Early detection of late-onset neonatal sepsis from noninvasive biosignals using deep learning: A multicenter prospective development and validation study.

BACKGROUND: Neonatal sepsis is responsible for significant morbidity and mortality worldwide. Its accurate and timely diagnosis is hindered by vague symptoms and the urgent necessity for early antibiotic intervention. The gold standard for diagnosing the condition is the identification of a pathogenic organism from normally sterile sites via laboratory testing. However, this method is resource-intensive and cannot be conducted continuously. OBJECTIVE: This study aimed to predict the onset of late-onset sepsis (LOS) with good diagnostic value as early as possible using non-invasive biosignal measurements from neonatal intensive care unit (NICU) monitors. METHODS: In this prospective multicenter study, we developed a multimodal machine learning algorithm based on a convolutional neural network (CNN) structure that uses the power spectral density (PSD) of recorded biosignals to predict the onset of LOS. This approach aimed to discern LOS-related pathogenic spectral signatures without labor-intensive manual artifact removal. RESULTS: The model achieved an area under the receiver operating characteristic score of 0.810 (95 % CI 0.698-0.922) on the validation dataset. With an optimal operating point, LOS detection had 83 % sensitivity and 73 % specificity. The median early detection was 44 h before clinical suspicion. The results highlighted the additive importance of electrocardiogram and respiratory impedance (RESP) signals in improving predictive accuracy. According to a more detailed analysis, the predictive power arose from the morphology of the electrocardiogram's R-wave and sudden changes in the RESP signal. CONCLUSION: Raw biosignals from NICU monitors, in conjunction with PSD transformation, as input to the CNN, can provide state-of-the-art prediction performance for LOS without the need for artifact removal. To the knowledge of the authors, this is the first study to highlight the independent and additive predictive potential of electrocardiogram R-wave morphology and concurrent, sudden changes in the RESP waveform in predicting the onset of LOS using non-invasive biosignals.

Increased blood coagulation is associated with poor neurological outcome in aneurysmal subarachnoid hemorrhage.

BACKGROUND AND PURPOSE: Patients with aneurysmal subarachnoid hemorrhage (aSAH) have demonstrated increased blood coagulation which is thought to contribute to delayed cerebral ischemia (DCI) and to a worse outcome. Therefore, we sought to determine whether this increased blood coagulation, detectable with rotational thromboelastometry (ROTEM), was associated with DCI and neurological outcome. METHODS: We conducted a prospective observational study of 60 consecutive adult aSAH patients. ROTEM's EXTEM and FIBTEM assays and D-dimer were analyzed at admission and post-bleed days (PBDs) 2-3, 4-5, 7-8, and 11-12. ROTEM's clot formation time (CFT) represents the stabilization of the clot, and the maximum clot firmness (MCF) the maximum clot strength. Glasgow Outcome Scale extended (GOSe) at three months determined the neurological outcome. RESULTS: DCI incidence was 41.7%. EXTEM-CFT was significantly shorter in patients with unfavorable neurological outcome (GOSe 1-4) on PBDs 4-5 and 7-8, p < 0.05, respectively. FIBTEM-MCF was significantly higher in patients with unfavorable neurological outcomes on PBD 4-5 (p < 0.05), PBD 7-8 (p < 0.05), and PBD 11-12 (p < 0.05). EXTEM-CFT decreased, and FIBTEM-MCF rose during the study period in all patients. Patients with unfavorable neurological outcome had a higher D-dimer at all studied time points, p < 0.05. No difference was found in the ROTEM parameters or D-dimer when assessing patients with and without DCI. CONCLUSIONS: Patients were in a state of increased blood coagulation after aSAH, with those with unfavorable neurological outcome being more coagulable than those with favorable outcome. However, increased blood coagulation was not associated with DCI. CLINICALTRIALS: gov, NCT03985176.

Trends in the national early warning score are associated with subsequent mortality - A prospective three-centre observational study with 11,331 general ward patients.

Aim: To investigate whether trends in the NEWS values are associated with patient mortality in general ward patients. Methods: A one-year prospective observational study in three hospitals in Finland. All data on patients' NEWS values during the first three days of general ward admissions were collected. The linear regression model was used to investigate the association of the NEWS trajectories with subsequent mortality. We used three outcome measures: 4-7-day, 4-14-day and 4-21-day mortality rates after the 0-3 days of initial hospitalization, respectively. Results: The study cohort consisted of 11,331 general ward patients. The non-survivors had higher initial NEWS score values in all outcome categories (all p < 0.001). The non-survivors had a rising trajectory in their NEWS values in all the outcome categories, whereas the survivors had a downward trajectory in their NEWS values in all outcome categories (data presented as first- and third-day's median values): an increase from 5.0 to 6.0 vs. a decrease from 1.5 to 1.0 (4-7-day non-survivors vs. survivors), an increase from 4.0 to 5.0 vs. a decrease from 1.5 to 1.0 (4-14-day non-survivors vs. survivors) and an increase from 4.0 to 5.0 vs. a decrease from 1.5 to 1.0 (4-21-day non-survivors vs. survivors). In the linear regression model, these differences in trends were statistically significant in all the outcome categories (p < 0.05). Conclusion: The NEWS score trajectory during the first three days of general ward admission is associated with patient outcome. Further studies are warranted to determine specific thresholds for clinically relevant changes in the NEWS trajectories.

Ability of the National Early Warning Score and its respiratory and haemodynamic subcomponents to predict short-term mortality on general wards: a prospective three-centre observational study in Finland.

OBJECTIVES: To validate the ability of the National Early Warning Score (NEWS) to predict short-term mortality on hospital wards, with a special reference to the NEWS's respiratory and haemodynamic subcomponents. DESIGN: A large, 1-year, prospective, observational three-centre study. First measured vital sign datasets on general wards were prospectively collected using a mobile solution system during routine patient care. Area under receiver operator characteristic curves were constructed, and comparisons between ROC curves were conducted with Delong's test for two correlated ROC curves. SETTING: One university hospital and two regional hospitals in Finland. PARTICIPANTS: All 19 001 adult patients admitted to 45 general wards in the three hospitals over the 1-year study period. After excluding 102/19 001 patients (0.53%) with data on some vital signs missing, the final cohort consisted of 18 889 patients with full datasets. PRIMARY AND SECONDARY OUTCOME MEASURES: The primary outcome measure was 1-day mortality and secondary outcomes were 2-day and 30-day mortality rates. RESULTS: Patients' median age was 70 years, 51% were male and 31% had a surgical reason for admission. The 1-day mortality was 0.36% and the 30-day mortality was 3.9%. The NEWS discriminated 1-day non-survivors with excellent accuracy (AUROC 0.91, 95% CI 0.87 to 0.95) and 30-day mortality with acceptable accuracy (0.75, 95% CI 0.73 to 0.77). The NEWS's respiratory rate component discriminated 1-day non-survivors better (0.78, 95% CI 0.72 to 0.84) as compared with the oxygen saturation (0.66, 95% CI 0.59 to 0.73), systolic blood pressure (0.65, 95% CI 0.59 to 0.72) and heart rate (0.67, 95% CI 0.61 to 0.74) subcomponents (p<0.01 in all ROC comparisons). As with the total NEWS, the discriminative performance of the individual score components decreased substantially for the 30-day mortality. CONCLUSIONS: NEWS discriminated general ward patients at risk for acute death with excellent statistical accuracy. The respiratory rate component is especially strongly associated with short-term mortality. TRIAL REGISTRATION NUMBER: NCT04055350.

Multiomics and digital monitoring during lifestyle changes reveal independent dimensions of human biology and health.

We explored opportunities for personalized and predictive health care by collecting serial clinical measurements, health surveys, genomics, proteomics, autoantibodies, metabolomics, and gut microbiome data from 96 individuals who participated in a data-driven health coaching program over a 16-month period with continuous digital monitoring of activity and sleep. We generated a resource of >20,000 biological samples from this study and a compendium of >53 million primary data points for 558,032 distinct features. Multiomics factor analysis revealed distinct and independent molecular factors linked to obesity, diabetes, liver function, cardiovascular disease, inflammation, immunity, exercise, diet, and hormonal effects. For example, ethinyl estradiol, a common oral contraceptive, produced characteristic molecular and physiological effects, including increased levels of inflammation and impact on thyroid, cortisol levels, and pulse, that were distinct from other sources of variability observed in our study. In total, this work illustrates the value of combining deep molecular and digital monitoring of human health. A record of this paper's transparent peer review process is included in the supplemental information.

Machine learning model predicts short-term mortality among prehospital patients: A prospective development study from Finland.

AIM: To show whether adding blood glucose to the National Early Warning Score (NEWS) parameters in a machine learning model predicts 30-day mortality more precisely than the standard NEWS in a prehospital setting. METHODS: In this study, vital sign data prospectively collected from 3632 unselected prehospital patients in June 2015 were used to compare the standard NEWS to random forest models for predicting 30-day mortality. The NEWS parameters and blood glucose levels were used to develop the random forest models. Predictive performance on an unknown patient population was estimated with a ten-fold stratified cross-validation method. RESULTS: All NEWS parameters and blood glucose levels were reported in 2853 (79%) eligible patients. Within 30 days after contact with ambulance staff, 97 (3.4%) of the analysed patients had died. The area under the receiver operating characteristic curve for the 30-day mortality of the evaluated models was 0.682 (95% confidence interval [CI], 0.619-0.744) for the standard NEWS, 0.735 (95% CI, 0.679-0.787) for the random forest-trained NEWS parameters only and 0.758 (95% CI, 0.705-0.807) for the random forest-trained NEWS parameters and blood glucose. The models predicted secondary outcomes similarly, but adding blood glucose into the random forest model slightly improved its performance in predicting short-term mortality. CONCLUSIONS: Among unselected prehospital patients, a machine learning model including blood glucose and NEWS parameters had a fair performance in predicting 30-day mortality.

Random forest machine learning method outperforms prehospital National Early Warning Score for predicting one-day mortality: A retrospective study.

AIM OF THE STUDY: The National Early Warning Score (NEWS) is a validated method for predicting clinical deterioration in hospital wards, but its performance in prehospital settings remains controversial. Modern machine learning models may outperform traditional statistical analyses for predicting short-term mortality. Thus, we aimed to compare the mortality prediction accuracy of NEWS and random forest machine learning using prehospital vital signs. METHODS: In this retrospective study, all electronic ambulance mission reports between 2008 and 2015 in a single EMS system were collected. Adult patients (≥ 18 years) were included in the analysis. Random forest models with and without blood glucose were compared to the traditional NEWS for predicting one-day mortality. A ten-fold cross-validation method was applied to train and validate the random forest models. RESULTS: A total of 26,458 patients were included in the study of whom 278 (1.0%) died within one day of ambulance mission. The area under the receiver operating characteristic curve for one-day mortality was 0.836 (95% CI, 0.810-0.860) for NEWS, 0.858 (95% CI, 0.832-0.883) for a random forest trained with NEWS variables only and 0.868 (0.843-0.892) for a random forest trained with NEWS variables and blood glucose. CONCLUSION: A random forest algorithm trained with NEWS variables was superior to traditional NEWS for predicting one-day mortality in adult prehospital patients, although the risk of selection bias must be acknowledged. The inclusion of blood glucose in the model further improved its predictive performance.