Dietary patterns among U.S. food insecure cancer survivors and the risk of mortality: NHANES 1999-2018.

PURPOSE: Food insecurity-the lack of unabated access to nutritious foods-is a consequence many cancer survivors face. Food insecurity is associated with adverse health outcomes and lower diet quality in the general public. The goal of this analysis was to extract major and prevailing dietary patterns among food insecure cancer survivors from observed 24-h recall data and evaluate their relationship to survival after a cancer diagnosis. METHODS: We implemented two dietary patterns analysis approaches: penalized logistic regression and principal components analysis. Using nationally representative data from the National Health and Nutrition Examination Survey (NHANES) study, we extracted three dietary patterns. Additionally, we evaluated the HEI-2015 for comparison. Cox proportional hazards models assessed the relationship between the diet quality indices and survival after a cancer diagnosis. RESULTS: There were 981 deaths from all causes and 343 cancer-related deaths. After multivariable adjustment, we found higher risks of all-cause mortality associated with higher adherence to Pattern #1 (HR 1.25; 95% CI 1.09-1.43) and Pattern #2 (HR 1.15; 95% CI 1.01-1.31) among cancer survivors. CONCLUSION: Among all cancer survivors, higher adherence to major and prevailing dietary patterns from the U.S. food insecure cancer survivor population may lead to worse survival outcomes.

Fecal Metagenomics to Identify Biomarkers of Food Intake in Healthy Adults: Findings from Randomized, Controlled, Nutrition Trials.

BACKGROUND: Undigested components of the human diet affect the composition and function of the microorganisms present in the gastrointestinal tract. Techniques like metagenomic analyses allow researchers to study functional capacity, thus revealing the potential of using metagenomic data for developing objective biomarkers of food intake. OBJECTIVES: As a continuation of our previous work using 16S and metabolomic datasets, we aimed to utilize a computationally intensive, multivariate, machine-learning approach to identify fecal KEGG (Kyoto encyclopedia of genes and genomes) Orthology (KO) categories as biomarkers that accurately classify food intake. METHODS: Data were aggregated from 5 controlled feeding studies that studied the individual impact of almonds, avocados, broccoli, walnuts, barley, and oats on the adult gastrointestinal microbiota. Deoxyribonucleic acid from preintervention and postintervention fecal samples underwent shotgun genomic sequencing. After preprocessing, sequences were aligned and functionally annotated with Double Index AlignMent Of Next-generation sequencing Data v2.0.11.149 and MEtaGenome ANalyzer v6.12.2, respectively. After the count normalization, the log of the fold change ratio for resulting KOs between pre- and postintervention of the treatment group against its corresponding control was utilized to conduct differential abundance analysis. Differentially abundant KOs were used to train machine-learning models examining potential biomarkers in both single-food and multi-food models. RESULTS: We identified differentially abundant KOs in the almond (n = 54), broccoli (n = 2474), and walnut (n = 732) groups (q < 0.20), which demonstrated classification accuracies of 80%, 87%, and 86% for the almond, broccoli, and walnut groups using a random forest model to classify food intake into each food group's respective treatment and control arms, respectively. The mixed-food random forest achieved 81% accuracy. CONCLUSIONS: Our findings reveal promise in utilizing fecal metagenomics to objectively complement self-reported measures of food intake. Future research on various foods and dietary patterns will expand these exploratory analyses for eventual use in feeding study compliance and clinical settings.

Efficient gradient boosting for prognostic biomarker discovery.

MOTIVATION: A gradient boosting decision tree (GBDT) is a powerful ensemble machine-learning method that has the potential to accelerate biomarker discovery from high-dimensional molecular data. Recent algorithmic advances, such as extreme gradient boosting (XGB) and light gradient boosting (LGB), have rendered the GBDT training more efficient, scalable and accurate. However, these modern techniques have not yet been widely adopted in discovering biomarkers for censored survival outcomes, which are key clinical outcomes or endpoints in cancer studies. RESULTS: In this paper, we present a new R package 'Xsurv' as an integrated solution that applies two modern GBDT training frameworks namely, XGB and LGB, for the modeling of right-censored survival outcomes. Based on our simulations, we benchmark the new approaches against traditional methods including the stepwise Cox regression model and the original gradient boosting function implemented in the package 'gbm'. We also demonstrate the application of Xsurv in analyzing a melanoma methylation dataset. Together, these results suggest that Xsurv is a useful and computationally viable tool for screening a large number of prognostic candidate biomarkers, which may facilitate future translational and clinical research. AVAILABILITY AND IMPLEMENTATION: 'Xsurv' is freely available as an R package at: https://github.com/topycyao/Xsurv. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Fecal Metabolites as Biomarkers for Predicting Food Intake by Healthy Adults.

BACKGROUND: The fecal metabolome is affected by diet and includes metabolites generated by human and microbial metabolism. Advances in -omics technologies and analytic approaches have allowed researchers to identify metabolites and better utilize large data sets to generate usable information. One promising aspect of these advancements is the ability to determine objective biomarkers of food intake. OBJECTIVES: We aimed to utilize a multivariate, machine learning approach to identify metabolite biomarkers that accurately predict food intake. METHODS: Data were aggregated from 5 controlled feeding studies in adults that tested the impact of specific foods (almonds, avocados, broccoli, walnuts, barley, and oats) on the gastrointestinal microbiota. Fecal samples underwent GC-MS metabolomic analysis; 344 metabolites were detected in preintervention samples, whereas 307 metabolites were detected postintervention. After removing metabolites that were only detected in either pre- or postintervention and those undetectable in ≥80% of samples in all study groups, changes in 96 metabolites relative concentrations (treatment postintervention minus preintervention) were utilized in random forest models to 1) examine the relation between food consumption and fecal metabolome changes and 2) rank the fecal metabolites by their predictive power (i.e., feature importance score). RESULTS: Using the change in relative concentration of 96 fecal metabolites, 6 single-food random forest models for almond, avocado, broccoli, walnuts, whole-grain barley, and whole-grain oats revealed prediction accuracies between 47% and 89%. When comparing foods with one another, almond intake was differentiated from walnut intake with 91% classification accuracy. CONCLUSIONS: Our findings reveal promise in utilizing fecal metabolites as objective complements to certain self-reported food intake estimates. Future research on other foods at different doses and dietary patterns is needed to identify biomarkers that can be applied in feeding study compliance and clinical settings.

Fecal Bacteria as Biomarkers for Predicting Food Intake in Healthy Adults.

BACKGROUND: Diet affects the human gastrointestinal microbiota. Blood and urine samples have been used to determine nutritional biomarkers. However, there is a dearth of knowledge on the utility of fecal biomarkers, including microbes, as biomarkers of food intake. OBJECTIVES: This study aimed to identify a compact set of fecal microbial biomarkers of food intake with high predictive accuracy. METHODS: Data were aggregated from 5 controlled feeding studies in metabolically healthy adults (n = 285; 21-75 y; BMI 19-59 kg/m2; 340 data observations) that studied the impact of specific foods (almonds, avocados, broccoli, walnuts, and whole-grain barley and whole-grain oats) on the human gastrointestinal microbiota. Fecal DNA was sequenced using 16S ribosomal RNA gene sequencing. Marginal screening was performed on all species-level taxa to examine the differences between the 6 foods and their respective controls. The top 20 species were selected and pooled together to predict study food consumption using a random forest model and out-of-bag estimation. The number of taxa was further decreased based on variable importance scores to determine the most compact, yet accurate feature set. RESULTS: Using the change in relative abundance of the 22 taxa remaining after feature selection, the overall model classification accuracy of all 6 foods was 70%. Collapsing barley and oats into 1 grains category increased the model accuracy to 77% with 23 unique taxa. Overall model accuracy was 85% using 15 unique taxa when classifying almonds (76% accurate), avocados (88% accurate), walnuts (72% accurate), and whole grains (96% accurate). Additional statistical validation was conducted to confirm that the model was predictive of specific food intake and not the studies themselves. CONCLUSIONS: Food consumption by healthy adults can be predicted using fecal bacteria as biomarkers. The fecal microbiota may provide useful fidelity measures to ascertain nutrition study compliance.

Constructing dynamic treatment regimes with shared parameters for censored data.

Dynamic treatment regimes are sequential decision rules that adapt throughout disease progression according to a patient's evolving characteristics. In many clinical applications, it is desirable that the format of the decision rules remains consistent over time. Unlike the estimation of dynamic treatment regimes in regular settings, where decision rules are formed without shared parameters, the derivation of the shared decision rules requires estimating shared parameters indexing the decision rules across different decision points. Estimation of such rules becomes more complicated when the clinical outcome of interest is a survival time subject to censoring. To address these challenges, we propose two novel methods: censored shared-Q-learning and censored shared-O-learning. Both methods incorporate clinical preferences into a qualitative rule, where the parameters indexing the decision rules are shared across different decision points and estimated simultaneously. We use simulation studies to demonstrate the superior performance of the proposed methods. The methods are further applied to the Framingham Heart Study to derive treatment rules for cardiovascular disease.

Bagging and deep learning in optimal individualized treatment rules.

An ENsemble Deep Learning Optimal Treatment (EndLot) approach is proposed for personalized medicine problems. The statistical framework of the proposed method is based on the outcome weighted learning (OWL) framework which transforms the optimal decision rule problem into a weighted classification problem. We further employ an ensemble of deep neural networks (DNNs) to learn the optimal decision rule. Utilizing the flexibility of DNNs and the stability of bootstrap aggregation, the proposed method achieves a considerable improvement over existing methods. An R package "ITRlearn" is developed to implement the proposed method. Numerical performance is demonstrated via simulation studies and a real data analysis of the Cancer Cell Line Encyclopedia data.

Integrating multidimensional omics data for cancer outcome.

In multidimensional cancer omics studies, one subject is profiled on multiple layers of omics activities. In this article, the goal is to integrate multiple types of omics measurements, identify markers, and build a model for cancer outcome. The proposed analysis is achieved in two steps. In the first step, we analyze the regulation among different types of omics measurements, through the construction of linear regulatory modules (LRMs). The LRMs have sound biological basis, and their construction differs from the existing analyses by modeling the regulation of sets of gene expressions (GEs) by sets of regulators. The construction is realized with the assistance of regularized singular value decomposition. In the second step, the proposed cancer outcome model includes the regulated GEs, "residuals" of GEs, and "residuals" of regulators, and we use regularized estimation to select relevant markers. Simulation shows that the proposed method outperforms the alternatives with more accurate marker identification. We analyze the The Cancer Genome Atlas data on cutaneous melanoma and lung adenocarcinoma and obtain meaningful results.

Greedy outcome weighted tree learning of optimal personalized treatment rules.

We propose a subgroup identification approach for inferring optimal and interpretable personalized treatment rules with high-dimensional covariates. Our approach is based on a two-step greedy tree algorithm to pursue signals in a high-dimensional space. In the first step, we transform the treatment selection problem into a weighted classification problem that can utilize tree-based methods. In the second step, we adopt a newly proposed tree-based method, known as reinforcement learning trees, to detect features involved in the optimal treatment rules and to construct binary splitting rules. The method is further extended to right censored survival data by using the accelerated failure time model and introducing double weighting to the classification trees. The performance of the proposed method is demonstrated via simulation studies, as well as analyses of the Cancer Cell Line Encyclopedia (CCLE) data and the Tamoxifen breast cancer data.

Identifying gene-environment and gene-gene interactions using a progressive penalization approach.

In genomic studies, identifying important gene-environment and gene-gene interactions is a challenging problem. In this study, we adopt the statistical modeling approach, where interactions are represented by product terms in regression models. For the identification of important interactions, we adopt penalization, which has been used in many genomic studies. Straightforward application of penalization does not respect the "main effect, interaction" hierarchical structure. A few recently proposed methods respect this structure by applying constrained penalization. However, they demand very complicated computational algorithms and can only accommodate a small number of genomic measurements. We propose a computationally fast penalization method that can identify important gene-environment and gene-gene interactions and respect a strong hierarchical structure. The method takes a stagewise approach and progressively expands its optimization domain to account for possible hierarchical interactions. It is applicable to multiple data types and models. A coordinate descent method is utilized to produce the entire regularized solution path. Simulation study demonstrates the superior performance of the proposed method. We analyze a lung cancer prognosis study with gene expression measurements and identify important gene-environment interactions.

The effects of state psychiatric hospital waitlist policies on length of stay and time to readmission.

This study examined the effects of a waitlist policy for state psychiatric hospitals on length of stay and time to readmission using data from North Carolina for 2004-2010. Cox proportional hazards models tested the hypothesis that patients were discharged "quicker-but-sicker" post-waitlist, as hospitals struggled to manage admission delays and quickly admit waitlisted patients. Results refute this hypothesis, indicating that waitlists were associated with increased length of stay and time to readmission. Further research is needed to evaluate patients' clinical outcomes directly and to examine the impact of state hospital waitlists in other areas, such as state hospital case mix, local emergency departments, and outpatient mental health agencies.

Environmental regulation, industrial structure and energy efficiency: Evidence from 30 provinces in China.

The government's environmental protection policy can significantly contribute to alleviating resource shortages and curbing environmental pollution, but the impact of various policy instruments implemented by the government on energy efficiency is unclear. Based on the panel data of 30 provinces in China from 2005 to 2021, this paper analyses the impact of environmental regulation and the industrial structure on energy efficiency from the perspective of resource taxes. The U-shaped relationship between environmental regulation and energy efficiency and between the optimization of industrial structure can significantly improve energy efficiency, and the optimization of industrial structure is conducive to weakening the initial inhibitory effect of environmental regulation. In addition, the analysis of regional heterogeneity showed that the impact of environmental regulation was stronger in the central and western regions, while the impact of industrial structure was stronger in the eastern and western regions. The conclusions of this study can help to expand the understanding of the relationship between environmental regulation and industrial structure on energy efficiency, provide policy enlightenment for the realization of green development and high-quality development, and provide Chinese examples and experiences for developing countries to improve energy efficiency.

Poplars with a PtDDM1-RNAi transgene have reduced DNA methylation and show aberrant post-dormancy morphology.

The Arabidopsis thaliana DDM1 (Decreased DNA Methylation) gene is necessary for the maintenance of DNA methylation and heterochromatin assembly. In Arabidopsis, ddm1 mutants exhibit strong but delayed morphological phenotypes. We used RNA interference (RNAi) to suppress transcripts of two orthologous DDM1 paralogs in Populus trichocarpa and examined effects on whole plant phenotypes during perennial growth and seasonal dormancy. The RNAi-PtDDM1 transgenic poplars showed a wide range of DDM1 transcript suppression; the most strongly suppressed line had 37.5 % of the expression of the non-transgenic control. Genomic cytosine methylation (mC %) was 11.1 % in the non-transgenic control, compared with 9.1 % for the transgenic event with lowest mC %, a reduction of 18.1 %. An evaluation of greenhouse growth directly after acclimation of in vitro grown plants showed no developmental or growth rate abnormalities associated with the decrease in PtDDM1 expression. However, after a dormancy cycle and growth outdoors, a mottled leaf phenotype appeared in some of the transgenic insertion events that had strongly reduced PtDDM1 expression and DNA methylation. The phenotypic consequences of reduced DDM1 activity and DNA methylation appears to increase with cumulative plant propagation and growth.

Population analysis of mortality risk: Predictive models from passive monitors using motion sensors for 100,000 UK Biobank participants.

Many studies have utilized physical activity for predicting mortality risk, using measures such as participant walk tests and self-reported walking pace. The rise of passive monitors to measure participant activity without requiring specific actions opens the possibility for population level analysis. We have developed novel technology for this predictive health monitoring, using limited sensor inputs. In previous studies, we validated these models in clinical experiments with carried smartphones, using only their embedded accelerometers as motion sensors. Using smartphones as passive monitors for population measurement is critically important for health equity, since they are already ubiquitous in high-income countries and increasingly common in low-income countries. Our current study simulates smartphone data by extracting walking window inputs from wrist worn sensors. To analyze a population at national scale, we studied 100,000 participants in the UK Biobank who wore activity monitors with motion sensors for 1 week. This national cohort is demographically representative of the UK population, and this dataset represents the largest such available sensor record. We characterized participant motion during normal activities, including daily living equivalent of timed walk tests. We then compute walking intensity from sensor data, as input to survival analysis. Simulating passive smartphone monitoring, we validated predictive models using only sensors and demographics. This resulted in C-index of 0.76 for 1-year risk decreasing to 0.73 for 5-year. A minimum set of sensor features achieves C-index of 0.72 for 5-year risk, which is similar accuracy to other studies using methods not achievable with smartphone sensors. The smallest minimum model uses average acceleration, which has predictive value independent of demographics of age and sex, similar to physical measures of gait speed. Our results show passive measures with motion sensors can achieve similar accuracy to active measures of gait speed and walk pace, which utilize physical walk tests and self-reported questionnaires.

Optimizing Health Coaching for Patients With Type 2 Diabetes Using Machine Learning: Model Development and Validation Study.

BACKGROUND: Health coaching is an emerging intervention that has been shown to improve clinical and patient-relevant outcomes for type 2 diabetes. Advances in artificial intelligence may provide an avenue for developing a more personalized, adaptive, and cost-effective approach to diabetes health coaching. OBJECTIVE: We aim to apply Q-learning, a widely used reinforcement learning algorithm, to a diabetes health-coaching data set to develop a model for recommending an optimal coaching intervention at each decision point that is tailored to a patient's accumulated history. METHODS: In this pilot study, we fit a two-stage reinforcement learning model on 177 patients from the intervention arm of a community-based randomized controlled trial conducted in Canada. The policy produced by the reinforcement learning model can recommend a coaching intervention at each decision point that is tailored to a patient's accumulated history and is expected to maximize the composite clinical outcome of hemoglobin A1c reduction and quality of life improvement (normalized to [ ​0, 1 ​], with a higher score being better). Our data, models, and source code are publicly available. RESULTS: Among the 177 patients, the coaching intervention recommended by our policy mirrored the observed diabetes health coach's interventions in 17.5% (n=31) of the patients in stage 1 and 14.1% (n=25) of the patients in stage 2. Where there was agreement in both stages, the average cumulative composite outcome (0.839, 95% CI 0.460-1.220) was better than those for whom the optimal policy agreed with the diabetes health coach in only one stage (0.791, 95% CI 0.747-0.836) or differed in both stages (0.755, 95% CI 0.728-0.781). Additionally, the average cumulative composite outcome predicted for the policy's recommendations was significantly better than that of the observed diabetes health coach's recommendations (tn-1=10.040; P<.001). CONCLUSIONS: Applying reinforcement learning to diabetes health coaching could allow for both the automation of health coaching and an improvement in health outcomes produced by this type of intervention.

Empirical Dietary Patterns Associated with Food Insecurity in U.S. Cancer Survivors: NHANES 1999-2018.

(1) Background: Food insecurity (FI) is a public health and sociodemographic phenomenon that besets many cancer survivors in the United States. FI in cancer survivors may arise as a consequence of financial toxicity stemming from treatment costs, physical impairment, labor force egress, or a combination of those factors. To our knowledge, an understanding of the dietary intake practices of this population has not been delineated but is imperative for addressing the needs of this vulnerable population; (2) Methods: Using data from NHANES, 1999-2018, we characterized major dietary patterns in the food insecure cancer survivor population using: i. penalized logistic regression (logit) and ii. principal components analysis (PCA). We validated these patterns by examining the association of those patterns with food insecurity in the cancer population; (3) Results: Four dietary patterns were extracted with penalized logit and two with PCA. In the pattern validation phase, we found several patterns exhibited strong associations with FI. The FI, SNAP, and Household Size patterns (all extracted with penalized logit) harbored the strongest associations and there was evidence of stronger associations in those moderately removed from a cancer diagnosis (≥2 and <6 years since diagnosis); (4) Conclusions: FI may play an influential role on the dietary intake patterns of cancer survivors in the U.S. The results highlight the relevance of FI screening and monitoring for cancer survivors.

A parsimonious personalized dose-finding model via dimension reduction.

Learning an individualized dose rule in personalized medicine is a challenging statistical problem. Existing methods often suffer from the curse of dimensionality, especially when the decision function is estimated nonparametrically. To tackle this problem, we propose a dimension reduction framework that effectively reduces the estimation to a lower-dimensional subspace of the covariates. We exploit that the individualized dose rule can be defined in a subspace spanned by a few linear combinations of the covariates, leading to a more parsimonious model. Also, our framework does not require the inverse probability of the propensity score under observational studies due to a direct maximization of the value function. This distinguishes us from the outcome weighted learning framework, which also solves decision rules directly. Under the same framework, we further propose a pseudo-direct learning approach focuses more on estimating the dimensionality-reduced subspace of the treatment outcome. Parameters in both approaches can be estimated efficiently using an orthogonality constrained optimization algorithm on the Stiefel manifold. Under mild regularity assumptions, the asymptotic normality results of the proposed estimators can are established, respectively. We also derive the consistency and convergence rate for the value function under the estimated optimal dose rule. We evaluate the performance of the proposed approaches through extensive simulation studies and a warfarin pharmacogenetic dataset.

Risk assessment of latent tuberculosis infection through a multiplexed cytokine biosensor assay and machine learning feature selection.

Accurate detection and risk stratification of latent tuberculosis infection (LTBI) remains a major clinical and public health problem. We hypothesize that multiparameter strategies that probe immune responses to Mycobacterium tuberculosis can provide new diagnostic insights into not only the status of LTBI infection, but also the risk of reactivation. After the initial proof-of-concept study, we developed a 13-plex immunoassay panel to profile cytokine release from peripheral blood mononuclear cells stimulated separately with Mtb-relevant and non-specific antigens to identify putative biomarker signatures. We sequentially enrolled 65 subjects with various risk of TB exposure, including 32 subjects with diagnosis of LTBI. Random Forest feature selection and statistical data reduction methods were applied to determine cytokine levels across different normalized stimulation conditions. Receiver Operator Characteristic (ROC) analysis for full and reduced feature sets revealed differences in biomarkers signatures for LTBI status and reactivation risk designations. The reduced set for increased risk included IP-10, IL-2, IFN-γ, TNF-α, IL-15, IL-17, CCL3, and CCL8 under varying normalized stimulation conditions. ROC curves determined predictive accuracies of > 80% for both LTBI diagnosis and increased risk designations. Our study findings suggest that a multiparameter diagnostic approach to detect normalized cytokine biomarker signatures might improve risk stratification in LTBI.

Lung epithelial and endothelial damage, loss of tissue repair, inhibition of fibrinolysis, and cellular senescence in fatal COVID-19.

Coronavirus disease 2019 (COVID-19), caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is characterized by respiratory distress, multiorgan dysfunction, and, in some cases, death. The pathological mechanisms underlying COVID-19 respiratory distress and the interplay with aggravating risk factors have not been fully defined. Lung autopsy samples from 18 patients with fatal COVID-19, with symptom onset-to-death times ranging from 3 to 47 days, and antemortem plasma samples from 6 of these cases were evaluated using deep sequencing of SARS-CoV-2 RNA, multiplex plasma protein measurements, and pulmonary gene expression and imaging analyses. Prominent histopathological features in this case series included progressive diffuse alveolar damage with excessive thrombosis and late-onset pulmonary tissue and vascular remodeling. Acute damage at the alveolar-capillary barrier was characterized by the loss of surfactant protein expression with injury to alveolar epithelial cells, endothelial cells, respiratory epithelial basal cells, and defective tissue repair processes. Other key findings included impaired clot fibrinolysis with increased concentrations of plasma and lung plasminogen activator inhibitor-1 and modulation of cellular senescence markers, including p21 and sirtuin-1, in both lung epithelial and endothelial cells. Together, these findings further define the molecular pathological features underlying the pulmonary response to SARS-CoV-2 infection and provide important insights into signaling pathways that may be amenable to therapeutic intervention.

Diagnostic and prognostic capabilities of a biomarker and EMR-based machine learning algorithm for sepsis.

Sepsis is a major cause of mortality among hospitalized patients worldwide. Shorter time to administration of broad-spectrum antibiotics is associated with improved outcomes, but early recognition of sepsis remains a major challenge. In a two-center cohort study with prospective sample collection from 1400 adult patients in emergency departments suspected of sepsis, we sought to determine the diagnostic and prognostic capabilities of a machine-learning algorithm based on clinical data and a set of uncommonly measured biomarkers. Specifically, we demonstrate that a machine-learning model developed using this dataset outputs a score with not only diagnostic capability but also prognostic power with respect to hospital length of stay (LOS), 30-day mortality, and 3-day inpatient re-admission both in our entire testing cohort and various subpopulations. The area under the receiver operating curve (AUROC) for diagnosis of sepsis was 0.83. Predicted risk scores for patients with septic shock were higher compared with patients with sepsis but without shock (p < 0.0001). Scores for patients with infection and organ dysfunction were higher compared with those without either condition (p < 0.0001). Stratification based on predicted scores of the patients into low, medium, and high-risk groups showed significant differences in LOS (p < 0.0001), 30-day mortality (p < 0.0001), and 30-day inpatient readmission (p < 0.0001). In conclusion, a machine-learning algorithm based on electronic medical record (EMR) data and three nonroutinely measured biomarkers demonstrated good diagnostic and prognostic capability at the time of initial blood culture.