Key insight is that if some control process keeps the geometric mean of R(t) near 1 one way or another, better it should happen via targeted measures than indiscriminately. Value is calculated without needing to trade off lives for livelihoods
The value of even seemingly unimpressive interventions, eg exposure notifications with modest adoption, can be large, in increasing the degree to which we return to normal while keeping R(t)~1. 4/5
Returns are smaller for endemic disease than for outbreaks, because preventing infection sometimes merely delays it and the accompanying immunity boost. But still significant, because it might also get you through to your next vaccine booster. Working on that next 5/5
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Many are torn between frustration and sympathy for @CDCgov. Given all the things CDC are failing to get done (eg rapid data dissemination and analysis), I am shocked by what they DO they spend time on. 🧵sharing observations from my interactions with CDC staff. 1/14
I was happy when CDC contacted me in Dec 2020 after this NYT story nytimes.com/2020/12/07/tec…, to urge rapid publication of the data supporting our claim of exposure notifications for up to 12% of transmission. The fact of CDC outreach got data release unblocked at @uarizona 2/14
We put a very informal preprint on github nearly immediately, and a more formal one on medrxiv soon after doi.org/10.1101/2021.0… . A CDC coauthor was added to shepherd it through @CDCMMWR 3/14
Thread on why the launch of @CovidWatch in Arizona yesterday is different from other launches. This GAEN app is superior re risk scoring, re rollout strategy, and re customizability (the last two related) blog.covidwatch.org/en/covid-watch… 1/10
Other apps try to reproduce 6 feet (or 1 meter or 2 meters) for 15 minutes. But how infectious someone is also varies ~10-fold as function of timing relative to the day their symptoms begin 2/10
@CovidWatch integrates timing, duration, and Bluetooth signal (a noisy correlate of distance) to estimate infection risk. 7 feet for 8 hours at most dangerous time is much riskier than 5 feet for 15 minutes at edge of infectious period
Longer viral shedding in asymptomatics, if true, would matter. With limited testing, it would mean very long quarantines. So I pulled the easily available data from this @NatureMedicine paper and tried to do a meta-analysis including other data 1/4
@mugecevik kindly pointed me to 3 other papers with similar comparisons for covid-19 (plus one for flu). None deposited their data. Two corresponding authors did not reply, the third refused, citing ethics. All I want is shedding durations
@JAMANetworkOpen's data sharing policy is risible. For clinical trials only, authors are required to state whether or not data will be available. How can such a journal have "Open" in its name!? jamanetwork.com/journals/jaman… 3/4