A Novel Adaptive Testing Scheme for Multidisease Testing

Testing for multiple diseases, which is essential for screening and diagnosis (particularly where symptoms overlap but treatments differ), is expensive. Two cost-saving strategies are multiplexing (using a single assay for multiple diseases) and pooling (testing multiple subjects simultaneously with one assay, with individual retesting for positive pools). Although laboratory automation enables the flexible integration of these strategies, optimal testing designs that fully exploit this potential remain underexplored. Existing approaches restrict retesting to the initial pooled multiplex (Dorfman testing), leading to inefficiencies. To address this gap, we introduce a novel adaptive testing design that tailors each subject’s retesting assay based on pooled test outcomes. This adaptivity induces dependencies among pooled test outcomes, resulting in a complex partition-type optimization problem significantly more challenging than the Dorfman design problem. We characterize key structural properties of optimal adaptive designs, which underpin a robust model under prevalence correlation uncertainty and effective solution methods. A case study of 18 respiratory diseases demonstrates that adaptive designs substantially reduce cost under forecast error, outperforming current practice benchmarks by 51.5%–60.9% and Dorfman designs by 32.4%–39.6%. Crucially, robust adaptive designs combine substantial cost reductions with a highly scalable solution method, offering a practical approach for real-world testing applications.

Supplemental Material: All supplemental materials, including the code, data, and files required to reproduce the results, are available at https://doi.org/10.1287/opre.2025.2006.