In emergency departments, the most common cause of death associated with suspected infected patients is sepsis. In this study, deep learning algorithms were used to predict the mortality of suspected infected patients in a hospital emergency department. During January 2007 and December 2013, 42,220 patients considered in this study were admitted to the emergency department due to suspected infection. In the present study, a deep learning structure for mortality prediction of septic patients was developed and compared with several machine learning methods as well as two sepsis screening tools: the systemic inflammatory response syndrome (SIRS) and quick sepsis-related organ failure assessment (qSOFA). The mortality predictions were explored for septic patients who died within 72 h and 28 days. Results demonstrated that the accuracy rate of deep learning methods, especially Convolutional Neural Network plus SoftMax (87.01% in 72 h and 81.59% in 28 d), exceeds that of the other machine learning methods, SIRS, and qSOFA. We expect that deep learning can effectively assist medical staff in early identification of critical patients.

Cadmium exposure is associated with chronic kidney disease (CKD), but the optimal biomarker for early cadmium-associated nephrotoxicity in low-level exposure has not yet been established. We conducted a cross-sectional investigation involving 167 CKD patients stratified according to tertiles of urinary cadmium levels (UCd), in which enzyme-linked immunosorbent assay (ELISA)-measured novel renal biomarkers were utilized to assess the extent of renal injury associated with cadmium burden. In the analyses, urinary kidney injury molecule-1 (KIM-1) levels and age were the independent factors positively correlated with UCd after adjusting for covariates in non-dialysis-dependent CKD patients (high vs. low UCd, odds ratio (95% confidence interval), 1.0016 (1.0001–1.0032), p = 0.043, and 1.0534 (1.0091–1.0997), p = 0.018). Other conventional and novel renal biomarkers, such as serum creatinine, estimated glomerular filtration rate, CKD staging, urinary protein/creatinine ratio, urinary 8-hydroxy-2-deoxyguanosine (8-OHdG), and urinary epidermal growth factor (EGF) were not independently correlated with UCd in the analyses. In conclusion, our study found that the ELISA-measured urinary KIM-1 level could serve as an early renal injury marker in low-level cadmium exposure for non-dialysis-dependent CKD patients. In addition, age was an independent factor positively associated with UCd in this population.

Low-level cadmium exposure has adverse effects on chronic kidney disease (CKD); however, the risk factors for elevated blood cadmium levels (BCLs) have not been studied in CKD. We conducted a cross-sectional investigation in 200 CKD patients and stratified them by the tertiles of BCL to compare their demographic, environmental, and biochemical data. The factors associated with BCL were identified, and their effects were examined in subgroups. In the analyses, female sex, smoking, and CKD stage 5D were associated with high BCL, and statin was inversely correlated with BCL (odds ratio [95% confidence interval, CI], 6.858 [2.381–19.746], p < 0.001, 11.719 [2.843–48.296], p = 0.001, 30.333 [2.252–408.520], p = 0.010, and 0.326 [0.122–0.873], p = 0.026; deviations of BCL [nmol/L, 95% CI], 2.66 [1.33–4.00], p < 0.001, 3.68 [1.81–5.56], p < 0.001, 3.38 [0.95–5.82], p = 0.007, and −2.07 [−3.35–−0.78], p = 0.002). These factors were also independently correlated with BCL in subgroups, including non-dialysis CKD, hypertensive patients, non-smokers, and male patients. In conclusion, female sex, smoking, and CKD stage 5D were the major risk factors for elevated BCL; additionally, statins were negatively associated with BCL in CKD.

Background: Hyperkalemia is a critical condition, especially in intensive care units. So far, there have been no accurate and noninvasive methods for recognizing hyperkalemia events on ambulatory electrocardiogram monitors. Objective: This study aimed to improve the accuracy of hyperkalemia predictions from ambulatory electrocardiogram (ECG) monitors using a personalized transfer learning method; this would be done by training a generic model and refining it with personal data. Methods: This retrospective cohort study used open source data from the Waveform Database Matched Subset of the Medical Information Mart From Intensive Care III (MIMIC-III). We included patients with multiple serum potassium test results and matched ECG data from the MIMIC-III database. A 1D convolutional neural network–based deep learning model was first developed to predict hyperkalemia in a generic population. Once the model achieved a state-of-the-art performance, it was used in an active transfer learning process to perform patient-adaptive heartbeat classification tasks. Results: The results show that by acquiring data from each new patient, the personalized model can improve the accuracy of hyperkalemia detection significantly, from an average of 0.604 (SD 0.211) to 0.980 (SD 0.078), when compared with the generic model. Moreover, the area under the receiver operating characteristic curve level improved from 0.729 (SD 0.240) to 0.945 (SD 0.094). Conclusions: By using the deep transfer learning method, we were able to build a clinical standard model for hyperkalemia detection using ambulatory ECG monitors. These findings could potentially be extended to applications that continuously monitor one’s ECGs for early alerts of hyperkalemia and help avoid unnecessary blood tests.

Background. Several risk scores have been developed to predict and analyze in-hospital mortality and short- and long-term outcomes of ST-elevation myocardial infarction (STEMI) patients after primary percutaneous coronary intervention (PPCI); these can classify patients as having a high or low risk of death or complications. Objective. To compare the prognostic precision of four risk scores for predicting in-hospital mortality in patients with STEMI treated with PPCI. Methods. We performed a retrospective cohort analysis of patients with STEMI who underwent PPCI between 2012 and 2019 (N=1346). GRACE (Global Registry of Acute Cardiac Events), CADILLAC (Controlled Abciximab and Device Investigation to Lower Late Angioplasty Complications), Zwolle, and TIMI (Thrombolysis in Myocardial Infarction) risk scores were calculated for each patient according to different variables. We evaluated the predictive accuracy of these scores for in-hospital mortality using the C statistic, which was obtained using logistic regression and receiver operating characteristic curves. Results. The GRACE, CADILLAC, Zwolle, and TIMI risk scores all had good predictive precision for in-hospital mortality, with C statistics ranging from 0.842 to 0.923. The GRACE and CADILLAC risk scores were found to be superior. Conclusions. All GRACE, CADILLAC, Zwolle, and TIMI risk scores showed a high predictive value for in-hospital mortality due to all causes in patients with STEMI treated with PPCI. The GRACE and CADILLAC risk scores revealed a better accuracy for predicting in-hospital mortality than the Zwolle and TIMI risk scores.