Prognostic Models for COVID-19 in Low- and Middle-Income Countries
Author Information
Author(s): Johanna A. A. Damen, Banafsheh Arshi, Maarten van Smeden, Silvia Bertagnolio, Janet V. Diaz, Ronaldo Silva, So Soe Thwin, Wynants Laure, Karel G. M. Moons
Primary Institution: Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht University
Hypothesis
Can existing prognostic models for predicting mortality or ICU admission in COVID-19 patients be validated in low- and middle-income countries?
Conclusion
Only a few COVID-19 prognostic models could be validated for use in low- and middle-income countries, and they showed poor calibration despite good discrimination.
Supporting Evidence
- Out of 144 eligible models, 140 were excluded due to high risk of bias or insufficient model description.
- Three models showed good discrimination for predicting in-hospital mortality with AUCs ranging from 0.76 to 0.84.
- All models exhibited signs of miscalibration and overfitting.
Takeaway
This study looked at how well certain models predict if COVID-19 patients will get worse in hospitals in poorer countries. They found that while the models can tell who might be in trouble, they often get it wrong.
Methodology
The study used a meta-analysis approach to evaluate the performance of existing prognostic models using data from the WHO Global Clinical Platform.
Potential Biases
High risk of bias was noted in many models due to incomplete reporting and unavailability of predictors.
Limitations
The models showed severe miscalibration and were developed using data from the early pandemic, which may not reflect current conditions.
Participant Demographics
The median age of participants was 43 years, with 58.8% being men, and data was collected from 9 low- and middle-income countries.
Statistical Information
Confidence Interval
95% CI 0.71–0.81
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website