Hospital Readmission Outcomes by Frailty Risk in Adults in Behavioral Health Acute Care.

The purpose of the current retrospective study was to determine whether frailty is predictive of 30-day readmission in adults aged ≥50 years who were admitted with a psychiatric diagnosis to a behavioral health hospital from 2013 to 2017. A total of 1,063 patients were included. A 26-item frailty risk score (FRS-26-ICD) was constructed from electronic health record (EHR) data. There were 114 readmissions. Cox regression modeling for demographic characteristics, emergent admission, comorbidity, and FRS-26-ICD determined prediction of time to readmission was modest (incremental area under the receiver operating characteristic curve = 0.671). The FRS-26-ICD was a significant predictor of readmission alone and in models with demographics and emergent admission; however, only the Elixhauser Comorbidity Index was significantly related to hazard of readmission adjusting for other factors (adjusted hazard ratio = 1.26, 95% confidence interval [1.17, 1.37]; p < 0.001), whereas FRS-26-ICD became non-significant. Frailty is a relevant syndrome in behavioral health that should be further studied in risk prediction and incorporated into care planning to prevent hospital readmissions. [Journal of Psychosocial Nursing and Mental Health Services, 59(10), 27-39.].

The digital revolution in phenotyping.

Phenotypes have gained increased notoriety in the clinical and biological domain owing to their application in numerous areas such as the discovery of disease genes and drug targets, phylogenetics and pharmacogenomics. Phenotypes, defined as observable characteristics of organisms, can be seen as one of the bridges that lead to a translation of experimental findings into clinical applications and thereby support 'bench to bedside' efforts. However, to build this translational bridge, a common and universal understanding of phenotypes is required that goes beyond domain-specific definitions. To achieve this ambitious goal, a digital revolution is ongoing that enables the encoding of data in computer-readable formats and the data storage in specialized repositories, ready for integration, enabling translational research. While phenome research is an ongoing endeavor, the true potential hidden in the currently available data still needs to be unlocked, offering exciting opportunities for the forthcoming years. Here, we provide insights into the state-of-the-art in digital phenotyping, by means of representing, acquiring and analyzing phenotype data. In addition, we provide visions of this field for future research work that could enable better applications of phenotype data.

Using the phenoscape knowledgebase to relate genetic perturbations to phenotypic evolution.

The abundance of phenotypic diversity among species can enrich our knowledge of development and genetics beyond the limits of variation that can be observed in model organisms. The Phenoscape Knowledgebase (KB) is designed to enable exploration and discovery of phenotypic variation among species. Because phenotypes in the KB are annotated using standard ontologies, evolutionary phenotypes can be compared with phenotypes from genetic perturbations in model organisms. To illustrate the power of this approach, we review the use of the KB to find taxa showing evolutionary variation similar to that of a query gene. Matches are made between the full set of phenotypes described for a gene and an evolutionary profile, the latter of which is defined as the set of phenotypes that are variable among the daughters of any node on the taxonomic tree. Phenoscape's semantic similarity interface allows the user to assess the statistical significance of each match and flags matches that may only result from differences in annotation coverage between genetic and evolutionary studies. Tools such as this will help meet the challenge of relating the growing volume of genetic knowledge in model organisms to the diversity of phenotypes in nature. The Phenoscape KB is available at http://kb.phenoscape.org.

Interestingness measures and strategies for mining multi-ontology multi-level association rules from gene ontology annotations for the discovery of new GO relationships.

The Gene Ontology (GO), a set of three sub-ontologies, is one of the most popular bio-ontologies used for describing gene product characteristics. GO annotation data containing terms from multiple sub-ontologies and at different levels in the ontologies is an important source of implicit relationships between terms from the three sub-ontologies. Data mining techniques such as association rule mining that are tailored to mine from multiple ontologies at multiple levels of abstraction are required for effective knowledge discovery from GO annotation data. We present a data mining approach, Multi-ontology data mining at All Levels (MOAL) that uses the structure and relationships of the GO to mine multi-ontology multi-level association rules. We introduce two interestingness measures: Multi-ontology Support (MOSupport) and Multi-ontology Confidence (MOConfidence) customized to evaluate multi-ontology multi-level association rules. We also describe a variety of post-processing strategies for pruning uninteresting rules. We use publicly available GO annotation data to demonstrate our methods with respect to two applications (1) the discovery of co-annotation suggestions and (2) the discovery of new cross-ontology relationships.

AgBase: supporting functional modeling in agricultural organisms.

AgBase (http://www.agbase.msstate.edu/) provides resources to facilitate modeling of functional genomics data and structural and functional annotation of agriculturally important animal, plant, microbe and parasite genomes. The website is redesigned to improve accessibility and ease of use, including improved search capabilities. Expanded capabilities include new dedicated pages for horse, cat, dog, cotton, rice and soybean. We currently provide 590 240 Gene Ontology (GO) annotations to 105 454 gene products in 64 different species, including GO annotations linked to transcripts represented on agricultural microarrays. For many of these arrays, this provides the only functional annotation available. GO annotations are available for download and we provide comprehensive, species-specific GO annotation files for 18 different organisms. The tools available at AgBase have been expanded and several existing tools improved based upon user feedback. One of seven new tools available at AgBase, GOModeler, supports hypothesis testing from functional genomics data. We host several associated databases and provide genome browsers for three agricultural pathogens. Moreover, we provide comprehensive training resources (including worked examples and tutorials) via links to Educational Resources at the AgBase website.

A deep semantic matching approach for identifying relevant messages for social media analysis.

There is a growing interest in using social media content for Natural Language Processing applications. However, it is not easy to computationally identify the most relevant set of tweets related to any specific event. Challenging semantics coupled with different ways for using natural language in social media make it difficult for retrieving the most relevant set of data from any social media outlet. This paper seeks to demonstrate a way to present the changing semantics of Twitter within the context of a crisis event, specifically tweets during Hurricane Irma. These methods can be used to identify the most relevant corpus of text for analysis in relevance to a specific incident such as a hurricane. Using an implementation of the Word2Vec method of Neural Network training mechanisms to create Word Embeddings, this paper will: discuss how the relative meaning of words changes as events unfold; present a mechanism for scoring tweets based upon dynamic, relative context relatedness; and show that similarity between words is not necessarily static. We present different methods for training the vector model in Word2Vec for identification of the most relevant tweets for any search query. The impact of tuning parameters such as Word Window Size, Minimum Word Frequency, Hidden Layer Dimensionality, and Negative Sampling on model performance was explored. The window containing the local maximum for AU_ROC for each parameter serves as a guide for other studies using the methods presented here for social media data analysis.

Using EHR Data to Identify Patient Frailty and Risk for ICU Transfer.

The predictive properties of four definitions of a frailty risk score (FRS) constructed using combinations of nursing flowsheet data, laboratory tests, and ICD-10 codes were examined for time to first intensive care unit (ICU) transfer in medical-surgical inpatients ≥50 years of age. Cox regression modeled time to first ICU transfer and Schemper-Henderson explained variance summarized predictive accuracy of FRS combinations. Modeling by age group and controlling for sex, all FRS measures significantly predicted time to first ICU transfer. Further multivariable modeling controlling for clinical characteristics substantially improved predictive accuracy. The effect of frailty on time to first ICU transfer depended on age, with highest risk in 50 to <60 years and ≥80 years age groups. Frailty prevalence ranged from 25.1% to 56.4%. Findings indicate that FRS-based frailty is a risk factor for time to first ICU transfer and should be considered in assessment and care-planning to address frailty in high-risk patients.Frailty prevalence was highest med-surg pts 60 to <70 years (56%); highest risk for time to first ICU transfer was in younger (50 to <60 years) and older (≥80 years) groups.

A Gated Recurrent Unit based architecture for recognizing ontology concepts from biological literature.

BACKGROUND: Annotating scientific literature with ontology concepts is a critical task in biology and several other domains for knowledge discovery. Ontology based annotations can power large-scale comparative analyses in a wide range of applications ranging from evolutionary phenotypes to rare human diseases to the study of protein functions. Computational methods that can tag scientific text with ontology terms have included lexical/syntactic methods, traditional machine learning, and most recently, deep learning. RESULTS: Here, we present state of the art deep learning architectures based on Gated Recurrent Units for annotating text with ontology concepts. We use the Colorado Richly Annotated Full Text Corpus (CRAFT) as a gold standard for training and testing. We explore a number of additional information sources including NCBI's BioThesauraus and Unified Medical Language System (UMLS) to augment information from CRAFT for increasing prediction accuracy. Our best model results in a 0.84 F1 and semantic similarity. CONCLUSION: The results shown here underscore the impact for using deep learning architectures for automatically recognizing ontology concepts from literature. The augmentation of the models with biological information beyond that present in the gold standard corpus shows a distinct improvement in prediction accuracy.

Frailty and In-Hospital Mortality Risk Using EHR Nursing Data.

PurposeThe purpose of this study was to evaluate four definitions of a Frailty Risk Score (FRS) derived from EHR data that includes combinations of biopsychosocial risk factors using nursing flowsheet data or International Classification of Disease, 10th revision (ICD-10) codes and blood biomarkers and its predictive properties for in-hospital mortality in adults ≥50 years admitted to medical-surgical units. Methods In this retrospective observational study and secondary analysis of an EHR dataset, survival analysis and Cox regression models were performed with sociodemographic and clinical covariates. Integrated area under the ROC curve (iAUC) across follow-up time based on Cox modeling was estimated. Results The 46,645 patients averaged 1.5 hospitalizations (SD = 1.1) over the study period and 63.3% were emergent admissions. The average age was 70.4 years (SD = 11.4), 55.3% were female, 73.0% were non-Hispanic White (73.0%), mean comorbidity score was 3.9 (SD = 2.9), 80.5% were taking 1.5 high risk medications, and 42% recorded polypharmacy. The best performing FRS-NF-26-LABS included nursing flowsheet data and blood biomarkers (Adj. HR = 1.30, 95% CI [1.28, 1.33]), with good accuracy (iAUC = .794); the reduced model with age, sex, and FRS only demonstrated similar accuracy. The poorest performance was the ICD-10 code-based FRS. Conclusion The FRS captures information about the patient that increases risk for in-hospital mortality not accounted for by other factors. Identification of frailty enables providers to enhance various aspects of care, including increased monitoring, applying more intensive, individualized resources, and initiating more informed discussions about treatments and discharge planning.

Machine learning for predicting readmission risk among the frail: Explainable AI for healthcare.

Healthcare costs due to unplanned readmissions are high and negatively affect health and wellness of patients. Hospital readmission is an undesirable outcome for elderly patients. Here, we present readmission risk prediction using five machine learning approaches for predicting 30-day unplanned readmission for elderly patients (age ≥ 50 years). We use a comprehensive and curated set of variables that include frailty, comorbidities, high-risk medications, demographics, hospital, and insurance utilization to build these models. We conduct a large-scale study with electronic health record (her) data with over 145,000 observations from 76,000 patients. Findings indicate that the category boost (CatBoost) model outperforms other models with a mean area under the curve (AUC) of 0.79. We find that prior readmissions, discharge to a rehabilitation facility, length of stay, comorbidities, and frailty indicators were all strong predictors of 30-day readmission. We present in-depth insights using Shapley additive explanations (SHAP), the state of the art in machine learning explainability.

Comparison of a Frailty Risk Score and Comorbidity Indices for Hospital Readmission Using Electronic Health Record Data.

The purpose of the current study was to investigate the predictive properties of five definitions of a frailty risk score (FRS) and three comorbidity indices using data from electronic health records (EHRs) of hospitalized adults aged ≥50 years for 3-day, 7-day, and 30-day readmission, and to identify an optimal model for a FRS and comorbidity combination. Retrospective analysis of the EHR dataset was performed, and multivariable logistic regression and area under the curve (AUC) were used to examine readmission for frailty and comorbidity. The sample (N = 55,778) was mostly female (53%), non-Hispanic White (73%), married (53%), and on Medicare (55%). Mean FRSs ranged from 1.3 (SD = 1.5) to 4.3 (SD = 2.1). FRS and comorbidity were independently associated with readmission. Predictive accuracy for FRS and comorbidity combinations ranged from AUC of 0.75 to 0.77 (30-day readmission) to 0.84 to 0.85 (3-day readmission). FRS and comorbidity combinations performed similarly well, whereas comorbidity was always independently associated with readmission. FRS measures were more associated with 30-day readmission than 7-day and 3-day readmission. [Research in Gerontological Nursing, 14(2), 91-103.].

GOModeler--a tool for hypothesis-testing of functional genomics datasets.

BACKGROUND: Functional genomics technologies that measure genome expression at a global scale are accelerating biological knowledge discovery. Generating these high throughput datasets is relatively easy compared to the downstream functional modelling necessary for elucidating the molecular mechanisms that govern the biology under investigation. A number of publicly available 'discovery-based' computational tools use the computationally amenable Gene Ontology (GO) for hypothesis generation. However, there are few tools that support hypothesis-based testing using the GO and none that support testing with user defined hypothesis terms.Here, we present GOModeler, a tool that enables researchers to conduct hypothesis-based testing of high throughput datasets using the GO. GOModeler summarizes the overall effect of a user defined gene/protein differential expression dataset on specific GO hypothesis terms selected by the user to describe a biological experiment. The design of the tool allows the user to complement the functional information in the GO with his/her domain specific expertise for comprehensive hypothesis testing. RESULTS: GOModeler tests the relevance of the hypothesis terms chosen by the user for the input gene dataset by providing the individual effects of the genes on the hypothesis terms and the overall effect of the entire dataset on each of the hypothesis terms. It matches the GO identifiers (ids) of the genes with the GO ids of the hypothesis terms and parses the names of those ids that match to assign effects. We demonstrate the capabilities of GOModeler with a dataset of nine differentially expressed cytokine genes and compare the results to those obtained through manual analysis of the dataset by an immunologist. The direction of overall effects on all hypothesis terms except one was consistent with the results obtained by manual analysis. The tool's editing capability enables the user to augment the information extracted. GOModeler is available as a part of the AgBase tool suite (http://www.agbase.msstate.edu). CONCLUSIONS: GOModeler allows hypothesis driven analysis of high throughput datasets using the GO. Using this tool, researchers can quickly evaluate the overall effect of quantitative expression changes of gene set on specific biological processes of interest. The results are provided in both tabular and graphical formats.

Using Manual and Computer-Based Text-Mining to Uncover Research Trends for Apis mellifera.

Honey bee research is believed to be influenced dramatically by colony collapse disorder (CCD) and the sequenced genome release in 2006, but this assertion has never been tested. By employing text-mining approaches, research trends were tested by analyzing over 14,000 publications during the period of 1957 to 2017. Quantitatively, the data revealed an exponential growth until 2010 when the number of articles published per year ceased following the trend. Analysis of author-assigned keywords revealed that changes in keywords occurred roughly every decade with the most fundamental change in 1991-1992, instead of 2006. This change might be due to several factors including the research intensification on the Varroa mite. The genome release and CCD had quantitively only minor effects, mainly on honey bee health-related topics post-2006. Further analysis revealed that computational topic modeling can provide potentially hidden information and connections between some topics that might be ignored in author-assigned keywords.

Comparing gene annotation enrichment tools for functional modeling of agricultural microarray data.

UNLABELLED: The widespread availability of microarray technology has driven functional genomics to the forefront as scientists seek to draw meaningful biological conclusions from their microarray results. Gene annotation enrichment analysis is a functional analysis technique that has gained widespread attention and for which many tools have been developed. Unfortunately, most of these tools have limited support for agricultural species. Here, we evaluate and compare four publicly available computational tools (Onto-Express, EasyGO, GOstat, and DAVID) that support analysis of gene expression datasets in agricultural species. We use AgBase as the functional annotation reference for agricultural species. The selected tools were evaluated based on i) available features, usage and accessibility, ii) implemented statistical computational methods, and iii) annotation and enrichment performance analysis. Annotation was assessed using a randomly selected test gene annotation set and an experimental differentially expressed gene-set--both from chicken. The experimental set was also used to evaluate identification of enriched functional groups.Comparison of the tools shows that they produce different sets of annotations for the two datasets and different functional groups for the experimental dataset. While DAVID, GOstat and Onto-Express annotate comparable numbers of genes, DAVID provides by far the most annotations per gene. However, many of DAVID's annotations appear to be redundant or are at very high levels in the GO hierarchy. The GOSlim distribution of annotations shows that GOstat, Onto-Express and EasyGO provide similar GO distributions to those found in AgBase while annotations from DAVID show a different GOSlim distribution, again probably due to duplication and many non-specific terms. No consistent trends were found in results of GO term over/under representation analysis applied to the experimental data using different tools. While GOstat, David and Onto-Express could retrieve some significantly enriched terms, EasyGO did not show any significantly enriched terms. There was little agreement about the enriched terms identified by the tools. CONCLUSION: Different tools for functionally annotating gene sets and identifying significantly enriched GO categories differ widely in their results when applied to a test annotation gene set and an experimental dataset from chicken. These results emphasize the need for care when interpreting the results of such analysis and the lack of standardization of approaches.

Avoiding "conflicts of interest": a computational approach to scheduling parallel conference tracks and its human evaluation.

Conferences with contributed talks grouped into multiple concurrent sessions pose an interesting scheduling problem. From an attendee's perspective, choosing which talks to visit when there are many concurrent sessions is challenging since an individual may be interested in topics that are discussed in different sessions simultaneously. The frequency of topically similar talks in different concurrent sessions is, in fact, a common cause for complaint in post-conference surveys. Here, we introduce a practical solution to the conference scheduling problem by heuristic optimization of an objective function that weighs the occurrence of both topically similar talks in one session and topically different talks in concurrent sessions. Rather than clustering talks based on a limited number of preconceived topics, we employ a topic model to allow the topics to naturally emerge from the corpus of contributed talk titles and abstracts. We then measure the topical distance between all pairs of talks. Heuristic optimization of preliminary schedules seeks to balance the topical similarity of talks within a session and the dissimilarity between concurrent sessions. Using an ecology conference as a test case, we find that stochastic optimization dramatically improves the objective function relative to the schedule manually produced by the program committee. Approximate Integer Linear Programming can be used to provide a partially-optimized starting schedule, but the final value of the discrimination ratio (an objective function used to estimate coherence within a session and disparity between concurrent sessions) is surprisingly insensitive to the starting schedule. Furthermore, we show that, in contrast to the manual process, arbitrary scheduling constraints are straightforward to include. We applied our method to a second biology conference with over 1,000 contributed talks plus scheduling constraints. In a randomized experiment, biologists responded similarly to a machine-optimized schedule and a highly modified schedule produced by domain experts on the conference program committee.

Annotation of phenotypes using ontologies: a gold standard for the training and evaluation of natural language processing systems.

Natural language descriptions of organismal phenotypes, a principal object of study in biology, are abundant in the biological literature. Expressing these phenotypes as logical statements using ontologies would enable large-scale analysis on phenotypic information from diverse systems. However, considerable human effort is required to make these phenotype descriptions amenable to machine reasoning. Natural language processing tools have been developed to facilitate this task, and the training and evaluation of these tools depend on the availability of high quality, manually annotated gold standard data sets. We describe the development of an expert-curated gold standard data set of annotated phenotypes for evolutionary biology. The gold standard was developed for the curation of complex comparative phenotypes for the Phenoscape project. It was created by consensus among three curators and consists of entity-quality expressions of varying complexity. We use the gold standard to evaluate annotations created by human curators and those generated by the Semantic CharaParser tool. Using four annotation accuracy metrics that can account for any level of relationship between terms from two phenotype annotations, we found that machine-human consistency, or similarity, was significantly lower than inter-curator (human-human) consistency. Surprisingly, allowing curatorsaccess to external information did not significantly increase the similarity of their annotations to the gold standard or have a significant effect on inter-curator consistency. We found that the similarity of machine annotations to the gold standard increased after new relevant ontology terms had been added. Evaluation by the original authors of the character descriptions indicated that the gold standard annotations came closer to representing their intended meaning than did either the curator or machine annotations. These findings point toward ways to better design software to augment human curators and the use of the gold standard corpus will allow training and assessment of new tools to improve phenotype annotation accuracy at scale.