ObjectivePrediabetes is a high-risk state for developing diabetes and associated complications. The purpose of this paper was to report trends in prevalence of prediabetes for individuals aged 16 and older in England without previously diagnosed diabetes.SettingData collected by the Health Survey for England (HSE) in England in the years 2003, 2006, 2009 and 2011.ParticipantsIndividuals aged 16 and older who participated in the HSE and provided a blood sample.Primary outcome variableIndividuals were classified as having prediabetes if glycated haemoglobin was between 5.7% and 6.4% and were not previously diagnosed with diabetes.ResultsThe prevalence rate of prediabetes increased from 11.6% to 35.3% from 2003 to 2011. By 2011, 50.6% of the population who were overweight (body mass index (BMI)>25) and ≥40 years of age had prediabetes. In bivariate relationships, individuals with greater socioeconomic deprivation were more likely to have prediabetes in 2003 (p=0.0008) and 2006 (p=0.0246), but the relationship was not significant in 2009 (p=0.213) and 2011 (p=0.3153). In logistic regressions controlling for age, sex, race/ethnicity, BMI and high blood pressure, the second most socioeconomically deprived had a significantly elevated risk of having prediabetes (2011, OR=1.45; 95% CI 1.26 to 1.88).ConclusionsThere has been a marked increase in the proportion of adults in England with prediabetes. The socioeconomically deprived are at substantial risk. In the absence of concerted and effective efforts to reduce risk, the number of people with diabetes is likely to increase steeply in coming years.

There is an increasing interest in developing artificial intelligence (AI) systems to process and interpret electronic health records (EHRs). Natural language processing (NLP) powered by pretrained language models is the key technology for medical AI systems utilizing clinical narratives. However, there are few clinical language models, the largest of which trained in the clinical domain is comparatively small at 110 million parameters (compared with billions of parameters in the general domain). It is not clear how large clinical language models with billions of parameters can help medical AI systems utilize unstructured EHRs. In this study, we develop from scratch a large clinical language model—GatorTron—using >90 billion words of text (including >82 billion words of de-identified clinical text) and systematically evaluate it on five clinical NLP tasks including clinical concept extraction, medical relation extraction, semantic textual similarity, natural language inference (NLI), and medical question answering (MQA). We examine how (1) scaling up the number of parameters and (2) scaling up the size of the training data could benefit these NLP tasks. GatorTron models scale up the clinical language model from 110 million to 8.9 billion parameters and improve five clinical NLP tasks (e.g., 9.6% and 9.5% improvement in accuracy for NLI and MQA), which can be applied to medical AI systems to improve healthcare delivery. The GatorTron models are publicly available at: https://catalog.ngc.nvidia.com/orgs/nvidia/teams/clara/models/gatortron_og.

Purpose: CYP2D6 bioactivates codeine and tramadol, with intermediate and poor metabolizers (IMs and PMs) expected to have impaired analgesia. This pragmatic proof-of-concept trial tested the effects of CYP2D6-guided opioid prescribing on pain control. Methods: Participants with chronic pain (94% on an opioid) from 7 clinics were enrolled into CYP2D6-guided (n=235) or usual care (n=135) arms using a cluster design. CYP2D6 phenotypes were assigned based on genotype and CYP2D6 inhibitor use, with recommendations for opioid prescribing made in the CYP2D6-guided arm. Pain was assessed at baseline and 3 months using PROMIS ® measures. Results: On stepwise multiple linear regression, the primary outcome of composite pain intensity (composite of current pain and worst and average pain in the past week) among IM/PMs initially prescribed tramadol/codeine (n=45) had greater improvement in the CYP2D6-guided versus usual care arm (−1.01±1.59 versus −0.40±1.20; adj- P =0.016); 24% of CYP2D6-guided versus 0% of usual care participants reported ≥30% (clinically meaningful) reduction in the composite outcome. In contrast, among normal metabolizers prescribed tramadol or codeine at baseline, there was no difference in the change in composite pain intensity at 3 months between CYP2D6-guided (−0.61±1.39) and usual care (−0.54±1.69) groups (adj-P=0.540). Conclusion: These data support the potential benefits of CYP2D6-guided pain management.

The current systematic review found that studies with more rigorous designs all reported benefits from HIE. Such benefits include fewer duplicated procedures, reduced imaging, lower costs, and improved patient safety. We also found that studies evaluating community HIEs were more likely to find benefits than studies that evaluated enterprise HIEs or vendor-mediated exchanges. Overall, these finding bode well for the HIEs ability to deliver on anticipated improvements in care delivery and reduction in costs.