Machine Learning Automated Detection of Large Vessel Occlusion From Mobile Stroke Unit Computed Tomography Angiography

Background and Purpose: Prehospital automated large vessel occlusion (LVO) detection in Mobile Stroke Units (MSUs) could accelerate identification and treatment of patients with LVO acute ischemic stroke. Here, we evaluate the performance of a machine learning (ML) model on CT angiograms (CTAs) obtained from 2 MSUs to detect LVO. Methods: Patients evaluated on MSUs in Houston and Los Angeles with out-of-hospital CTAs were identified. Anterior circulation LVO was defined as an occlusion of the intracranial internal carotid artery, middle cerebral artery (M1 or M2), or anterior cerebral artery vessels and determined by an expert human reader. A ML model to detect LVO was trained and tested on independent data sets consisting of in-hospital CTAs and then tested on MSU CTA images. Model performance was determined using area under the receiver-operator curve statistics. Results: Among 68 patients with out-of-hospital MSU CTAs, 40% had an LVO. The most common occlusion location was the middle cerebral artery M1 segment (59%), followed by the internal carotid artery (30%), and middle cerebral artery M2 (11%). Median time from last known well to CTA imaging was 88.0 (interquartile range, 59.5–196.0) minutes. After training on 870 in-hospital CTAs, the ML model performed well in identifying LVO in a separate in-hospital data set of 441 images with area under receiver-operator curve of 0.84 (95% CI, 0.80–0.87). ML algorithm analysis time was under 1 minute. The performance of the ML model on the MSU CTA images was comparable with area under receiver-operator curve 0.80 (95% CI, 0.71–0.89). There was no significant difference in performance between the Houston and Los Angeles MSU CTA cohorts. Conclusions: In this study of patients evaluated on MSUs in 2 cities, a ML algorithm was able to accurately and rapidly detect LVO using prehospital CTA acquisitions.

Circulating proteins to predict adverse COVID-19 outcomes

Predicting COVID-19 severity is difficult, and the biological pathways involved are not fully understood. To approach this problem, we measured 4,701 circulating human protein abundances in two independent cohorts totaling 986 individuals. We then trained prediction models including protein abundances and clinical risk factors to predict adverse COVID-19 outcomes in 417 subjects and tested these models in a separate cohort of 569 individuals. For severe COVID-19, a baseline model including age and sex provided an area under the receiver operator curve (AUC) of 65% in the test cohort. Selecting 92 proteins from the 4,701 unique protein abundances improved the AUC to 88% in the training cohort, which remained relatively stable in the testing cohort at 86%, suggesting good generalizability. Proteins selected from different adverse COVID-19 outcomes were enriched for cytokine and cytokine receptors, but more than half of the enriched pathways were not immune-related. Taken together, these findings suggest that circulating proteins measured at early stages of disease progression are reasonably accurate predictors of adverse COVID-19 outcomes. Further research is needed to understand how to incorporate protein measurement into clinical care.

Machine Learning Applied to Electronic Health Records: Identification of Chemotherapy Patients at High Risk for Preventable Emergency Department Visits and Hospital Admissions

PURPOSE Acute care use (ACU) is a major driver of oncologic costs and is penalized by a Centers for Medicare & Medicaid Services quality measure, OP-35. Targeted interventions reduce preventable ACU; however, identifying which patients might benefit remains challenging. Prior predictive models have made use of a limited subset of the data in the electronic health record (EHR). We aimed to predict risk of preventable ACU after starting chemotherapy using machine learning (ML) algorithms trained on comprehensive EHR data. METHODS Chemotherapy patients treated at an academic institution and affiliated community care sites between January 2013 and July 2019 who met inclusion criteria for OP-35 were identified. Preventable ACU was defined using OP-35 criteria. Structured EHR data generated before chemotherapy treatment were obtained. ML models were trained to predict risk for ACU after starting chemotherapy using 80% of the cohort. The remaining 20% were used to test model performance by the area under the receiver operator curve. RESULTS Eight thousand four hundred thirty-nine patients were included, of whom 35% had preventable ACU within 180 days of starting chemotherapy. Our primary model classified patients at risk for preventable ACU with an area under the receiver operator curve of 0.783 (95% CI, 0.761 to 0.806). Performance was better for identifying admissions than emergency department visits. Key variables included prior hospitalizations, cancer stage, race, laboratory values, and a diagnosis of depression. Analyses showed limited benefit from including patient-reported outcome data and indicated inequities in outcomes and risk modeling for Black and Medicaid patients. CONCLUSION Dense EHR data can identify patients at risk for ACU using ML with promising accuracy. These models have potential to improve cancer care outcomes, patient experience, and costs by allowing for targeted, preventative interventions.

Predicting clinical outcomes among hospitalized COVID-19 patients using both local and published models

Background Many models are published which predict outcomes in hospitalized COVID-19 patients. The generalizability of many is unknown. We evaluated the performance of selected models from the literature and our own models to predict outcomes in patients at our institution. Methods We searched the literature for models predicting outcomes in inpatients with COVID-19. We produced models of mortality or criticality (mortality or ICU admission) in a development cohort. We tested external models which provided sufficient information and our models using a test cohort of our most recent patients. The performance of models was compared using the area under the receiver operator curve (AUC). Results Our literature review yielded 41 papers. Of those, 8 were found to have sufficient documentation and concordance with features available in our cohort to implement in our test cohort. All models were from Chinese patients. One model predicted criticality and seven mortality. Tested against the test cohort, internal models had an AUC of 0.84 (0.74–0.94) for mortality and 0.83 (0.76–0.90) for criticality. The best external model had an AUC of 0.89 (0.82–0.96) using three variables, another an AUC of 0.84 (0.78–0.91) using ten variables. AUC’s ranged from 0.68 to 0.89. On average, models tested were unable to produce predictions in 27% of patients due to missing lab data. Conclusion Despite differences in pandemic timeline, race, and socio-cultural healthcare context some models derived in China performed well. For healthcare organizations considering implementation of an external model, concordance between the features used in the model and features available in their own patients may be important. Analysis of both local and external models should be done to help decide on what prediction method is used to provide clinical decision support to clinicians treating COVID-19 patients as well as what lab tests should be included in order sets.

A Vasoactive Inotropic Score Predicts the Severity of Hypotensive Shock and Mortality in Preterm Infants

Objective: To validate the prediction of the severity of hypotensive shock and mortality using the vasoactive inotropic score in preterm infants.Methods: In this retrospective study we calculated the vasoactive inotropic score (VIS) and cumulative exposure to cardiovascular medications over time (VISct) in a cohort of preterm infants with hypotensive shock who received a cardiovascular support. Receiver operator curve was constructed to predict the primary outcome which was death due to hypotensive shock. Results: VIS had an area under the curve of 0.73 (95% CI 0.85-0.98, p < 0.001). A VIS cut off of 25 has sensitivity and specificity of 66% and 92%, and positive and negative predictive values of 78.5% and 83%, respectively.Conclusion: High VIS predicts high mortality rate due to irreversible shock in preterm infants

Quantitative morphometrics reveals glomerular changes in patients with infrequent segmentally sclerosed glomeruli

AimsDetection of one segmentally sclerosed glomerulus (SSG) identifies patients with focal segmental glomerulosclerosis (FSGS) but rare SSGs may be missed in kidney biopsies. It is unknown whether alterations of unaffected glomeruli in patients with infrequent SSG can be detected by quantitative morphometrics.MethodsWe determined SSG frequency and obtained quantitative morphometrics in glomeruli without a pathologic phenotype in large kidney sections of non-involved kidney tissue from 137 patients undergoing total nephrectomy. We used multivariate modelling to identify morphometrics independently associated with increasing frequency of SSG and Receiver Operator Curve (ROC) analysis to determine the ability of quantitative morphometrics to identify patients with FSGS. We used the geometric distribution to estimate the sensitivity and specificity of a needle biopsy to identify patients with FSGS.ResultsIn seventy-one patients (51.8%), at least one SSG was observed, and of those, 39 (54.9%) had an SSG lesion in less than 2% of all glomeruli (mean of 249 glomeruli per specimen). Increasing percent of SSG was independently associated with decreasing podocyte density and increasing mesangial index in multivariate modelling. For infrequent SSG lesions (<1% of glomeruli), kidney biopsy could miss FSGS diagnosis more than 74% of the time, and podocyte density had an area under the curve (AUC) of 0.77, and mesangial index, an AUC of 0.79 to identify patients with FSGS.ConclusionsMore than half of patients had FSGS, although 30% had infrequent SSG. Quantitative morphometrics in glomeruli without pathology, such as podocyte density and mesangial index, identified patients with infrequent SSG and may serve as clinical markers to identify patients with FSGS.

A retrospective study on transabdominal ultrasound measurements of the rumen wall thickness to evaluate chronic rumen acidosis in beef cattle

Background Chronic and subacute rumen acidosis are economically important in the beef industry. The aim of this study was to evaluate the potential suitability of the transabdominal ultrasonographic examination of the ruminal wall to diagnose chronic rumen acidosis in beef cattle compared to direct measurement of ruminal pH, as a fast non-invasive tool to be used in field condition. Ultrasonographic examination of the rumen was conducted in 478 beef cattle before rumenocentesis (chronic rumen acidosis group = pH ≤ 5.8; healthy group = pH ≥ 5.9). Rumen wall ultrasound measurements included rumen wall thickness (RWT) and rumen mucosa and submucosa thickness (RMST). Results The Analysis of Variance showed the high significant effect of the pH class for RWT and RMST (P < 0.001). Spearman RANK correlation analysis showed interaction between rumen pH and RWT (− 0.71; P < 0.0001) and RMST (− 0.75; P < 0.0001). A significant Spearman’s correlations were found between volatile fatty acids (VFA) and RWT and RMST. The differentiation efficiency of RWT between healthy and chronic rumen acidosis groups, as a result of the receiver operator curve (ROC) analysis, was quite good with an area under the receiver operator curve (AUROC) of 0.88: P < 0.0001; 95% CI: 0.83–0.98. Using a cut-off value of > 8.2 mm. The differentiation efficiency of RMST between healthy and chronic rumen acidosis groups, as a result of ROC curve analysis, was good with an AUROC of 0.90: p < 0.0001; 95% CI: 0.85–0.94. Using a cut-off value of > 5.3 mm. Conclusions In this study, the thickening of RWT and RMST is correlated with the changes of ruminal pH. Transabdominal rumen ultrasound has the potential to become a powerful diagnostic tool useful to identify fattening bulls affected by chronic rumen acidosis.

P847Performance of the GRACE score in patients with type 1 and type 2 myocardial infarction

Introduction The Global Registry of Acute Coronary Events (GRACE) score is a widely used risk stratification tool in the evaluation of patients with myocardial infarction. However, the performance of GRACE in patients with myocardial infarction secondary to oxygen supply-demand imbalance in the absence of atherothrombosis (type 2 myocardial infarction), is uncertain. Outcomes in patients with type 2 myocardial infarction are poor, and a risk stratification tool is urgently required. Methods We assessed the GRACE score in two cohorts of consecutive patients presenting to the Emergency Department with suspected acute coronary syndrome. One cohort was recruited as part of a stepped wedge cluster randomised controlled trial across ten hospitals in Scotland, and one cohort from a large tertiary centre in Sweden. All diagnoses were adjudicated according to the Fourth Universal Definition. We applied the GRACE 2.0 algorithm to estimate death at one year. We assessed model discrimination using the area under the receiver-operator-curve (AUC), and compared AUC between type 1 and type 2 myocardial infarction using the DeLong test. Calibration was assessed using the Hosmer-Lemeshow (HL) goodness of fit test. Results We identified 2,538 and 1,080 patients with type 1 myocardial infarction from the Scottish and Swedish cohorts, with death from any cause occurring in 378 (14.9%) and 112 (10.4%) patients, respectively. The AUC for the GRACE score was 0.843 (0.823–0.864) and 0.848 (0.810–0.886). There were 642 and 247 patients with type 2 myocardial infarction in the Scottish and Swedish cohorts, respectively, with death occurring in 144 (22.4%) and 57 (23.1%) patients. The AUC was 0.708 (0.662–0.754) and 0.733 (0.657–0.808), (P<0.001 for both compared to type 1 myocardial infarction). The results of the HL Test suggest that the calibration of the GRACE 2.0 score needs further improvement (Table). Evaluation of GRACE 2.0 algorithm Type 1 Myocardial Infarction Type 2 Myocardial Infarction Scotland (n=2,538) Sweden (n=1,080) Scotland (n=642) Sweden (n=247) Deaths 378 (14.9%) 112 (10.4%) 144 (22.4%) 57 (23.0%) AUC (C-statistic) 0.843 (0.823–0.864) 0.848 (0.810–0.886) 0.708 (0.662–0.754) 0.733 (0.657–0.808) P-value for HL <0.001 <0.001 <0.001 <0.001 AUC: Area Under the receiver-operator Curve; HL: Hosmer-Lemeshow test. Figure 1. ROC curves Conclusions The GRACE score provided excellent discrimination for all cause death at one year in two contemporary consecutive patient cohorts with tye 1 myocardial infarction. In patients with type 2 myocardial infarction, GRACE performed well, but recalibration or the development of novel risk scores has the potential to improve risk stratification.

Factors Associated With Intraoperative Conversion to Total Thyroidectomy in Benign Goiters

Background: Decision to undertake total thyroidectomy when gross inspection of the gland raises suspicion of widespread degenerative changes is often intraoperative. Knowing the factors associated with intraoperative conversion to total thyroidectomy may assist preoperative counselling. This study describes the probability of conversion to total thyroidectomy and factors associated with con-version among patients hitherto planned for partial thyroidectomy. Methods: We reviewed 191 records and extracted data on patient demographics, the pre-operative radiograph findings, the weight of excised gland and the operation performed. Descriptive and inferential statistics were performed. Receiver operator curve was used to assess for cut-off point. P-value was set at 0.05. Results: A total of 191 records was reviewed consisting of 181 females (94.8% 95% CI 90.6-97.5) and 10 males (5.2%, 95%CI 2.5-9.4). Only nodular goiters required conversion to total thyroidectomy. The over-all probability of total thyroidectomy was 11%(95% CI 7.0-16.3). The probability of total thyroidectomy in female was 10.5%(95% CI 6.4-16.9) while in male was 20%(95% CI2.5-55.6). The probability of total thyroidectomy in a female with nodular goiter was 8.1%(95% CI 4.8-13.5), compared to 28.6%(95% CI 3.7-71) in males. The risk of total thyroidectomy was associated with the weight of the excised gland. Conclusion: Only nodular goiters required intraoperative conversion to total thyroidecto-my and the probability of conversion was higher in males.