We provide an accessible description of a peer-reviewed generalizable causal machine learning pipeline to (i) discover latent causal sources of large-scale electronic health records observations, and (ii) quantify the source causal effects on clinical outcomes. We illustrate how imperfect multimodal clinical data can be processed, decomposed into probabilistic independent latent sources, and used to train taskspecific causal models from which individual causal effects can be estimated. We summarize the findings of the two real-world applications of the approach to date as a demonstration of its versatility and utility for medical discovery at scale.

Graph Convolutional Networks (GCNs) have emerged as a promising approach to machine learning on Electronic Health Records (EHRs). By constructing a graph representation of patient data and performing convolutions on neighborhoods of nodes, GCNs can capture complex relationships and extract meaningful insights to support medical decision making. This survey provides an overview of the current research in applying GCNs to EHR data. We identify the key medical domains and prediction tasks where these models are being utilized, common benchmark datasets, and architectural patterns to provide a comprehensive survey of this field. While this is a nascent area of research, GCNs demonstrate strong potential to leverage the complex information hidden in EHRs. Challenges and opportunities for future work are also discussed.

Biomedical relation extraction (RE) is the task of automatically identifying and characterizing relations between biomedical concepts from free text. RE is a central task in biomedical natural language processing (NLP) research and plays a critical role in many downstream applications, such as literature-based discovery and knowledge graph construction. State-of-the-art methods were used primarily to train machine learning models on individual RE datasets, such as protein-protein interaction and chemical-induced disease relation. Manual dataset annotation, however, is highly expensive and time-consuming, as it requires domain knowledge. Existing RE datasets are usually domain-specific or small, which limits the development of generalized and high-performing RE models. In this work, we present a novel framework for systematically addressing the data heterogeneity of individual datasets and combining them into a large dataset. Based on the framework and dataset, we report on BioREx, a data-centric approach for extracting relations. Our evaluation shows that BioREx achieves significantly higher performance than the benchmark system trained on the individual dataset, setting a new SOTA from 74.4% to 79.6% in F-1 measure on the recently released BioRED corpus. We further demonstrate that the combined dataset can improve performance for five different RE tasks. In addition, we show that on average BioREx compares favorably to current best-performing methods such as transfer learning and multi-task learning. Finally, we demonstrate BioREx's robustness and generalizability in two independent RE tasks not previously seen in training data: drug-drug N-ary combination and document-level gene-disease RE. The integrated dataset and optimized method have been packaged as a stand-alone tool available at https://github.com/ncbi/BioREx.

Clinical semantic parsing (SP) is an important step toward identifying the exact information need (as a machine-understandable logical form) from a natural language query aimed at retrieving information from electronic health records (EHRs). Current approaches to clinical SP are largely based on traditional machine learning and require hand-building a lexicon. The recent advancements in neural SP show a promise for building a robust and flexible semantic parser without much human effort. Thus, in this paper, we aim to systematically assess the performance of two such neural SP models for EHR question answering (QA). We found that the performance of these advanced neural models on two clinical SP datasets is promising given their ease of application and generalizability. Our error analysis surfaces the common types of errors made by these models and has the potential to inform future research into improving the performance of neural SP models for EHR QA.

Within precision medicine, and specifically rare diseases, clinicians and researchers rely on genetic and diagnostic testing to help drive accurate diagnosis and treatment. However, genomic data alone are often insufficient to unlock the life-changing diagnoses of rare diseases. Well-curated and accurate phenotype data, which may include quantified observable traits such as short stature, low set ears, and blood biochemistry, along with genetic and diagnostic test results, are vital for shortening the diagnostic journey of these patients and identifying the most effective treatments available. The need for accurate patient phenotyping is not a new concept. In fact, over 20 years ago, Isaac Kohane, Chair of the Department of Biomedical Informatics and the Marion V. Nelson Professor of Biomedical Informatics at Harvard Medical School, predicted that the accurate practice of patient phenotyping would become essential as the volume of genomic information continued to surge.

Improving the speed and accuracy of clinical diagnosis, augmenting clinical decision-making, reducing human error in clinical care, individualizing therapies based on a patient's genomic and metabolomic profiles, differentiating benign from cancerous lesions with impeccable accuracy, identifying likely conditions a person may develop years down the road, spotting early tell-tale signs of an ultrarare disease, intercepting dangerous drug interactions before a patient is given a new medication, yielding real-time insights amidst a raging pandemic to inform optimal treatment of patients infected with a novel human pathogen. These are some of the promises that physicians and researchers look to fulfill using artificial intelligence -- promises poised to transform clinical care, lead to better patient outcomes, and, ultimately, improve human lives. Yet, AI is no silver bullet. It can fall prey to the cognitive fallibilities and blind spots of the humans who design it. AI models can be as imperfect as the data and clinical practices that the machine-learning algorithms are trained on, propagating the very same biases AI was designed to eliminate in the first place. Beyond conceptual and design pitfalls, realizing the potential of AI also requires overcoming systemic hurdles that stand in the way of integrating AI-based technologies into clinical practice.

A machine learning algorithm that predicts suicide attempt recently underwent a prospective trial at the institution where it was developed, Vanderbilt University Medical Center. Over the 11 consecutive months concluding in April 2020, predictions ran silently in the background as adult patients were seen at VUMC. The algorithm, dubbed the Vanderbilt Suicide Attempt and Ideation Likelihood (VSAIL) model, uses routine information from electronic health records (EHRs) to calculate 30-day risk of return visits for suicide attempt, and, by extension, suicidal ideation. Suicide has been on the rise in the U.S. for a generation and is estimated to claim the lives of 14 in 100,000 Americans each year, making it the nation's tenth leading cause of death. Nationally, some 8.5% of suicide attempts end in death.

Artificial Intelligence's deployment is once again helping healthcare by achieving a demonstrated utility within suicide prediction and clinical management of patients at high risk. It is impossible to know if those who committed suicide would have changed their mind. If something had predicted their suicide attempt, perhaps. Many, though, have survived the suicidal thoughts and the attempt. After receiving help, treatment, and being fully recovered, many have recognized that causing their death would have been a mistake.

A machine learning algorithm that predicts suicide attempt recently underwent a prospective trial at the institution where it was developed, Vanderbilt University Medical Center. Over the 11 consecutive months concluding in April 2020, predictions ran silently in the background as adult patients were seen at VUMC. The algorithm, dubbed the Vanderbilt Suicide Attempt and Ideation Likelihood (VSAIL) model, uses routine information from electronic health records (EHRs) to calculate 30-day risk of return visits for suicide attempt, and, by extension, suicidal ideation. Suicide has been on the rise in the U.S. for a generation and is estimated to claim the lives of 14 in 100,000 Americans each year, making it the nation's tenth leading cause of death. Nationally, some 8.5% of suicide attempts end in death.

A machine learning algorithm that predicts suicide attempt recently underwent a prospective trial at the institution where it was developed, Vanderbilt University Medical Center. Over the 11 consecutive months concluding in April 2020, predictions ran silently in the background as adult patients were seen at VUMC. The algorithm, dubbed the Vanderbilt Suicide Attempt and Ideation Likelihood (VSAIL) model, uses routine information from electronic health records (EHRs) to calculate the 30-day risk of return visits for a suicide attempt, and, by extension, suicidal ideation. Suicide has been on the rise in the U.S. for a generation and is estimated to claim the lives of 14 in 100,000 Americans each year, making it the nation's tenth leading cause of death. Nationally, some 8.5 per cent of suicide attempts end in death.