Nursing Notes May Predict ICU Patient Survival
New research reports that sentiments included in nursing notes can serve as an indicator of whether ICU patients will survive.
Health care providers in hospitals typically use predictive models comprised of lab results, vital signs, and physiological and demographic information to predict the 30-day survival of intensive care unit (ICU) patients. However, new research suggests that the notes of ICU nurses should be included when predicting survival rates.
Researchers at the University of Waterloo in Ontario, Canada, have found that sentiments included in nursing notes can serve as a predictor of whether or not intensive care unit (ICU) patients will survive.
In previous studies, researchers had determined that sentiments of clinicians toward patients could be measured through a sentiment analysis, or “a method to quantify or categorize subjective properties of written text.” The sentiments were then polarized from very negative (-1) to very positive (1.)
In the new study, published in the journal PLoS ONE, the researchers developed a sentiment analysis algorithm to examine nursing notes from patients included in the Medical Information Mart for Intensive Care III database.
The researchers used a syntactic analysis approach referred to as the Pattern module which makes inferences based on the structure of the test. Polarity and subjectivity scores were computed for the adjectives used in the notes and researchers established whether each note was positive, neutral, or negative.
Using a multiple logistic regression model, the data collected from nursing notes of 27,477 patients showed a relationship between the measured sentiment and 30-day mortality—the overall mortality was 11.02%.
Additionally, the researchers reported that the nursing notes of patients who survived the 30-day post-admission period demonstrated high mean polarity scores compared with the notes of patients who had not survived (means: 0.0717 and 0.0407 respectively; t-test: p < 0.001). However, in contrast, the group that survived exhibited lower mean sentiment subjectivity scores than the group that did not survive (means: 0.3674 and 0.3700, respectively; t-test: p = 0.0172).
“Maybe we shouldn’t just focus on the objective components of a patient’s health status,” said Joel Dubin, MD, associate professor, department of Statistics and Actuarial Science and the School of Public Health and Health Systems, University of Waterloo, and an author of the study, in a statement. “It turns out that there is some added predictive value to including nursing notes as opposed to excluding them.”
The researchers indicate that further research is warranted to study and make use of the wealth of data that clinical notes have to offer. “Mortality is not the only outcome that nursing notes could potentially predict,” expressed Dr Dubin, “They might also be used to predict readmission, or recovery from infection while in the ICU.”
Next steps include translating the findings into assisting health care providers in making better decisions in the ICU. Possible future steps could include developing automated prediction models with the capability of identifying high-risk patients so that appropriate resources can be used to prevent adverse outcomes.
In regard to the studying of sentiment analysis in patient’s notes, the researchers indicate an area of interest could be if nurses assigned their notes a score of “negative,” “positive,” or “neutral,” to produce a “labeled corpus of nursing notes.”