TL;DR
This paper introduces and validates a multivariable risk prediction model to identify inflammatory bowel disease (IBD) patients at higher risk for severe COVID-19 outcomes. The model, available as a free online calculator, demonstrates excellent predictive power and aims to aid clinicians in patient counseling and risk stratification.
What Problem Does This Paper Address?
The COVID-19 pandemic highlighted the need for tools to predict severe outcomes, especially in vulnerable populations. Patients with Inflammatory Bowel Disease (IBD) are often on immunosuppressive therapies, which could potentially increase their risk of severe COVID-19. However, there was a lack of specific, validated risk prediction models tailored to this patient group that could help healthcare providers assess individual risk and guide clinical decisions. This paper aims to fill that gap by developing and validating such a model.
Methods
The researchers developed penalized multivariable logistic regression models. They used a training dataset of 2009 IBD patients with COVID-19 and validated the models on an independent test dataset of 700 patients. To handle missing data, they employed multiple imputation by chained equations (MICE), creating 30 imputed datasets. Model performance was assessed using Receiver Operator Characteristic (ROC) curves and calculating the Area Under the Curve (AUC). The study also developed a publicly accessible online risk calculator based on these models.
Datasets
The study utilized data from the Surveillance Epidemiology of Coronavirus Under Research Exclusion for Inflammatory Bowel Disease (SECURE-IBD) database. This is an international voluntary registry of IBD patients with COVID-19. The data was collected from March 13, 2020, to October 20, 2020. The training set comprised 85% of the data available as of September 15, 2020 (2009 cases), and the test set included the remaining 15% of data from September 15, plus all cases reported between September 16 and October 20, 2020 (700 cases).
Key Insights and Ideas
- IBD-Specific Risk Factors Matter: The study confirms that IBD patients have unique risk profiles for COVID-19, necessitating tailored prediction models beyond general population tools.
- Online Tools Facilitate Clinical Practice: The development of a free, user-friendly online risk calculator is a significant contribution, making complex predictive models accessible to clinicians for real-time patient discussions.
- Corticosteroids are a Major Risk Driver: Oral corticosteroid use emerged as a particularly strong predictor of adverse COVID-19 outcomes in this population, underscoring the importance of medication management.
- Biologics May Offer Protection: Conversely, biologic therapies were associated with a lower risk of adverse outcomes, suggesting a potential protective effect.
Key Results
The developed multivariable risk models demonstrated excellent discrimination:
- Hospitalization+ (hospitalization, ICU, mechanical ventilation, or death): Test set AUC of 0.79 (95% CI: 0.75, 0.83)
- ICU+ (ICU admission, mechanical ventilation, or death): Test set AUC of 0.88 (95% CI: 0.82, 0.95)
- Death: Test set AUC of 0.94 (95% CI: 0.89, 0.99)
Key predictors for death included older age, male gender, and comorbidities. Corticosteroid use was a significant predictor for all three outcomes, with a daily dose equivalent of 40mg of prednisone associated with a ten-fold increase in the odds of death. Biologic therapies, particularly TNF inhibitors, were associated with a lower risk of death.
Limitations and Future Work
The authors acknowledge several limitations:
- Voluntary Registry: The data comes from a voluntary registry, meaning the sampling process is not random, and reported cases may underrepresent both low-risk asymptomatic cases and severely ill patients who might not be reported.
- Associational, Not Causal: The risk calculator provides associational predictions, and users should not interpret it for "what-if" scenarios regarding medication changes, as these are causal questions.
- Limited Social Determinants of Health: The registry does not collect granular data on many social determinants of health, which can influence health outcomes.
- Insurance Status Not Collected: For US patients, insurance status, which can affect healthcare access, was not collected.
- No Comparison to General Population: The study cannot directly compare the risk of adverse COVID-19 outcomes in IBD patients to the general population.
Future work could involve adapting this model for risk stratification in other chronic disease populations. The authors also emphasize the importance of this tool for informing public health efforts and rational vaccine allocation.
Summary updated