Partial Verification Bias Correction Using Scaled Inverse Probability Resampling for Binary Diagnostic Tests

Diagnostic accuracy studies are crucial for evaluating new tests before their clinical application. These tests are compared against their respective gold standard tests, and accuracy measures such as sensitivity (Sn) and specificity (Sp) are often calculated. However, these studies frequently suffer from partial verification bias (PVB) due to selective verification of patients. PVB eventually leads to biased accuracy estimates in such studies. Among methods developed for PVB correction under the missing at random assumption for binary diagnostic tests, a bootstrap-based method known as the inverse probability bootstrap (IPB) was proposed. Despite showing low bias for estimating Sn and Sp, the IPB method exhibited higher standard errors than other PVB correction methods. This paper introduces two new methods: scaled inverse probability weighted resampling (SIPW) and scaled inverse probability weighted balanced resampling (SIPW-B), which build upon the IPB approach. Through simulations and clinical data, SIPW and SIPW-B were compared against IPB and other methods. The results demonstrated that the new methods outperformed IPB by showing lower bias and standard errors in Sn and Sp estimation. Specifically, SIPW-B outperformed IPB in Sn estimation, while SIPW performed better in Sp estimation, particularly when disease prevalence is low. These methods offer advantages such as complete data restoration and calculations independent of disease prevalence. Although computationally demanding, this limitation becomes less significant with the increasing power of modern computing resources.