An Interpretable [ML] Model for Predicting the Effectiveness of Low-Dose Methylprednisolone in Treating [LC]: A Retrospective Study, 2025, Zhang et al

forestglip

forestglip

Moderator
Staff member
An Interpretable Machine Learning Model for Predicting the Effectiveness of Low-Dose Methylprednisolone in Treating Long COVID: A Retrospective Study

Zhang, Jisheng; Chen, Yang; Zhang, Aijun; Yang, Yi; Ma, Liqian; Yu, Kewei; Zhang, Weitao; Ye, Xiaojing; Zhang, Jiangsong; Lin, Ke; Lin, Xianming

[Line breaks added]

Abstract
Long COVID is a chronic, multisystem disease with limited response to conventional treatments. While low-dose methylprednisolone has shown effectiveness in some patients, individual responses vary, and accurate predictive tools are lacking.

This retrospective study included 330 Long COVID patients who received low-dose methylprednisolone treatment across three hospitals. Patients were divided into training (n=202), test (n=33), and external validation sets (n=53, n=42). Using Least Absolute Shrinkage and Selection Operator (LASSO) regression, 38 variables were analyzed to develop six machine learning models.

The Logistic Regression (LR) model showed stable performance across all datasets (AUCs: 0.8715, 0.7198, 0.8419, 0.8676), making it the final model selected. Shapley Additive Explanations (SHAP) analysis identified seven key variables, which were used to construct a nomogram for predicting treatment efficacy.

The LR model and nomogram demonstrated strong predictive performance and clinical interpretability, offering a valuable decision-support tool for individualized treatment of Long COVID with low-dose methylprednisolone.

Web | DOI | PDF | iScience | Open Access
 
Looking at data from Chinese hospitals to try to make a model that predicts methylprednisolone effectiveness.
we developed a predictive model for the efficacy of low-dose methylprednisolone in Long COVID treatment using six ML algorithms, based on patient data recruited from the Third Affiliated Hospital of Zhejiang Chinese Medical University (TAHZCMU) between 2022 and 2024. The model was tested internally, followed by external validation using data from patients recruited at the First Hospital of Jiaxing (FHJ) and Haining People’s Hospital (HPH) from 2023 to 2024.
Click to expand...

This study ultimately included 330 cases, of which 266 were classified as treatment-effective and 64 as treatment-ineffective.
Click to expand...

Based on preliminary analysis, they selected these variables for training several types of machine learning models:
Fatigue (binary, I think)
SF-36 survey total score
Ratio of Forced Expiratory Volume in 1 Second to Forced Vital Capacity (FEV1/FVC)
Pittsburgh Sleep Quality Index (PSQI
IL-6
CRP
Neutrophil percentage
Click to expand...

They trained 6 types of machine learning models, and based on results on a test set, they selected the logistic regression (LR) model as the best one. They then tested again on two different validation sets, and still got good results with the LR model.
In contrast, the Logistic Regression (LR) model demonstrated more stable performance across the training set and test set with minimal fluctuations in AUC and F1 score, and overall excellent performance (Table 2 and 3, Figure 3A and 3B). Therefore, the LR model was selected for downstream analysis. Additionally, the external validation set was used to assess the robustness of the final selected model (Table 4 and 5, Figure 3C and 3D), and the LR model maintained stable performance, validating its effectiveness.
Click to expand...

And they do some interpretation of the model:
For example, patients with higher levels of CRP and IL-6 are associated with a better therapeutic response
Click to expand...
Similarly, patients with fatigue symptoms have a stronger association with treatment effectiveness compared to those without fatigue symptoms
Click to expand...
 
You must log in or register to reply here.
Back
Top Bottom