Developing a deep learning natural language processing algorithm for automated reporting of adverse drug reactions.

The detection of adverse drug reactions (ADRs) is critical to our understanding of the safety and risk-benefit profile of medications. With an incidence that has not changed over the last 30 years, ADRs are a significant source of patient morbidity, responsible for 5%-10% of acute care hospital admissions worldwide. Spontaneous reporting of ADRs has long been the standard method of reporting, however this approach is known to have high rates of under-reporting, a problem that limits pharmacovigilance efforts. Automated ADR reporting presents an alternative pathway to increase reporting rates, although this may be limited by over-reporting of other drug-related adverse events. We developed a deep learning natural language processing algorithm to identify ADRs in discharge summaries at a single academic hospital centre. Our model was developed in two stages: first, a pre-trained model (DeBERTa) was further pre-trained on 1.1 million unlabelled clinical documents; secondly, this model was fine-tuned to detect ADR mentions in a corpus of 861 annotated discharge summaries. This model was compared to a version without the pre-training step, and a previously published RoBERTa model pretrained on MIMIC III, which has demonstrated strong performance on other pharmacovigilance tasks. To ensure that our algorithm could differentiate ADRs from other drug-related adverse events, the annotated corpus was enriched for both validated ADR reports and confounding drug-related adverse events using. The final model demonstrated good performance with a ROC-AUC of 0.955 (95% CI 0.933 - 0.978) for the task of identifying discharge summaries containing ADR mentions, significantly outperforming the two comparator models.

Medicines stewardship.

Medicines stewardship refers to coordinated strategies and interventions to optimise medicines use, usually within a specific therapeutic area. Medicines stewardship programs can reduce variations in practice and improve patient outcomes. Therapeutic domains for medicines stewardship are chosen to address frequently used drug classes associated with a high risk of adverse outcomes. Some examples include antimicrobial, opioid analgesic, anticoagulation and psychotropic stewardship. Common elements of successful stewardship programs include multidisciplinary leadership, stakeholder engagement, tailored communication strategies, behavioural changes, implementation science methodologies, and ongoing program monitoring, evaluation and reporting. Medicines stewardship is a continual quality improvement process that requires ongoing support and resources, as well as clinician and consumer engagement, to remain sustainable. It is critical for hospital-based medicines stewardship programs to consider impacts on care in the community when making and communicating changes to patient therapy. This ensures that stewardship efforts are sustained across transitions of care.

Paracetamol dosing in hospital and on discharge for older people who are frail or have low body weight.

AIMS: To describe paracetamol dosing and liver function test (LFT) monitoring in older hospital inpatients who are frail or have low body weight. METHODS: Retrospective observational study, at a 790-bed metropolitan public health service in Australia. Patients aged ≥70 years, with body weight <50 kg or frailty index based on laboratory data (FI-Lab) score ≥0.3, who were administered paracetamol during an admission with length-of-stay >72 hours, were included. Data were extracted from electronic medical records. Paracetamol doses administered in hospital, and doses prescribed on discharge, were compared against consensus guidelines that recommended ≤60 mg/kg/d for older people weighing <50 kg, and ≤3000 mg/d for frail older people. RESULTS: In total, 240 admissions (n = 229 patients, mean age 84.7 years) were analysed. During 150 (62.5%) admissions, higher than recommended paracetamol doses were prescribed. On 138 (57.5%) occasions, patients were prescribed paracetamol on discharge, and 112/138 (81.2%) doses were higher than recommended. Most discharge prescriptions (97/138, 70.3%) were for regular administration. The median daily dose on discharge for patients <50 kg was 83.7 mg/kg (interquartile range 73.6-90.9 mg/kg). For frail patients ≥50 kg, the median daily discharge dose was 3990 mg (interquartile range 3000-4000 mg). LFTs were measured in hospital for 151/200 (75.5%) and 93/166 (56.0%) patients who received paracetamol for >48 hours and >5 days, respectively. CONCLUSION: Majority of paracetamol doses prescribed for frail or low-weight older patients in hospital and on discharge were higher than recommended in consensus guidelines. LFTs were not measured for 44% patients who received paracetamol regularly for >5 days. Further studies are needed to explore long-term outcomes of this practice.

Targeted Depletion of Bacteria from Mixed Populations by Programmable Adhesion with Antagonistic Competitor Cells.

Selective and targeted removal of individual species or strains of bacteria from complex communities can be desirable over traditional, broadly acting antibacterials in several contexts. However, generalizable strategies that accomplish this with high specificity have been slow to emerge. Here we develop programmed inhibitor cells (PICs) that direct the potent antibacterial activity of the type VI secretion system (T6SS) against specified target cells. The PICs express surface-displayed nanobodies that mediate antigen-specific cell-cell adhesion to effectively overcome the barrier to T6SS activity in fluid conditions. We demonstrate the capacity of PICs to efficiently deplete low-abundance target bacteria without significant collateral damage to complex microbial communities. The only known requirements for PIC targeting are a Gram-negative cell envelope and a unique cell surface antigen; therefore, this approach should be generalizable to a wide array of bacteria and find application in medical, research, and environmental settings.