Por favor, use este identificador para citar o enlazar este ítem:
http://conacyt.repositorioinstitucional.mx/jspui/handle/1000/4539| On the Bias Arising from Relative Time Lag in COVID-19 Case Fatality Rate Estimation | |
| Anastasios Nikolas Angelopoulos. Reese Pathak. Rohit Varma. Michael I. Jordan. | |
| Acceso Abierto | |
| Atribución-NoComercial-SinDerivadas | |
| https://arxiv.org/pdf/2003.08592v4.pdf | |
| The relative CFRs between groups and countries are key ratios that guide policy decisions regarding scarce medical resource allocation. In the middle of an active outbreak, estimating this measure involves correcting for time- and severity- dependent reporting of cases as well as time-lags in observed patient outcomes. In this work, we argue that we must make up for lost information about time when estimating the relative CFR: without inferring the time-dependent balance between reporting rates of fatal and non-fatal cases, CFR estimators can perform badly. To make this argument rigorous, we carry out a theoretical analysis of some current estimators of CFR. We then adapt a previously developed method -- based on the well known expectation-maximization (EM) technique -- for COVID-19 reporting. Our analysis is supplemented by numerical results and an open-source implementation https://github.com/aangelopoulos/cfr-covid-19 . This should enable epidemiologists and other analysts to fit likelihood-based models similar to the ones we propose as remedies for the biased nature of naive CFR estimates, permitting more accurate planning of medical resource distribution. | |
| arxiv.org | |
| 2020 | |
| Artículo | |
| https://arxiv.org/pdf/2003.08592v4.pdf | |
| Inglés | |
| VIRUS RESPIRATORIOS | |
| Aparece en las colecciones: | Artículos científicos |
Cargar archivos:
| Fichero | Tamaño | Formato | |
|---|---|---|---|
| 1106754.pdf | 433.07 kB | Adobe PDF | Visualizar/Abrir |