Construction of a new reflex fluorescence platelet count confirmation model based on red blood cell and platelet parameters.

INTRODUCTION

We sought to clarify the interference of red blood cell (RBC) and platelet (PLT) parameters on PLT count by impedance method results and establish an intelligent reflex fluorescence PLT count (PLT-F) confirmation model using PLT-F as the reference to quantify the model's clinical efficacy.

METHODS

A retrospective analysis was performed on routine blood data from 11 283 patients (June 2023 to February 2024) at our hospital. Patients were stratified into a training set (n = 7898) and verification set (n = 3385) using Laboman AI, version 1.0-312, software (Shanghai Hyson Meikang Medical Electronics Co Ltd). A decision tree algorithm was employed to construct a new model that integrates RBC and PLT parameters for reflex PLT-F confirmation, and the model's effectiveness was evaluated.

RESULTS

The combined 3-factor model achieved optimal performance for PLT-F reflex confirmation. Its overall false-negative proportion, release rate, total effective rate, and negative predictive value reached 2.9%, 55.2%, 71.1%, and 94.7%, respectively, in the training set, with matching values of 4.2%, 52.4%, 66.1%, and 91.9%, respectively, in the validation set.

DISCUSSION: This study established a novel reflex PLT-F confirmation model based on RBC and PLT parameters. The model achieved a 52.4% sample release rate, with a corresponding overall false-negative proportion of 4.2% in the validation set, presenting a clear trade-off between laboratory operational efficiency and diagnostic safety.