Big data and machine learning helped create a model that efficiently predicts critical care patients' risk for pressure injuries, according to a study published in the American Journal of Critical Care.
The researchers analyzed five years of data on patients admitted to the adult surgical or surgical cardiovascular intensive care units at Salt Lake City-based University of Utah Hospital.
The study included 6,376 patients and looked at stages of hospital-acquired pressure injuries, which are classified as stage one, two, three or four based on the deepest tissue type exposed. Among these patients, hospital-acquired pressure injuries of stage one or greater developed in 516 patients, and injuries of stage two or greater developed in 257 patients.
After identifying these two outcome variables, the researchers used machine learning to analyze the large amount of clinical data available in patient records and assess the relationships among the variables that predict a patient's pressure injury risk.
The machine learning technique, called random forest, is relatively unaffected by slight correlations among variables, meaning it can create a more accurate and stable prediction.
The approach is different from other models because it does not require clinicians to input information into a tool, the researchers said. Rather, the technique uses information readily available in EHRs, making it easier for clinicians to get results.
The most important variables in predicting pressure injuries included time required for surgery, body mass index, age and hemoglobin level, according to the model.
"Current risk-assessment tools classify most critical care patients as high risk for developing pressure injuries and therefore do not provide a way to differentiate among critical care patients in terms of pressure injury risk," said principal investigator Jenny Alderden, PhD. "Eventually, our model may offer additional insight to clinicians as they develop a plan of care for patients at highest risk and identify those who would benefit most from interventions that are not financially feasible for every patient."