Risk algorithms are the keystones supporting value-based contracts.  They enable the evaluation of the relative health risks of individuals within populations, thus driving the compensation of healthcare systems.  In a value-based arrangement, healthcare systems are paid for the care of populations based on the projected needs of those populations.

But the accuracy of the risk calculated by an algorithm (and the accuracy with which the healthcare system is compensated) depends not only on the design of the calculation, but also on the inputs to that calculation.  

  • At the population level: Is member attribution correct? Or more practically, do all parties in the contract trust that the right members are being evaluated by the risk algorithm used to determine compensation?
  • At the patient level: Is all relevant clinical data are supplied to the algorithm?

Traditionally, patient-level clinical input to a risk algorithm comes from claims.  But because claims-based data provides only a limited amount of clinical information as an input, the risk algorithm is unable to generate an accurate output.   This can have significant financial implications for a healthcare system – from $100-200 PMPY retrospectively to $1,000-2,000 prospectively.

To understand why claims-only risk algorithms are deficient, it helps to understand risk, risk algorithms, and how they are used by a healthcare system.

How do healthcare systems need to think about risk?    

In a recent blog post, our chief medical officer Rich Parker, MD explained the three ways healthcare organizations need to think about risk:

  • Patients likely to have a negative outcome. Healthcare organizations need to identify patients at high, intermediate, or low probabilities of certain outcomes such as admission to the hospital, dying within 12 months, or accruing high levels of medical utilization and costs
  • Financial consequences for clinical performance. Healthcare organizations take on financial responsibility for a population and need to provide appropriate, quality care within a certain budget.
  • A patient’s aggregate conditions. Related to both of the above, an individual patient’s likelihood of negative outcomes determines how much a provider organization is compensated for the care of that patient.

Each way of thinking about risk relies on a calculation – a risk algorithm.

What is a risk algorithm?

At a very basic level, a risk algorithm takes an input – data about the health of an individual patient – and processes that data by assigning each piece of information a different weight.  The risk algorithm then provides an output in the form of a risk score for that specific patient.  By reducing complex, individual medical histories to a simple number, the risk score facilitates easier comparison of patients and analysis of populations.

There are many risk algorithms, and each one uses different health information and applies a different relative importance to that health information.   This can be a challenge for healthcare systems, which frequently manage multiple risk-based contracts.  Each contract might calculate risk using a different algorithm, meaning that even if two patients have very similar clinical conditions the healthcare system might be paid quite differently for their care.

Risk algorithms and financial performance

Given this challenge, healthcare systems need analytics that calculate risk using:

  • the right risk algorithm
  • for the right patient population
  • covered by the right contract
  • with the right data

If a healthcare system has incorrect attribution, uses the wrong risk algorithm, or does not have complete data, financial performance will be incorrectly projected.

For example, a given healthcare organization might have a Medicare Advantage contract, a Medicaid contract, and a commercial contract that all use different risk algorithms.  If that system can can see accurate risk scoring using the appropriate algorithm for each contract, it can forecast financial performance.  On the one hand, the system can project its risk-adjusted compensation, and on the other hand, it can understand the associated utilization.

The stakes are high

For a healthcare system engaged in risk-based contracting, accurate risk adjustment is critical to financial performance.  But accurate risk adjustment depends on accurate documentation, and documentation gaps can mean healthcare systems miss out on $100-200 PMPY in retrospective risk adjustment.  Documentation gaps can also prevent health systems from capturing $1000-2000 PMPY in current year opportunities.

Conversely, excessive documentation unsupported by clinical evidence could also result in inaccurate risk adjustment, where risk is overstated.  This could lead to an audit and potentially, to significant financial penalties for a healthcare system.

With a lot of money at stake, healthcare systems need to ensure the accuracy of their risk adjusted premiums.

Alyssa Drew

Alyssa Drew is the Strategic Marketing Director at Arcadia, where she helps healthcare systems understand and unlock the value of their data to enable their success in value-based care.   Her background bridges both strategy and technology.   In over five years at Arcadia, she has managed complex analytics and transformation projects for Arcadia clients across the country and served as Arcadia’s Business Practice Leader.  Before joining Arcadia, she held management roles in enterprise analysis, strategic planning, and financial analysis for a $1B organization.

Alyssa has an undergraduate degree in Visual and Environmental Studies from Harvard University.  She has tremendous enthusiasm for the incredible work her Arcadia colleagues do on a daily basis, and is excited to host the Arcadia Healthcare Datathon annually.

September 20, 2016

Why claims-based approaches to risk are deficient – and what to do about it

The traditional approach to risk adjustment uses claims-based data, but this approach to calculating clinical risk is deficient – and it can put an organization at financial risk.  Claims-based information is incomplete, slow, misaligned with provider data, and often not relevant at the point of care.   Luckily, EHR data can close the gaps.

Read Our Blog Post