Machine learning improves mortality risk prediction after cardiac surgery: Systematic review and meta-analysis.

BACKGROUND: Interest in the usefulness of machine learning (ML) methods for outcomes prediction has continued to increase in recent years. However, the advantage of advanced ML model over traditional logistic regression (LR) remains controversial. We performed a systematic review and meta-analysis of studies comparing the discrimination accuracy between ML models versus LR in predicting operative mortality following cardiac surgery. METHODS: The present systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analysis statement. Discrimination ability was assessed using the C-statistic. Pooled C-statistics and its 95% credibility interval for ML models and LR were obtained were obtained using a Bayesian framework. Pooled estimates for ML models and LR were compared to inform on difference between the 2 approaches. RESULTS: We identified 459 published citations of which 15 studies met inclusion criteria and were used for the quantitative and qualitative analysis. When the best ML model from individual study was used, meta-analytic estimates showed that ML were associated with a significantly higher C-statistic (ML, 0.88; 95% credibility interval, 0.83-0.93 vs LR, 0.81; 95% credibility interval, 0.77-0.85; P = .03). When individual ML algorithms were instead selected, we found a nonsignificant trend toward better prediction with each of ML algorithms. We found no evidence of publication bias (P = .70). CONCLUSIONS: The present findings suggest that when compared with LR, ML models provide better discrimination in mortality prediction after cardiac surgery. However, the magnitude and clinical influence of such an improvement remains uncertain.

Development of a Telemedicine Screening Program During the COVID-19 Pandemic.

Background: Telemedicine use increased during the COVID-19 pandemic due to concerns for patient and provider safety. Given the lack of testing resources initially and the large geographical range served by Augusta University (AU), a telemedicine platform with up-to-date screening guidelines was implemented for COVID-19 testing in March 2020. Our objective was to understand the level of adherence to telemedicine screening guidelines for COVID-19. Methods: The study population included health care providers and population who participated in an encounter in the AU Health Express Care virtual care program from March 22 to May 21, 2020. All encounters were intended to be for COVID-19 screening, free, and available 24 h per day, 7 days per week. Screening guidelines were developed by AU based on information from the Centers for Disease Control and Prevention and the Georgia Department of Public Health. Results: Among 17,801 total encounters, 13,600 were included in the final analysis. Overall adherence to screening guidelines was 71% in the adult population and 57% in the pediatric population. When providers did not follow guidelines, 72% determined that the patient should have a positive screen. Guidelines themselves determined that only 52% of encounters should have a positive screen. Providers' specialty significantly correlated with guideline adherence (p = 0.002). Departments with the highest adherence were psychiatry, neurology, and ophthalmology. No significant correlation was found between guideline adherence and provider degree/position. Conclusions: This study provides proof of concept of a free telehealth screening platform during an ongoing pandemic. Our screening experience was effective and different specialties participated. Our patient population lived in lower than average income zip codes, suggesting that our free telemedicine screening program successfully reached populations with higher financial barriers to health care. Early training and a posteriori knowledge of telemedicine was likely key to screening guideline adherence.

EpiGraphDB: a database and data mining platform for health data science.

MOTIVATION: The wealth of data resources on human phenotypes, risk factors, molecular traits and therapeutic interventions presents new opportunities for population health sciences. These opportunities are paralleled by a growing need for data integration, curation and mining to increase research efficiency, reduce mis-inference and ensure reproducible research. RESULTS: We developed EpiGraphDB (https://epigraphdb.org/), a graph database containing an array of different biomedical and epidemiological relationships and an analytical platform to support their use in human population health data science. In addition, we present three case studies that illustrate the value of this platform. The first uses EpiGraphDB to evaluate potential pleiotropic relationships, addressing mis-inference in systematic causal analysis. In the second case study, we illustrate how protein-protein interaction data offer opportunities to identify new drug targets. The final case study integrates causal inference using Mendelian randomization with relationships mined from the biomedical literature to 'triangulate' evidence from different sources. AVAILABILITY AND IMPLEMENTATION: The EpiGraphDB platform is openly available at https://epigraphdb.org. Code for replicating case study results is available at https://github.com/MRCIEU/epigraphdb as Jupyter notebooks using the API, and https://mrcieu.github.io/epigraphdb-r using the R package. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Can machine learning improve mortality prediction following cardiac surgery?

OBJECTIVES: Interest in the clinical usefulness of machine learning for risk prediction has bloomed recently. Cardiac surgery patients are at high risk of complications and therefore presurgical risk assessment is of crucial relevance. We aimed to compare the performance of machine learning algorithms over traditional logistic regression (LR) model to predict in-hospital mortality following cardiac surgery. METHODS: A single-centre data set of prospectively collected information from patients undergoing adult cardiac surgery from 1996 to 2017 was split into 70% training set and 30% testing set. Prediction models were developed using neural network, random forest, naive Bayes and retrained LR based on features included in the EuroSCORE. Discrimination was assessed using area under the receiver operating characteristic curve, and calibration analysis was undertaken using the calibration belt method. Model calibration drift was assessed by comparing Goodness of fit χ2 statistics observed in 2 equal bins from the testing sample ordered by procedure date. RESULTS: A total of 28 761 cardiac procedures were performed during the study period. The in-hospital mortality rate was 2.7%. Retrained LR [area under the receiver operating characteristic curve 0.80; 95% confidence interval (CI) 0.77-0.83] and random forest model (0.80; 95% CI 0.76-0.83) showed the best discrimination. All models showed significant miscalibration. Retrained LR proved to have the weakest calibration drift. CONCLUSIONS: Our findings do not support the hypothesis that machine learning methods provide advantage over LR model in predicting operative mortality after cardiac surgery.

Gender-specific structural abnormalities in major depressive disorder revealed by fixel-based analysis.

BACKGROUND: Major depressive disorder (MDD) is a chronic disease with a large global impact. There are currently no clinically useful predictors of treatment outcome, and the development of biomarkers to inform clinical treatment decisions is highly desirable. METHODS: In this exploratory study we performed fixel-based analysis of diffusion MRI data from the International Study to Predict Optimized Treatment in Depression with the aim of identifying novel biomarkers at baseline that may relate to diagnosis and outcome to treatment with antidepressant medications. Analyses used MR data from individuals with MDD (n = 221) and healthy controls (n = 67). RESULTS: We show focal, gender-specific differences in the anterior limb of the internal capsule (males) and bilaterally in the genu of the corpus callosum (females) associated with diagnosis. Lower fibre cross-section in the tapetum, the conduit between the right and left hippocampi, were also associated with a decreased probability of remission. Analysis of conventional fractional anisotropy showed scattered abnormalities in the corona radiata, cerebral peduncles and mid-brain which were much lower in total volume compared to fixel-based analysis. CONCLUSIONS: Fixel-based analysis appeared to identify different underlying abnormalities than conventional tensor-based metrics, with almost no overlap between significant regions. We show that MDD is associated with gender specific abnormalities in the genu of the corpus callosum (females) and in the anterior limb of the internal capsule (males), as well as gender-independent differences in the tapetum that predict remission. Diffusion MRI may play a key role in future guidance of clinical decision-making for MDD.

Progressing Preemptive Genotyping of CYP2C19 Allelic Variants for Sickle Cell Disease Patients.

AIMS: Interindividual variability in drug response and adverse effects have been described for proton pump inhibitors, anticonvulsants, selective serotonin reuptake inhibitors, tricyclic antidepressants, and anti-infectives, but little is known about the safety and efficacy of these medications in patients with sickle cell disease (SCD). We genotyped the CYP2C19 gene which has been implicated in the metabolism of these drugs in an SCD patient cohort to determine the frequencies of reduced function, increased function, or complete loss-of-function variants. MATERIALS AND METHODS: DNAs from 165 unrelated SCD patients were genotyped for nine CYP2C19 (*2, *3, *4, *5, *6, *7,*8, *12, and *17) alleles using the iPLEX® ADME PGx multiplex panel. RESULTS: Three CYP2C19 alleles (*2, *12, and *17) were detected with the following frequencies: 0.209, 0.006, and 0.236, respectively. The predicted phenotype frequencies were distributed as extensive (31.5%), intermediate (24.8%), poor (5.5%), ultrarapid (30.3%), and unknown metabolizers (7.9%). DISCUSSION: Prognostic genotyping is potentially useful for identifying SCD patients with allelic variants linked to proven clinical pharmacokinetic consequences for several drugs metabolized by the CYP2C19 gene. However, the main challenge to implementing a genetics-guided prescribing practice is ensuring concordance between CYP2C19 genotypes and metabolic phenotypes in SCD patients.

Preemptive Genotyping of CYP2C8 and CYP2C9 Allelic Variants Involved in NSAIDs Metabolism for Sickle Cell Disease Pain Management.

Interindividual variability in analgesic effects of nonsteroidal anti-inflammatory drugs prescribed for sickle cell disease (SCD) pain is attributed to polymorphisms in the CYP2C8 and CYP2C9 enzymes. We described CYP2C8 and CYP2C9 genotype/phenotype profiles and frequency of emergency department (ED) visits for pain management in an African American SCD patient cohort. DNA from 165 unrelated patients was genotyped for seven CYP2C8 and 15 CYP2C9 alleles using the iPLEX ADME PGx multiplexed panel. CYP2C8*1 (0.806),*2 (0.164), *3 (0.018), and *4 (0.012) alleles were identified. Genotype frequencies were distributed as homozygous wild type (66.7%), heterozygous (27.8%), and homozygous variant/compound heterozygous (5.4%), respectively. CYP2C9*1 (0.824), *2 (0.027), *3 (0.012), *5 (0.009), *6 (0.009), *8 (0.042), *9 (0.061), and *11(0.015) were observed with extensive (68.5%), intermediate (18.1%) and poor predicted metabolizers (0.6%), respectively. Fifty-two and 55 subjects, respectively had at least one variant CYP2C8 or CYP2C9 allele. Although the distribution of the CYP2C9 (p = 0.0515) phenotypes was marginally significantly in high and low ED users; some CYP2C8 and CYP2C9 allelic combinations observed in 15.2% (25) of the cohort are associated with higher risks for analgesic failure. CYP2C8 and CYP2C9 preemptive genotyping could potentially enable clinicians to identify patients with impaired metabolic phenotypes.

Use of a novel abdominal aortic tourniquet to reduce or eliminate flow in the common femoral artery in human subjects.

BACKGROUND: Penetrating injuries of proximal femoral and iliac vessels are a common cause of death on the battlefield. Previous studies have shown that by applying 80 lb to 140 lb of pressure externally over the distal abdominal aorta, flow can be ceased in the common femoral artery (CFA). It has also been demonstrated that in a porcine model, an externally applied pneumatic abdominal aortic tourniquet (AAT) can occlude the aorta and inferior vena cava for 60 minutes without bowel injury or significant potassium elevations.The objectives of this study were (1) determine if AAT use in humans results in flow cessation in the CFA, (2) measure the pressure required to cease flow in the CFA, and (3) measure discomfort associated with application of the AAT. METHODS: Pulse wave Doppler measurements were taken in supine volunteers at the right CFA. The AAT was placed just above the iliac crests over the anterior abdomen. The AAT was inflated using a hand pump with an integrated manometer. Measurements were taken every 30 mm Hg. Discomfort was measured using a 10-point pain scale. RESULTS: In all subjects, flow was reduced in the CFA. Flow ceased in seven of nine subjects at a median pressure of 180 mm Hg (150-230 mm Hg). Median discomfort at ceasing of flow was 7 (3-10), returning to 0 after device removal. CONCLUSION: The AAT device was effective at reducing flow in the CFA and ceased flow in most of the subjects. Application of the device was associated with discomfort varying from moderate to severe and resolving with device removal.

Detection of soft tissue foreign bodies in the presence of soft tissue gas.

OBJECTIVE: To determine the effect of soft tissue gas on the accuracy of foreign body detection by realtime sonography. METHODS: This was a prospective randomized study using glass, metal, and bone inserted into turkey breasts to simulate human soft tissue foreign bodies. Air was subsequently injected around a random selection of the foreign bodies to simulate soft tissue gas that can accompany a blast or high-force injury. Using a linear transducer, physicians credentialed in the use of sonography were each asked to scan the breasts, identify the location of any foreign body, and describe whether the object located was bone, metal, or glass. They were also asked to describe the characteristics of the foreign body, including surface echogenicity, visibility, and artifacts, if any. RESULTS: The sensitivity for localization of each foreign body by each sonographer was 100% (48 of 48) and was unaffected by the presence of soft tissue gas. The accuracy of classifying the foreign body was poor except with bone. Glass and metal were often confused with each other. With the addition of soft tissue gas over the foreign bodies, the sensitivity of classifying the foreign body was decreased further from a combined 58% to 28%. The presence of soft tissue gas decreased the amount of reflection of the foreign body and obscured the subtle differences in the brightness of each foreign body, leading to a decrease in the accuracy of identification but not localization of the foreign body. CONCLUSIONS: In an experimental model, soft tissue gas does not affect the localization of soft tissue foreign bodies. However, correct identification of the type of foreign body is limited by soft tissue gas because of loss of the typical sonographic characteristics.

Blistering distal dactylitis: a case series in children under nine months of age.

Blistering distal dactylitis (BDD) is a distinct clinical entity that is infrequently reported in the literature. Characteristically, blistering distal dactylitis is described as a localized infection involving the volar fat pad of the distal phalanx of the digits, and it usually presents as a fluid-filled blister. The usual causative organism is group A beta-hemolytic Streptococcus, but less commonly, Staphyloccous aureus and Staphyloccoccus epidermis are present. The normal age range is reported to be 2 to 16 years old, but there are case reports of this infection in adults. Only one case has been reported in the literature in a child younger than 24 months of age. In this report we describe three cases in children younger than 9 months old. These three cases indicate that BDD can and does occur in children younger than 2 years of age.

Evaluation of extremity trauma with sonography.

OBJECTIVE: To report on the use of bedside sonography in the diagnosis and treatment of penetrating extremity trauma. METHODS: Sonography was performed in the emergency department of a level 1 trauma center with both curved and linear array transducers. RESULTS: In both cases, foreign bodies and fractures resulting from gunshot wounds to the extremities were found. In addition, intra-abdominal and thoracic injuries were evaluated for using sonography. CONCLUSIONS: The increasing mobility and portability of sonography has led to its increasing use at the patient' bedside to diagnose a variety of conditions. It provides potential immediate diagnosis and has the flexibility for evaluating a multitude of injuries normally requiring several different imaging technologies. In circumstances in which conventional radiography is unavailable, such as at trauma scenes, mass casualty situations, or at the bedside with unstable patients, sonography can provide unique and essential information about these patients.