Concept recognition as a machine translation problem.

BACKGROUND: Automated assignment of specific ontology concepts to mentions in text is a critical task in biomedical natural language processing, and the subject of many open shared tasks. Although the current state of the art involves the use of neural network language models as a post-processing step, the very large number of ontology classes to be recognized and the limited amount of gold-standard training data has impeded the creation of end-to-end systems based entirely on machine learning. Recently, Hailu et al. recast the concept recognition problem as a type of machine translation and demonstrated that sequence-to-sequence machine learning models have the potential to outperform multi-class classification approaches. METHODS: We systematically characterize the factors that contribute to the accuracy and efficiency of several approaches to sequence-to-sequence machine learning through extensive studies of alternative methods and hyperparameter selections. We not only identify the best-performing systems and parameters across a wide variety of ontologies but also provide insights into the widely varying resource requirements and hyperparameter robustness of alternative approaches. Analysis of the strengths and weaknesses of such systems suggest promising avenues for future improvements as well as design choices that can increase computational efficiency with small costs in performance. RESULTS: Bidirectional encoder representations from transformers for biomedical text mining (BioBERT) for span detection along with the open-source toolkit for neural machine translation (OpenNMT) for concept normalization achieve state-of-the-art performance for most ontologies annotated in the CRAFT Corpus. This approach uses substantially fewer computational resources, including hardware, memory, and time than several alternative approaches. CONCLUSIONS: Machine translation is a promising avenue for fully machine-learning-based concept recognition that achieves state-of-the-art results on the CRAFT Corpus, evaluated via a direct comparison to previous results from the 2019 CRAFT shared task. Experiments illuminating the reasons for the surprisingly good performance of sequence-to-sequence methods targeting ontology identifiers suggest that further progress may be possible by mapping to alternative target concept representations. All code and models can be found at: https://github.com/UCDenver-ccp/Concept-Recognition-as-Translation .

Inflammatory profile in a canine model of hypothermic circulatory arrest.

BACKGROUND: Hypothermic circulatory arrest (HCA) is a technique used for complex repair of the aorta, but it can be associated with neurologic morbidity. To better understand the molecular changes that underlie ischemic brain injury, we assessed gene expression and cytokine/chemokine polypeptide concentration in brain tissue and cerebrospinal fluid (CSF) of canines that underwent two hours of HCA. MATERIALS AND METHODS: Adult male canines were cannulated peripherally for cardiopulmonary bypass, cooled to 18°C, and arrested for two hours. Animals were euthanized two, eight, or 24 hours post-HCA (n = 8 per group), and their brains were compared to brains from eight normal canines, using gene expression microarray analysis, cytokine assay, and histopathology. RESULTS: Two to eight hours after HCA, pro-inflammatory cytokine mRNAs increased markedly, and gene expression was enriched within signaling pathways related to neuroinflammation or ischemic injury. Concentrations of pro-inflammatory cytokine polypeptides IL-6, IL-8, IL-1ß, and CCL2 were very low in normal canine brain, whereas anti-inflammatory IL-10 and TGF-ß1 were expressed at moderate levels. Pro-inflammatory cytokine concentrations rose robustly in cerebral tissue and CSF after HCA. IL-6 and IL-8 peaked at eight hours and declined at 24 hours, while IL-1ß and CCL2 remained elevated. Concentrations of anti-inflammatory IL-10 and TGF-ß1 were maintained after HCA, with a significant increase in TGF-ß1 at 24 hours. CONCLUSIONS: These cytokines represent potential diagnostic markers for ischemic neurologic injury that could be used to assess neurologic injury in patients undergoing HCA. The cellular mechanisms underlying this pro-inflammatory, ischemic-induced injury represent potential targets for neuroprotection in the future.

Correlating Oxygen Delivery During Cardiopulmonary Bypass With the Neurologic Injury Biomarker Ubiquitin C-Terminal Hydrolase L1 (UCH-L1).

OBJECTIVE: The authors sought to assess the relationship between low oxygen delivery (DO2) during cardiopulmonary bypass (CPB) and a neuron-specific biomarker of neurologic injury, ubiquitin C-terminal hydrolase L1 (UCH-L1). DESIGN: Retrospective analysis of patient charts and prospectively collected blood samples. SETTING: University-affiliated tertiary care hospital. PARTICIPANTS: Adult patients undergoing cardiac surgery on CPB. INTERVENTIONS: Serum UCH-L1 levels were drawn at baseline and 6 and 24 hours after CPB cessation. DO2 was computed from perfusion records, with area-under-the-curve (AUC) computations performed to account for distance of DO2 excursions below predefined DO2 thresholds and the amount of time spent below them. Strokes were defined radiographically using computed tomography and magnetic resonance imaging. MEASUREMENTS AND MAIN RESULTS: Forty-three adults were included (median age 65 y, interquartile range 59-72). Three patients experienced strokes (imaged at 2, 7, and 8 d postoperatively). Most patients underwent isolated coronary artery bypass grafting (41%, 18 patients) or isolated aortic valve replacement (30%, 13). Median UCH-L1 levels differed from baseline to 6 and 24 hours after CPB (40, 232, and 166 pg/mL, respectively; p < 0.001). On multivariable linear regression analysis controlling for baseline and surgical variables, only DO2 AUC <225 was significantly associated with 6-hour UCH-L1 levels (p = 0.001), whereas only DO2 AUC <300 was significantly associated with 24- hour levels (p < 0.001). The 3 patients who experienced radiographic strokes had nonsignificantly elevated 24-hour UCH-L1 levels compared with control patients (585 v 151 pg/mL, p = 0.11). CONCLUSIONS: This is the first study to demonstrate an independent association between DO2 during CPB and elevations of a brain injury biomarker; additional study is needed to clarify the clinical significance of these results.

A Novel Risk Score to Predict New Onset Atrial Fibrillation in Patients Undergoing Isolated Coronary Artery Bypass Grafting.

BACKGROUND: Atrial fibrillation (AF) is common after cardiac surgery and contributes to increased morbidity and mortality. Our objective was to derive and validate a predictive model for AF after CABG in patients, incorporating novel echocardiographic and laboratory values. METHODS: We retrospectively reviewed patients at our institution without preexisting dysrhythmia who underwent on-pump, isolated CABG from 2011-2015. The primary outcome was new onset AF lasting >1 hour on continuous telemetry or requiring medical treatment. Patients with a preoperative echocardiographic measurement of left atrial diameter were included in a risk model, and were randomly divided into derivation (80%) and validation (20%) cohorts. The predictors of AF after CABG (PAFAC) score was derived from a multivariable logistic regression model by multiplying the adjusted odds ratios of significant risk factors (P < .05) by a factor of 4 to derive an integer point system. RESULTS: 1307 patients underwent isolated CABG, including 762/1307 patients with a preoperative left atrial diameter measurement. 209/762 patients (27%) developed new onset AF including 165/611 (27%) in the derivation cohort. We identified four risk factors independently associated with postoperative AF which comprised the PAFAC score: age > 60 years (5 points), White race (5 points), baseline GFR < 90 mL/min (4 points) and left atrial diameter > 4.5 cm (4 points). Scores ranged from 0-18. The PAFAC score was then applied to the validation cohort and predicted incidence of AF strongly correlated with observed incidence (r = 0.92). CONCLUSION: The PAFAC score is easy to calculate and can be used upon ICU admission to reliably identify patients at high risk of developing AF after isolated CABG.

Coreference annotation and resolution in the Colorado Richly Annotated Full Text (CRAFT) corpus of biomedical journal articles.

BACKGROUND: Coreference resolution is the task of finding strings in text that have the same referent as other strings. Failures of coreference resolution are a common cause of false negatives in information extraction from the scientific literature. In order to better understand the nature of the phenomenon of coreference in biomedical publications and to increase performance on the task, we annotated the Colorado Richly Annotated Full Text (CRAFT) corpus with coreference relations. RESULTS: The corpus was manually annotated with coreference relations, including identity and appositives for all coreferring base noun phrases. The OntoNotes annotation guidelines, with minor adaptations, were used. Interannotator agreement ranges from 0.480 (entity-based CEAF) to 0.858 (Class-B3), depending on the metric that is used to assess it. The resulting corpus adds nearly 30,000 annotations to the previous release of the CRAFT corpus. Differences from related projects include a much broader definition of markables, connection to extensive annotation of several domain-relevant semantic classes, and connection to complete syntactic annotation. Tool performance was benchmarked on the data. A publicly available out-of-the-box, general-domain coreference resolution system achieved an F-measure of 0.14 (B3), while a simple domain-adapted rule-based system achieved an F-measure of 0.42. An ensemble of the two reached F of 0.46. Following the IDENTITY chains in the data would add 106,263 additional named entities in the full 97-paper corpus, for an increase of 76% percent in the semantic classes of the eight ontologies that have been annotated in earlier versions of the CRAFT corpus. CONCLUSIONS: The project produced a large data set for further investigation of coreference and coreference resolution in the scientific literature. The work raised issues in the phenomenon of reference in this domain and genre, and the paper proposes that many mentions that would be considered generic in the general domain are not generic in the biomedical domain due to their referents to specific classes in domain-specific ontologies. The comparison of the performance of a publicly available and well-understood coreference resolution system with a domain-adapted system produced results that are consistent with the notion that the requirements for successful coreference resolution in this genre are quite different from those of the general domain, and also suggest that the baseline performance difference is quite large.

Perioperative Low Arterial Oxygenation Is Associated With Increased Stroke Risk in Cardiac Surgery.

BACKGROUND: Both patient characteristics and intraoperative factors have been associated with a higher risk of stroke after cardiac surgery. We hypothesized that poor systemic oxygenation in the perioperative period is associated with increased risk of stroke following cardiopulmonary bypass. METHODS: In this study of 251 adult patients who underwent cardiopulmonary bypass procedures at a single center from 2003 to 2006, cases (patients with a postoperative stroke at least 24 hours after surgery) were matched 1:2 to controls without stroke. Minimum and average partial pressure of oxygen in arterial blood (PaO2) values, from arterial blood gas values during and up to 24 hours after surgery, were evaluated as continuous and categorical predictors. Conditional logistic regression models adjusted for potential confounders (demographics, comorbidities, and intraoperative variables) were used to evaluate associations between PaO2 variables and stroke status. RESULTS: Lower nadir PaO2 values were associated with postoperative stroke, with estimated odds of stroke increasing over 20% (adjusted odds ratio [OR], 1.23; 95% confidence interval [CI], 1.07-1.41) per 10 mm Hg lower nadir PaO2, and similarly increased odds of stroke per lower quartile of nadir PaO2 (OR, 1.60; 95% CI, 1.19-2.16). When average PaO2 was considered, odds of stroke was also increased (adjusted OR, 1.39 per lower quartile of mean PaO2; 95% CI, 1.05-1.83). Having a nadir PaO2 value in the lowest versus any other quartile was associated with an estimated 2.41-fold increased odds of stroke (95% CI, 1.22-4.78). Quartile of nadir but not average PaO2 results remained significant after adjustment for multiple comparisons. CONCLUSIONS: Odds of stroke after cardiac surgery are increased in patients with a low minimum PaO2 within 24 hours of surgery. Results should be validated in an independent cohort. Further characterizing the underlying etiology of hypoxic episodes will be important to improve patient outcomes.

Management of sickle cell disease in patients undergoing cardiac surgery.

Sickle cell disease is a life-limiting inherited hemoglobinopathy that poses inherent risk for surgical complications following cardiac operations. In this review, we discuss preoperative considerations, intraoperative decision-making, and postoperative strategies to optimize the care of a patient with sickle cell disease undergoing cardiac surgery.

Predictors of Late-Onset Atrial Fibrillation Following Isolated Mitral Valve Repairs in Patients With Preserved Ejection Fraction.

BACKGROUND: We sought to determine the incidence of and risk factors for late-onset atrial fibrillation (LOAF) in patients with preserved ejection fractions undergoing mitral valve repair METHODS: We included patients undergoing isolated mitral valve repair (MVR) for degenerative disease between 1997-2014 at our institution with EF ≥60%. Patients who had AF preoperatively were excluded from the final analysis. Our primary outcome, LOAF, was defined as AF occurring after discharge following MVR (≥9 days). RESULTS: 223 patients were included in the study with a mean follow-up of 4.8 ± 4.6 years. A total of 25 patients developed LOAF, and freedom from LOAF was 93.9% at one year, and 87.3% at five years. Patients developing LOAF were of similar mean age (58 vs. 63 years in controls, p = 0.08) and had similar preoperative comorbidities, but did show a trend toward larger left atrial diameter (5.1 vs. 4.7 cm, p = 0.11). After risk adjustment with Cox regression analysis, only increasing left atrial size was associated with LOAF (HR 1.63, p = 0.04). On follow-up, 29 patients (10.8%) developed moderate or greater mitral regurgitation at a mean of 2.2 years. Using a mixed-effects model, we were unable to detect an association between recurrent mitral regurgitation following MVR and LOAF (OR 1.36, p = 0.42). CONCLUSIONS: LOAF occurs in about 13% of preserved ejection fraction patients undergoing MVR by five years. Increasing left atrial diameter is an independent predictor of LOAF. Concomitant anti-arrhythmic procedures may warrant further investigation in patients with preserved ejection fraction and enlarged left atria undergoing MVR. doi: 10.1111/jocs.12774 (J Card Surg 2016;31:486-492).

KaBOB: ontology-based semantic integration of biomedical databases.

BACKGROUND: The ability to query many independent biological databases using a common ontology-based semantic model would facilitate deeper integration and more effective utilization of these diverse and rapidly growing resources. Despite ongoing work moving toward shared data formats and linked identifiers, significant problems persist in semantic data integration in order to establish shared identity and shared meaning across heterogeneous biomedical data sources. RESULTS: We present five processes for semantic data integration that, when applied collectively, solve seven key problems. These processes include making explicit the differences between biomedical concepts and database records, aggregating sets of identifiers denoting the same biomedical concepts across data sources, and using declaratively represented forward-chaining rules to take information that is variably represented in source databases and integrating it into a consistent biomedical representation. We demonstrate these processes and solutions by presenting KaBOB (the Knowledge Base Of Biomedicine), a knowledge base of semantically integrated data from 18 prominent biomedical databases using common representations grounded in Open Biomedical Ontologies. An instance of KaBOB with data about humans and seven major model organisms can be built using on the order of 500 million RDF triples. All source code for building KaBOB is available under an open-source license. CONCLUSIONS: KaBOB is an integrated knowledge base of biomedical data representationally based in prominent, actively maintained Open Biomedical Ontologies, thus enabling queries of the underlying data in terms of biomedical concepts (e.g., genes and gene products, interactions and processes) rather than features of source-specific data schemas or file formats. KaBOB resolves many of the issues that routinely plague biomedical researchers intending to work with data from multiple data sources and provides a platform for ongoing data integration and development and for formal reasoning over a wealth of integrated biomedical data.

General and acute care surgical procedures in patients with left ventricular assist devices.

BACKGROUND: Left ventricular assist devices (LVADs) have become common as a bridge to heart transplant as well as destination therapy. Acute care surgical (ACS) problems in this population are prevalent but remain ill-defined. Therefore, we reviewed our experience with ACS interventions in LVAD patients. METHODS: A total of 173 patients who received HeartMate(®) XVE or HeartMate(®) II (HMII) LVADs between December 2001 and March 2010 were studied. Patient demographics, presentation of ACS problem, operative intervention, co-morbidities, transplantation, complications, and survival were analyzed. RESULTS: A total of 47 (27 %) patients underwent 67 ACS procedures at a median of 38 days after device implant (interquartile range 15-110), with a peri-operative mortality rate of 5 % (N = 3). Demographics, device type, and acuity were comparable between the ACS and non-ACS groups. A total of 21 ACS procedures were performed emergently, eight were urgent, and 38 were elective. Of 29 urgent and emergent procedures, 28 were for abdominal pathology. In eight patients, the cause of the ACS problem was related to LVADs or anticoagulation. Cumulative survival estimates revealed no survival differences if patients underwent ACS procedures (p = 0.17). Among HMII patients, transplantation rates were unaffected by an ACS intervention (p = 0.2). CONCLUSIONS: ACS problems occur frequently in LVAD patients and are not associated with adverse outcomes in HMII patients. The acute care surgeon is an integral member of a comprehensive approach to effective LVAD management.

An open source knowledge graph ecosystem for the life sciences.

Translational research requires data at multiple scales of biological organization. Advancements in sequencing and multi-omics technologies have increased the availability of these data, but researchers face significant integration challenges. Knowledge graphs (KGs) are used to model complex phenomena, and methods exist to construct them automatically. However, tackling complex biomedical integration problems requires flexibility in the way knowledge is modeled. Moreover, existing KG construction methods provide robust tooling at the cost of fixed or limited choices among knowledge representation models. PheKnowLator (Phenotype Knowledge Translator) is a semantic ecosystem for automating the FAIR (Findable, Accessible, Interoperable, and Reusable) construction of ontologically grounded KGs with fully customizable knowledge representation. The ecosystem includes KG construction resources (e.g., data preparation APIs), analysis tools (e.g., SPARQL endpoint resources and abstraction algorithms), and benchmarks (e.g., prebuilt KGs). We evaluated the ecosystem by systematically comparing it to existing open-source KG construction methods and by analyzing its computational performance when used to construct 12 different large-scale KGs. With flexible knowledge representation, PheKnowLator enables fully customizable KGs without compromising performance or usability.

Simple score to assess the risk of rejection after orthotopic heart transplantation.

BACKGROUND: The aim of this study was to derive and validate a risk score for rejection after orthotopic heart transplantation. METHODS AND RESULTS: The United Network for Organ Sharing registry was used to identify patients undergoing orthotopic heart transplantation between 1998 and 2008. A total of 14 265 eligible patients were randomly divided into derivation (80%; n=11 412) and validation (20%; n=2853) cohorts. The primary outcome was drug-treated rejection within 1 year of orthotopic heart transplantation. Covariates found to be associated (exploratory univariate P<0.2) with rejection were entered into a multivariable logistic regression model. Inclusion of each variable in the model was assessed by improvement in the McFadden pseudo-R(2), likelihood ratio test, and c index. A risk score was then generated through the use of relative magnitudes of the odds ratios from the derivation cohort, and its ability to predict rejection was tested independently in the validation cohort. A 13-point risk score incorporating 4 variables (age, race, sex, HLA matching) was created. The mean scores in the derivation and validation cohorts were 8.3±2.2 and 8.4±2.1, respectively. Predicted 1-year rejection rates based on the derivation cohort ranged from 16.2% (score=0) to 50.7% (score=13; P<0.001). In weighted regression analysis, there was a strong correlation between these predicted rates of rejection and actual, observed rejection rates in the validation cohort (r(2)=0.96, P<0.001). Logistic regression analysis also demonstrated a significant association (odds ratio, 1.13; P<0.001). The c index of the composite score was equivalent in both the derivation and validation cohorts (c=0.67). CONCLUSIONS: This novel 13-point risk score is highly predictive of clinically significant rejection episodes within 1 year of orthotopic heart transplantation. It has potential utility in tailoring immunosuppressive regimens and in research stratification in orthotopic heart transplantation.

Ventricular assist device implantation in the elderly: nationwide outcomes in the United States.

BACKGROUND: The aim of this study was to evaluate nationwide outcomes of ventricular assist device (VAD) implantation in elderly patients in the United States. METHODS: Patients undergoing VAD implantation between 2003 and 2008 were identified in the Nationwide Inpatient Sample. The primary outcome was inpatient mortality following VAD implantation. Secondary outcomes included disposition following discharge and costs of care. After stratification based on primary versus postcardiotomy VAD support, outcomes were compared between controls aged 60-69 years and elderly patients aged ≥70 years. RESULTS: A total of 2787 patients aged 60-69 years and 1472 patients aged ≥70 years underwent VAD implantation during the study period. Unadjusted mortality rates were comparable between elderly and control patients in both primary support (35.7% vs. 32.1%, p = 0.61) and postcardiotomy support (58.1% vs. 56.1%, p = 0.70). Similarly, in risk-adjusted multivariable logistic regression analysis incorporating clinically relevant variables, age ≥70 did not exert an independent effect on inpatient mortality for either indication. Inpatient costs in the elderly were lower than controls in the primary support cohort, although costs per day were similar, with comparable overall costs between age groups in the postcardiotomy cohort. Elderly survivors were discharged to a facility more frequently than control survivors (primary: 49.9% vs. 29.6%, p = 0.007; postcardiotomy: 67.4% vs. 45.7%, p = 0.03). CONCLUSIONS: This large-cohort population-based analysis provides a useful framework for inpatient prognosis and resource utilization in elderly patients undergoing VAD implantation. Although mortality rates and costs were found to be comparable between elderly patients and those aged 60-69 years, these rates were nonetheless significant. This combined with more frequent discharge-to-facility in elderly survivors underscores the importance of careful patient selection in this population.

Knowledge-Driven Mechanistic Enrichment of the Preeclampsia Ignorome.

Preeclampsia is a leading cause of maternal and fetal morbidity and mortality. Currently, the only definitive treatment of preeclampsia is delivery of the placenta, which is central to the pathogenesis of the disease. Transcriptional profiling of human placenta from pregnancies complicated by preeclampsia has been extensively performed to identify differentially expressed genes (DEGs). The decisions to investigate DEGs experimentally are biased by many factors, causing many DEGs to remain uninvestigated. A set of DEGs which are associated with a disease experimentally, but which have no known association to the disease in the literature are known as the ignorome. Preeclampsia has an extensive body of scientific literature, a large pool of DEG data, and only one definitive treatment. Tools facilitating knowledge-based analyses, which are capable of combining disparate data from many sources in order to suggest underlying mechanisms of action, may be a valuable resource to support discovery and improve our understanding of this disease. In this work we demonstrate how a biomedical knowledge graph (KG) can be used to identify novel preeclampsia molecular mechanisms. Existing open source biomedical resources and publicly available high-throughput transcriptional profiling data were used to identify and annotate the function of currently uninvestigated preeclampsia-associated DEGs. Experimentally investigated genes associated with preeclampsia were identified from PubMed abstracts using text-mining methodologies. The relative complement of the text-mined- and meta-analysis-derived lists were identified as the uninvestigated preeclampsia-associated DEGs (n=445), i.e., the preeclampsia ignorome. Using the KG to investigate relevant DEGs revealed 53 novel clinically relevant and biologically actionable mechanistic associations.

Ketamine Mitigates Neurobehavioral Deficits in a Canine Model of Hypothermic Circulatory Arrest.

Hypothermic circulatory arrest is a protective technique used when complete cessation of circulation is required during cardiac surgery. Prior efforts to decrease neurologic injury with the NMDA receptor antagonist MK801 were limited by unacceptable side effects. We hypothesized that ketamine would provide neuroprotection without dose-limiting side effects. Canines were peripherally cannulated for cardiopulmonary bypass, cooled to 18°C, and underwent 90 minutes of circulatory arrest. Ketamine-treated canines (n = 5; total dose 2.85 mg/kg) were compared to untreated controls (n = 10). A validated neurobehavioral deficit score was obtained at 24, 48, and 72 hours (0 = no deficits/normal exam; higher score represents increasing deficits). Biomarkers of neuronal injury in the cerebrospinal fluid were examined at baseline and at 8, 24, 48, and 72 hours. Brain histopathologic injury was scored at 72 hours (higher score indicates more necrosis and apoptosis). Ketamine-treated canines had significantly improved, lower neurobehavioral deficit scores compared to controls (overall P = 0.003; 24 hours: median 72 vs 112, P = 0.030; 48 hours: 47 vs 90, P = 0.021; 72 hours: 30 vs 89, P = 0.069). Although the histopathologic injury scores of ketamine-treated canines (median 12) were lower than controls (16), there was no statistical difference (P = 0.10). Levels of phosphorylated neurofilament-H and neuron specific enolase, markers of neuronal injury, were significantly lower in ketamine-treated animals (P = 0.010 and = 0.039, respectively). Ketamine significantly reduced neurologic deficits and biomarkers of injury in canines after hypothermic circulatory arrest. Ketamine represents a safe and approved medication that may be useful as a pharmacologic neuroprotectant during cardiac surgery with circulatory arrest.

Reproducible variability: assessing investigator discordance across 9 research teams attempting to reproduce the same observational study.

OBJECTIVE: Observational studies can impact patient care but must be robust and reproducible. Nonreproducibility is primarily caused by unclear reporting of design choices and analytic procedures. This study aimed to: (1) assess how the study logic described in an observational study could be interpreted by independent researchers and (2) quantify the impact of interpretations' variability on patient characteristics. MATERIALS AND METHODS: Nine teams of highly qualified researchers reproduced a cohort from a study by Albogami et al. The teams were provided the clinical codes and access to the tools to create cohort definitions such that the only variable part was their logic choices. We executed teams' cohort definitions against the database and compared the number of subjects, patient overlap, and patient characteristics. RESULTS: On average, the teams' interpretations fully aligned with the master implementation in 4 out of 10 inclusion criteria with at least 4 deviations per team. Cohorts' size varied from one-third of the master cohort size to 10 times the cohort size (2159-63 619 subjects compared to 6196 subjects). Median agreement was 9.4% (interquartile range 15.3-16.2%). The teams' cohorts significantly differed from the master implementation by at least 2 baseline characteristics, and most of the teams differed by at least 5. CONCLUSIONS: Independent research teams attempting to reproduce the study based on its free-text description alone produce different implementations that vary in the population size and composition. Sharing analytical code supported by a common data model and open-source tools allows reproducing a study unambiguously thereby preserving initial design choices.

Ontologizing health systems data at scale: making translational discovery a reality.

Common data models solve many challenges of standardizing electronic health record (EHR) data but are unable to semantically integrate all of the resources needed for deep phenotyping. Open Biological and Biomedical Ontology (OBO) Foundry ontologies provide computable representations of biological knowledge and enable the integration of heterogeneous data. However, mapping EHR data to OBO ontologies requires significant manual curation and domain expertise. We introduce OMOP2OBO, an algorithm for mapping Observational Medical Outcomes Partnership (OMOP) vocabularies to OBO ontologies. Using OMOP2OBO, we produced mappings for 92,367 conditions, 8611 drug ingredients, and 10,673 measurement results, which covered 68-99% of concepts used in clinical practice when examined across 24 hospitals. When used to phenotype rare disease patients, the mappings helped systematically identify undiagnosed patients who might benefit from genetic testing. By aligning OMOP vocabularies to OBO ontologies our algorithm presents new opportunities to advance EHR-based deep phenotyping.

Concept annotation in the CRAFT corpus.

BACKGROUND: Manually annotated corpora are critical for the training and evaluation of automated methods to identify concepts in biomedical text. RESULTS: This paper presents the concept annotations of the Colorado Richly Annotated Full-Text (CRAFT) Corpus, a collection of 97 full-length, open-access biomedical journal articles that have been annotated both semantically and syntactically to serve as a research resource for the biomedical natural-language-processing (NLP) community. CRAFT identifies all mentions of nearly all concepts from nine prominent biomedical ontologies and terminologies: the Cell Type Ontology, the Chemical Entities of Biological Interest ontology, the NCBI Taxonomy, the Protein Ontology, the Sequence Ontology, the entries of the Entrez Gene database, and the three subontologies of the Gene Ontology. The first public release includes the annotations for 67 of the 97 articles, reserving two sets of 15 articles for future text-mining competitions (after which these too will be released). Concept annotations were created based on a single set of guidelines, which has enabled us to achieve consistently high interannotator agreement. CONCLUSIONS: As the initial 67-article release contains more than 560,000 tokens (and the full set more than 790,000 tokens), our corpus is among the largest gold-standard annotated biomedical corpora. Unlike most others, the journal articles that comprise the corpus are drawn from diverse biomedical disciplines and are marked up in their entirety. Additionally, with a concept-annotation count of nearly 100,000 in the 67-article subset (and more than 140,000 in the full collection), the scale of conceptual markup is also among the largest of comparable corpora. The concept annotations of the CRAFT Corpus have the potential to significantly advance biomedical text mining by providing a high-quality gold standard for NLP systems. The corpus, annotation guidelines, and other associated resources are freely available at http://bionlp-corpora.sourceforge.net/CRAFT/index.shtml.

A corpus of full-text journal articles is a robust evaluation tool for revealing differences in performance of biomedical natural language processing tools.

BACKGROUND: We introduce the linguistic annotation of a corpus of 97 full-text biomedical publications, known as the Colorado Richly Annotated Full Text (CRAFT) corpus. We further assess the performance of existing tools for performing sentence splitting, tokenization, syntactic parsing, and named entity recognition on this corpus. RESULTS: Many biomedical natural language processing systems demonstrated large differences between their previously published results and their performance on the CRAFT corpus when tested with the publicly available models or rule sets. Trainable systems differed widely with respect to their ability to build high-performing models based on this data. CONCLUSIONS: The finding that some systems were able to train high-performing models based on this corpus is additional evidence, beyond high inter-annotator agreement, that the quality of the CRAFT corpus is high. The overall poor performance of various systems indicates that considerable work needs to be done to enable natural language processing systems to work well when the input is full-text journal articles. The CRAFT corpus provides a valuable resource to the biomedical natural language processing community for evaluation and training of new models for biomedical full text publications.

Institutional factors beyond procedural volume significantly impact center variability in outcomes after orthotopic heart transplantation.

OBJECTIVE: To evaluate the contribution of institutional volume and other unmeasured institutional factors beyond volume to the between-center variability in outcomes after orthotopic heart transplantation (OHT). BACKGROUND: It is unclear if institutional factors beyond volume have a significant impact on OHT outcomes. METHODS: The United Network for Organ Sharing registry was used to identify OHTs performed between 2000 and 2010. Separate mixed-effect logistic regression models were constructed, with the primary endpoint being post-OHT mortality. Model A included only individual centers, model B added validated recipient and donor risk indices as well as the year of transplantation, and model C added institutional volume as a continuous variable to model B. The reduction in between-center variability in mortality between models B and C was used to define the contribution of institutional volume. Kaplan-Meier survival curves were also compared after stratifying patients into equal-size tertiles based on center volume. RESULTS: A total of 119 centers performed OHT in 19,156 patients. After adjusting for transplantation year and differences in recipient and donor risk, decreasing center volume was associated with an increased risk of 1-year mortality (P < 0.001). However, procedural volume only accounted for 16.7% of the variability in mortality between centers, and significant between-center variability persisted after adjusting for institutional volume (P<0.001). In Kaplan-Meier analysis, there was significant variability in 1-year survival between centers within each volume category: low-volume (66.7%-96.6%), intermediate-volume (80.7%-97.3%), and high-volume (83.8%-93.9%). These trends were also observed with 5-year mortality. CONCLUSIONS: This large-cohort analysis demonstrates that although institutional volume is a significant predictor of post-OHT outcomes, there are other unmeasured institutional factors that contribute substantially to the between-center variability in outcomes. Institutional volume should therefore not be the sole indicator of "center quality" in OHT.