Systems Approach to Integrating Preclinical Apolipoprotein E-Knockout Investigations Reveals Novel Etiologic Pathways and Master Atherosclerosis Network in Humans.

OBJECTIVE: Animal models of atherosclerosis are used extensively to interrogate molecular mechanisms in serial fashion. We tested whether a novel systems biology approach to integration of preclinical data identifies novel pathways and regulators in human disease. Approach and Results: Of 716 articles published in ATVB from 1995 to 2019 using the apolipoprotein E knockout mouse to study atherosclerosis, data were extracted from 360 unique studies in which a gene was experimentally perturbed to impact plaque size or composition and analyzed using Ingenuity Pathway Analysis software. TREM1 (triggering receptor expressed on myeloid cells) signaling and LXR/RXR (liver X receptor/retinoid X receptor) activation were identified as the top atherosclerosis-associated pathways in mice (both P<1.93×10-4, TREM1 implicated early and LXR/RXR in late atherogenesis). The top upstream regulatory network in mice (sc-58125, a COX2 inhibitor) linked 64.0% of the genes into a single network. The pathways and networks identified in mice were interrogated by testing for associations between the genetically predicted gene expression of each mouse pathway-identified human homolog with clinical atherosclerosis in a cohort of 88 660 human subjects. Homologous human pathways and networks were significantly enriched for gene-atherosclerosis associations (empirical P<0.01 for TREM1 and LXR/RXR pathways and COX2 network). This included 12(60.0%) TREM1 pathway genes, 15(53.6%) LXR/RXR pathway genes, and 67(49.3%) COX2 network genes. Mouse analyses predicted, and human study validated, the strong association of COX2 expression (PTGS2) with increased likelihood of atherosclerosis (odds ratio, 1.68 per SD of genetically predicted gene expression; P=1.07×10-6). CONCLUSIONS: PRESCIANT (Preclinical Science Integration and Translation) leverages published preclinical investigations to identify high-confidence pathways, networks, and regulators of human disease.

Phenotyping coronavirus disease 2019 during a global health pandemic: Lessons learned from the characterization of an early cohort.

From the start of the coronavirus disease 2019 (COVID-19) pandemic, researchers have looked to electronic health record (EHR) data as a way to study possible risk factors and outcomes. To ensure the validity and accuracy of research using these data, investigators need to be confident that the phenotypes they construct are reliable and accurate, reflecting the healthcare settings from which they are ascertained. We developed a COVID-19 registry at a single academic medical center and used data from March 1 to June 5, 2020 to assess differences in population-level characteristics in pandemic and non-pandemic years respectively. Median EHR length, previously shown to impact phenotype performance in type 2 diabetes, was significantly shorter in the SARS-CoV-2 positive group relative to a 2019 influenza tested group (median 3.1 years vs 8.7; Wilcoxon rank sum P = 1.3e-52). Using three phenotyping methods of increasing complexity (billing codes alone and domain-specific algorithms provided by an EHR vendor and clinical experts), common medical comorbidities were abstracted from COVID-19 EHRs, defined by the presence of a positive laboratory test (positive predictive value 100%, recall 93%). After combining performance data across phenotyping methods, we observed significantly lower false negative rates for those records billed for a comprehensive care visit (p = 4e-11) and those with complete demographics data recorded (p = 7e-5). In an early COVID-19 cohort, we found that phenotyping performance of nine common comorbidities was influenced by median EHR length, consistent with previous studies, as well as by data density, which can be measured using portable metrics including CPT codes. Here we present those challenges and potential solutions to creating deeply phenotyped, acute COVID-19 cohorts.

Leveraging medical context to recommend semantically similar terms for chart reviews.

BACKGROUND: Information retrieval (IR) help clinicians answer questions posed to large collections of electronic medical records (EMRs), such as how best to identify a patient's cancer stage. One of the more promising approaches to IR for EMRs is to expand a keyword query with similar terms (e.g., augmenting cancer with mets). However, there is a large range of clinical chart review tasks, such that fixed sets of similar terms is insufficient. Current language models, such as Bidirectional Encoder Representations from Transformers (BERT) embeddings, do not capture the full non-textual context of a task. In this study, we present new methods that provide similar terms dynamically by adjusting with the context of the chart review task. METHODS: We introduce a vector space for medical-context in which each word is represented by a vector that captures the word's usage in different medical contexts (e.g., how frequently cancer is used when ordering a prescription versus describing family history) beyond the context learned from the surrounding text. These vectors are transformed into a vector space for customizing the set of similar terms selected for different chart review tasks. We evaluate the vector space model with multiple chart review tasks, in which supervised machine learning models learn to predict the preferred terms of clinically knowledgeable reviewers. To quantify the usefulness of the predicted similar terms to a baseline of standard word2vec embeddings, we measure (1) the prediction performance of the medical-context vector space model using the area under the receiver operating characteristic curve (AUROC) and (2) the labeling effort required to train the models. RESULTS: The vector space outperformed the baseline word2vec embeddings in all three chart review tasks with an average AUROC of 0.80 versus 0.66, respectively. Additionally, the medical-context vector space significantly reduced the number of labels required to learn and predict the preferred similar terms of reviewers. Specifically, the labeling effort was reduced to 10% of the entire dataset in all three tasks. CONCLUSIONS: The set of preferred similar terms that are relevant to a chart review task can be learned by leveraging the medical context of the task.

Delirium Variability is Influenced by the Sound Environment (DEVISE Study): How Changes in the Intensive Care Unit soundscape affect delirium incidence.

Quantitative data on the sensory environment of intensive care unit (ICU) patients and its potential link to increased risk of delirium is limited. We examined whether higher average sound and light levels in ICU environments are associated with delirium incidence. Over 111 million sound and light measurements from 143 patient stays in the surgical and trauma ICUs were collected using Quietyme® (Neshkoro, Wisconsin) sensors from May to July 2018 and analyzed. Sensory data were grouped into time of day, then normalized against their ICU environments, with Confusion Assessment Method (CAM-ICU) scores measured each shift. We then performed logistic regression analysis, adjusting for possible confounding variables. Lower morning sound averages (8 am-12 pm) (OR = 0.835, 95% OR CI = [0.746, 0.934], p = 0.002) and higher daytime sound averages (12 pm-6 pm) (OR = 1.157, 95% OR CI = [1.036, 1.292], p = 0.011) were associated with an increased odds of delirium incidence, while nighttime sound averages (10 pm-8 am) (OR = 0.990, 95% OR CI = [0.804, 1.221], p = 0.928) and the ICU light environment did not show statistical significance. Our results suggest an association between the ICU soundscape and the odds of developing delirium. This creates a future paradigm for studies of the ICU soundscape and lightscape.

Understanding the Information Needs and Context of Trauma Handoffs to Design Automated Sensing Clinical Documentation Technologies: Qualitative Mixed-Method Study of Military and Civilian Cases.

BACKGROUND: Current methods of communication between the point of injury and receiving medical facilities rely on verbal communication, supported by brief notes and the memory of the field medic. This communication can be made more complete and reliable with technologies that automatically document the actions of field medics. However, designing state-of-the-art technology for military field personnel and civilian first responders is challenging due to the barriers researchers face in accessing the environment and understanding situated actions and cognitive models employed in the field. OBJECTIVE: To identify design insights for an automated sensing clinical documentation (ASCD) system, we sought to understand what information is transferred in trauma cases between prehospital and hospital personnel, and what contextual factors influence the collection, management, and handover of information in trauma cases, in both military and civilian cases. METHODS: Using a multi-method approach including video review and focus groups, we developed an understanding of the information needs of trauma handoffs and the context of field documentation to inform the design of an automated sensing documentation system that uses wearables, cameras, and environmental sensors to passively infer clinical activity and automatically produce documentation. RESULTS: Comparing military and civilian trauma documentation and handoff, we found similarities in the types of data collected and the prioritization of information. We found that military environments involved many more contextual factors that have implications for design, such as the physical environment (eg, heat, lack of lighting, lack of power) and the potential for active combat and triage, creating additional complexity. CONCLUSIONS: An ineffectiveness of communication is evident in both the civilian and military worlds. We used multiple methods of inquiry to study the information needs of trauma care and handoff, and the context of medical work in the field. Our findings informed the design and evaluation of an automated documentation tool. The data illustrated the need for more accurate recordkeeping, specifically temporal aspects, during transportation, and characterized the environment in which field testing of the developed tool will take place. The employment of a systems perspective in this project produced design insights that our team would not have identified otherwise. These insights created exciting and interesting challenges for the technical team to resolve.

Crowdsourcing to delineate skin affected by chronic graft-vs-host disease.

BACKGROUND: Estimating the extent of affected skin is an important unmet clinical need both for research and practical management in many diseases. In particular, cutaneous burden of chronic graft-vs-host disease (cGVHD) is a primary outcome in many trials. Despite advances in artificial intelligence and 3D photography, progress toward reliable automated techniques is hindered by limited expert time to delineate cGVHD patient images. Crowdsourcing may have potential to provide the requisite expert-level data. MATERIALS AND METHODS: Forty-one three-dimensional photographs of three cutaneous cGVHD patients were delineated by a board-certified dermatologist. 410 two-dimensional projections of the raw photos were each annotated by seven crowd workers, whose consensus performance was compared to the expert. RESULTS: The consensus delineation by four of seven crowd workers achieved the highest agreement with the expert, measured by a median Dice index of 0.7551 across all 410 images, outperforming even the best worker from the crowd (Dice index 0.7216). For their internal agreement, crowd workers achieved a median Fleiss's kappa of 0.4140 across the images. The time a worker spent marking an image had only weak correlation with the surface area marked, and very low correlation with accuracy. Percent of pixels selected by the consensus exhibited good correlation (Pearson R = 0.81) with the patient's affected surface area. CONCLUSION: Crowdsourcing may be an efficient method for obtaining demarcations of affected skin, on par with expert performance. Crowdsourced data generally agreed with the current clinical standard of percent body surface area to assess cGVHD severity in the skin.

A clinical risk prediction model to identify patients with hepatorenal syndrome at hospital admission.

BACKGROUND: Hepatorenal syndrome (HRS) is a life-threatening complication of cirrhosis and early detection of evolving HRS may provide opportunities for early intervention. We developed a HRS risk model to assist early recognition of inpatient HRS. METHODS: We analysed a retrospective cohort of patients hospitalised from among 122 medical centres in the US Department of Veterans Affairs between 1 January 2005 and 31 December 2013. We included cirrhotic patients who had Kidney Disease Improving Global Outcomes criteria based acute kidney injury on admission. We developed a logistic regression risk prediction model to detect HRS on admission using 10 variables. We calculated 95% confidence intervals on the model building dataset and, subsequently, calculated performance on a 1000 sample holdout test set. We report model performance with area under the curve (AUC) for discrimination and several calibration measures. RESULTS: The cohort included 19 368 patients comprising 32 047 inpatient admissions. The event rate for hospitalised HRS was 2810/31 047 (9.1%) and 79/1000 (7.9%) in the model building and validation datasets, respectively. The variable selection procedure designed a parsimonious model involving ten predictor variables. Final model performance in the validation dataset had an AUC of 0.87, Brier score of 0.05, slope of 1.10 and intercept of 0.04. CONCLUSIONS: We developed a probabilistic risk model to diagnose HRS within 24 hours of hospital admission using routine clinical variables in the largest ever published HRS cohort. The performance was excellent and this model may help identify high-risk patients for HRS and promote early intervention.

Extracting similar terms from multiple EMR-based semantic embeddings to support chart reviews.

OBJECTIVE: Word embeddings project semantically similar terms into nearby points in a vector space. When trained on clinical text, these embeddings can be leveraged to improve keyword search and text highlighting. In this paper, we present methods to refine the selection process of similar terms from multiple EMR-based word embeddings, and evaluate their performance quantitatively and qualitatively across multiple chart review tasks. MATERIALS AND METHODS: Word embeddings were trained on each clinical note type in an EMR. These embeddings were then combined, weighted, and truncated to select a refined set of similar terms to be used in keyword search and text highlighting. To evaluate their quality, we measured the similar terms' information retrieval (IR) performance using precision-at-K (P@5, P@10). Additionally a user study evaluated users' search term preferences, while a timing study measured the time to answer a question from a clinical chart. RESULTS: The refined terms outperformed the baseline method's information retrieval performance (e.g., increasing the average P@5 from 0.48 to 0.60). Additionally, the refined terms were preferred by most users, and reduced the average time to answer a question. CONCLUSIONS: Clinical information can be more quickly retrieved and synthesized when using semantically similar term from multiple embeddings.

Development of an automated phenotyping algorithm for hepatorenal syndrome.

OBJECTIVE: Hepatorenal Syndrome (HRS) is a devastating form of acute kidney injury (AKI) in advanced liver disease patients with high morbidity and mortality, but phenotyping algorithms have not yet been developed using large electronic health record (EHR) databases. We evaluated and compared multiple phenotyping methods to achieve an accurate algorithm for HRS identification. MATERIALS AND METHODS: A national retrospective cohort of patients with cirrhosis and AKI admitted to 124 Veterans Affairs hospitals was assembled from electronic health record data collected from 2005 to 2013. AKI was defined by the Kidney Disease: Improving Global Outcomes criteria. Five hundred and four hospitalizations were selected for manual chart review and served as the gold standard. Electronic Health Record based predictors were identified using structured and free text clinical data, subjected through NLP from the clinical Text Analysis Knowledge Extraction System. We explored several dimension reduction techniques for the NLP data, including newer high-throughput phenotyping and word embedding methods, and ascertained their effectiveness in identifying the phenotype without structured predictor variables. With the combined structured and NLP variables, we analyzed five phenotyping algorithms: penalized logistic regression, naïve Bayes, support vector machines, random forest, and gradient boosting. Calibration and discrimination metrics were calculated using 100 bootstrap iterations. In the final model, we report odds ratios and 95% confidence intervals. RESULTS: The area under the receiver operating characteristic curve (AUC) for the different models ranged from 0.73 to 0.93; with penalized logistic regression having the best discriminatory performance. Calibration for logistic regression was modest, but gradient boosting and support vector machines were superior. NLP identified 6985 variables; a priori variable selection performed similarly to dimensionality reduction using high-throughput phenotyping and semantic similarity informed clustering (AUC of 0.81 - 0.82). CONCLUSION: This study demonstrated improved phenotyping of a challenging AKI etiology, HRS, over ICD-9 coding. We also compared performance among multiple approaches to EHR-derived phenotyping, and found similar results between methods. Lastly, we showed that automated NLP dimension reduction is viable for acute illness.

A System for Automated Determination of Perioperative Patient Acuity.

The widely used American Society of Anesthesiologists Physical Status (ASA PS) classification is subjective, requires manual clinician review to score, and has limited granularity. Our objective was to develop a system that automatically generates an ASA PS with finer granularity by creating a continuous ASA PS score. Supervised machine learning methods were used to create a model that predicts a patient's ASA PS on a continuous scale using the patient's home medications and comorbidities. Three different types of predictive models were trained: regression models, ordinal models, and classification models. The performance and agreement of each model to anesthesiologists were compared by calculating the mean squared error (MSE), rounded MSE and Cohen's Kappa on a holdout set. To assess model performance on continuous ASA PS, model rankings were compared to two anesthesiologists on a subset of ASA PS 3 case pairs. The random forest regression model achieved the best MSE and rounded MSE. A model consisting of three random forest classifiers (split model) achieved the best Cohen's Kappa. The model's agreement with our anesthesiologists on the ASA PS 3 case pairs yielded fair to moderate Kappa values. The results suggest that the random forest split classification model can predict ASA PS with agreement similar to that of anesthesiologists reported in literature and produce a continuous score in which agreement in accurately judging granularity is fair to moderate.

Complexity of medical decision-making in care provided by surgeons through patient portals.

BACKGROUND: Patient portals are online applications that allow patients to interact with healthcare organizations and view information. Portal messages exchanged between patients and providers contain diverse types of communications, including delivery of medical care. The types of communications and complexity of medical decision-making in portal messages sent to surgeons have not been studied. MATERIALS AND METHODS: We obtained all message threads initiated by patients and exchanged with surgical providers through the Vanderbilt University Medical Center patient portal from June 1 to December 31, 2014. Five hundred randomly selected messages were manually analyzed by two research team members to determine the types of communication (i.e., informational, medical, logistical, or social), whether medical care was delivered, and complexity of medical decision-making as defined for outpatient billing in each message thread. RESULTS: A total of 9408 message threads were sent to 401 surgical providers during the study period. In the 500 threads selected for detailed analysis, 1293 distinct issues were communicated, with an average of 2.6 issues per thread. Medical needs were communicated in 453 message threads (90.6%). Further, 339 message threads (67.8%) contained medical decision-making. Overall complexity of medical decision-making was straightforward in 210 messages (62%), low in 102 messages (30%), and moderate in 27 messages (8%). No highly complex decisions were made over portal messaging. CONCLUSIONS: Through patient portal messages, surgeons deliver substantial medical care with varied levels of medical complexity. Models for compensation of online care must be developed as consumer and surgeon adoption of these technologies increases.

Classifying patient portal messages using Convolutional Neural Networks.

OBJECTIVE: Patients communicate with healthcare providers via secure messaging in patient portals. As patient portal adoption increases, growing messaging volumes may overwhelm providers. Prior research has demonstrated promise in automating classification of patient portal messages into communication types to support message triage or answering. This paper examines if using semantic features and word context improves portal message classification. MATERIALS AND METHODS: Portal messages were classified into the following categories: informational, medical, social, and logistical. We constructed features from portal messages including bag of words, bag of phrases, graph representations, and word embeddings. We trained one-versus-all random forest and logistic regression classifiers, and convolutional neural network (CNN) with a softmax output. We evaluated each classifier's performance using Area Under the Curve (AUC). RESULTS: Representing the messages using bag of words, the random forest detected informational, medical, social, and logistical communications in patient portal messages with AUCs: 0.803, 0.884, 0.828, and 0.928, respectively. Graph representations of messages outperformed simpler features with AUCs: 0.837, 0.914, 0.846, 0.884 for informational, medical, social, and logistical communication, respectively. Representing words with Word2Vec embeddings, and mapping features using a CNN had the best performance with AUCs: 0.908 for informational, 0.917 for medical, 0.935 for social, and 0.943 for logistical categories. DISCUSSION AND CONCLUSION: Word2Vec and graph representations improved the accuracy of classifying portal messages compared to features that lacked semantic information such as bag of words, and bag of phrases. Furthermore, using Word2Vec along with a CNN model, which provide a higher order representation, improved the classification of portal messages.

A Scalable Framework to Detect Personal Health Mentions on Twitter.

BACKGROUND: Biomedical research has traditionally been conducted via surveys and the analysis of medical records. However, these resources are limited in their content, such that non-traditional domains (eg, online forums and social media) have an opportunity to supplement the view of an individual's health. OBJECTIVE: The objective of this study was to develop a scalable framework to detect personal health status mentions on Twitter and assess the extent to which such information is disclosed. METHODS: We collected more than 250 million tweets via the Twitter streaming API over a 2-month period in 2014. The corpus was filtered down to approximately 250,000 tweets, stratified across 34 high-impact health issues, based on guidance from the Medical Expenditure Panel Survey. We created a labeled corpus of several thousand tweets via a survey, administered over Amazon Mechanical Turk, that documents when terms correspond to mentions of personal health issues or an alternative (eg, a metaphor). We engineered a scalable classifier for personal health mentions via feature selection and assessed its potential over the health issues. We further investigated the utility of the tweets by determining the extent to which Twitter users disclose personal health status. RESULTS: Our investigation yielded several notable findings. First, we find that tweets from a small subset of the health issues can train a scalable classifier to detect health mentions. Specifically, training on 2000 tweets from four health issues (cancer, depression, hypertension, and leukemia) yielded a classifier with precision of 0.77 on all 34 health issues. Second, Twitter users disclosed personal health status for all health issues. Notably, personal health status was disclosed over 50% of the time for 11 out of 34 (33%) investigated health issues. Third, the disclosure rate was dependent on the health issue in a statistically significant manner (P<.001). For instance, more than 80% of the tweets about migraines (83/100) and allergies (85/100) communicated personal health status, while only around 10% of the tweets about obesity (13/100) and heart attack (12/100) did so. Fourth, the likelihood that people disclose their own versus other people's health status was dependent on health issue in a statistically significant manner as well (P<.001). For example, 69% (69/100) of the insomnia tweets disclosed the author's status, while only 1% (1/100) disclosed another person's status. By contrast, 1% (1/100) of the Down syndrome tweets disclosed the author's status, while 21% (21/100) disclosed another person's status. CONCLUSIONS: It is possible to automatically detect personal health status mentions on Twitter in a scalable manner. These mentions correspond to the health issues of the Twitter users themselves, but also other individuals. Though this study did not investigate the veracity of such statements, we anticipate such information may be useful in supplementing traditional health-related sources for research purposes.

Use of noisy labels as weak learners to identify incompletely ascertainable outcomes: A Feasibility study with opioid-induced respiratory depression.

Objective: Assigning outcome labels to large observational data sets in a timely and accurate manner, particularly when outcomes are rare or not directly ascertainable, remains a significant challenge within biomedical informatics. We examined whether noisy labels generated from subject matter experts' heuristics using heterogenous data types within a data programming paradigm could provide outcomes labels to a large, observational data set. We chose the clinical condition of opioid-induced respiratory depression for our use case because it is rare, has no administrative codes to easily identify the condition, and typically requires at least some unstructured text to ascertain its presence. Materials and methods: Using de-identified electronic health records of 52,861 post-operative encounters, we applied a data programming paradigm (implemented in the Snorkel software) for the development of a machine learning classifier for opioid-induced respiratory depression. Our approach included subject matter experts creating 14 labeling functions that served as noisy labels for developing a probabilistic Generative model. We used probabilistic labels from the Generative model as outcome labels for training a Discriminative model on the source data. We evaluated performance of the Discriminative model with a hold-out test set of 599 independently-reviewed patient records. Results: The final Discriminative classification model achieved an accuracy of 0.977, an F1 score of 0.417, a sensitivity of 1.0, and an AUC of 0.988 in the hold-out test set with a prevalence of 0.83% (5/599). Discussion: All of the confirmed Cases were identified by the classifier. For rare outcomes, this finding is encouraging because it reduces the number of manual reviews needed by excluding visits/patients with low probabilities. Conclusion: Application of a data programming paradigm with expert-informed labeling functions might have utility for phenotyping clinical phenomena that are not easily ascertainable from highly-structured data.

Use of Noisy Labels as Weak Learners to Identify Incompletely Ascertainable Outcomes: A Feasibility Study with Opioid-Induced Respiratory Depression.

Objective: Assigning outcome labels to large observational data sets in a timely and accurate manner, particularly when outcomes are rare or not directly ascertainable, remains a significant challenge within biomedical informatics. We examined whether noisy labels generated from subject matter experts' heuristics using heterogenous data types within a data programming paradigm could provide outcomes labels to a large, observational data set. We chose the clinical condition of opioid-induced respiratory depression for our use case because it is rare, has no administrative codes to easily identify the condition, and typically requires at least some unstructured text to ascertain its presence. Materials and Methods: Using de-identified electronic health records of 52,861 post-operative encounters, we applied a data programming paradigm (implemented in the Snorkel software) for the development of a machine learning classifier for opioid-induced respiratory depression. Our approach included subject matter experts creating 14 labeling functions that served as noisy labels for developing a probabilistic Generative model. We used probabilistic labels from the Generative model as outcome labels for training a Discriminative model on the source data. We evaluated performance of the Discriminative model with a hold-out test set of 599 independently-reviewed patient records. Results: The final Discriminative classification model achieved an accuracy of 0.977, an F1 score of 0.417, a sensitivity of 1.0, and an AUC of 0.988 in the hold-out test set with a prevalence of 0.83% (5/599). Discussion: All of the confirmed Cases were identified by the classifier. For rare outcomes, this finding is encouraging because it reduces the number of manual reviews needed by excluding visits/patients with low probabilities. Conclusion: Application of a data programming paradigm with expert-informed labeling functions might have utility for phenotyping clinical phenomena that are not easily ascertainable from highly-structured data.

Strengths, weaknesses, opportunities, and threats of using AI-enabled technology in sleep medicine: a commentary.

Over the past few years, artificial intelligence (AI) has emerged as a powerful tool used to efficiently automate several tasks across multiple domains. Sleep medicine is perfectly positioned to leverage this tool due to the wealth of physiological signals obtained through sleep studies or sleep tracking devices and abundance of accessible clinical data through electronic medical records. However, caution must be applied when utilizing AI, due to intrinsic challenges associated with novel technology. The Artificial Intelligence in Sleep Medicine Committee of the American Academy of Sleep Medicine reviews advancements in AI within the sleep medicine field. In this article, the Artificial Intelligence in Sleep Medicine committee members provide a commentary on the scope of AI technology in sleep medicine. The commentary identifies 3 pivotal areas in sleep medicine that can benefit from AI technologies: clinical care, lifestyle management, and population health management. This article provides a detailed analysis of the strengths, weaknesses, opportunities, and threats associated with using AI-enabled technologies in each pivotal area. Finally, the article broadly reviews barriers and challenges associated with using AI-enabled technologies and offers possible solutions. CITATION: Bandyopadhyay A, Oks M, Sun H, et al. Strengths, weaknesses, opportunities, and threats of using AI-enabled technology in sleep medicine: a commentary. J Clin Sleep Med. 2024;20(7):1183-1191.

Prediction of Effectiveness and Toxicities of Immune Checkpoint Inhibitors Using Real-World Patient Data.

PURPOSE: Although immune checkpoint inhibitors (ICIs) have improved outcomes in certain patients with cancer, they can also cause life-threatening immunotoxicities. Predicting immunotoxicity risks alongside response could provide a personalized risk-benefit profile, inform therapeutic decision making, and improve clinical trial cohort selection. We aimed to build a machine learning (ML) framework using routine electronic health record (EHR) data to predict hepatitis, colitis, pneumonitis, and 1-year overall survival. METHODS: Real-world EHR data of more than 2,200 patients treated with ICI through December 31, 2018, were used to develop predictive models. Using a prediction time point of ICI initiation, a 1-year prediction time window was applied to create binary labels for the four outcomes for each patient. Feature engineering involved aggregating laboratory measurements over appropriate time windows (60-365 days). Patients were randomly partitioned into training (80%) and test (20%) sets. Random forest classifiers were developed using a rigorous model development framework. RESULTS: The patient cohort had a median age of 63 years and was 61.8% male. Patients predominantly had melanoma (37.8%), lung cancer (27.3%), or genitourinary cancer (16.4%). They were treated with PD-1 (60.4%), PD-L1 (9.0%), and CTLA-4 (19.7%) ICIs. Our models demonstrate reasonably strong performance, with AUCs of 0.739, 0.729, 0.755, and 0.752 for the pneumonitis, hepatitis, colitis, and 1-year overall survival models, respectively. Each model relies on an outcome-specific feature set, though some features are shared among models. CONCLUSION: To our knowledge, this is the first ML solution that assesses individual ICI risk-benefit profiles based predominantly on routine structured EHR data. As such, use of our ML solution will not require additional data collection or documentation in the clinic.

Objective Pharyngeal Phenotyping in Obstructive Sleep Apnea With High-Resolution Manometry.

OBJECTIVE: Drug-induced sleep endoscopy (DISE) is a commonly used diagnostic tool for surgical procedural selection in obstructive sleep apnea (OSA), but it is expensive, subjective, and requires sedation. Here we present an initial investigation of high-resolution pharyngeal manometry (HRM) for upper airway phenotyping in OSA, developing a software system that reliably predicts pharyngeal sites of collapse based solely on manometric recordings. STUDY DESIGN: Prospective cross-sectional study. SETTING: An academic sleep medicine and surgery practice. METHODS: Forty participants underwent simultaneous HRM and DISE. A machine learning algorithm was constructed to estimate pharyngeal level-specific severity of collapse, as determined by an expert DISE reviewer. The primary outcome metrics for each level were model accuracy and F1-score, which balances model precision against recall. RESULTS: During model training, the average F1-score across all categories was 0.86, with an average weighted accuracy of 0.91. Using a holdout test set of 9 participants, a K-nearest neighbor model trained on 31 participants attained an average F1-score of 0.96 and an average accuracy of 0.97. The F1-score for prediction of complete concentric palatal collapse was 0.86. CONCLUSION: Our findings suggest that HRM may enable objective and dynamic mapping of the pharynx, opening new pathways toward reliable and reproducible assessment of this complex anatomy in sleep.

Accelerated curation of checkpoint inhibitor-induced colitis cases from electronic health records.

Objective: Automatically identifying patients at risk of immune checkpoint inhibitor (ICI)-induced colitis allows physicians to improve patientcare. However, predictive models require training data curated from electronic health records (EHR). Our objective is to automatically identify notes documenting ICI-colitis cases to accelerate data curation. Materials and Methods: We present a data pipeline to automatically identify ICI-colitis from EHR notes, accelerating chart review. The pipeline relies on BERT, a state-of-the-art natural language processing (NLP) model. The first stage of the pipeline segments long notes using keywords identified through a logistic classifier and applies BERT to identify ICI-colitis notes. The next stage uses a second BERT model tuned to identify false positive notes and remove notes that were likely positive for mentioning colitis as a side-effect. The final stage further accelerates curation by highlighting the colitis-relevant portions of notes. Specifically, we use BERT's attention scores to find high-density regions describing colitis. Results: The overall pipeline identified colitis notes with 84% precision and reduced the curator note review load by 75%. The segment BERT classifier had a high recall of 0.98, which is crucial to identify the low incidence (<10%) of colitis. Discussion: Curation from EHR notes is a burdensome task, especially when the curation topic is complicated. Methods described in this work are not only useful for ICI colitis but can also be adapted for other domains. Conclusion: Our extraction pipeline reduces manual note review load and makes EHR data more accessible for research.