Surgical site infections are incredibly costly to hospitals. So when the University of Iowa managed to reduce patient infection rates by 74%, people paid attention. By leveraging EHR data in combination with machine learning, surgeons were able to better assess patient risk and implement the appropriate treatment pathway. This had a big payoff-- ultimately saving the hospital an estimated $1.2 Million.
How they did it
In 2012, the University of Iowa Hospital and their analytics consultant constructed a data warehouse to leverage both patient data from the EHR as well as real-time procedural and risk assessment data. Building the data warehouse was the necessary first step to achieving a structured backend for data modeling
As the project grew, the Iowa team then introduced a prescriptive analytics model, the end result of which was a risk score to assess the likelihood of surgical site infection at the point-of-care.
Here's how it worked:
As the surgeon is preparing to suture the patient, the surgical site infection reduction module has already aggregated patient data from the EHR, such as case duration or estimated blood loss flows. That data has been combined with surgical data at critical junctures throughout the surgery, as well as historical data on the patient.
The machine learning model then calculates the infection risk score and recommends specific interventions that the surgeon may take at the time of closure to reduce the risk of SSI. Using the risk score, the surgeon can course correct and ensure that appropriate care and preventative steps are taken. Similarly, patients with low risk scores won’t receive unnecessary intervention.
Setting a new standard
Using the predictive model platform, the attending staff are tracking (with one single click) whether or not the surgeon followed the recommendations correlated with the risk score. The patient outcome is also reported back to the platform, and an aggregate report is generated for the surgeons review to display how outcomes were correlated with the risk scores. This feedback loop produces an environment of continual learning, accountability, and improvement, where the use of appropriate decision-making is reinforced by comprehensive reporting.
Reducing surgical variation means adherence to best practices. With the sheer volume of data being collected now, it is simply unrealistic for surgeons to assess, process, and act on all the information available to them; augmenting their skill sets with data analytics models means that doctors can provide patients with a higher level of customized care, tailored to their specific needs or risk profiles.
AI- Enabling the Future
This is just one example of how AI and machine learning are changing the hospital. Predictive medicine that draws on AI is being used to treat a whole host of diseases ranging from pancreatic cancer to depression. And with value-based care now using outcomes to determine payment, there has never been a greater push to improve diagnostic tools and patient care. The University of Iowa’s million-dollar idea is an early indicator of a much larger trend -- healthcare professionals estimate that by 2030, the industry will be almost entirely digitized.