A Novel Hybrid Resampling Approach for Analyzing Immune Thrombocytopenia

A Platelet Variability Index (PVI), reflecting both platelet count fluctuations and thrombocytopenia severity, distinguishes immune thrombocytopenia (ITP) from other thrombocytopenic causes (Na Li et al, 2021). Higher PVI scores correlated with definite ITP, even at initial assessment. Prior studies have used various metrics (e.g., average Standard Deviation, Coefficient of Variation (CV), Median Absolute Deviation, Mean Absolute Error) to represent individual patient platelet levels. This work focuses on the average CV. Given ITP's rarity and the fact that extreme variations in platelet counts are a key characteristic of the disease, we hypothesize that the distribution of average CVs may be modeled by an extreme value distribution. Specifically, we employ the Gumbel distribution, commonly used to model the maximum or minimum of a set of independent and identically distributed random variables, which is appropriate for capturing the extreme fluctuations in platelet counts observed in ITP. We then apply Bootstrapping, Jackknife resampling, and their convex combination to analyze the CV data.