We constructed an automated predictive analytics framework for machine-learning algorithm with high discriminatory ability for assessing the risk of surgical complications and death using readily available preoperative electronic health records data. The feasibility of this novel algorithm implemented in real time clinical workflow requires further testing.

Data in computer-based patient records (CPRs) have many uses beyond their primary role in patient care, including research and health-system management. Although the accuracy of CPR data directly affects these applications, there has been only sporadic interest in, and no previous review of, data accuracy in CPRs. This paper reviews the published studies of data accuracy in CPRs. These studies report highly variable levels of accuracy. This variability stems from differences in study design, in types of data studied, and in the CPRs themselves. These differences confound interpretation of this literature. We conclude that our knowledge of data accuracy in CPRs is not commensurate with its importance and further studies are needed. We propose methodological guidelines for studying accuracy that address shortcomings of the current literature. As CPR data are used increasingly for research, methods used in research databases to continuously monitor and improve accuracy should be applied to CPRs.

The threat of bioterrorism has stimulated interest in enhancing public health surveillance to detect disease outbreaks more rapidly than is currently possible. To advance research on improving the timeliness of outbreak detection, the Defense Advanced Research Project Agency sponsored the Bio-event Advanced Leading Indicator Recognition Technology (BioALIRT) project beginning in 2001. The purpose of this paper is to provide a synthesis of research on outbreak detection algorithms conducted by academic and industrial partners in the BioALIRT project. We first suggest a practical classification for outbreak detection algorithms that considers the types of information encountered in surveillance analysis. We then present a synthesis of our research according to this classification. The research conducted for this project has examined how to use spatial and other covariate information from disparate sources to improve the timeliness of outbreak detection. Our results suggest that use of spatial and other covariate information can improve outbreak detection performance. We also identified, however, methodological challenges that limited our ability to determine the benefit of using outbreak detection algorithms that operate on large volumes of data. Future research must address challenges such as forecasting expected values in high-dimensional data and generating spatial and multivariate test data sets.

There is an increasing interest in developing artificial intelligence (AI) systems to process and interpret electronic health records (EHRs). Natural language processing (NLP) powered by pretrained language models is the key technology for medical AI systems utilizing clinical narratives. However, there are few clinical language models, the largest of which trained in the clinical domain is comparatively small at 110 million parameters (compared with billions of parameters in the general domain). It is not clear how large clinical language models with billions of parameters can help medical AI systems utilize unstructured EHRs. In this study, we develop from scratch a large clinical language model—GatorTron—using >90 billion words of text (including >82 billion words of de-identified clinical text) and systematically evaluate it on five clinical NLP tasks including clinical concept extraction, medical relation extraction, semantic textual similarity, natural language inference (NLI), and medical question answering (MQA). We examine how (1) scaling up the number of parameters and (2) scaling up the size of the training data could benefit these NLP tasks. GatorTron models scale up the clinical language model from 110 million to 8.9 billion parameters and improve five clinical NLP tasks (e.g., 9.6% and 9.5% improvement in accuracy for NLI and MQA), which can be applied to medical AI systems to improve healthcare delivery. The GatorTron models are publicly available at: https://catalog.ngc.nvidia.com/orgs/nvidia/teams/clara/models/gatortron_og.