OBJECTIVES: Within cost effectiveness analyses (CEAs), descriptions of decision uncertainty often rely on cost effectiveness acceptability curves (CEACs) which report the probability that each alternative is cost-effective. An extension of decision uncertainty analysis in the form of net loss curves, as described in seminal contributions to the literature by Eckermann, offers distinct advantages to also inform the magnitude of losses if a sub-optimal alternative is selected. We demonstrate the advantage of net loss curves in an example of a CEA examining treatment alternatives for opioid use disorder (OUD).

METHODS: We perform a probabilistic sensitivity analysis for a CEA comparing office-based methadone (MO), clinic-based methadone (MC), and office-base buprenorphine (BO) to treat patients with OUD. We report not only the probability of alternatives’ cost effectiveness via the CEAC but also quantify and report, across a range of willingness to pay (WTP) thresholds, the per-patient net loss that is incurred conditional on the adoption of each respective treatment alternative.

RESULTS: Given a WTP of $7,000 per additional patient retained in six-month treatment (based on previous publications), MO has the lowest expected net loss at $230/patient, followed by MC at $334/patient and BO at $1,255/patient. Choice of a non-optimal alternative (MC or BO) thus implies losses of $104 or $1,025, respectively per patient - and much larger in aggregate.

CONCLUSIONS: Net loss curves illustrate the substantial additional net loss associated with choosing alternatives other than MO, particularly BO. This example demonstrates the potential for net loss curves to serve as an impactful visualization tool to emphasize the value of the cost-effective alternative over ranges of WTP by quantifying the magnitude of consequences when implementing suboptimal alternatives. This quantification of consequences is absent when the analysis is limited to CEACs; thus, net loss curves serve to build a stronger case for the cost-effective alternative.