A VOI framework for personalizing the timing of biomarker collection (306611)*Aasthaa Bansal, The Comparative Health Outcomes, Policy, and Economics (CHOICE) Institute, University of Washington
Anirban Basu, The Comparative Health Outcomes, Policy, and Economics (CHOICE) Institute, University of Washington
Patrick Heagerty, Department of Biostatistics, University of Washington
Lurdes Inoue, Department of Biostatistics, University of Washington
David Veenstra, The Comparative Health Outcomes, Policy, and Economics (CHOICE) Institute, University of Washington
Keywords: dynamic decision-making, prediction, value of information
Patient surveillance with repeated biomarker measurements presents an opportunity to detect and treat disease progression early. However, frequent biomarker collection can be costly and inconvenient. Alternatively, predictions based on patients’ earlier biomarker values may be used to inform decision-making, but predicted biomarker levels are uncertain, leading to decision uncertainty. We extend standard VOI methods to account for prediction uncertainty over time and determine the time point where more precise information on biomarker levels would be most valuable. We illustrate our methods using longitudinal data on cystic fibrosis patients. Current practice is to regularly measure FEV1% and use the last measured value of this biomarker to determine the need for lung transplantation. We contrast this with the alternative approach of using FEV1% predictions and the proposed VOI approach to determine the optimal time interval between updating FEV1% data. Our results show that patients with poorer FEV1% values benefit more from frequent testing. Furthermore, using more than one past biomarker measurement for prediction substantially reduces the value of annual biomarker collection.