Researchers create AI algorithm to improve ti
HAMILTON, ON (November 25, 2021) – Each year, sepsis affects more than 30 million people worldwide, causing approximately six million deaths. Sepsis is the body’s extreme response to infection and is often life threatening.
Since each hour of delayed treatment can increase the risk of death by four to eight percent, accurate and timely predictions of sepsis are crucial in reducing morbidity and mortality. To this end, various healthcare organizations have deployed predictive analytics to help identify patients with sepsis using data from electronic medical records (EMRs).
An international research team, including data scientists, physicians and engineers from McMaster University and St. Joseph’s Healthcare Hamilton, has created an artificial intelligence (AI) predictive algorithm that dramatically improves speed and accuracy data-based sepsis predictions.
âSepsis can be predicted very accurately and very early using AI with clinical data, but the key questions for the clinician and data scientists are how much historical data these algorithms need to make accurate predictions and how far they can accurately predict sepsis, âsaid Manaf Zargoush, study co-author and assistant professor of health policy and management at McMaster’s DeGroote School of Business.
To predict sepsis in clinical care settings, some systems use EMR data with disease scoring tools to determine sepsis risk scores – essentially acting as digital, automated assessment tools. More advanced systems use predictive analytics, such as AI algorithms, to go beyond risk assessment and identify sepsis itself.
Using AI predictive analytics, the researchers created an algorithm called the Long-term bi-directional memory (BiLSTM). It examines several variables in four key areas: administrative variables (e.g. length of stay in intensive care unit (ICU), hours between hospital and ICU admission, etc.), vital signs (eg, heart rate and pulse oximetry, etc.), demographic data (eg, age and sex) and laboratory tests (eg, blood sugar, creatinine, platelet count , etc.). Compared to other algorithms, BiLSTM is a more complex subset of machine learning – called deep learning – that uses neural networks to increase its predictive power.
The study compared the BiLSTM to six other machine learning algorithms and found that it was superior to the others in terms of accuracy. Improving accuracy by reducing false positives is the key to a successful algorithm, as these errors not only waste medical resources, but they also erode doctors’ confidence in the algorithm.
Interestingly, the study found that predictive accuracy can be increased through algorithms that focus more on a patient’s recent data points, instead of looking back to include as many data points as possible.
The researchers noted that it is understandable that clinicians are inclined to populate the algorithm with as many data points as possible over a long period of time. However, their results suggest that when the goal of prediction is accurate and timely with respect to sepsis predictions, physicians with long prediction horizons should rely more on fewer but more recent patient clinical data.
âSt. Joe’s will launch a cognitive computing pilot project in late November that includes understanding how AI can be used to help predict sepsis in real patients and in real time,â said Dan Perri, co -Study author, physician and chief information officer at St. Joseph’s Healthcare Hamilton. He is also an associate professor of medicine at McMaster.
âUnderstanding the breadth and scope of the data that can predict sepsis is important for any organization looking to use AI to save lives against serious infections,â added Perri.
âThe lessons learned from sepsis models translate into the creation of better machine learning tools that lead to appropriate early intervention for some of the sickest patients, while avoiding unnecessary warnings that could lead to fatigue for hospital workers. health. “
The study was published in the review Scientific reports on nature.
Photos of Manaf Zargoush and Dan Perri are available here: https://flic.kr/s/aHsmXeFted
Marketing and communications strategist
DeGroote School of Business
Computer simulation / modeling
The title of the article
The impact of recency and adequacy of historical information on sepsis predictions using machine learning
Publication date of the article
The authors declare no competing interests.
Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of any press releases posted on EurekAlert! by contributing institutions or for the use of any information via the EurekAlert system.