on January 12th, 2016 No Comments
Nigam Shah, MBBS, PhD, a Stanford associate professor of medicine, took another one of his deep dives into patient medical records to find out. The result is a creative proof-of-concept model that can predict which wounds need special care.
Earlier work has shown that even very simple models of wound healing can help caregivers pay attention to the wounds most likely to take 15 weeks or longer to heal, the definition of delayed healing.
For this work, Shah, an expert in biomedical informatics; first author Kenneth Jung, a research scientist at Stanford; and a national team of researchers turned to a dataset consisting of more than 150,000 wounds from more than 53,000 people. The team looked at hundreds of variables from patient records, including, for example the length, breadth, area, and depth of wounds, and how old patients were. The wounds in the study ranged from bed sores or diabetic ulcers to surgical or trauma wounds.
The researchers randomly assigned patients to one of two groups. One group constituted the raw data on which computers could learn which factors predicted slow wound healing in order to create a predictive model. The other group was the test that showed that the model worked on a new, and previously unseen, separate set of data.
They found that the best 100 predictors accounted for 95 percent of the influence on whether wounds were slow to heal. The single most important predictor of poor wound healing was whether a patient was receiving palliative care. Other good predictors of poor wound healing were the patient’s age, the size of the wound and how quickly it began healing in the first week.
The model’s strengths are that it works regardless of the kind of wound and it can be customized for different situations. However, as noted in the paper, the model was developed within the confines of a single company — a chain of specialty wound-care clinics called Healogics — so the model may not necessarily apply to wound healing at other institutions or for patients at home.
The paper, which made the front cover of the journal Wound Repair and Regeneration yesterday, is accessible for an academic paper — so if you’re interested in learning more about using patient records to create predictive health-care models, check it out.
Previously: Stanford researchers investigate source of scarring and To boldly go into a scar-free future: Stanford researchers tackle wound healing
Art — The Incredulity of Saint Thomas by Caravaggio — from Wikimedia