AI Pressure Injury Prediction

The Braden Scale, developed by Barbara Braden and Nancy Bergstrom in the 1980s, scores six subscales: sensory perception, moisture, activity, mobility, nutrition, and friction and shear. Totals run from 6 to 23, and lower scores mean higher risk. A score of 18 or below is commonly treated as at risk. The scale is quick and familiar, but it depends on subjective ratings and predicts only moderately well. It also leaves out factors that matter a lot in acute care, such as vasopressors, low perfusion, long surgeries and medical devices. In 2016 the National Pressure Ulcer Advisory Panel, now the NPIAP, changed its term from pressure ulcer to pressure injury, because early stages may involve intact skin. Staging covers Stages 1 through 4, unstageable, and deep tissue pressure injury, and medical device-related injuries are also recognized. Machine learning models trained on EHR data can use vasopressor doses, mechanical ventilation, time in surgery, BMI, incontinence, lab values, mobility documentation and repositioning records. Sensors add direct measurements. Sub-epidermal moisture (SEM) scanners, such as the Provizio SEM Scanner, detect fluid changes under the skin that can come before visible damage. This is especially useful because early redness is harder to see on darker skin. Wearable sensors like the Leaf system track patient position and turning, and pressure mapping mats show where pressure builds up. Wound photo apps, such as Swift Medical's, measure wound size and classify tissue from smartphone images. One misconception is that a photo app stages the wound. Staging still requires clinical assessment, including depth, palpation and tissue that cannot be seen. Another is that prediction prevents injuries. Prevention comes from repositioning, support surfaces, moisture management, nutrition and device checks. CMS treats hospital-acquired Stage 3 and 4 pressure injuries as hospital-acquired conditions, which keeps them a priority for quality programs.