Artificial intelligence (AI) could help drive accurate and effective risk adjustment in value-based care plans.
Risk capture of patient disease is a critical process that health plans must optimize to control care costs and obtain accurate reimbursement in value-based care agreements.
Properly capturing patient risk typically involves combing through patients’ electronic health records to extract key clinical information that is often indicative of underlying, undiagnosed conditions. However, because about 80% of health care data is unstructured and not in suitable form for analysis, payers often struggle to gain a comprehensive view of patient risk.
In traditional methods of risk capture, payers and managed care organizations have relied on expensive and time-consuming chart reviews performed by clinicians to spot and extract important unstructured information from the often massive amounts of data in patient records and claims. In recent years, though, as artificial intelligence (AI)–based tools such as natural language processing (NLP) have grown more sophisticated, health plans have looked to this technology to capture risk in a way that is more consistent and cost-effective than traditional chart reviews.
Critical for Value-Based Care Success
While progress toward value-based care has been relatively slow—caused in part by the understandable focus on the COVID-19 pandemic—payers and providers continue to explore this emerging alternative payment model. For example, a recent joint survey from the Medical Group Management Association and Humana found that in 15% of responding providers, more than half of their revenue was from value-based care contracts. Alternatively, 63% of respondents said that less than one-quarter of their practice’s revenue is based on value-based contracts.
Separately, recent data from The Health Care Payment Learning & Action Network (LAN) show that 40.9% of US health care payments, representing approximately 238.8 million Americans and over 80% of the covered population, stemmed from value-based reimbursement models in 2020. Data from LAN also show progress toward value-based care by Medicare and Medicare Advantage.
In traditional Medicare, 24.2% of payments in 2020 were part of some 2-sided risk model compared with 20.2% in 2019. In Medicare Advantage, the percentage of payments in 2-sided risk models increased to 29.3% in 2020 from 28.6% the prior year.
So, clearly, both payers and providers will continue to invest in value-based care. Here’s where risk-adjustment comes into play: In value-based arrangements, such as Medicare Advantage, for example, health plans contract with CMS to offer benefits to enrollees and are reimbursed via a capitated system, in which plans receive a per-month, per-member predetermined payment from CMS. These monthly capitated payments are risk-adjusted for each member to reflect their health status and project an appropriate level of monthly spending for Medicare-covered services.
As a result, accurate risk adjustment is necessary to ensure that health plans receive adequate funding to care for their patients, based on each patient’s disease status and risk. Rather than performing chart reviews to find and extract key unstructured data from patient records and claims, payers are turning to AI-based technologies such as NLP.
What NLP Can Do for Payers
NLP is an essential tool for payers that engage in value-based arrangements because it enables them to process patient records and extract risk-adjustable diagnoses and the supporting documentation typically buried in unstructured data, driving more accurate risk adjustment.
Cost control: Payers often facilitate care for complex patients with multiple chronic conditions and comorbidities who require more expensive care. Accurate risk capture is essential for health plans to be properly reimbursed for these patients’ care. NLP combines with predictive analytics to enable payers to identify patients, for example, with diabetes and stage 1 kidney disease who are at risk of progressing to stage 2.
To control cost requires foresight and prevention. By enabling payers to gain a more accurate picture of overall patient health, NLP helps identify patients who need proactive intervention to control disease progression, or, conversely, spot patients who have been overtreated.
Longitudinal patient view: Many patients, particularly those with chronic conditions and comorbidities, see multiple providers and specialists, which poses a challenge for payers in assembling a full picture of patient health, because patient data are spread across multiple medical records systems. A patient with diabetes, for example, may visit regularly with a primary care physician, endocrinologist, dietician, and other providers. Payers can use NLP to scour unstructured data patient records, then synthesize the information from disparate providers to calculate the patient’s total cost of care.
Collaborative data sharing: When payers analyze patient records, they may uncover various data points that could help providers close patients’ care gaps. Health plans can share these insights with providers to collaborate around the shared goal of better patient health.
As value-based care plans, including Medicare Advantage, continue to spread throughout the market, risk adjustment will become an increasingly important cost-control mechanism for payers. AI-based technologies like NLP will help health plans comb through unstructured data to surface the insights needed to drive more accurate, effective risk adjustment.