AI is rapidly transforming how healthcare systems treat patients and handle information and has already proven to be an invaluable tool. EHRs, or electronic health records, are one of the systems with the most to gain from artificial intelligence, with risk detection features cropping up in research to stop diseases early.
The AI risk detection model shows promise in oncology, as a recent Harvard research study successfully used AI to determine the risk of pancreatic cancer from a registry of patients. It combed data from the Danish National Patient Registry, including 8.6 million patient records, to find patients with a higher risk of the disease. Then, using four machine learning algorithms, researchers successfully predicted cancer occurrence in intervals of 3 months to 60 months after risk assessment with an accuracy of .88 (a measurement of accuracy that increases as the value approaches 1), according to News Medical.
These results indicate the potential of advanced computational technologies, such as AI and deep learning, to make increasingly accurate predictions based on each person's health and disease history.
Using a similar tactic as the Harvard pancreatic cancer study, researchers with the University of Utah leveraged AI to find patterns of outcomes among type II diabetes patients. With 83 percent accuracy, the algorithm predicted successful outcomes and even anticipated medication treatment options.
Often, patients nearing their end of life (EoL) receive unnecessary and dangerous treatment. To address this, researchers used AI to identify cancer patients with short-term mortality risk, increasing its effectiveness using social determinants of health information, including health habits, racial background, and geographic location. Their algorithm showed promising results, accurately predicting 30-day mortality among its database of patients, causing the researchers to conclude that the AI's most influential role is determining those at short-term clinical risk so providers can take corrective action.
AI risk detection proves to be accurate and reliable and reduces the guesswork of providers. However, the fraction of misdiagnoses means trouble in a potential lawsuit. Many EHR AIs do not leave an "audit trail" or documentation of how they came to a medical conclusion. Technology companies behind AI do not want to be held liable for a malfunction, so often, they refuse to elaborate on their product's inner workings. Defense attorney Matthew Keris elaborated on the risk of AI in healthcare.
Problems arise when healthcare systems buy a product created for them. If they were to get a new product, the old EMR vendor doesn't want anything to do with the healthcare provider. They don't want to get involved in the litigation, and if there's a real issue, they don't want to have to turn over their proprietary information.
While EHR AI benefits care in many ways, it still adds to providers' workload. According to Health IT Analytics, reviewing EHR increases the already staggering 62 percent of the time caregivers spend looking at data. AI risk detection, if not handled correctly, adds to the mountains of data that healthcare providers must sift through between patients and potentially to the risk of burnout. The WHO warned of another AI risk: lack of privacy and biased algorithms.
It remains clear that AI tools like clinical prediction and risk detection give providers an upper hand when treating patients, alerting them to threats they may not have seen otherwise. Altogether, healthcare workers, AI vendors, and researchers will have to collaborate to create tools that work for all parties involved. While current models show promise, AI will still need to mature to best serve healthcare.