**3.3 Step 3: risk assessment**

In Step 2 we created the current and proposed patient pathways, estimated the value created by the HCM and the basic pricing parameters. However, this estimate is a mean; we do not know the variance around the estimated outcome. Variance estimation is important for healthcare models: healthcare claims are highly variable for two reasons. First, the distribution of healthcare claims itself is a convolution of two highly-variable distributions, frequency and severity. Second, outcomes of a healthcare program are subject to performance risk. Step 3 begins with modeling the distribution of the predicted outcome. Additionally, there are multiple variables involved in the predicted outcome; many of these variables can be controlled in order to limit the contract risk. The Risk Assessment step helps the analyst to understand the contribution of individual variables to the predicted outcome and to choose values in such as way as to mitigate some of the inherent stochastic risk of the contracted outcome. **Figure 6** shows some of the variables that comprise a value-based contract that an analyst should consider when modeling contract risk.

**Figure 6** shows that designing a value-based contract is a complex undertaking. While we will not discuss all the variables in **Figure 6**, we will discuss some key variables and use them to illustrate the complexity of the modeling that is required as part of the Value-based Contract pricing.

• **Attribution:** it is important to define precisely those patients for which the HCM or provider will accept risk, and at what point the patient is triggered into the risk group. Attribution can occur on a population basis (for example patients with diabetes) or an episode basis (for example knee surgery). Triggers for these patients generally occur within claims datasets. Occasionally triggers may also be found in electronic medical records (although the lack of integrated medical record/claims data makes modeling difficult in this context). Attribution may also be triggered by the use of a derived marker, for example a grouper model (in the US, Hierarchical Condition Categories (HCCs) or Episode Groupers (for example ETGs)). It is also necessary to use triggers to determine which provider should have accountability for a given patient.

**Figure 6.** *Key parameters for a value-based contract.*
