Major postoperative complications (PC) are common and lead to increase in mortality and healthcare cost. Some cost-effective strategies, implemented in a timely fashion, can ameliorate the risk for PC but the ability to use them depends on the timely and accurate identification of those patients at greatest risk. Assessment of that risk requires timely, accurate and dynamic synthesis of the large amount of clinical information obtained throughout the perioperative period. Today it is not possible to predict and quantify, for a given patient, a personal and real-time risk for PC that integrates preoperative risk with the risk incurred by the physiologic response to events during surgery. The interventions that could prevent PC are applied without consideration of a patient's personal risk profile or often not applied at all because the risk is underestimated. There is an abundance of physiologic, laboratory and other clinical data in the perioperative electronic health records (EHR), but their magnitude and complexity often overwhelms a physicians' ability to comprehend and use the information in an optimal and timely way. The objective is to develop an intelligent system, composed of high- performance computers, algorithms and physicians interacting in real time, which can generate usable medical knowledge with both increased speed and accuracy using complex clinical data. Our multidisciplinary team of scientific experts in medicine and engineering will address methodological challenges related to implementation of real-time data integration and processing, data analytics and knowledge exchange between computers and physicians in the clinical environment. There are three specific aims: 1. Refine and validate predictive risk models for major complications using EHR integrated with intraoperative physiologic time series using a temporal database for 10,000 surgical patients. 2. Implement and validate two-way knowledge exchange between predictive risk models and physicians. We will design an interactive knowledge exchange application that presents the knowledge behind predictive models to physicians, while allowing them to input their own assessment into the models. 3. Implement and evaluate an intelligent perioperative system for automated risk analysis using real-time EHR data. In a prospective clinical study of 60 physicians we will validate the diagnostic performance of predictive risk models, compare them with the physicians' risk assessment and measure change in physicians' risk perception after knowledge exchange with the system. This methodology will provide a significant step towards personalized perioperative medicine by modeling and quantifying the body's responses to surgery while using clinical data acquired during routine medical care.

Public Health Relevance

Each year complications after surgery affect more than 1.5 million of patients imposing an acute burden of death and suffering. In alignment with NIH's mission to advance disease diagnosis through medical applications of new tools and technologies this proposal applies computational methods to timely identify patients at the greatest risk of complications using readily available medical data in electronic health records. Preventive interventions to improve outcomes could then be tailored to each patient's personal risk profile.

National Institute of Health (NIH)
National Institute of General Medical Sciences (NIGMS)
Research Project (R01)
Project #
Application #
Study Section
Surgery, Anesthesiology and Trauma Study Section (SAT)
Program Officer
Somers, Scott D
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
University of Florida
Internal Medicine/Medicine
Schools of Medicine
United States
Zip Code
Shickel, Benjamin; Tighe, Patrick James; Bihorac, Azra et al. (2018) Deep EHR: A Survey of Recent Advances in Deep Learning Techniques for Electronic Health Record (EHR) Analysis. IEEE J Biomed Health Inform 22:1589-1604
Stortz, Julie A; Murphy, Tyler J; Raymond, Steven L et al. (2018) Evidence for Persistent Immune Suppression in Patients Who Develop Chronic Critical Illness After Sepsis. Shock 49:249-258
Hobson, Charles; Lysak, Nicholas; Huber, Matthew et al. (2018) Epidemiology, outcomes, and management of acute kidney injury in the vascular surgery patient. J Vasc Surg 68:916-928
Zhu, Qile; Li, Xiaolin; Conesa, Ana et al. (2018) GRAM-CNN: a deep learning approach with local context for named entity recognition in biomedical text. Bioinformatics 34:1547-1554
Sundaram, Laksshman; Gao, Hong; Padigepati, Samskruthi Reddy et al. (2018) Predicting the clinical impact of human mutation with deep neural networks. Nat Genet 50:1161-1170
Horiguchi, Hiroyuki; Loftus, Tyler J; Hawkins, Russell B et al. (2018) Innate Immunity in the Persistent Inflammation, Immunosuppression, and Catabolism Syndrome and Its Implications for Therapy. Front Immunol 9:595
Nickerson, Paul; Baharloo, Raheleh; Davoudi, Anis et al. (2018) Comparison of Gaussian Processes Methods to Linear methods for Imputation of Sparse Physiological Time Series. Conf Proc IEEE Eng Med Biol Soc 2018:4106-4109
Efron, Philip A; Mohr, Alicia M; Bihorac, Azra et al. (2018) Persistent inflammation, immunosuppression, and catabolism and the development of chronic critical illness after surgery. Surgery 164:178-184
Bihorac, Azra (2018) Response to: ""MySurgeryRisk: Development and Validation of a Machine-learning Risk Algorithm for Major Complications and Death After Surgery"". Ann Surg :
Nadim, Mitra K; Forni, Lui G; Bihorac, Azra et al. (2018) Cardiac and Vascular Surgery-Associated Acute Kidney Injury: The 20th International Consensus Conference of the ADQI (Acute Disease Quality Initiative) Group. J Am Heart Assoc 7:

Showing the most recent 10 out of 37 publications