OBJECTIVES: Test value is driven by how the clinical information provided determines disease management and through this individuals’ outcomes. Understanding the link between misclassification and outcomes is not straightforward for tests that classify individuals into multiple categories. We illustrate how disaggregation of cost-effectiveness results makes value drivers explicit for such tests, using the example of software fusion versus cognitive fusion biopsy for individuals with suspected prostate cancer (PCa).

METHODS: We present a two-component model, which estimates quality-adjusted life-years (QALYs) and 2021 costs (GBP) from a healthcare payer perspective, discounted at 3.5% per annum. The decision tree component captures the diagnostic pathway and adverse events, and classifies individuals according to biopsy results and true disease status (15 possible categories). The state transition component links final classification to clinical management and this to longer-term outcomes over a lifetime horizon. Cost-effectiveness results are expressed as net health benefit (NHB) at £20,000/QALY. We disaggregate NHB by final classification and true disease status and then further disaggregate results by model component and health state in which they were generated.

RESULTS: Software fusion biopsy increases the correct detection of PCa, particularly at cancer grade 1 and 2, and has an incremental NHB of 0.0081 QALYs versus cognitive fusion. Disaggregated long-term model results suggest that correctly identifying grade 1 cancer results in NHB loss, which is offset by gains from increased correct detection at grade 2; this balance can shift if prevalence is varied.

CONCLUSIONS: Although results suggest that software fusion is cost-effective versus cognitive fusion, its value is driven by sparse prevalence and diagnostic accuracy evidence. Presenting results disaggregated by final classification and true disease status allows decision makers to explore plausible alternative values of prevalence by grade and diagnostic accuracy for software fusion and cognitive fusion can impact on cost-effectiveness estimates.