Moving Biosurveillance Beyond Coded Data Using AI for Symptom Detection From Physician Notes: Retrospective Cohort Study.

BACKGROUND: Real-time surveillance of emerging infectious diseases necessitates a dynamically evolving, computable case definition, which frequently incorporates symptom-related criteria. For symptom detection, both population health monitoring platforms and research initiatives primarily depend on structured data extracted from electronic health records. OBJECTIVE: This study sought to validate and test an artificial intelligence (AI)-based natural language processing (NLP) pipeline for detecting COVID-19 symptoms from physician notes in pediatric patients. We specifically study patients presenting to the emergency department (ED) who can be sentinel cases in an outbreak. METHODS: Subjects in this retrospective cohort study are patients who are 21 years of age and younger, who presented to a pediatric ED at a large academic children's hospital between March 1, 2020, and May 31, 2022. The ED notes for all patients were processed with an NLP pipeline tuned to detect the mention of 11 COVID-19 symptoms based on Centers for Disease Control and Prevention (CDC) criteria. For a gold standard, 3 subject matter experts labeled 226 ED notes and had strong agreement (F1-score=0.986; positive predictive value [PPV]=0.972; and sensitivity=1.0). F1-score, PPV, and sensitivity were used to compare the performance of both NLP and the International Classification of Diseases, 10th Revision (ICD-10) coding to the gold standard chart review. As a formative use case, variations in symptom patterns were measured across SARS-CoV-2 variant eras. RESULTS: There were 85,678 ED encounters during the study period, including 4% (n=3420) with patients with COVID-19. NLP was more accurate at identifying encounters with patients that had any of the COVID-19 symptoms (F1-score=0.796) than ICD-10 codes (F1-score =0.451). NLP accuracy was higher for positive symptoms (sensitivity=0.930) than ICD-10 (sensitivity=0.300). However, ICD-10 accuracy was higher for negative symptoms (specificity=0.994) than NLP (specificity=0.917). Congestion or runny nose showed the highest accuracy difference (NLP: F1-score=0.828 and ICD-10: F1-score=0.042). For encounters with patients with COVID-19, prevalence estimates of each NLP symptom differed across variant eras. Patients with COVID-19 were more likely to have each NLP symptom detected than patients without this disease. Effect sizes (odds ratios) varied across pandemic eras. CONCLUSIONS: This study establishes the value of AI-based NLP as a highly effective tool for real-time COVID-19 symptom detection in pediatric patients, outperforming traditional ICD-10 methods. It also reveals the evolving nature of symptom prevalence across different virus variants, underscoring the need for dynamic, technology-driven approaches in infectious disease surveillance.

Cumulus: A federated EHR-based learning system powered by FHIR and AI.

Objective: To address challenges in large-scale electronic health record (EHR) data exchange, we sought to develop, deploy, and test an open source, cloud-hosted app 'listener' that accesses standardized data across the SMART/HL7 Bulk FHIR Access application programming interface (API). Methods: We advance a model for scalable, federated, data sharing and learning. Cumulus software is designed to address key technology and policy desiderata including local utility, control, and administrative simplicity as well as privacy preservation during robust data sharing, and AI for processing unstructured text. Results: Cumulus relies on containerized, cloud-hosted software, installed within a healthcare organization's security envelope. Cumulus accesses EHR data via the Bulk FHIR interface and streamlines automated processing and sharing. The modular design enables use of the latest AI and natural language processing tools and supports provider autonomy and administrative simplicity. In an initial test, Cumulus was deployed across five healthcare systems each partnered with public health. Cumulus output is patient counts which were aggregated into a table stratifying variables of interest to enable population health studies. All code is available open source. A policy stipulating that only aggregate data leave the institution greatly facilitated data sharing agreements. Discussion and Conclusion: Cumulus addresses barriers to data sharing based on (1) federally required support for standard APIs (2), increasing use of cloud computing, and (3) advances in AI. There is potential for scalability to support learning across myriad network configurations and use cases.

Improving the Transferability of Clinical Note Section Classification Models with BERT and Large Language Model Ensembles.

Text in electronic health records is organized into sections, and classifying those sections into section categories is useful for downstream tasks. In this work, we attempt to improve the transferability of section classification models by combining the dataset-specific knowledge in supervised learning models with the world knowledge inside large language models (LLMs). Surprisingly, we find that zero-shot LLMs out-perform supervised BERT-based models applied to out-of-domain data. We also find that their strengths are synergistic, so that a simple ensemble technique leads to additional performance gains.

Improving model transferability for clinical note section classification models using continued pretraining.

OBJECTIVE: The classification of clinical note sections is a critical step before doing more fine-grained natural language processing tasks such as social determinants of health extraction and temporal information extraction. Often, clinical note section classification models that achieve high accuracy for 1 institution experience a large drop of accuracy when transferred to another institution. The objective of this study is to develop methods that classify clinical note sections under the SOAP ("Subjective," "Object," "Assessment," and "Plan") framework with improved transferability. MATERIALS AND METHODS: We trained the baseline models by fine-tuning BERT-based models, and enhanced their transferability with continued pretraining, including domain-adaptive pretraining and task-adaptive pretraining. We added in-domain annotated samples during fine-tuning and observed model performance over a varying number of annotated sample size. Finally, we quantified the impact of continued pretraining in equivalence of the number of in-domain annotated samples added. RESULTS: We found continued pretraining improved models only when combined with in-domain annotated samples, improving the F1 score from 0.756 to 0.808, averaged across 3 datasets. This improvement was equivalent to adding 35 in-domain annotated samples. DISCUSSION: Although considered a straightforward task when performing in-domain, section classification is still a considerably difficult task when performing cross-domain, even using highly sophisticated neural network-based methods. CONCLUSION: Continued pretraining improved model transferability for cross-domain clinical note section classification in the presence of a small amount of in-domain labeled samples.

Natural Language Processing to Automatically Extract the Presence and Severity of Esophagitis in Notes of Patients Undergoing Radiotherapy.

PURPOSE: Radiotherapy (RT) toxicities can impair survival and quality of life, yet remain understudied. Real-world evidence holds potential to improve our understanding of toxicities, but toxicity information is often only in clinical notes. We developed natural language processing (NLP) models to identify the presence and severity of esophagitis from notes of patients treated with thoracic RT. METHODS: Our corpus consisted of a gold-labeled data set of 1,524 clinical notes from 124 patients with lung cancer treated with RT, manually annotated for Common Terminology Criteria for Adverse Events (CTCAE) v5.0 esophagitis grade, and a silver-labeled data set of 2,420 notes from 1,832 patients from whom toxicity grades had been collected as structured data during clinical care. We fine-tuned statistical and pretrained Bidirectional Encoder Representations from Transformers-based models for three esophagitis classification tasks: task 1, no esophagitis versus grade 1-3; task 2, grade ≤1 versus >1; and task 3, no esophagitis versus grade 1 versus grade 2-3. Transferability was tested on 345 notes from patients with esophageal cancer undergoing RT. RESULTS: Fine-tuning of PubMedBERT yielded the best performance. The best macro-F1 was 0.92, 0.82, and 0.74 for tasks 1, 2, and 3, respectively. Selecting the most informative note sections during fine-tuning improved macro-F1 by ≥2% for all tasks. Silver-labeled data improved the macro-F1 by ≥3% across all tasks. For the esophageal cancer notes, the best macro-F1 was 0.73, 0.74, and 0.65 for tasks 1, 2, and 3, respectively, without additional fine-tuning. CONCLUSION: To our knowledge, this is the first effort to automatically extract esophagitis toxicity severity according to CTCAE guidelines from clinical notes. This provides proof of concept for NLP-based automated detailed toxicity monitoring in expanded domains.

A computable case definition for patients with SARS-CoV2 testing that occurred outside the hospital.

Objective: To identify a cohort of COVID-19 cases, including when evidence of virus positivity was only mentioned in the clinical text, not in structured laboratory data in the electronic health record (EHR). Materials and Methods: Statistical classifiers were trained on feature representations derived from unstructured text in patient EHRs. We used a proxy dataset of patients with COVID-19 polymerase chain reaction (PCR) tests for training. We selected a model based on performance on our proxy dataset and applied it to instances without COVID-19 PCR tests. A physician reviewed a sample of these instances to validate the classifier. Results: On the test split of the proxy dataset, our best classifier obtained 0.56 F1, 0.6 precision, and 0.52 recall scores for SARS-CoV2 positive cases. In an expert validation, the classifier correctly identified 97.6% (81/84) as COVID-19 positive and 97.8% (91/93) as not SARS-CoV2 positive. The classifier labeled an additional 960 cases as not having SARS-CoV2 lab tests in hospital, and only 177 of those cases had the ICD-10 code for COVID-19. Discussion: Proxy dataset performance may be worse because these instances sometimes include discussion of pending lab tests. The most predictive features are meaningful and interpretable. The type of external test that was performed is rarely mentioned. Conclusion: COVID-19 cases that had testing done outside of the hospital can be reliably detected from the text in EHRs. Training on a proxy dataset was a suitable method for developing a highly performant classifier without labor-intensive labeling efforts.

Improving Model Transferability for Clinical Note Section Classification Models Using Continued Pretraining.

Objective: The classification of clinical note sections is a critical step before doing more fine-grained natural language processing tasks such as social determinants of health extraction and temporal information extraction. Often, clinical note section classification models that achieve high accuracy for one institution experience a large drop of accuracy when transferred to another institution. The objective of this study is to develop methods that classify clinical note sections under the SOAP ("Subjective", "Object", "Assessment" and "Plan") framework with improved transferability. Materials and methods: We trained the baseline models by fine-tuning BERT-based models, and enhanced their transferability with continued pretraining, including domain adaptive pretraining (DAPT) and task adaptive pretraining (TAPT). We added out-of-domain annotated samples during fine-tuning and observed model performance over a varying number of annotated sample size. Finally, we quantified the impact of continued pretraining in equivalence of the number of in-domain annotated samples added. Results: We found continued pretraining improved models only when combined with in-domain annotated samples, improving the F1 score from 0.756 to 0.808, averaged across three datasets. This improvement was equivalent to adding 50.2 in-domain annotated samples. Discussion: Although considered a straightforward task when performing in-domain, section classification is still a considerably difficult task when performing cross-domain, even using highly sophisticated neural network-based methods. Conclusion: Continued pretraining improved model transferability for cross-domain clinical note section classification in the presence of a small amount of in-domain labeled samples.

Natural Language Processing Methods to Empirically Explore Social Contexts and Needs in Cancer Patient Notes.

PURPOSE: There is an unmet need to empirically explore and understand drivers of cancer disparities, particularly social determinants of health. We explored natural language processing methods to automatically and empirically extract clinical documentation of social contexts and needs that may underlie disparities. METHODS: This was a retrospective analysis of 230,325 clinical notes from 5,285 patients treated with radiotherapy from 2007 to 2019. We compared linguistic features among White versus non-White, low-income insurance versus other insurance, and male versus female patients' notes. Log odds ratios with an informative Dirichlet prior were calculated to compare words over-represented in each group. A variational autoencoder topic model was applied, and topic probability was compared between groups. The presence of machine-learnable bias was explored by developing statistical and neural demographic group classifiers. RESULTS: Terms associated with varied social contexts and needs were identified for all demographic group comparisons. For example, notes of non-White and low-income insurance patients were over-represented with terms associated with housing and transportation, whereas notes of White and other insurance patients were over-represented with terms related to physical activity. Topic models identified a social history topic, and topic probability varied significantly between the demographic group comparisons. Classification models performed poorly at classifying notes of non-White and low-income insurance patients (F1 of 0.30 and 0.23, respectively). CONCLUSION: Exploration of linguistic differences in clinical notes between patients of different race/ethnicity, insurance status, and sex identified social contexts and needs in patients with cancer and revealed high-level differences in notes. Future work is needed to validate whether these findings may play a role in cancer disparities.

The SMART Text2FHIR Pipeline.

Objective: To implement an open source, free, and easily deployable high throughput natural language processing module to extract concepts from clinician notes and map them to Fast Healthcare Interoperability Resources (FHIR). Materials and Methods: Using a popular open-source NLP tool (Apache cTAKES), we create FHIR resources that use modifier extensions to represent negation and NLP sourcing, and another extension to represent provenance of extracted concepts. Results: The SMART Text2FHIR Pipeline is an open-source tool, released through standard package managers, and publicly available container images that implement the mappings, enabling ready conversion of clinical text to FHIR. Discussion: With the increased data liquidity because of new interoperability regulations, NLP processes that can output FHIR can enable a common language for transporting structured and unstructured data. This framework can be valuable for critical public health or clinical research use cases. Conclusion: Future work should include mapping more categories of NLP-extracted information into FHIR resources and mappings from additional open-source NLP tools.

A computable phenotype for patients with SARS-CoV2 testing that occurred outside the hospital.

Objective: To identify a cohort of COVID-19 cases, including when evidence of virus positivity was only mentioned in the clinical text, not in structured laboratory data in the electronic health record (EHR). Materials and Methods: Statistical classifiers were trained on feature representations derived from unstructured text in patient electronic health records (EHRs). We used a proxy dataset of patients with COVID-19 polymerase chain reaction (PCR) tests for training. We selected a model based on performance on our proxy dataset and applied it to instances without COVID-19 PCR tests. A physician reviewed a sample of these instances to validate the classifier. Results: On the test split of the proxy dataset, our best classifier obtained 0.56 F1, 0.6 precision, and 0.52 recall scores for SARS-CoV2 positive cases. In an expert validation, the classifier correctly identified 90.8% (79/87) as COVID-19 positive and 97.8% (91/93) as not SARS-CoV2 positive. The classifier identified an additional 960 positive cases that did not have SARS-CoV2 lab tests in hospital, and only 177 of those cases had the ICD-10 code for COVID-19. Discussion: Proxy dataset performance may be worse because these instances sometimes include discussion of pending lab tests. The most predictive features are meaningful and interpretable. The type of external test that was performed is rarely mentioned. Conclusion: COVID-19 cases that had testing done outside of the hospital can be reliably detected from the text in EHRs. Training on a proxy dataset was a suitable method for developing a highly performant classifier without labor intensive labeling efforts.

The SMART Text2FHIR Pipeline.

Objective: To implement an open source, free, and easily deployable high throughput natural language processing module to extract concepts from clinician notes and map them to Fast Healthcare Interoperability Resources (FHIR). Materials and Methods: Using a popular open-source NLP tool (Apache cTAKES), we create FHIR resources that use modifier extensions to represent negation and NLP sourcing, and another extension to represent provenance of extracted concepts. Results: The SMART Text2FHIR Pipeline is an open-source tool, released through standard package managers, and publicly available container images that implement the mappings, enabling ready conversion of clinical text to FHIR. Discussion: With the increased data liquidity because of new interoperability regulations, NLP processes that can output FHIR can enable a common language for transporting structured and unstructured data. This framework can be valuable for critical public health or clinical research use cases. Conclusion: Future work should include mapping more categories of NLP-extracted information into FHIR resources and mappings from additional open-source NLP tools.

Classifying unstructured electronic consult messages to understand primary care physician specialty information needs.

OBJECTIVE: Electronic consultation (eConsult) content reflects important information about referring clinician needs across an organization, but is challenging to extract. The objective of this work was to develop machine learning models for classifying eConsult questions for question type and question content. Another objective of this work was to investigate the ability to solve this task with constrained expert time resources. MATERIALS AND METHODS: Our data source is the San Francisco Health Network eConsult system, with over 700 000 deidentified questions from the years 2008-2017, from gastroenterology, urology, and neurology specialties. We develop classifiers based on Bidirectional Encoder Representations from Transformers, experimenting with multitask learning to learn when information can be shared across classifiers. We produce learning curves to understand when we may be able to reduce the amount of human labeling required. RESULTS: Multitask learning shows benefits only in the neurology-urology pair where they shared substantial similarities in the distribution of question types. Continued pretraining of models in new domains is highly effective. In the neurology-urology pair, near-peak performance is achieved with only 10% of the urology training data given all of the neurology data. DISCUSSION: Sharing information across classifier types shows little benefit, whereas sharing classifier components across specialties can help if they are similar in the balance of procedural versus cognitive patient care. CONCLUSION: We can accurately classify eConsult content with enough labeled data, but only in special cases do methods for reducing labeling effort apply. Future work should explore new learning paradigms to further reduce labeling effort.

Clinical Natural Language Processing for Radiation Oncology: A Review and Practical Primer.

Natural language processing (NLP), which aims to convert human language into expressions that can be analyzed by computers, is one of the most rapidly developing and widely used technologies in the field of artificial intelligence. Natural language processing algorithms convert unstructured free text data into structured data that can be extracted and analyzed at scale. In medicine, this unlocking of the rich, expressive data within clinical free text in electronic medical records will help untap the full potential of big data for research and clinical purposes. Recent major NLP algorithmic advances have significantly improved the performance of these algorithms, leading to a surge in academic and industry interest in developing tools to automate information extraction and phenotyping from clinical texts. Thus, these technologies are poised to transform medical research and alter clinical practices in the future. Radiation oncology stands to benefit from NLP algorithms if they are appropriately developed and deployed, as they may enable advances such as automated inclusion of radiation therapy details into cancer registries, discovery of novel insights about cancer care, and improved patient data curation and presentation at the point of care. However, challenges remain before the full value of NLP is realized, such as the plethora of jargon specific to radiation oncology, nonstandard nomenclature, a lack of publicly available labeled data for model development, and interoperability limitations between radiation oncology data silos. Successful development and implementation of high quality and high value NLP models for radiation oncology will require close collaboration between computer scientists and the radiation oncology community. Here, we present a primer on artificial intelligence algorithms in general and NLP algorithms in particular; provide guidance on how to assess the performance of such algorithms; review prior research on NLP algorithms for oncology; and describe future avenues for NLP in radiation oncology research and clinics.

Incorporating Risk Factor Embeddings in Pre-trained Transformers Improves Sentiment Prediction in Psychiatric Discharge Summaries.

Reducing rates of early hospital readmission has been recognized and identified as a key to improve quality of care and reduce costs. There are a number of risk factors that have been hypothesized to be important for understanding re-admission risk, including such factors as problems with substance abuse, ability to maintain work, relations with family. In this work, we develop RoBERTa-based models to predict the sentiment of sentences describing readmission risk factors in discharge summaries of patients with psychosis. We improve substantially on previous results by a scheme that shares information across risk factors while also allowing the model to learn risk factor-specific information.

Classifying Electronic Consults for Triage Status and Question Type.

Electronic consult (eConsult) systems allow specialists more flexibility to respond to referrals more efficiently, thereby increasing access in under-resourced healthcare settings like safety net systems. Understanding the usage patterns of eConsult system is an important part of improving specialist efficiency. In this work, we develop and apply classifiers to a dataset of eConsult questions from primary care providers to specialists, classifying the messages for how they were triaged by the specialist office, and the underlying type of clinical question posed by the primary care provider. We show that pre-trained transformer models are strong baselines, with improving performance from domain-specific training and shared representations.

Rethinking domain adaptation for machine learning over clinical language.

Building clinical natural language processing (NLP) systems that work on widely varying data is an absolute necessity because of the expense of obtaining new training data. While domain adaptation research can have a positive impact on this problem, the most widely studied paradigms do not take into account the realities of clinical data sharing. To address this issue, we lay out a taxonomy of domain adaptation, parameterizing by what data is shareable. We show that the most realistic settings for clinical use cases are seriously under-studied. To support research in these important directions, we make a series of recommendations, not just for domain adaptation but for clinical NLP in general, that ensure that data, shared tasks, and released models are broadly useful, and that initiate research directions where the clinical NLP community can lead the broader NLP and machine learning fields.

Experiences implementing scalable, containerized, cloud-based NLP for extracting biobank participant phenotypes at scale.

OBJECTIVE: To develop scalable natural language processing (NLP) infrastructure for processing the free text in electronic health records (EHRs). MATERIALS AND METHODS: We extend the open-source Apache cTAKES NLP software with several standard technologies for scalability. We remove processing bottlenecks by monitoring component queue size. We process EHR free text for patients in the PrecisionLink Biobank at Boston Children's Hospital. The extracted concepts are made searchable via a web-based portal. RESULTS: We processed over 1.2 million notes for over 8000 patients, extracting 154 million concepts. Our largest tested configuration processes over 1 million notes per day. DISCUSSION: The unique information represented by extracted NLP concepts has great potential to provide a more complete picture of patient status. CONCLUSION: NLP large EHR document collections can be done efficiently, in service of high throughput phenotyping.