Every day, companies like Google, Amazon, Facebook, and Netflix are using data to better care for their customers, to personalize ads and recommendations, and ultimately, to make money. They pull any data they can get their hands on and push them through modern machine-learning algorithms to create accurate predictions about future customer behavior.
In medicine, however, we often rely on sleep-deprived providers to cull through an enormous amount of data by hand in order to make decisions, which can lead to errors that can cost patients their lives. The prediction algorithms we do use were often created using modeling techniques developed over 100 years ago and value simplicity over accuracy. In 2016, where EHRs are ubiquitous and false alarms are everywhere, I think we owe it to our patients to develop the most accurate models possible and to use the data we collect on our patients every day. I’m excited for our field to be inspired by what the private sector is doing with data and to develop ways to use real-time information to improve patient care.
On October 23, from 7:30 to 8:30 am, I will be the presenter at the Critical Care NetWork Featured Lecture and will give a talk entitled: “Using Big Data to Detect Clinical Deterioration.” I will discuss how we can use the data collected on patients in the hospital, such as vital signs and laboratory results, to identify patients who will become critically ill. Some of the talk will focus on patients on the wards who later deteriorate and then are transferred to the ICU or suffer a cardiac arrest. I will also discuss how big data can be used to identify high-risk patients with sepsis outside the ICU.
This lecture is open to everyone. Data and technology are ubiquitous in today’s world. Using the data that clinicians collect every day on their patients to develop decision support systems will turn our electronic health records into something that actually works for our patients and can make our lives easier. In addition, because I will focus on early critical illness and sepsis, my talk will have broad applicability to clinicians caring for sick patients in the hospital.
If you want to read more on this topic, our recent paper comparing different prediction modeling methods would be of interest. In this paper, we found that modern machine- learning techniques were more accurate than logistic regression. Several well-written articles on the topics of big data in health care, how big data can be integrated with clinical trials, aspects of big data relevant to critical care, and how data can be used from the EHR to identify patients with delays in follow-up from abnormal chest imaging studies are other excellent resources in this topic area.
I hope to see you in October!
Dr. Churpek is an Assistant Professor of Medicine in the Section of Pulmonary and Critical Care at the University of Chicago. His research involves using electronic health record data to develop machine-learning models to detect clinical deterioration in hospitalized patients. Dr. Churpek’s work is supported by an NIH K08 award and an American Thoracic Society Foundation Recognition Award for Outstanding Early Career Investigators.