Deep Neural Networks for Fine-Grained Surveillance of Overdose Mortality.

Surveillance of drug overdose deaths relies on death certificates for identification of the substances that caused death. Drugs and drug classes can be identified through the International Classification of Diseases, Tenth Revision (ICD-10), codes present on death certificates. However, ICD-10 codes do not always provide high levels of specificity in drug identification. To achieve more fine-grained identification of substances on death certificate, the free-text cause-of-death section, completed by the medical certifier, must be analyzed. Current methods for analyzing free-text death certificates rely solely on lookup tables for identifying specific substances, which must be frequently updated and maintained. To improve identification of drugs on death certificates, a deep-learning named-entity recognition model was developed, utilizing data from the Kentucky Drug Overdose Fatality Surveillance System (2014-2019), which achieved an F1-score of 99.13%. This model can identify new drug misspellings and novel substances that are not present on current surveillance lookup tables, enhancing the surveillance of drug overdose deaths.

Improved biomedical word embeddings in the transformer era.

BACKGROUND: Recent natural language processing (NLP) research is dominated by neural network methods that employ word embeddings as basic building blocks. Pre-training with neural methods that capture local and global distributional properties (e.g., skip-gram, GLoVE) using free text corpora is often used to embed both words and concepts. Pre-trained embeddings are typically leveraged in downstream tasks using various neural architectures that are designed to optimize task-specific objectives that might further tune such embeddings. OBJECTIVE: Despite advances in contextualized language model based embeddings, static word embeddings still form an essential starting point in BioNLP research and applications. They are useful in low resource settings and in lexical semantics studies. Our main goal is to build improved biomedical word embeddings and make them publicly available for downstream applications. METHODS: We jointly learn word and concept embeddings by first using the skip-gram method and further fine-tuning them with correlational information manifesting in co-occurring Medical Subject Heading (MeSH) concepts in biomedical citations. This fine-tuning is accomplished with the transformer-based BERT architecture in the two-sentence input mode with a classification objective that captures MeSH pair co-occurrence. We conduct evaluations of these tuned static embeddings using multiple datasets for word relatedness developed by previous efforts. RESULTS: Both in qualitative and quantitative evaluations we demonstrate that our methods produce improved biomedical embeddings in comparison with other static embedding efforts. Without selectively culling concepts and terms (as was pursued by previous efforts), we believe we offer the most exhaustive evaluation of biomedical embeddings to date with clear performance improvements across the board. CONCLUSION: We repurposed a transformer architecture (typically used to generate dynamic embeddings) to improve static biomedical word embeddings using concept correlations. We provide our code and embeddings for public use for downstream applications and research endeavors: https://github.com/bionlproc/BERT-CRel-Embeddings.

International Changes in COVID-19 Clinical Trajectories Across 315 Hospitals and 6 Countries: Retrospective Cohort Study.

BACKGROUND: Many countries have experienced 2 predominant waves of COVID-19-related hospitalizations. Comparing the clinical trajectories of patients hospitalized in separate waves of the pandemic enables further understanding of the evolving epidemiology, pathophysiology, and health care dynamics of the COVID-19 pandemic. OBJECTIVE: In this retrospective cohort study, we analyzed electronic health record (EHR) data from patients with SARS-CoV-2 infections hospitalized in participating health care systems representing 315 hospitals across 6 countries. We compared hospitalization rates, severe COVID-19 risk, and mean laboratory values between patients hospitalized during the first and second waves of the pandemic. METHODS: Using a federated approach, each participating health care system extracted patient-level clinical data on their first and second wave cohorts and submitted aggregated data to the central site. Data quality control steps were adopted at the central site to correct for implausible values and harmonize units. Statistical analyses were performed by computing individual health care system effect sizes and synthesizing these using random effect meta-analyses to account for heterogeneity. We focused the laboratory analysis on C-reactive protein (CRP), ferritin, fibrinogen, procalcitonin, D-dimer, and creatinine based on their reported associations with severe COVID-19. RESULTS: Data were available for 79,613 patients, of which 32,467 were hospitalized in the first wave and 47,146 in the second wave. The prevalence of male patients and patients aged 50 to 69 years decreased significantly between the first and second waves. Patients hospitalized in the second wave had a 9.9% reduction in the risk of severe COVID-19 compared to patients hospitalized in the first wave (95% CI 8.5%-11.3%). Demographic subgroup analyses indicated that patients aged 26 to 49 years and 50 to 69 years; male and female patients; and black patients had significantly lower risk for severe disease in the second wave than in the first wave. At admission, the mean values of CRP were significantly lower in the second wave than in the first wave. On the seventh hospital day, the mean values of CRP, ferritin, fibrinogen, and procalcitonin were significantly lower in the second wave than in the first wave. In general, countries exhibited variable changes in laboratory testing rates from the first to the second wave. At admission, there was a significantly higher testing rate for D-dimer in France, Germany, and Spain. CONCLUSIONS: Patients hospitalized in the second wave were at significantly lower risk for severe COVID-19. This corresponded to mean laboratory values in the second wave that were more likely to be in typical physiological ranges on the seventh hospital day compared to the first wave. Our federated approach demonstrated the feasibility and power of harmonizing heterogeneous EHR data from multiple international health care systems to rapidly conduct large-scale studies to characterize how COVID-19 clinical trajectories evolve.

Authorship Correction: International Changes in COVID-19 Clinical Trajectories Across 315 Hospitals and 6 Countries: Retrospective Cohort Study.

[This corrects the article DOI: 10.2196/31400.].

Generalizing biomedical relation classification with neural adversarial domain adaptation.

Motivation: Creating large datasets for biomedical relation classification can be prohibitively expensive. While some datasets have been curated to extract protein-protein and drug-drug interactions (PPIs and DDIs) from text, we are also interested in other interactions including gene-disease and chemical-protein connections. Also, many biomedical researchers have begun to explore ternary relationships. Even when annotated data are available, many datasets used for relation classification are inherently biased. For example, issues such as sample selection bias typically prevent models from generalizing in the wild. To address the problem of cross-corpora generalization, we present a novel adversarial learning algorithm for unsupervised domain adaptation tasks where no labeled data are available in the target domain. Instead, our method takes advantage of unlabeled data to improve biased classifiers through learning domain-invariant features via an adversarial process. Finally, our method is built upon recent advances in neural network (NN) methods. Results: We experiment by extracting PPIs and DDIs from text. In our experiments, we show domain invariant features can be learned in NNs such that classifiers trained for one interaction type (protein-protein) can be re-purposed to others (drug-drug). We also show that our method can adapt to different source and target pairs of PPI datasets. Compared to prior convolutional and recurrent NN-based relation classification methods without domain adaptation, we achieve improvements as high as 30% in F1-score. Likewise, we show improvements over state-of-the-art adversarial methods. Availability and implementation: Experimental code is available at https://github.com/bionlproc/adversarial-relation-classification. Supplementary information: Supplementary data are available at Bioinformatics online.

Cross-registry neural domain adaptation to extract mutational test results from pathology reports.

OBJECTIVE: We study the performance of machine learning (ML) methods, including neural networks (NNs), to extract mutational test results from pathology reports collected by cancer registries. Given the lack of hand-labeled datasets for mutational test result extraction, we focus on the particular use-case of extracting Epidermal Growth Factor Receptor mutation results in non-small cell lung cancers. We explore the generalization of NNs across different registries where our goals are twofold: (1) to assess how well models trained on a registry's data port to test data from a different registry and (2) to assess whether and to what extent such models can be improved using state-of-the-art neural domain adaptation techniques under different assumptions about what is available (labeled vs unlabeled data) at the target registry site. MATERIALS AND METHODS: We collected data from two registries: the Kentucky Cancer Registry (KCR) and the Fred Hutchinson Cancer Research Center (FH) Cancer Surveillance System. We combine NNs with adversarial domain adaptation to improve cross-registry performance. We compare to other classifiers in the standard supervised classification, unsupervised domain adaptation, and supervised domain adaptation scenarios. RESULTS: The performance of ML methods varied between registries. To extract positive results, the basic convolutional neural network (CNN) had an F1 of 71.5% on the KCR dataset and 95.7% on the FH dataset. For the KCR dataset, the CNN F1 results were low when trained on FH data (Positive F1: 23%). Using our proposed adversarial CNN, without any labeled data, we match the F1 of the models trained directly on each target registry's data. The adversarial CNN F1 improved when trained on FH and applied to KCR dataset (Positive F1: 70.8%). We found similar performance improvements when we trained on KCR and tested on FH reports (Positive F1: 45% to 96%). CONCLUSION: Adversarial domain adaptation improves the performance of NNs applied to pathology reports. In the unsupervised domain adaptation setting, we match the performance of models that are trained directly on target registry's data by using source registry's labeled data and unlabeled examples from the target registry.

Exploiting semantic patterns over biomedical knowledge graphs for predicting treatment and causative relations.

BACKGROUND: Identifying new potential treatment options for medical conditions that cause human disease burden is a central task of biomedical research. Since all candidate drugs cannot be tested with animal and clinical trials, in vitro approaches are first attempted to identify promising candidates. Likewise, identifying different causal relations between biomedical entities is also critical to understand biomedical processes. Generally, natural language processing (NLP) and machine learning are used to predict specific relations between any given pair of entities using the distant supervision approach. OBJECTIVE: To build high accuracy supervised predictive models to predict previously unknown treatment and causative relations between biomedical entities based only on semantic graph pattern features extracted from biomedical knowledge graphs. METHODS: We used 7000 treats and 2918 causes hand-curated relations from the UMLS Metathesaurus to train and test our models. Our graph pattern features are extracted from simple paths connecting biomedical entities in the SemMedDB graph (based on the well-known SemMedDB database made available by the U.S. National Library of Medicine). Using these graph patterns connecting biomedical entities as features of logistic regression and decision tree models, we computed mean performance measures (precision, recall, F-score) over 100 distinct 80-20% train-test splits of the datasets. For all experiments, we used a positive:negative class imbalance of 1:10 in the test set to model relatively more realistic scenarios. RESULTS: Our models predict treats and causes relations with high F-scores of 99% and 90% respectively. Logistic regression model coefficients also help us identify highly discriminative patterns that have an intuitive interpretation. We are also able to predict some new plausible relations based on false positives that our models scored highly based on our collaborations with two physician co-authors. Finally, our decision tree models are able to retrieve over 50% of treatment relations from a recently created external dataset. CONCLUSIONS: We employed semantic graph patterns connecting pairs of candidate biomedical entities in a knowledge graph as features to predict treatment/causative relations between them. We provide what we believe is the first evidence in direct prediction of biomedical relations based on graph features. Our work complements lexical pattern based approaches in that the graph patterns can be used as additional features for weakly supervised relation prediction.

Ordinal convolutional neural networks for predicting RDoC positive valence psychiatric symptom severity scores.

BACKGROUND: The CEGS N-GRID 2016 Shared Task in Clinical Natural Language Processing (NLP) provided a set of 1000 neuropsychiatric notes to participants as part of a competition to predict psychiatric symptom severity scores. This paper summarizes our methods, results, and experiences based on our participation in the second track of the shared task. OBJECTIVE: Classical methods of text classification usually fall into one of three problem types: binary, multi-class, and multi-label classification. In this effort, we study ordinal regression problems with text data where misclassifications are penalized differently based on how far apart the ground truth and model predictions are on the ordinal scale. Specifically, we present our entries (methods and results) in the N-GRID shared task in predicting research domain criteria (RDoC) positive valence ordinal symptom severity scores (absent, mild, moderate, and severe) from psychiatric notes. METHODS: We propose a novel convolutional neural network (CNN) model designed to handle ordinal regression tasks on psychiatric notes. Broadly speaking, our model combines an ordinal loss function, a CNN, and conventional feature engineering (wide features) into a single model which is learned end-to-end. Given interpretability is an important concern with nonlinear models, we apply a recent approach called locally interpretable model-agnostic explanation (LIME) to identify important words that lead to instance specific predictions. RESULTS: Our best model entered into the shared task placed third among 24 teams and scored a macro mean absolute error (MMAE) based normalized score (100·(1-MMAE)) of 83.86. Since the competition, we improved our score (using basic ensembling) to 85.55, comparable with the winning shared task entry. Applying LIME to model predictions, we demonstrate the feasibility of instance specific prediction interpretation by identifying words that led to a particular decision. CONCLUSION: In this paper, we present a method that successfully uses wide features and an ordinal loss function applied to convolutional neural networks for ordinal text classification specifically in predicting psychiatric symptom severity scores. Our approach leads to excellent performance on the N-GRID shared task and is also amenable to interpretability using existing model-agnostic approaches.

Predicting mental conditions based on "history of present illness" in psychiatric notes with deep neural networks.

BACKGROUND: Applications of natural language processing to mental health notes are not common given the sensitive nature of the associated narratives. The CEGS N-GRID 2016 Shared Task in Clinical Natural Language Processing (NLP) changed this scenario by providing the first set of neuropsychiatric notes to participants. This study summarizes our efforts and results in proposing a novel data use case for this dataset as part of the third track in this shared task. OBJECTIVE: We explore the feasibility and effectiveness of predicting a set of common mental conditions a patient has based on the short textual description of patient's history of present illness typically occurring in the beginning of a psychiatric initial evaluation note. MATERIALS AND METHODS: We clean and process the 1000 records made available through the N-GRID clinical NLP task into a key-value dictionary and build a dataset of 986 examples for which there is a narrative for history of present illness as well as Yes/No responses with regards to presence of specific mental conditions. We propose two independent deep neural network models: one based on convolutional neural networks (CNN) and another based on recurrent neural networks with hierarchical attention (ReHAN), the latter of which allows for interpretation of model decisions. We conduct experiments to compare these methods to each other and to baselines based on linear models and named entity recognition (NER). RESULTS: Our CNN model with optimized thresholding of output probability estimates achieves best overall mean micro-F score of 63.144% for 11 common mental conditions with statistically significant gains (p<0.05) over all other models. The ReHAN model with interpretable attention mechanism scored 61.904% mean micro-F1 score. Both models' improvements over baseline models (support vector machines and NER) are statistically significant. The ReHAN model additionally aids in interpretation of the results by surfacing important words and sentences that lead to a particular prediction for each instance. CONCLUSIONS: Although the history of present illness is a short text segment averaging 300 words, it is a good predictor for a few conditions such as anxiety, depression, panic disorder, and attention deficit hyperactivity disorder. Proposed CNN and RNN models outperform baseline approaches and complement each other when evaluating on a per-label basis.

Toward automated e-cigarette surveillance: Spotting e-cigarette proponents on Twitter.

BACKGROUND: Electronic cigarettes (e-cigarettes or e-cigs) are a popular emerging tobacco product. Because e-cigs do not generate toxic tobacco combustion products that result from smoking regular cigarettes, they are sometimes perceived and promoted as a less harmful alternative to smoking and also as means to quit smoking. However, the safety of e-cigs and their efficacy in supporting smoking cessation is yet to be determined. Importantly, the federal drug administration (FDA) currently does not regulate e-cigs and as such their manufacturing, marketing, and sale is not subject to the rules that apply to traditional cigarettes. A number of manufacturers, advocates, and e-cig users are actively promoting e-cigs on Twitter. OBJECTIVE: We develop a high accuracy supervised predictive model to automatically identify e-cig "proponents" on Twitter and analyze the quantitative variation of their tweeting behavior along popular themes when compared with other Twitter users (or tweeters). METHODS: Using a dataset of 1000 independently annotated Twitter profiles by two different annotators, we employed a variety of textual features from latest tweet content and tweeter profile biography to build predictive models to automatically identify proponent tweeters. We used a set of manually curated key phrases to analyze e-cig proponent tweets from a corpus of over one million e-cig tweets along well known e-cig themes and compared the results with those generated by regular tweeters. RESULTS: Our model identifies e-cig proponents with 97% precision, 86% recall, 91% F-score, and 96% overall accuracy, with tight 95% confidence intervals. We find that as opposed to regular tweeters that form over 90% of the dataset, e-cig proponents are a much smaller subset but tweet two to five times more than regular tweeters. Proponents also disproportionately (one to two orders of magnitude more) highlight e-cig flavors, their smoke-free and potential harm reduction aspects, and their claimed use in smoking cessation. CONCLUSIONS: Given FDA is currently in the process of proposing meaningful regulation, we believe our work demonstrates the strong potential of informatics approaches, specifically machine learning, for automated e-cig surveillance on Twitter.

Context-driven automatic subgraph creation for literature-based discovery.

BACKGROUND: Literature-based discovery (LBD) is characterized by uncovering hidden associations in non-interacting scientific literature. Prior approaches to LBD include use of: (1) domain expertise and structured background knowledge to manually filter and explore the literature, (2) distributional statistics and graph-theoretic measures to rank interesting connections, and (3) heuristics to help eliminate spurious connections. However, manual approaches to LBD are not scalable and purely distributional approaches may not be sufficient to obtain insights into the meaning of poorly understood associations. While several graph-based approaches have the potential to elucidate associations, their effectiveness has not been fully demonstrated. A considerable degree of a priori knowledge, heuristics, and manual filtering is still required. OBJECTIVES: In this paper we implement and evaluate a context-driven, automatic subgraph creation method that captures multifaceted complex associations between biomedical concepts to facilitate LBD. Given a pair of concepts, our method automatically generates a ranked list of subgraphs, which provide informative and potentially unknown associations between such concepts. METHODS: To generate subgraphs, the set of all MEDLINE articles that contain either of the two specified concepts (A, C) are first collected. Then binary relationships or assertions, which are automatically extracted from the MEDLINE articles, called semantic predications, are used to create a labeled directed predications graph. In this predications graph, a path is represented as a sequence of semantic predications. The hierarchical agglomerative clustering (HAC) algorithm is then applied to cluster paths that are bounded by the two concepts (A, C). HAC relies on implicit semantics captured through Medical Subject Heading (MeSH) descriptors, and explicit semantics from the MeSH hierarchy, for clustering. Paths that exceed a threshold of semantic relatedness are clustered into subgraphs based on their shared context. Finally, the automatically generated clusters are provided as a ranked list of subgraphs. RESULTS: The subgraphs generated using this approach facilitated the rediscovery of 8 out of 9 existing scientific discoveries. In particular, they directly (or indirectly) led to the recovery of several intermediates (or B-concepts) between A- and C-terms, while also providing insights into the meaning of the associations. Such meaning is derived from predicates between the concepts, as well as the provenance of the semantic predications in MEDLINE. Additionally, by generating subgraphs on different thematic dimensions (such as Cellular Activity, Pharmaceutical Treatment and Tissue Function), the approach may enable a broader understanding of the nature of complex associations between concepts. Finally, in a statistical evaluation to determine the interestingness of the subgraphs, it was observed that an arbitrary association is mentioned in only approximately 4 articles in MEDLINE on average. CONCLUSION: These results suggest that leveraging the implicit and explicit semantics provided by manually assigned MeSH descriptors is an effective representation for capturing the underlying context of complex associations, along multiple thematic dimensions in LBD situations.

Leveraging output term co-occurrence frequencies and latent associations in predicting medical subject headings.

Trained indexers at the National Library of Medicine (NLM) manually tag each biomedical abstract with the most suitable terms from the Medical Subject Headings (MeSH) terminology to be indexed by their PubMed information system. MeSH has over 26,000 terms and indexers look at each article's full text while assigning the terms. Recent automated attempts focused on using the article title and abstract text to identify MeSH terms for the corresponding article. Most of these approaches used supervised machine learning techniques that use already indexed articles and the corresponding MeSH terms. In this paper, we present a new indexing approach that leverages term co-occurrence frequencies and latent term associations computed using MeSH term sets corresponding to a set of nearly 18 million articles already indexed with MeSH terms by indexers at NLM. The main goal of our study is to gauge the potential of output label co-occurrences, latent associations, and relationships extracted from free text in both unsupervised and supervised indexing approaches. In this paper, using a novel and purely unsupervised approach, we achieve a micro-F-score that is comparable to those obtained using supervised machine learning techniques. By incorporating term co-occurrence and latent association features into a supervised learning framework, we also improve over the best results published on two public datasets.

COVID-19 Event Extraction from Twitter via Extractive Question Answering with Continuous Prompts.

As COVID-19 ravages the world, social media analytics could augment traditional surveys in assessing how the pandemic evolves and capturing consumer chatter that could help healthcare agencies in addressing it. This typically involves mining disclosure events that mention testing positive for the disease or discussions surrounding perceptions and beliefs in preventative or treatment options. The 2020 shared task on COVID-19 event extraction (conducted as part of the W-NUT workshop during the EMNLP conference) introduced a new Twitter dataset for benchmarking event extraction from COVID-19 tweets. In this paper, we cast the problem of event extraction as extractive question answering using recent advances in continuous prompting in language models. On the shared task test dataset, our approach leads to over 5% absolute micro-averaged F1-score improvement over prior best results, across all COVID-19 event slots. Our ablation study shows that continuous prompts have a major impact on the eventual performance.

End-to-End n-ary Relation Extraction for Combination Drug Therapies.

Combination drug therapies are treatment regimens that involve two or more drugs, administered more commonly for patients with cancer, HIV, malaria, or tuberculosis. Currently there are over 350K articles in PubMed that use the combination drug therapy MeSH heading with at least 10K articles published per year over the past two decades. Extracting combination therapies from scientific literature inherently constitutes an n-ary relation extraction problem. Unlike in the general n-ary setting where n is fixed (e.g., drug-gene-mutation relations where n = 3), extracting combination therapies is a special setting where n ≥ 2 is dynamic, depending on each instance. Recently, Tiktinsky et al. (NAACL 2022) introduced a first of its kind dataset, CombDrugExt, for extracting such therapies from literature. Here, we use a sequence-to-sequence style end-to-end extraction method to achieve an F1-Score of 66.7% on the CombDrugExt test set for positive (or effective) combinations. This is an absolute ≈ 5% F1-score improvement even over the prior best relation classification score with spotted drug entities (hence, not end-to-end). Thus our effort introduces a state-of-the-art first model for end-to-end extraction that is already superior to the best prior non end-to-end model for this task. Our model seamlessly extracts all drug entities and relations in a single pass and is highly suitable for dynamic n-ary extraction scenarios.

End-to-End Models for Chemical-Protein Interaction Extraction: Better Tokenization and Span-Based Pipeline Strategies.

End-to-end relation extraction (E2ERE) is an important task in information extraction, more so for biomedicine as scientific literature continues to grow exponentially. E2ERE typically involves identifying entities (or named entity recognition (NER)) and associated relations, while most RE tasks simply assume that the entities are provided upfront and end up performing relation classification. E2ERE is inherently more difficult than RE alone given the potential snowball effect of errors from NER leading to more errors in RE. A complex dataset in biomedical E2ERE is the ChemProt dataset (BioCreative VI, 2017) that identifies relations between chemical compounds and genes/proteins in scientific literature. ChemProt is included in all recent biomedical natural language processing benchmarks including BLUE, BLURB, and BigBio. However, its treatment in these benchmarks and in other separate efforts is typically not end-to-end, with few exceptions. In this effort, we employ a span-based pipeline approach to produce a new state-of-the-art E2ERE performance on the ChemProt dataset, resulting in > 4% improvement in F1-score over the prior best effort. Our results indicate that a straightforward fine-grained tokenization scheme helps span-based approaches excel in E2ERE, especially with regards to handling complex named entities. Our error analysis also identifies a few key failure modes in E2ERE for ChemProt.

Acquisition of a Lexicon for Family History Information: Bidirectional Encoder Representations From Transformers-Assisted Sublanguage Analysis.

BACKGROUND: A patient's family history (FH) information significantly influences downstream clinical care. Despite this importance, there is no standardized method to capture FH information in electronic health records and a substantial portion of FH information is frequently embedded in clinical notes. This renders FH information difficult to use in downstream data analytics or clinical decision support applications. To address this issue, a natural language processing system capable of extracting and normalizing FH information can be used. OBJECTIVE: In this study, we aimed to construct an FH lexical resource for information extraction and normalization. METHODS: We exploited a transformer-based method to construct an FH lexical resource leveraging a corpus consisting of clinical notes generated as part of primary care. The usability of the lexicon was demonstrated through the development of a rule-based FH system that extracts FH entities and relations as specified in previous FH challenges. We also experimented with a deep learning-based FH system for FH information extraction. Previous FH challenge data sets were used for evaluation. RESULTS: The resulting lexicon contains 33,603 lexicon entries normalized to 6408 concept unique identifiers of the Unified Medical Language System and 15,126 codes of the Systematized Nomenclature of Medicine Clinical Terms, with an average number of 5.4 variants per concept. The performance evaluation demonstrated that the rule-based FH system achieved reasonable performance. The combination of the rule-based FH system with a state-of-the-art deep learning-based FH system can improve the recall of FH information evaluated using the BioCreative/N2C2 FH challenge data set, with the F1 score varied but comparable. CONCLUSIONS: The resulting lexicon and rule-based FH system are freely available through the Open Health Natural Language Processing GitHub.

DeepPhe-CR: Natural Language Processing Software Services for Cancer Registrar Case Abstraction.

Objective: The manual extraction of case details from patient records for cancer surveillance efforts is a resource-intensive task. Natural Language Processing (NLP) techniques have been proposed for automating the identification of key details in clinical notes. Our goal was to develop NLP application programming interfaces (APIs) for integration into cancer registry data abstraction tools in a computer-assisted abstraction setting. Methods: We used cancer registry manual abstraction processes to guide the design of DeepPhe-CR, a web-based NLP service API. The coding of key variables was done through NLP methods validated using established workflows. A container-based implementation including the NLP wasdeveloped. Existing registry data abstraction software was modified to include results from DeepPhe-CR. An initial usability study with data registrars provided early validation of the feasibility of the DeepPhe-CR tools. Results: API calls support submission of single documents and summarization of cases across multiple documents. The container-based implementation uses a REST router to handle requests and support a graph database for storing results. NLP modules extract topography, histology, behavior, laterality, and grade at 0.79-1.00 F1 across common and rare cancer types (breast, prostate, lung, colorectal, ovary and pediatric brain) on data from two cancer registries. Usability study participants were able to use the tool effectively and expressed interest in adopting the tool. Discussion: Our DeepPhe-CR system provides a flexible architecture for building cancer-specific NLP tools directly into registrar workflows in a computer-assisted abstraction setting. Improving user interactions in client tools, may be needed to realize the potential of these approaches. DeepPhe-CR: https://deepphe.github.io/.

A Comparative Effectiveness Study on Opioid Use Disorder Prediction Using Artificial Intelligence and Existing Risk Models.

Opioid use disorder (OUD) is a leading cause of death in the United States placing a tremendous burden on patients, their families, and health care systems. Artificial intelligence (AI) can be harnessed with available healthcare data to produce automated OUD prediction tools. In this retrospective study, we developed AI based models for OUD prediction and showed that AI can predict OUD more effectively than existing clinical tools including the unweighted opioid risk tool (ORT). Data include 474,208 patients' data over 10 years; 269,748 were females with an average age of 56.78 years. Cases are prescription opioid users with at least one diagnosis of OUD or at least one prescription for buprenorphine or methadone. Controls are prescription opioid users with no OUD diagnoses or buprenorphine or methadone prescriptions. On 100 randomly selected test sets including 47,396 patients, our proposed transformer-based AI model can predict OUD more efficiently (AUC = 0.742 ± 0.021) compared to logistic regression (AUC = 0.651 ± 0.025), random forest (AUC = 0.679 ± 0.026), xgboost (AUC = 0.690 ± 0.027), long short-term memory model (AUC = 0.706 ± 0.026), transformer (AUC = 0.725 ± 0.024), and unweighted ORT model (AUC = 0.559 ± 0.025). Our results show that embedding AI algorithms into clinical care may assist clinicians in risk stratification and management of patients receiving opioid therapy.

DeepPhe-CR: Natural Language Processing Software Services for Cancer Registrar Case Abstraction.

PURPOSE: Manual extraction of case details from patient records for cancer surveillance is a resource-intensive task. Natural Language Processing (NLP) techniques have been proposed for automating the identification of key details in clinical notes. Our goal was to develop NLP application programming interfaces (APIs) for integration into cancer registry data abstraction tools in a computer-assisted abstraction setting. METHODS: We used cancer registry manual abstraction processes to guide the design of DeepPhe-CR, a web-based NLP service API. The coding of key variables was performed through NLP methods validated using established workflows. A container-based implementation of the NLP methods and the supporting infrastructure was developed. Existing registry data abstraction software was modified to include results from DeepPhe-CR. An initial usability study with data registrars provided early validation of the feasibility of the DeepPhe-CR tools. RESULTS: API calls support submission of single documents and summarization of cases across one or more documents. The container-based implementation uses a REST router to handle requests and support a graph database for storing results. NLP modules extract topography, histology, behavior, laterality, and grade at 0.79-1.00 F1 across multiple cancer types (breast, prostate, lung, colorectal, ovary, and pediatric brain) from data of two population-based cancer registries. Usability study participants were able to use the tool effectively and expressed interest in the tool. CONCLUSION: The DeepPhe-CR system provides an architecture for building cancer-specific NLP tools directly into registrar workflows in a computer-assisted abstraction setting. Improved user interactions in client tools may be needed to realize the potential of these approaches.