Due to the intricate pathogenesis of ITP and the absence of definitive diagnostic indicators and gold standards, this study integrated a variety of common laboratory parameters that are routinely measurable in primary hospital to construct a clinical diagnostic model for ITP. In this study, we initially considered relevant indicators from the general baseline data, such as age and sex, because according to the large-scale sample, the prevalence of ITP was greater in women and elderly people aged older than 60 years. The complete blood count can be used to determine whether anemia is caused by insufficient or impaired utilization of hematopoietic materials (vitamin B12, folate and/or iron deficiency), as indicated by laboratory parameters such as the mean red blood cell volume and red blood cell distribution width. Notably, we calculated an index to evaluate the variability in platelet counts based on three consecutive platelet count measurements; this index has value in assessing platelet counts variability in patients and holds diagnostic significance for ITP (Li et al. 2021). The unstable platelet counts in ITP patients may be related to premature destruction of platelet or impaired platelet production caused by immune regulation disorders caused by autoantibodies, cytotoxicity, complement or other immune mechanisms (Audia et al. 2017; Zufferey et al. 2017, Toltl et al. 2011; Shrestha et al. 2020). Rapid fluctuations in platelet counts can be observed in patients infected, vaccinated, or receiving other immune stimuli (Rinaldi et al. 2014). Conversely, patients with platelet reduction caused by non-immune factors usually exhibit more stable platelet counts, with stable rates of peripheral platelet clearance and platelet production. The PVI can be calculated based solely on routine platelet count values, making it highly clinically applicable compared to other laboratory parameters. Moreover, the PVI is a dynamic metric that shows improved accuracy as more platelet count values accumulate over time. Therefore, incorporating PVI score along with other relevant laboratory parameters is clinically important when developing clinical prediction models (Li et al. 2021).

We selected 6 variables from the aforementioned 34 laboratory parameters, namely RBC, MCHC, RDW-SD, PVI score, IPF, lymphocyte absolute value. In other thrombocytopenic conditions, a decrease in platelet count is typically not isolated. For example, patients with myelodysplastic neoplasms, aplastic anemia, and liver diseases often exhibit concurrent anemia making the RBC and MCHC valuable discriminative indicators. The IPF can reflect the functionality of bone marrow megakaryocytes (Arshad et al. 2021). In patients with ITP, bone marrow proliferation is usually normal, while IPF levels are elevated. In some cases of ITP, the platelet can be severely reduced, but relying solely on the lowest platelet count is inadequate for diagnosis. The pathogenesis of ITP is complex and diverse and involves humoral immune disorders, cellular immune disorders, abnormal cytokine secretion, platelet apoptosis, and genetic and environmental factors (Miltiadous et al. 2020). Since the first description of regulatory T cell abnormalities in ITP was provided (Liu et al. 2007), a large body of literature has described new aspects of T cell, B cell, and dendritic cell biology in this disorder, and animal models are continuing to yield relevant results. However, to our surprise, lymphocyte absolute value was helpful in the diagnosis of ITP. We suspect that this difference partly reflects the number of T cells in the peripheral blood, but since the pathogenesis of ITP involves the interaction of multiple immune cells, our hypothesis needs to be verified in a larger cohort.

This study also has certain limitations. Firstly, glycoprotein-platelet antibody and detailed immunoglobulin profiles are potentially informative indicators for ITP, but because of the high proportion of missing values, we excluded these variables from our current model. Secondly, relying solely on a diagnostic model containing these 6 variables in clinical practice may overlook important clinical information compared to the laboratory diagnostic indicators of ITP mentioned in the existing scientific literature. Thirdly, while the study included a substantial number of patients, the relatively small size of the ITP group may affect the performance of the diagnostic model. In the next stage, additional specialized diagnostic tests will be incorporated, along with relevant medical history information such as transfusion history, megakaryocyte platelet generation capacity, and previous treatment history. Our model will be further optimized with larger and more balanced sample sizes.

In conclusion, we used the aforementioned 6 variables to construct a diagnostic model that had a strong ability to distinguish between patients with ITP and those with other thrombocytopenic conditions, with a sensitivity and specificity in the validation set of 0.90 and 0.87, respectively, suggesting that applying this model to the initial diagnosis of patients with thrombocytopenia can provide strong indications for suspected ITP, shorten the clinical diagnostic process, and facilitate prompt initiation of ITP-specific therapies.