Can Machine Learning Personalize Cardiovascular Therapy in Sepsis?

Large randomized trials in sepsis have generally failed to find effective novel treatments. This is increasingly attributed to patient heterogeneity, including heterogeneous cardiovascular changes in septic shock. We discuss the potential for machine learning systems to personalize cardiovascular resuscitation in sepsis. While the literature is replete with proofs of concept, the technological readiness of current systems is low, with a paucity of clinical trials and proven patient benefit. Systems may be vulnerable to confounding and poor generalization to new patient populations or contemporary patterns of care. Typical electronic health records do not capture rich enough data, at sufficient temporal resolution, to produce systems that make actionable treatment suggestions. To resolve these issues, we recommend a simultaneous focus on technical challenges and removing barriers to translation. This will involve improving data quality, adopting causally grounded models, prioritizing safety assessment and integration into healthcare workflows, conducting randomized clinical trials and aligning with regulatory requirements.

Acute kidney injury in COVID-19: multicentre prospective analysis of registry data.

BACKGROUND: Acute kidney injury (AKI) is a common and important complication of coronavirus disease 2019 (COVID-19). Further characterization is required to reduce both short- and long-term adverse outcomes. METHODS: We examined registry data including adults with confirmed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection admitted to five London Hospitals from 1 January to 14 May 2020. Prior end-stage kidney disease was excluded. Early AKI was defined by Kidney Disease: Improving Global Outcomes creatinine criteria within 7 days of admission. Independent associations of AKI and survival were examined in multivariable analysis. Results are given as odds ratios (ORs) or hazard ratios (HRs) with 95% confidence intervals. RESULTS: Among 1855 admissions, 455 patients (24.5%) developed early AKI: 200 (44.0%) Stage 1, 90 (19.8%) Stage 2 and 165 (36.3%) Stage 3 (74 receiving renal replacement therapy). The strongest risk factor for AKI was high C-reactive protein [OR 3.35 (2.53-4.47), P < 0.001]. Death within 30 days occurred in 242 (53.2%) with AKI compared with 255 (18.2%) without. In multivariable analysis, increasing severity of AKI was incrementally associated with higher mortality: Stage 3 [HR 3.93 (3.04-5.08), P < 0.001]. In 333 patients with AKI surviving to Day 7, 134 (40.2%) recovered, 47 (14.1%) recovered then relapsed and 152 (45.6%) had persistent AKI at Day 7; an additional 105 (8.2%) patients developed AKI after Day 7. Persistent AKI was strongly associated with adjusted mortality at 90 days [OR 7.57 (4.50-12.89), P < 0.001]. CONCLUSIONS: AKI affected one in four hospital in-patients with COVID-19 and significantly increased mortality. Timing and recovery of COVID-19 AKI is a key determinant of outcome.

The pattern of prescribing of glucose modulating agents for type 2 diabetes in general practices in England 2016/17.

BACKGROUND: In the financial year 2016/17 there were 52.0 million items prescribed for diabetes at a total net ingredient cost of £983.7 million - up from 28.9 million prescription items and £572.4 million in 2006/07. Anti-diabetes drugs (British National Formulary section 6.1.2) make up 45.1 per cent of the total £983.7 million net ingredient cost of drugs used in diabetes and account for 72.0 per cent of prescription items for all diabetes prescribing. METHODS: We examined the way that agents licensed to treat type 2 diabetes were used across GP practices in England in the year 2016/2017. Analysis was at a GP practice level not at the level of patient data. RESULTS: Annual prescribing costs / patient / medication type for monotherapy varied considerable from £11/year for gliclazide and glimepiride to £885/year for Liraglutide. The use of SGLT-2i agents grew strongly at 70% per annum to around 100,000 DDD with prescriptions seen in 95% of GP practices. Liraglutide expenditure (11% of total) was high for a relatively small number of patients (1.3% of Defined Daily Doses), with still significant spend on exenatide. Liraglutide use significantly exceeded that of other glucagon-like peptide-1 (GLP-1) agonists. CONCLUSIONS: Our work demonstrates the significant cost of medication to modulate tissue glucose levels in type 2 diabetes and the dominance of some non-generic preparations in terms of number of prescriptions and overall spend. There are some older sulphonylureas in use, which should not generally be prescribed. Regular audit of patient treatment at a general practice level will ensure appropriate targeted use of licensed medications and of their cost effectiveness.