The Second Panel on Cost Effectiveness in Health and Medicine made a number of recommendations of how to improve cost effectiveness analysis (CEA). Yet, CEA is far from “solved”. In a recent article by Neumann et al. (2018), the Second Panel identifies seven CEA areas where additional research is needed. This include:
- CEA and perspective: “Many issues require work before the field can reach agreement on summary measures for a societal perspective, particularly which elements to include in a summary, how these should be determined, and how to value them.” While the Second Panel recommended taking a societal perspective and–if appropriate for the question at hand–also the health system perspective, it was unclear whether the goal should be to maximize health benefits or overall benefits (e.g., health but also labor market outcomes, schooling, or other non-health factors.). The authors mention that cost-benefit analysis (CBA) may be appropriate. As the name indicates, CBA considers the value of costs and benefits to the losers and gainers of an intervention based on market prices [or if these are not available, then shadow prices.
- Modeling. While many models conduct sensitivity analyses around models parameters, few CEAs conduct sensitvity analysis across model types. The Innovation and Value Initiative (IVI) has done this with their IVI-RA Value Toll Model, but few others do this. An exception is Cancer Intervention and Surveillance Modeling Network (CISNET), which does perform comparative model analysis. Also, best practices recommend that modelers provide sufficient detail about the model structure and parameterization to allow other researchers to reproduce it, other groups–such as IVI, have made their models itself available online.
- Valuing health outcomes. Measure of quality of life often can be translated into QALYs assuming that the health state persists over some fixed duration. But what about temporary health states? Chaining methods may be used to estimate the value of temporary health states (see Wright et al. 2009, Locadia et al. 2004, and McNamee et al. 2004). In one case, “subjects were asked to compare health states associated with the process of prenatal diagnosis to a temporary health state of the same duration based on a description of the experience of undergoing chemotherapy (but not so labeled).” QALYs are also problematic because they assume that individuals are indifferent to the order of when the events occur. The Neumann et al. paper even asks whether using virtual reality would be helpful to better model health states.
- Valuing non-health outcomes. What happens if a medicine–such as an antipsychotic–is able to reduce crime rates? Should this be included in a CEA model? The answer is likely yes. Also, the Second Panel argued that “the effects of morbidity on productivity in the labor market and in household production are not captured by standard utility measures and therefore should be assessed in pecuniary terms and included in the numerator of the CEA.”
- Evidence Synthesis. Before building a CEA model, one must know the clinical benefits of different treatments. Synthesizing available evidence when there is not a head to head trial is problematic. Neumann et al. write that “Currently, there is no rigorous, internally consistent set of premises and theorem-based derivative propositions that motivates and justifies the practice of evidence synthesis. The exception may be the mathematical foundations of quantitative synthesis (meta-analysis).”
- Estimating CE thresholds. There are two options here, supply side or demand side. Supply side calculates the CE threshold based on the opportunity cost; what would be the value of these funds if they were allocated to other activities. The demand side looks at consumer willingness to pay for health gains. Some CEAs–such as ICERs–have said that ‘a given intervention is high value, but not affordable as the CEA and budget impact analyses are done separately. If this is the case, then the CEA criteria clearly do not reflect the scale and value of the opportunity costs. While some argue that the supply side works better in single payer systems with fixed budgets, others would argue that the share of the government’s budget allocated to health is in fact a choice variable, and thus higher WTP would argue for more resources directed to health expenditures. Empirically measures of WTP for a QALY do vary greatly across countries and based on the methodology used to estimate these value (see Ryen and Svennson 2005).
- CEA communication. Should I do a perfect CEA or do a good CEA and get it out fast? As all diseases and treatments have their own idiosyncrasies, one could spend a nearly unlimited amount of time collecting information to make a CEA high quality. Yet, the Second Panel does make a few key recomendations including: (i) having a written protocol, (ii) having an impact inventory, and (iii) presenting the societal and health system perspective. It would be helpful if CEAs could be graded using a scoring system such as the Grading of Recommendations Assessment, Development and Evaluation (GRADE) working group, but applying these principles in practice may be difficult and focus more on process/documentation rather than the actual scientific quality of the CEA.
Overall, cost-effectiveness analysis has come a long way. Yet, there is still much work to be done to insured that treatment benefits, risks and costs can be adequately captured to inform stakeholder decision-making.
- Neumann, Peter J., David D. Kim, Thomas A. Trikalinos, Mark J. Sculpher, Joshua A. Salomon, Lisa A. Prosser, Douglas K. Owens et al. “Future Directions for Cost-effectiveness Analyses in Health and Medicine.” Medical Decision Making38, no. 7 (2018): 767-777.