Unstructured clinical nursing notes, including sentiments of clinicians, can help predict patient outcomes in the ICU, according to a study published in PLOS One.
“Nursing notes have not been widely used in prediction models for clinical outcomes, despite containing rich information,” wrote Joon Lee, PhD, of the University of Waterloo in Ontario, Canada, and colleagues. “Advances in natural language processing have made it possible to extract information from large scale unstructured data like nursing notes.”
Hospitals generally use severity of illness scores to predict 30-day survival of ICU patient. The scores include laboratory results, vital signs, and physiological and demographic characteristics gathered within 24 hours of admission.
“The physiological information collected in those first 24 hours of a patient's ICU stay is really good at predicting 30-day mortality," said co-author Joel Dubin, PhD, of the University of Waterloo in a statement. "But maybe we shouldn't just focus on the objective components of a patient's health status. It turns out that there is some added predictive value to including nursing notes as opposed to excluding them."
The researchers extracted the sentiment to examine how nurses’ impressions and attitudes correlated to 30-day mortality.
Using an ICU database, the researchers reviewed nursing notes from more than 27,000 ICU patients. The patients had an overall mortality rate of 11 percent. Adjectives from the nursing notes were analyzed using an algorithm, being designated as positive, neutral or negative.
Lee et al. found sentiment analysis improved predicting 30-day mortality in ICU patients. They found a well-defined difference between patients with the most positive messages who exhibited the highest survival rates and patients with the most negative messages who experienced the lowest survival rates.
Additionally, aside from mortality, nursing notes sentiment analysis can predict readmission or recovery from an infection while in the ICU.