Skip to navigation Skip to main content Skip to footer

Approved research

Development of Perioperative Risk Quantification Algorithm and Intraoperative Clinical Support System using Integrated Medical History and Patient Genotype

Principal Investigator: Dr Nicholas Douville
Approved Research ID: 44870
Approval date: November 8th 2018

Lay summary

Surgery imposes unique stresses, risks, and costs on patients and the health care system. Furthermore, vast amounts of medical data including laboratory studies, diagnostic tests, and medication administrations are recorded around the time of surgery. The perioperative period, represents a unique opportunity to use big data analytical techniques, such as machine learning, in the delivery of personalized medicine. Perioperative physicians, including anesthesiologists and intensivists, must adapt their practice to the growing percentage of patients, who have genetic information available at the time they initially present for surgery. Our goal is to assist physicians in improving patient outcomes through a unified platform that identifies patient attributes that may affect their care and stratifies the risk of key perioperative complications. The steps for translation from the basic science of genetic discovery to an integrated clinic support system are: (Aim 1) identification of comorbidities that might influence intraoperative management and (Aim 2) development of a predictive risk algorithm for key post-operative complications such as myocardial infarction and pneumonia. Aim 1: Validate our ability to identify conditions impacting anesthetic care. We have developed a list of genetically-linked conditions ('comorbidities') that directly impact intraoperative management. These are impacted by patient demographics (age, ethnicity, and sex), past medical history, and genetic data, which will be combined into a generalizable probabilistic framework for each comorbidity. The relative weight assigned to each predictive node will be determined using machine learning on individual-level UK Biobank data (initially GWAS, with expansion into exome sequencing as data becomes available) and validated on an independent dataset. Aim 2: Expand the algorithm to predict adverse medical outcomes following surgery. The first perioperative outcome we will attempt to predict is adverse cardiac events. This was selected because of (i) relatively high frequency, (ii) mature field of prediction using non-genetic factors, (iii) high-fidelity, objective data, and (iv) high morbidity, mortality, and overall cost to the health care system. As in Aim 1, we will combine genetic data with patient demographics and past medical history. Translational: Our ultimate goal is to develop a web-based application capable of processing spectrum of genetic data that a patient or institution may have: (i) Genome Wide Association Studies (GWAS), (ii) Exome Sequencing, and (iii) Whole Genome Sequencing. This development will be independent of the data obtained from the UK Biobank, which will only be used in the development and validation of our predictive model.