Skip to content


Links for 2020-11-03

  • COVID-19 IFR is estimated at 0.97%

    Florian Krammer on Twitter: “Our NYC serosurvey paper is now out in Nature: if extrapolated to the NYC population we found [more than] 1.7 million infected and IFR at 0.97” That’s high! Nearly 1 in 100.

    (tags: ifr covid-19 florian-krammer mortality deaths pandemics)

  • Fault in NHS Covid app meant thousands at risk did not quarantine

    Somebody, somewhere, will have died needlessly due to this bug.

    The root of the error, the Guardian has learned, was a decision to incorporate a measure of “infectiousness” into the app’s code. While the app was undergoing testing in the Isle of Wight, it used a simple metric that recommended isolation for anyone who had been in contact – closer than 2 metres – with a potentially infectious person for 15 minutes or more in a single day. But shortly before the app was launched nationally, it was updated to account for the fact that people are most infectious shortly after their symptoms show. The maths was changed so that people outside that period of peak infectiousness counted for just two-fifths of the risk. Since that meant the overall score was likely to be lower, the intention was to reduce the risk threshold correspondingly to ensure that someone of maximum infectiousness would need just three minutes of contact before they triggered an alert. But that change never happened, and as a result, users were only told to isolate if they had spent 15 minutes close to a very infectious person, or nearly 40 minutes near someone who was pre-symptomatic but still thought to be shedding the virus. The error was only discovered when a new version of the contact-tracing app, which can better account for exposures at mid-range (over a metre away) was created. The unfeasibly high risk score also explained another problem plaguing the app: “ghost notifications” warning users that they may have been exposed to someone with Covid, but which never resulted in advice to isolate. The app’s initial advice to users was that these notifications could be safely ignored, since they reflected a contact below the risk threshold; now that the NHS risk threshold is known to have been artificially low, one insider said, it is likely that the vast majority of those ghost notifications should in fact have been advice to self-isolate.

    (tags: bluetooth nhs bugs failure ble covid-19 uk)

Comments closed