Modrák et al., 2021
Disease progression of 213 patients hospitalized with COVID-19 in the Czech Republic in March-October 2020: An exploratory analysis
COVID-19
Trajectory
Hidden Markov Models
At a glance
- Objective
- To explore associations between COVID-19 treatments and patient outcomes.
- Related articles
- Currently none. This article may be related to other articles modeling disease trajectories over time.
- DOI
- DOI: https://doi.org/10.1371/journal.pone.0245103
Background
- Lack of knowledge about effective treatments for COVID-19
- Existing methods for predicting COVID-19 severity are at high risk of bias
Methods
- Convenience sample from 10 sites
- Data were available on daily resolution
- Multiverse analysis, which reports and compare different models to capture uncertainty about different models and its impact on the conclusions
- Among the models run is a Bayesian HMM using rates, restricted transitions, and terminal states implemented in
brms
with some added special sauce (for more detail, see the GitHub repo and a post on the Stan discourse)
Results
- Mostly inconclusive results
- Adjusted models suggest that effect of some candidate treatments is spurious
Conclusion
- Other studies on the analyzed treatments likely overestimated their effectiveness