One-Liners, Part Three

And, we're back to live!

A novel, standardised approach to balancing effectiveness, efficiency and utility of surveillance AI prediction models for hospitalised patients using sepsis prediction as an exemplar

A nice project formally evaluating something we all know, deep in our hearts, to be true: sepsis prediction models are rubbish in the clinical setting. The authors nail it in their concluding paragraph: “Although many [AI/ML] surveillance prediction tools have been implemented, evidence of improved care remains limited, which in part may be due to poor performance and utility of implemented models”.

Large-Scale Evaluation of Machine Learning Models in Identifying Follow-Up Recommendations in Radiology Reports

Transitions of care between one service or one setting and another are always fraught – one of the most pervasive being between radiology and the ordering clinical team. This study takes a look at a few ML and LLM approaches for gleaning actionable findings from radiology reports, primarily as a quality assurance mechanism. It’s all just decent performance on just one well-trod data set, but it certainly addresses a real need.

Caffeinated Coffee Consumption or Abstinence to Reduce Atrial Fibrillation - The DECAF Randomized Clinical Trial

Ack, my atrial fibrillation is acting up! Must be the coffee, right? In this trial, abstinence from coffee was not protective – although, the caffeine-allowed cohort was exposed to merely homeopathic doses (only 1 cup a day).

Cardiac Biomarkers, Echocardiography, and Outpatient Cardiac Monitoring for Evaluation of Emergency Department Patients With Syncope: A Systematic Review and Analysis of Direct Evidence for SAEM GRACE

How should patients suffering syncope be evaluated? This systematic review tells us there’s no simple answer to such a simple question – because, as we all know, the patient population presenting with syncope is grossly heterogeneous. Differing diagnostic approaches are required in order to catch the various needles in entirely separate haystacks.

Vaccine effectiveness of a bivalent respiratory syncytial virus (RSV) pre-F vaccine against RSV-associated hospital admission among adults aged 75–79 years in England: a multicentre, test-negative, case–control study

Sometimes we tell worried patients “it’s just a virus”, as if that should be reassuring. In actuality, viral illness can be just as deadly as bacterial illness, but there are fewer treatments available. Better to prevent them with vaccines – and the RSV vaccine is a fairly decent candidate. However, hospitalization from RSV is still rare enough during seasonal outbreaks that the cost of this new vaccine remains problematic.

Revisions to the Safety Assurance Factors for Electronic Health Record Resilience (SAFER) Guides to update national recommendations for safe use of electronic health records

Hey, the SAFER Guides have been updated! A lot has happened in the past ten years, and they’ve been revised a wee bit in response to feedback and changes in the digital landscape. Considering CMS requires institutional attestation to having reviewed this guide, modernization was certainly needed.