Pinaki Dasgupta is the founder and CEO of Hindsait, Inc.
The United States spends more on healthcare than any other developed country in the world. In 2021, U.S. national health spending was 18.3% of GDP—and by 2030, it is projected to reach nearly 20%, or approximately $6 trillion.
Yet data suggests that 25% to 30% of healthcare spending is considered waste and could actually be avoided. Fraud, waste and abuse, as well as inefficiencies, including human errors and administrative bottlenecks, add up to a significant portion of healthcare costs.
Risk adjustment in healthcare is a method used to assess patients’ health needs, estimate their likelihood of using healthcare services and calculate the cost of their care. It helps ensure that healthcare providers are compensated fairly for the patients they treat.
For example, people age 65 and older are eligible for Medicare, a government-funded health insurance program operated by the Centers for Medicare & Medicaid Services (CMS). As of March 2023, over 65 million people were enrolled in Medicare. CMS uses risk adjustment to estimate the cost to treat a patient in a given year, based on risks such as diagnoses and comorbidities. Providers receive higher reimbursements for patients with more health problems than for patients deemed healthier, who are less likely to need services.
Risk Adjustment Challenges And Solutions
Risk adjustment is a complex process that must take into account numerous data sources and patient factors, which presents tremendous challenges when applied to millions of patients. CMS and health insurance companies traditionally employ various tools, including zip code analysis and algorithmic assessment of population risks, for risk adjustment. But this approach often still requires manual steps with human oversight; doctors and nurses may have to review thousands of pages of medical records to find the information they need—a process that is painstaking, time-consuming and prone to errors.
This is an area where artificial intelligence offers powerful solutions that can streamline the risk adjustment process. Rather than asking people to search for a needle in a haystack, AI-driven algorithms can quickly and accurately identify relevant information in medical records. By automating the review process, AI can eliminate or significantly reduce the burden on human reviewers and minimize the risk of errors.
AI can be trained to recognize specific data points in massive datasets. For example, instead of a person spending hours flipping pages looking for pulmonary function test results and determining whether they correspond to documented comorbidities or risk codes, AI could find the right PFT result in seconds with a single click. An AI tool could highlight evidence, or lack of evidence, immediately for humans to make better informed decisions.
In my experience developing AI solutions for healthcare, I have seen the benefits of AI-driven risk adjustment in action: It saves valuable time and resources, while enhancing accuracy and minimizing bias. If you are looking to implement AI solutions in your healthcare organization, I recommend considering the following factors:
• Healthcare data privacy and security: Medical data is uniquely sensitive and must be handled with privacy, security and compliance in mind. Publicly available large language models are not equipped for this function, and you will need to use clinical natural language processing (CNLP) in combination with a narrow large language model (LLM) that can be trained on private healthcare datasets.
• Training on diverse datasets: Building AI models in-house can be resource- and time-consuming, in part because models must be trained on multiple and diverse datasets. Healthcare data can vary across regions, and medical providers might use different terminology. For example, one doctor or health system could describe heart failure in a variety of ways: cardiac failure, congestive heart failure, CF, HF or CHF. You need to train your model on a wide range of datasets from across the country to ensure accuracy and consistency.
• Contextualization: Context matters in healthcare data. If you’re reviewing a patient’s chart for congestive heart failure, for instance, it’s important to understand if an event occurred in 2023, 2022 or 2021. The context and temporal time frame are critical factors for documentation and risk adjustment. AI tools should go beyond simple optical character recognition and incorporate advanced clinical natural language processing to understand temporal aspects and contextualize information.
Improving the accuracy and efficiency of risk adjustment is a big undertaking, but it is achievable with the right tools and partners. By embracing AI-driven solutions and collaborating with experts in the field, your healthcare organization can transform its processes, ultimately improving the chances of delivering cost savings and better outcomes for patients.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?
Read the full article here
