Hybrid Deep Learning for Medication-Related Information Extraction From Clinical Texts in French: MedExt Algorithm Development Study.

BACKGROUND: Information related to patient medication is crucial for health care; however, up to 80% of the information resides solely in unstructured text. Manual extraction is difficult and time-consuming, and there is not a lot of research on natural language processing extracting medical information from unstructured text from French corpora. OBJECTIVE: We aimed to develop a system to extract medication-related information from clinical text written in French. METHODS: We developed a hybrid system combining an expert rule-based system, contextual word embedding (embedding for language model) trained on clinical notes, and a deep recurrent neural network (bidirectional long short term memory-conditional random field). The task consisted of extracting drug mentions and their related information (eg, dosage, frequency, duration, route, condition). We manually annotated 320 clinical notes from a French clinical data warehouse to train and evaluate the model. We compared the performance of our approach to those of standard approaches: rule-based or machine learning only and classic word embeddings. We evaluated the models using token-level recall, precision, and F-measure. RESULTS: The overall F-measure was 89.9% (precision 90.8; recall: 89.2) when combining expert rules and contextualized embeddings, compared to 88.1% (precision 89.5; recall 87.2) without expert rules or contextualized embeddings. The F-measures for each category were 95.3% for medication name, 64.4% for drug class mentions, 95.3% for dosage, 92.2% for frequency, 78.8% for duration, and 62.2% for condition of the intake. CONCLUSIONS: Associating expert rules, deep contextualized embedding, and deep neural networks improved medication information extraction. Our results revealed a synergy when associating expert knowledge and latent knowledge.

Association of blood bicarbonate and pH with mineral metabolism disturbance and outcome after kidney transplantation.

In kidney transplant recipients (KTRs), scarce evidence has associated low blood bicarbonate levels with mineral metabolic disturbance and reduced allograft survival. However, the contribution of the blood pH to these observations remains unassessed. Equally, little is known about the influence of the blood provenance (arteriovenous fistula vs peripheral vein) on bicarbonate values. We analyzed blood gas parameters in a single-center cohort of 1260 stable KTRs, 3 months after transplantation. Inspection of pO2 distribution allowed the unambiguous identification of the arterial (N = 914) or venous (N = 346) origin of the samples. In patients with arterial blood samples, 435 (46%) had bicarbonate levels below 22 mmol/L. Among them, 196 (40%) were acidemic (blood pH <7.38). In multivariate analysis, low arterial blood pH was associated with increased blood ionized calcium and phosphate and reduced fibroblast growth factor 23 and calcitriol, but not with outcome. In contrast, low bicarbonate concentration predicted allograft loss independently of measured glomerular filtration rate and other potential confounders (hazard ratio [HR] 1.70; 95% confidence interval [CI] 1.04-2.80). In KTRs, reduced arterial blood bicarbonate levels predict outcome while acidemia is associated with altered mineral metabolism.

Osmoregulation Performance and Kidney Transplant Outcome.

BACKGROUND: Kidney transplant recipients have an impaired ability to dilute urine but seldom develop baseline hyponatremia before ESRD. Although hyponatremia is a risk factor for adverse events in CKD and in kidney transplant recipients, it remains unclear whether subtler alterations in osmoregulation performance are associated with outcome. METHODS: We studied a single-center prospective cohort of 1258 kidney transplant recipients who underwent a water-loading test 3 months after transplant to determine osmoregulation performance. Measured GFR (mGFR) was performed at the same visit. A group of 164 healthy candidates for kidney donation served as controls. We further evaluated the association of osmoregulation performance with transplantation outcomes and subsequent kidney function. RESULTS: Unlike controls, most kidney transplant recipients failed to maintain plasma sodium during water loading (plasma sodium slope of -0.6±0.4 mmol/L per hour in transplant recipients versus -0.12±0.3 mmol/L per hour in controls; P<0.001). Steeper plasma sodium reduction during the test independently associated with the composite outcome of all-cause mortality and allograft loss (hazard ratio [HR], 1.73 per 1 mmol/L per hour decrease in plasma sodium; 95% confidence interval [95% CI], 1.23 to 2.45; P=0.002) and allograft loss alone (HR, 2.04 per 1 mmol/L per hour decrease in plasma sodium; 95% CI, 1.19 to 3.51; P=0.01). The association remained significant in a prespecified sensitivity analysis excluding patients with hyperglycemia. In addition, a steeper plasma sodium slope 3 months after transplantation independently correlated with lower mGFR at 12 months (ß=1.93; 95% CI, 0.46 to 3.41; P=0.01). CONCLUSIONS: Reduced osmoregulation performance occurs frequently in kidney transplant recipients and is an independent predictor of renal outcome.