AI is here to stay.
Photo/Thumb have been modified. Courtesy of Sergey - stock.adobe.com.
In the transition to value-based care, healthcare organizations are under tremendous pressure to reduce costs while improving patient outcomes, care efficiency and overall population health. Data analytics can unlock insights that can help us achieve those goals — but to date, reality hasn’t lived up to expectations.
The challenge is this: There’s plenty of healthcare data but plans and providers aren’t able to actually use them in a meaningful way. Much of health data lie scattered across clinical notes, electronic health records (EHRs), images and lab results, claims and other documents, making it extremely difficult to derive meaningful insights for patient care. Conventionally, it requires analysts to pore over documents looking for meaningful information and correlations. There simply aren’t enough human resources to make this approach affordable or scalable.
The inability to tap into these data-driven insights is holding the healthcare industry back from massive advances in care delivery, administrative efficiencies and new treatments and medications. The only viable solution to healthcare’s perpetual data problem is to use artificial intelligence (AI) to turn this treasure-trove of information into actionable intelligence.
Healthcare’s data problem has two main facets. The first is data access. Most health plan and provider data exist within siloed systems, stored in different formats and platforms. The lack of integration makes it impossible to share data or analyze trends across larger populations, which prevents healthcare organizations from learning from their own data or outside data sets. EHR vendors are also not incentivized to allow data sharing from their proprietary platforms.
The second facet is that most high-value patient and healthcare information is stored as unstructured data such as encounter notes, which can’t be mined using traditional computational techniques that require labeled data. Because human analysis of these documents is far too expensive and time-consuming, most providers and payers have no way to efficiently extract intelligence from these documents when making decisions about patient care. In addition, patient information is largely shared via faxes and printers, contributing to a huge problem to transform data from PDF to a structured, machine-readable format.
The net result: Healthcare organizations are unable to — or prohibited from — using the most critical, informative data to improve operations or care.
AI turns both structured and unstructured data into actionable information. By leveraging AI techniques, healthcare organizations can quickly analyze and surface accurate insights to inform operational and clinical decision-making. As an example, an AI model could look at the encounter notes in a patient’s medical record to identify conditions that need to be assessed and documented at the point of care.
Unlike conventional analytic approaches, AI systems can process millions of documents faster and more accurately than humans. For instance, a health system could deploy AI to flag anecdotal evidence of at-risk behaviors such as smoking in encounter notes that might trigger preventive screenings for COPD. This is the kind of targeted intelligence that will drive improvements in care across the board.
Organizations leveraging AI at scale are improving the quality of care and patient outcomes.
With access to fast, accurate analyses, providers can tap into population data to inform clinical decisions at the point of care — even during a patient encounter. For example, if a patient presents with specific symptoms of illness or a chronic disease, the physician can review trends in disease progression or comorbid conditions to predict the best course of treatment.
This can substantially improve the speed and accuracy of diagnosis by pointing physicians in the right direction earlier, reducing unnecessary diagnostics and placing the patient on the path to better health more quickly.
If a patient with diabetes needs a knee replacement, the primary care physician (PCP) can use an AI model to determine trends in treatment options and outcomes for orthopedic specialists and rehabilitation facilities in the area. The PCP can then refer the patient to the providers who have the most experience with certain procedures or demonstrate the best outcomes for patients with diabetes.
AI can also point to improvements in treatment modalities based on historical data. Providers can see over time what’s worked and what hasn’t for a large patient population to adjust treatment protocols and optimize results.
On the financial side, AI can reduce wasteful spending by identifying which diagnostics are most effective and economical for specific diagnoses. This means patients and payers can reduce the 30% waste spent every year on ineffective tests and treatments.
AI helps organizations contain costs on the operational side by reducing the time it takes to complete administrative tasks such as coding, documentation and reporting; automating manual processes such as prior authorizations and prospective programs; and reducing retrospective activities such as chart reviews by getting data to the right people earlier to inform decision-making.
Make no mistake about it: AI is here to stay.
In order to drive real, sustainable change in our healthcare system, plans and providers must create improvements at every level. AI can deliver the insights needed to make positive changes across the board. As the industry adopts more personalized medicine and value-based care, AI is the only solution that can deliver the analytical speed, scalability and accuracy needed to extract valuable intelligence from healthcare data. Leveraging AI is the only realistic way to quickly bring about the cost reductions, quality improvements and efficiency required to improve healthcare for all Americans.
Get the best insights in digital health directly to your inbox.
The Role of AI with Tumor Boards
AI Accurately Predicts Risk of A-Fib, Mortality from ECGs
AI Performs Similarly to Humans in Identifying Cancerous Lesions