Current Latest News
Technology

Researchers use predictive models to calculate more accurate COVID-19 case fatality ratios

Researchers use predictive models to calculate more accurate COVID-19 case fatality ratios

A preprint study published by researchers at the Southern California Eye Institute, CHA Hollywood Presbyterian Medical Center, and the University of California, Berkeley investigated the biases that exist in COVID-19 case fatality ratios between groups and countries, like time- and severity-dependent reporting of cases and time-lags in outcomes. They found that randomized data obtained through contact tracing might help to mitigate many of the confounding factors and that certain biases, like time-varying reporting rates, could be corrected for among fatal and non-fatal cases.

Case fatality ratios, or CFR, often guide policy decisions about distributing resources during public health crises. CFR average out the effects of medical care, age, geography, genetics, and more, but they’re imperfect in that they’re susceptible to under-ascertainment of mild cases, time lags, interventions, group characteristics, imperfect reporting and attribution, and other biases. Indeed, one recent paper pegged the under-ascertainment of mild COVID-19 cases at 50- to 80-fold based on serological testing.

In a similar vein, the coauthors of this paper attempted to mitigate some of the confounders by studying the processes relating CFR to populations. Among other models, they developed a parametric algorithm that accounted for two biases in particular: time-varying reporting and disease-delayed mortality. Similar models have been used for CFR estimation of influenza because they assume all non-fatal cases eventually recover, which eliminates the need to use a time series of recoveries.

All models were run in Hypernet Labs‘ Galileo, a platform-agnostic app that expedites code deployment for compute-intensive work, across a cluster of machines provided by Hypernet. (The startup began offering virologists, epidemiologists, and public health officials access to its resources shortly after the pandemic began.) Each was applied to open source COVID-19 data from Johns Hopkins, calculating the corrected relative CFRs for combinations of nine countries — Austria, South Korea, Switzerland, Germany, Italy, Spain, Belgium, Iran, and the U.K. — for April 2 to April 16.

VB Transform 2020 Online – July 15-17. Join leading AI executives: Register for the free livestream.

The researchers found that the relative CFR for Spain (which has 27,940 recorded fatalities as of publication time) versus South Korea (264), for instance, was lower than estimated at 30.27. Comparing the CFR for the U.K. (36,042) and Italy (32,486), it was 0.68.

To minimize bias in CFR going forward, the coauthors advocate contact tracing, which they say might expand the sampling frame to include a larger portion of populations, specifically mild cases. (Contact tracing refers to reaching out to all individuals recently exposed to a known COVID-19 positive individual, removing them from circulation, and monitoring their health.) Furthermore, they suggest all contacts be tested for COVID-19 one incubation period after exposure, regardless of whether or not they’re symptomatic.

“The number of data points gleaned from this strategy will be lower than the number of data points from surveillance data,” they wrote. “[T]here is no issue with time lag, since these cases can be tracked systematically.”

Source link

Related posts

Leave a Comment

* By using this form you agree with the storage and handling of your data by this website.

This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More

Privacy & Cookies Policy