The Pros and Cons of Using a Single Economic Model for Both Budget Impact and Cost-Effectiveness Analyses

Those with experience in health economics will likely have their own perspective on this topic. There are multiple advantages and disadvantages to using a hybrid model that accommodates both a Budget Impact (BI) Analysis and a Cost-Effectiveness (CE) Analysis. Similarly, using two separate models comes with its own set of challenges and opportunities.

1. Perspective of the Payer

One of the key factors influencing model design is local guidelines, which often specify which costs and perspectives to include. For example, BI analyses frequently require a strict payer (e.g., government) perspective, while CE analyses may also account for patient contributions or private insurance. This doesn’t just apply to drug costs but extends to hospital stays, medical devices, physician visits and all other cost items.

A hybrid model must be flexible enough to reflect these differences. From experience, we recommend making the user explicitly select a BI or CE setting before running the analysis. Without this safeguard, errors are likely to creep in. While advanced solutions like VBA scripting can help, we’ll set those aside for now.

2. Inclusion of Indirect Costs

Analysts often have the option to include or exclude indirect (societal) costs. However, not all models allow for separate inclusion/exclusion settings between BI and CE analyses. If not addressed properly, this can lead to inconsistent results and misinterpretation.

3. Comparators

In a BI analysis, the goal is to simulate real-world clinical practice—first without, then with, the introduction of the new technology. We frequently receive pre-designed models that include a predefined set of comparators, yet these may not reflect the most relevant or locally applicable options.

By contrast, CE analyses typically use one comparator—ideally the one used in clinical studies. This may not always be the most relevant in current practice. In some cases, indirect comparisons through network meta-analysis (NMA) are needed, adding further complexity.

4. Discounting

Future costs and health outcomes are usually discounted in CE analyses if the time horizon exceeds one year. This is not typically required for BI analyses. Many models fail to offer sufficient flexibility to either apply or omit discounting as needed, which can affect the validity of conclusions.

5. Half-Cycle Correction

Half-cycle corrections are often applied in CE analyses—especially when model cycles are longer (e.g., 3 months). This technique corrects for the fact that not all patients transition between health states at the start of each cycle.

In BI models, however, half-cycle correction adds unnecessary complexity. Separating CE and BI models can help avoid this issue entirely.

6. Mortality Adjustments

CE models often include proportional mortality adjustments based on background mortality. Integrating such corrections into BI models is both complex and often unnecessary—whether the model is incidence- or prevalence-based.

7. Sensitivity Analyses

Deterministic sensitivity analyses help quantify the impact of changes in input parameters. The parameters relevant for BI may differ from those used in CE, yet hybrid models don’t always allow for this separation. As a result, manual intervention is sometimes needed, increasing the risk of errors and inefficiencies.

8. VBA Scripts and Model Transparency

Some models rely heavily on complex VBA scripting. While these tools can automate calculations, they often obscure how the model works. We regularly question who ultimately benefits from this complexity—the client, the payer, the user, or the developer?

In our view, simpler models that are easy to understand, explain, and adjust—even by non-consultants—are usually more effective. Flashy does not equal functional.

Conclusion

As an analyst, you must weigh the convenience of a hybrid model against the clarity and flexibility of separate BI and CE models. At our firm, we prefer separate models. Many times, a client has asked for an updated BI analysis in the context of a managed entry agreement, or a health authority has requested model access. Untangling a hybrid model at that point is no fun—and you’ll quickly understand why this blog was written in the first place. 😊