New Data Suggests Machine Learning Algorithms

New data suggests machine learning algorithms can accurately predict C difference. Infection in hospitalized patients

Findings Published in the American Journal of Infection Control May Help Reduce Clinical and Economic Impact of Common and Serious HAIs

Arlington, Virginia, January 20, 2022 – New data published today suggest that several commonly used machine learning algorithms (MLA) can effectively predict which hospital patients will be infected with Clostridiodes difficult (C difference.). The conclusions, which appear in the American Journal of Infection Control (AJIC), the journal of the Association of Professionals in Infection Control and Epidemiology (APIC), may support infection prevention and early diagnosis, as well as faster implementation of infection control measures to minimize C difference. spread.

“The results of our study suggest that MLAs could play an important role in reducing the clinical and economic impact of nosocomial infections such as C difference. by providing early predictions about at-risk patients before they develop serious complications,” said Jana Hoffman, Vice President of Science, Dascena, Inc. “These data are consistent with a growing body of evidence that validates the intelligence and MLAs as essential components of healthcare management that can improve patient outcomes and help time-pressed clinicians provide the best patient care.

C difference. (CDI) is the leading cause of nosocomial diarrhea and is associated with significant morbidity, mortality and healthcare costs. There is currently no gold standard tool to assess the risk of contracting CDI in each patient. Hoffman and his colleagues have already published Data which demonstrate that MLAs can predict patients at risk of developing other high-impact HAIs.

For the study published today, researchers used a database comprising electronic health record (EHR) patient data from more than 700 hospitals nationwide to train and then systematically evaluate three different methods of conventional machine learning and deep learning. They first evaluated various models of each of these methods to determine whether they could effectively predict CDI in hospitalized patients using early hospitalized patient data, and then used a separate external dataset to assess the generalizability of the best performing MLA models.

The results suggest that ALMs can predict CDI with excellent discrimination using only the first six hours of hospitalized patient data. Of the three methods studied, a machine learning method called XGBoost provided the highest overall accuracy in predicting CDI, despite being the least complex model. XGBoost has also demonstrated its generalizability by maintaining its predictive performance in an external data set. The other two methods evaluated by the researchers, neural networks known as Deep Long Short Term Memory (D-LSTM) and one-dimensional convolutional neural network (1D-CNN), also demonstrated high levels of predictive accuracy, although they are less generalizable.

The best performing XGBoost, D-LSTM and 1D-CNN models used similar characteristics to predict CDI in patients, all of which have been previously identified as risk factors. In this study, age was the primary risk factor for CDI, followed by clinical measures such as sodium, body mass index, white blood cell count, and heart rate; active treatment with antibiotics or proton pump inhibitors; glycated hemoglobin; and race.

“This study supports previous research suggesting that MLAs provide a reliable prediction of infection risk that may allow clinical teams to implement appropriate infection control measures at earlier times and thus improve patient outcomes. healthcare,” said Linda Dickey, RN, MPH, CIC, FAPIC and APIC 2022 President.

About APIC

Founded in 1972, the Association of Professionals in Infection Control and Epidemiology (APIC) is the leading association of infection preventionists and epidemiologists. With more than 15,000 members, APIC advances the science and practice of infection prevention and control. APIC fulfills its mission through research, advocacy and patient safety; education, accreditation and certification; and fostering the development of tomorrow’s infection prevention and control workforce. Together with our members and partners, we work for a safer world through infection prevention. Join us and learn more about apic.org.

About AJIC

As the official peer-reviewed journal of APIC, The American Journal of Infection Control (AJIC) is the leading resource on infection control, epidemiology, infectious disease, quality management, occupational health and disease prevention. published by ElsevierAJIC also publishes APIC and CDC infection control guidelines. AJIC is included in Index Medicus and CINAHL. Visit AJIC at ajicjournal.org.

NOTES FOR EDITORS

“A Comparative Analysis of Machine Learning Approaches to Predict C. Difficile Infection in Hospitalized Patients”, by Saarang Panchavati, Nicole S. Zelin, MD, Anurag Garikipati, MS, Emily Pellegrini, MEng, Zohora Iqbal, PhD , Gina Barnes, MPH, Jana Hoffman, PhD, Jacob Calvert, MSc, Qingqing Mao, PhD, and Ritankar Das, MSc, were published online in AJIC January 20, 2022. The article can be viewed online at: https://doi.org/10.1016/j.ajic.2021.11.012

AUTHORS

Zohora Iqbal, PhD (Corresponding author: [email protected])

Dascena, Inc.

Houston, Texas, United States

Saarang Panchavati

Dascena, Inc.

Houston, Texas, United States

Nicole S. Zelin, MD

Dascena, Inc.

Houston, Texas, United States

Anurag Garikipati, MS

Dascena, Inc.

Houston, Texas, United States

Emily Pellegrini, MEng

Dascena, Inc.

Houston, Texas, United States

Gina Barnes, MPH

Dascena, Inc.

Houston, Texas, United States

Jana Hoffman, PhD

Dascena, Inc.

Houston, Texas, United States

Jacob Calvert, MSc

Dascena, Inc.

Houston, Texas, United States

Qingqing Mao, PhD

Dascena, Inc.

Houston, Texas, United States

Ritankar Das, MSc

Dascena, Inc.

Houston, Texas, United States

# # #


Warning: AAAS and EurekAlert! are not responsible for the accuracy of press releases posted on EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.

Comments are closed.