Dataset Information

Machine learning prediction of the adverse outcome for nontraumatic subarachnoid hemorrhage patients.

ABSTRACT: OBJECTIVE:Subarachnoid hemorrhage (SAH) is often devastating with increased early mortality, particularly in those with presumed delayed cerebral ischemia (DCI). The ability to accurately predict survival for SAH patients during the hospital course would provide valuable information for healthcare providers, patients, and families. This study aims to utilize electronic health record (EHR) data and machine learning approaches to predict the adverse outcome for nontraumatic SAH adult patients. METHODS:The cohort included nontraumatic SAH patients treated with vasopressors for presumed DCI from a large EHR database, the Cerner Health Facts® EMR database (2000-2014). The outcome of interest was the adverse outcome, defined as death in hospital or discharged to hospice. Machine learning-based models were developed and primarily assessed by area under the receiver operating characteristic curve (AUC). RESULTS:A total of 2467 nontraumatic SAH patients (64% female; median age [interquartile range]: 56 [47-66]) who were treated with vasopressors for presumed DCI were included in the study. 934 (38%) patients died or were discharged to hospice. The model achieved an AUC of 0.88 (95% CI, 0.84-0.92) with only the initial 24 h EHR data, and 0.94 (95% CI, 0.92-0.96) after the next 24 h. INTERPRETATION:EHR data and machine learning models can accurately predict the risk of the adverse outcome for critically ill nontraumatic SAH patients. It is possible to use EHR data and machine learning techniques to help with clinical decision-making.

SUBMITTER: Yu D

PROVIDER: S-EPMC7664270 | biostudies-literature | 2020 Nov

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Machine learning prediction of the adverse outcome for nontraumatic subarachnoid hemorrhage patients.

Yu Duo D Williams George W GW Aguilar David D Yamal José-Miguel JM Maroufy Vahed V Wang Xueying X Zhang Chenguang C Huang Yuefan Y Gu Yuxuan Y Talebi Yashar Y Wu Hulin H

Annals of clinical and translational neurology 20200929 11

<h4>Objective</h4>Subarachnoid hemorrhage (SAH) is often devastating with increased early mortality, particularly in those with presumed delayed cerebral ischemia (DCI). The ability to accurately predict survival for SAH patients during the hospital course would provide valuable information for healthcare providers, patients, and families. This study aims to utilize electronic health record (EHR) data and machine learning approaches to predict the adverse outcome for nontraumatic SAH adult patie ...[more]

PMID: 32990362

Similar Datasets

Project description:Reliable prediction of outcomes of aneurysmal subarachnoid hemorrhage (aSAH) based on factors available at patient admission may support responsible allocation of resources as well as treatment decisions. Radiographic and clinical scoring systems may help clinicians estimate disease severity, but their predictive value is limited, especially in devising treatment strategies. In this study, we aimed to examine whether a machine learning (ML) approach using variables available on admission may improve outcome prediction in aSAH compared to established scoring systems. Combined clinical and radiographic features as well as standard scores (Hunt & Hess, WFNS, BNI, Fisher, and VASOGRADE) available on patient admission were analyzed using a consecutive single-center database of patients that presented with aSAH (n = 388). Different ML models (seven algorithms including three types of traditional generalized linear models, as well as a tree bosting algorithm, a support vector machine classifier (SVMC), a Naive Bayes (NB) classifier, and a multilayer perceptron (MLP) artificial neural net) were trained for single features, scores, and combined features with a random split into training and test sets (4:1 ratio), ten-fold cross-validation, and 50 shuffles. For combined features, feature importance was calculated. There was no difference in performance between traditional and other ML applications using traditional clinico-radiographic features. Also, no relevant difference was identified between a combined set of clinico-radiological features available on admission (highest AUC 0.78, tree boosting) and the best performing clinical score GCS (highest AUC 0.76, tree boosting). GCS and age were the most important variables for the feature combination. In this cohort of patients with aSAH, the performance of functional outcome prediction by machine learning techniques was comparable to traditional methods and established clinical scores. Future work is necessary to examine input variables other than traditional clinico-radiographic features and to evaluate whether a higher performance for outcome prediction in aSAH can be achieved.

Project description:Background Previous studies of patients with nontraumatic subarachnoid hemorrhage (SAH) suggest better outcomes at hospitals with higher case and procedural volumes, but the shape of the volume-outcome curve has not been defined. We sought to establish minimum volume criteria for SAH and aneurysm obliteration procedures that could be used for comprehensive stroke center certification. Methods and Results Data from 8512 discharges in the National Inpatient Sample (NIS) from 2010 to 2011 were analyzed using logistic regression models to evaluate the association between clinical outcomes (in-hospital mortality and the NIS-SAH Outcome Measure [NIS-SOM]) and measures of hospital annual case volume (nontraumatic SAH discharges, coiling, and clipping procedures). Sensitivity and specificity analyses for the association of desirable outcomes with different volume thresholds were performed. During 8512 SAH hospitalizations, 28.7% of cases underwent clipping and 20.1% underwent coiling with rates of 21.2% for in-hospital mortality and 38.6% for poor outcome on the NIS-SOM. The mean (range) of SAH, coiling, and clipping annual case volumes were 30.9 (1-195), 8.7 (0-94), and 6.1 (0-69), respectively. Logistic regression demonstrated improved outcomes with increasing annual case volumes of SAH discharges and procedures for aneurysm obliteration, with attenuation of the benefit beyond 35 SAH cases/year. Analysis of sensitivity and specificity using different volume thresholds confirmed these results. Analysis of previously proposed volume thresholds, including those utilized as minimum standards for comprehensive stroke center certification, showed that hospitals with more than 35 SAH cases annually had consistently superior outcomes compared with hospitals with fewer cases, although some hospitals below this threshold had similar outcomes. The adjusted odds ratio demonstrating lower risk of poor outcomes with SAH annual case volume ≥35 compared with 20 to 34 was 0.82 for the NIS-SOM (95% CI, 0.71-094; P=0.0054) and 0.80 (95% CI, 0.68-0.93; P=0.0055) for in-hospital mortality. Conclusions Outcomes for patients with SAH improve with increasing hospital case volumes and procedure volumes, with consistently better outcomes for hospitals with more than 35 SAH cases per year.

Project description:Background and purposeAnimal and observational studies indicate that smoking is a risk factor for aneurysm formation and rupture, leading to nontraumatic subarachnoid hemorrhage (SAH). However, a definitive causal relationship between smoking and the risk of SAH has not been established. Using Mendelian randomization (MR) analyses, we tested the hypothesis that smoking is causally linked to the risk of SAH.MethodsWe conducted a 1-sample MR study using data from the UK Biobank, a large cohort study that enrolled over 500 000 Britons aged 40 to 69 from 2006 to 2010. Participants of European descent were included. SAH cases were ascertained using a combination of self-reported, electronic medical record, and death registry data. As the instrument, we built a polygenic risk score using independent genetic variants known to associate (P<5×10-8) with smoking behavior. This polygenic risk score represents the genetic susceptibility to smoking initiation. The primary MR analysis utilized the ratio method. Secondary MR analyses included the inverse variance weighted and weighted median methods.ResultsA total of 408 609 study participants were evaluated (mean age, 57 [SD 8], female sex, 220 937 [54%]). Among these, 132 566 (32%) ever smoked regularly, and 904 (0.22%) had a SAH. Each additional SD of the smoking polygenic risk score was associated with 21% increased risk of smoking (odds ratio [OR], 1.21 [95% CI, 1.20-1.21]; P<0.001) and a 10% increased risk of SAH (OR, 1.10 [95% CI, 1.03-1.17]; P=0.006). In the primary MR analysis, genetic susceptibility to smoking was associated with a 63% increase in the risk of SAH (OR, 1.63 [95% CI, 1.15-2.31]; P=0.006). Secondary analyses using the inverse variance weighted method (OR, 1.57 [95% CI, 1.13-2.17]; P=0.007) and the weighted median method (OR, 1.74 [95% CI, 1.06-2.86]; P=0.03) yielded similar results. There was no significant pleiotropy (MR-Egger intercept P=0.39; MR Pleiotropy Residual Sum and Outlier global test P=0.69).ConclusionsThese findings provide evidence for a causal link between smoking and the risk of SAH.

Project description:BACKGROUND:Prognostication is an essential tool for risk adjustment and decision making in the intensive care unit (ICU). Research into prognostication in ICU has so far been limited to data from admission or the first 24 hours. Most ICU admissions last longer than this, decisions are made throughout an admission, and some admissions are explicitly intended as time-limited prognostic trials. Despite this, temporal changes in prognostic ability during ICU admission has received little attention to date. Current predictive models, in the form of prognostic clinical tools, are typically derived from linear models and do not explicitly handle incremental information from trends. Machine learning (ML) allows predictive models to be developed which use non-linear predictors and complex interactions between variables, thus allowing incorporation of trends in measured variables over time; this has made it possible to investigate prognosis throughout an admission. METHODS AND FINDINGS:This study uses ML to assess the predictability of ICU mortality as a function of time. Logistic regression against physiological data alone outperformed APACHE-II and demonstrated several important interactions including between lactate & noradrenaline dose, between lactate & MAP, and between age & MAP consistent with the current sepsis definitions. ML models consistently outperformed logistic regression with Deep Learning giving the best results. Predictive power was maximal on the second day and was further improved by incorporating trend data. Using a limited range of physiological and demographic variables, the best machine learning model on the first day showed an area under the receiver-operator characteristic curve (AUC) of 0.883 (σ = 0.008), compared to 0.846 (σ = 0.010) for a logistic regression from the same predictors and 0.836 (σ = 0.007) for a logistic regression based on the APACHE-II score. Adding information gathered on the second day of admission improved the maximum AUC to 0.895 (σ = 0.008). Beyond the second day, predictive ability declined. CONCLUSION:This has implications for decision making in intensive care and provides a justification for time-limited trials of ICU therapy; the assessment of prognosis over more than one day may be a valuable strategy as new information on the second day helps to differentiate outcomes. New ML models based on trend data beyond the first day could greatly improve upon current risk stratification tools.

Project description:The diagnosis of cerebral vasospasm after Subarachnoid-Hemorrhage is currently very difficult, additional tools such as blood biomarkers are necessary. We tested the ability of gene expression profiles of blood cells to predict vasospasm. 32 patients suffering subarachnoid-hemorrhage were included in this prospective monocentre study. They were grouped according to have a complicated cerebral vasospasm (Vasospasm) or not (Control) and Paired according to age (+/- 10 years), sex, Fisher grade (+/- 1), location, smoking (at least 3 first parameters). Gene expression profiles of blood cells were determined using 25,000~gene microarray. Blood sample: 2.5 mL harvested in PAXgene® Blood RNA tubes (PreAnalytix) RNA extraction: PAXgene® Blood RNA kit (Qiagen). We used a Universel Reference RNA (Stratagene). RNA amplification and labelling: kit Amino Allyl MessageAmp II (Ambion). We hybridized 4 microarrays per patient using pangenomic microarrays from the "Réseau National des Génopôles" (Illkirch, France). 2 slides were hybridized with reference RNA labelled Cy3 and patient RNA labelled Cy5, and 2 slides were hybridized with reference RNA labelled Cy5 and patient RNA labelled Cy3. Hybridation : Agilent protocol with few modifications : 750 ng of each labelled RNA were hubriddized at 60°C during 17 hours in an Aglient hybridization oven. After washings, Slides were scanned with a GenePix 4000B scanner (Molecular Devices). Image intensity data were extracted with GenePix Pro 6.0 analysis software. Quantification of Cy3 and Cy5 and selection of good spots were performed using the MAIA software (Novikov E and Barillot E. Software package for automatic microarray image analysis (MAIA). The ACUITY software was then used to normalize log ratios Cy3/Cy5 with Lowess non linear normalization, to filter out genes not present in at least 3 slides out of 4, to evaluate the reproducibility of the 4 microarrays of each patient (hierarchical clustering, Self Organizing Maps). Statistical analyses to insure reproducibility was performed using Excel (correlation coefficients, ANOVA). Only slides that passed all reprocubility tests were validated.

Dataset Information

Machine learning prediction of the adverse outcome for nontraumatic subarachnoid hemorrhage patients.

Publications

Machine learning prediction of the adverse outcome for nontraumatic subarachnoid hemorrhage patients.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets