device performance. Common practices include manual recalibration requiring a specialist research team, semi-automatic recalibration which could be performed by a non-specialist caregiver, and automatic recalibration methods whereby patients did not require assistance. Devices utilizing electrocorticography (ECoG) recording arrays generally required less frequent recalibration compared to those using microelectrode arrays (MEAs). Extended independent use was more frequently reported with ECoG-based iBCIs.Significance.Reducing recalibration frequency or complexity can improve patient autonomy, which is crucial for enhancing long-term independent iBCI use in home and clinical settings. ECoG iBCIs typically have a low recalibration burden due to inherent signal stability. Conversely, MEA iBCIs typically involve a higher recalibration burden, though recent studies have reduced this by incorporating spectral data and continuously updating models. Despite this progress, recalibration procedures are often not fully defined in iBCI studies, and where they are, they usually relate to the study protocol rather than the clinically meaningful recalibration requirement due to worsening device performance. Future studies should continue to develop user-friendly recalibration procedures and outline the clinically relevant recalibration requirements where possible.

Abstract Introduction NHS England has expanded access to hybrid-closed-loop (HCL) insulin delivery systems for children and young-people (CYP) with type-1 diabetes mellitus (T1DM) through a five-year national implementation programme. Since January-2023, our district general hospital has adopted HCL therapy for CYP with T1DM. This study evaluates the real-world impact of HCL systems on glycaemic control, including HbA1c, time-in-range (TIR), and hypoglycaemia. Method CYP with T1DM initiated on HCL therapy between January-2023 and June-2024 were identified using electronic patient records. HbA1c values were obtained from clinical records, and continuous glucose monitoring (CGM) metrics were extracted from Glooko platform. Patients with incomplete datasets, insufficient CGM uploads, or lost follow-up were excluded. Paired pre- and post-HCL HbA1c measurements and early post-initiation CGM metrics were analysed. Results Of 120 CYP commenced on HCL therapy, 72 had complete data suitable for analysis. Median age at HCL initiation was 13.5 years (IQR 4-20), 53% were female, and median diabetes duration was 7 years (IQR 3-9). Omnipod-5 was the most commonly used system (75%). All participants used CGM and multiple daily injections prior. Mean HbA1c improved significantly from 72.34 ± 26.37 mmol/mol pre-HCL to 56.01 ± 8.41 mmol/mol at up to 18 months post-initiation (mean paired reduction 16.49 mmol/mol; 95% CI −22.88 to −10.10; P < .001). Early CGM outcomes were favourable. Mean 90-day time-in-range (3.9-10.0 mmol/L) was 65.55 ± 9.52%. Mean time below range (<3.9 mmol/L) was 1.53 ± 1.06%, and time in very-low range (<3.0 mmol/L) was 0.39 ± 0.60%.VariableMeasureTotalAge, yearMedian, IQR13.5 (4-20)Gender, Malen(%)34 (47%)Gender, Femalen(%)38 (53%)Diabetes Duration pre-HCL, YearsMedian, IQR7 (3-9)Pre-HCL HbA1c, mmol/molMean ± SD72.34 ± 26.37Post HCL MeasuresFirst 90-day Time-In-Range, TIR, % (3.9-10 mmol/L)Mean ± SD65.55 ± 9.52First 90-day Time-Below-Range, TBR, %(<3.9 mol/L)Mean ± SD1.53 ± 1.06First 90-day Time In Very Low-Range, %(<3 mol/L)Mean ± SD0.39 ± 0.60Post-HCL Mean HbA1c (Paired), mmol/molMean ± SD56.01 ± 8.41Pre-Post HbA1c difference, mmol/molMean ± SD16.49 ± 27.47 Conclusions Hybrid closed-loop insulin delivery systems delivered substantial and clinically meaningful improvements in glycaemic control within a district general hospital setting, with achievement of recommended CGM targets and low exposure to hypoglycaemia. These real-world data demonstrate that HCL technologies can be implemented safely and effectively outside specialist tertiary centres. Our findings highlight the capability of DGH-based multidisciplinary teams to deliver high-quality, technology-enabled diabetes care, reinforcing the scalability and sustainability of HCL systems within national diabetes pathways. Continued longitudinal evaluation will clarify long-term outcomes and inform service planning.