Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Statistical and Predictive Analysis to Identify Risk Factors and Effects of Post COVID-19 Syndrome

Milad Leyli-abadi, Jean-Patrick Brunet, Axel Tahmasebimoradi

TL;DR

The paper tackles the problem of identifying risk factors and predicting the intensity of long COVID using the Lifelines cohort. It defines Post COVID-19 Symptom Intensity (PCSI) as a continuous measure derived from post-infection symptom trajectories and evaluates multiple predictive models (LR, RF, GB, MLP) with a focus on explainability. Neural networks deliver the best predictive accuracy (MAPE ≈ $0.26$), while linear and tree-based models provide complementary insights via coefficients, feature importances, and SHAP explanations. Key predictors include loss of smell, headache, muscle pain, vaccination timing, chronic disease presence, and sex, with vaccination and good baseline health reducing risk. The work offers practical guidance for targeting interventions and contributes a reproducible benchmarking framework and Python package for health-data analyses on Lifelines data.

Abstract

Based on recent studies, some COVID-19 symptoms can persist for months after infection, leading to what is termed long COVID. Factors such as vaccination timing, patient characteristics, and symptoms during the acute phase of infection may contribute to the prolonged effects and intensity of long COVID. Each patient, based on their unique combination of factors, develops a specific risk or intensity of long COVID. In this work, we aim to achieve two objectives: (1) conduct a statistical analysis to identify relationships between various factors and long COVID, and (2) perform predictive analysis of long COVID intensity using these factors. We benchmark and interpret various data-driven approaches, including linear models, random forests, gradient boosting, and neural networks, using data from the Lifelines COVID-19 cohort. Our results show that Neural Networks (NN) achieve the best performance in terms of MAPE, with predictions averaging 19\% error. Additionally, interpretability analysis reveals key factors such as loss of smell, headache, muscle pain, and vaccination timing as significant predictors, while chronic disease and gender are critical risk factors. These insights provide valuable guidance for understanding long COVID and developing targeted interventions.

Statistical and Predictive Analysis to Identify Risk Factors and Effects of Post COVID-19 Syndrome

TL;DR

The paper tackles the problem of identifying risk factors and predicting the intensity of long COVID using the Lifelines cohort. It defines Post COVID-19 Symptom Intensity (PCSI) as a continuous measure derived from post-infection symptom trajectories and evaluates multiple predictive models (LR, RF, GB, MLP) with a focus on explainability. Neural networks deliver the best predictive accuracy (MAPE ≈ ), while linear and tree-based models provide complementary insights via coefficients, feature importances, and SHAP explanations. Key predictors include loss of smell, headache, muscle pain, vaccination timing, chronic disease presence, and sex, with vaccination and good baseline health reducing risk. The work offers practical guidance for targeting interventions and contributes a reproducible benchmarking framework and Python package for health-data analyses on Lifelines data.

Abstract

Based on recent studies, some COVID-19 symptoms can persist for months after infection, leading to what is termed long COVID. Factors such as vaccination timing, patient characteristics, and symptoms during the acute phase of infection may contribute to the prolonged effects and intensity of long COVID. Each patient, based on their unique combination of factors, develops a specific risk or intensity of long COVID. In this work, we aim to achieve two objectives: (1) conduct a statistical analysis to identify relationships between various factors and long COVID, and (2) perform predictive analysis of long COVID intensity using these factors. We benchmark and interpret various data-driven approaches, including linear models, random forests, gradient boosting, and neural networks, using data from the Lifelines COVID-19 cohort. Our results show that Neural Networks (NN) achieve the best performance in terms of MAPE, with predictions averaging 19\% error. Additionally, interpretability analysis reveals key factors such as loss of smell, headache, muscle pain, and vaccination timing as significant predictors, while chronic disease and gender are critical risk factors. These insights provide valuable guidance for understanding long COVID and developing targeted interventions.
Paper Structure (14 sections, 9 figures, 2 tables)

This paper contains 14 sections, 9 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Preprocessing and data analysis
  3. Data description
  4. Definition of Post COVID-19 symptoms intensity (PCSI)
  5. Data cleaning and processing
  6. Preliminary statistics
  7. Methodology and results
  8. Predictive Analysis and Benchmark Methodology
  9. Evaluated Methods
  10. Evaluation criteria
  11. Experimental setup
  12. Results
  13. Interpretation
  14. Discussion and Conclusion

Figures (9)

  • Figure 1: The overall process for defining Post COVID-19 symptom intensity (PCSI) using symptoms (symp) scores. All analyses were centered around the time of the first reported COVID-19 infection.
  • Figure 2: Data pre-processing and extraction strategy
  • Figure 3: Bar chart demonstrating the gender proportion with respect to low ($\leq 2$) and high ($\geq 3$) intensities of long COVID.
  • Figure 4: Chi-square test between vaccination and long COVID intensity. The test results indicate a significant relationship ($p < 0.05$) between vaccination and long COVID intensity.
  • Figure 5: Multiple Correspondence Analysis applied to static and vaccination variables. Each axis represents a component, with the corresponding variance explained. The Post-COVID (PC) intensity variable is discretized by rounding continuous values to the nearest integer (1, 2, 3, 4, and 5 modalities, shown in clear blue).
  • ...and 4 more figures