Search

Infection-related consultations on intensive care units (ICU) have a positive impact on quality of care and clinical outcome. However, timing of these consultations is essential and to date they are typically event-triggered and reactive. Here, we investigate a proactive approach to identify patients in need for infection-related consultations by machine learning models using routine electronic health records. Data was retrieved from a mixed ICU at a large academic tertiary care hospital including 9684 admissions. Infection-related consultations were predicted using logistic regression, random forest, gradient boosting machines, and long short-term memory neural networks (LSTM). Overall, 7.8% of admitted patients received an infection-related consultation. Time-sensitive modelling approaches performed better than static approaches. Using LSTM resulted in the prediction of infection-related consultations in the next clinical shift (up to eight hours in advance) with an area under the receiver operating curve (AUROC) of 0.921 and an area under the precision recall curve (AUPRC) of 0.541. The successful prediction of infection-related consultations for ICU patients was done without the use of classical triggers, such as (interim) microbiology reports. Predicting this key event can potentially streamline ICU and consultant workflows and improve care as well as outcome for critically ill patients with (suspected) infections. [ABSTRACT FROM AUTHOR]

Despite having a similar post-operative complication profile, cardiac valve operations are associated with a higher mortality rate compared to coronary artery bypass grafting (CABG) operations. For long-term mortality, few predictors are known. In this study, we applied an ensemble machine learning (ML) algorithm to 88 routinely collected peri-operative variables to predict 5-year mortality after different types of cardiac operations. The Super Learner algorithm was trained using prospectively collected peri-operative data from 8241 patients who underwent cardiac valve, CABG and combined operations. Model performance and calibration were determined for all models, and variable importance analysis was conducted for all peri-operative parameters. Results showed that the predictive accuracy was the highest for solitary mitral (0.846 [95% CI 0.812–0.880]) and solitary aortic (0.838 [0.813–0.864]) valve operations, confirming that ensemble ML using routine data collected perioperatively can predict 5-year mortality after cardiac operations with high accuracy. Additionally, post-operative urea was identified as a novel and strong predictor of mortality for several types of operation, having a seemingly additive effect to better known risk factors such as age and postoperative creatinine. [ABSTRACT FROM AUTHOR]

Intensive insulin therapy is associated with the risk of hypoglycemia and increased costs of material and personnel. We therefore evaluated the safety and efficiency of a computer-assisted glucose control protocol in a large population of critically ill patients. Observational cohort study in three intensive care units (32 beds) in a 1,300-bed university teaching hospital. All 2,800 patients admitted to the surgical, neurosurgical, and cardiothoracic units; the study period started at each ICU after implementation of Glucose Regulation for Intensive Care Patients (GRIP), a freely available computer-assisted glucose control protocol. We analysed compliance in relation to recommended insulin pump rates and glucose measurement frequency. Patients were on GRIP-ordered pump rates 97% of time. Median measurement time was 5 min late (IQR 20 min early to 34 min late). Hypoglycemia was uncommon (7% of patients for mild hypoglycemia, < 3.5 mmol/l; 0.86% for severe hypoglycemia, < 2.2 mmol/l). Our predefined target range (4.0–7.5 mmol/l) was reached after a median of 5.6 h (IQR 0.2–11.8) and maintained for 89% (70–100%) of the remaining stay at the ICU. The number of measurements needed was 5.9 (4.8–7.3) per patient per day. In-hospital mortality was 10.1%. Our computer-assisted glucose control protocol provides safe and efficient glucose regulation in routine intensive care practice. A low rate of hypoglycemic episodes was achieved with a considerably lower number of glucose measurements than used in most other schemes. [ABSTRACT FROM AUTHOR]

Objective: Various studies have documented a markedly high incidence of contrast-induced nephropathy (CIN). Most of these studies were conducted in patients not in the ICU. In ICU patients intravenous contrast may be withheld for fear of CIN. We investigated the incidence of CIN in ICU patients.Design and Setting: Retrospective cohort study in a 12-bed tertiary surgical ICU.Patients: Were evaluated all contrast-enhanced abdominal computed tomography (CT) scans between 1995 and 2003 in patients not on renal replacement therapy (RRT) before the CT. Patients received prophylactic prehydration and, since 2000, acetylcysteine. Low-osmolarity, nonionic contrast was used. CIN was defined as an increase in serum creatinine of more than 44 micromol/l (0.5 mg/dl) within 48 h after contrast administration, with no increase in creatinine of 44 micromol/l during the preceding 2 days. RRT initiated after the CT was also recorded.Measurements and Results: The patient was not on RRT before CT in 486 of 589 cases (16% diabetics). In these 486 cases the median (IQR) creatinine decreased significantly from 88 micromol/l (66-124) on the day of the CT-scan to 84 micromol/l (63-118) 2 days later. Only 7 of the 486 cases (1.4%) fulfilled the criteria of CIN, and in another 17 (3.5%) RRT was started after the CT. Important coexisting causes of renal failure were present in these patients, and in all survivors renal function recovered.Conclusions: CT with modern contrast is associated with a very low incidence of nephropathy in predominantly nondiabetic surgical ICU patients. Intravenous contrast should only rarely be withheld in these patients. [ABSTRACT FROM AUTHOR]

Objective: To find out if there is an association between hyperglycaemia and mortality in mixed ICU patients.Design and Setting: Retrospective cohort study over a 2-year period at the medical ICU of a university hospital.Measurements: Admission glucose, maximum and mean glucose, length of stay, mortality, insulin therapy and Apache-II score.Results: In 1085 consecutive patients, ICU- and hospital mortality were 20 and 25%, respectively. The total number of blood glucose measurements was 10.012. Admission glucose was 7.9 +/- 4.5 mmol/l (mean +/- SD), mean glucose 7.5 +/- 2.9 and maximum glucose 10.0 +/- 5.4 mmol/l. Median ICU length of stay (LOS) was 3.0 days (range 2.0-6.0 days, IQR), and hospital LOS was 16 days (range 7-32 days). In 28% of patients insulin treatment was started. Median Apache-II score was 13. 68% of patients were mechanically ventilated. Univariate analysis showed an association with ICU mortality for mean glucose (non-survivors 8.6 +/- 4.3 vs 7.2 +/- 2.4 survivors), maximum glucose (11.7 +/- 5.9 vs 9.6 +/- 5.2, non-survivors vs survivors, respectively), use of insulin (mortality 29 vs 17% in patients not using insulin) and age (61 vs 55.7 years). Gender and a history of diabetes mellitus were not associated with mortality. In a multivariate model, the Apache-II score was the only variable associated with mortality independent of other variables, including mean blood glucose.Conclusion: In this retrospective study mean glucose level was not an independent risk factor for mortality in mixed ICU patients. [ABSTRACT FROM AUTHOR]

Presents a letter to the editor about irreversible ischemia of the hand after peripheral administration of tromethamol.