Antibody tests, early lockdown advice and European deaths

Episode Summary

Executive Summary: This episode examines how early UK COVID-19 policy relied on overly slow epidemic-doubling assumptions, likely delaying lockdown, and then turns to the misuse of statistics in comparing UK and EU deaths and in marketing antibody tests as “100% accurate.” It also revisits government testing targets, showing how definitions were stretched to claim progress without equivalent published results.

Main Topics: Early COVID-19 doubling-time estimates and lockdown timing (Priority: 5/5): The episode investigates why government advisors believed UK cases were doubling every 5-7 days in March 2020, despite real-world data suggesting a much faster spread. It argues the slow estimate may have contributed to delayed suppression measures. How modelling consensus may have misled SAGE (Priority: 5/5): The discussion explains that SPI-M and SAGE relied heavily on early Imperial College modelling using Wuhan-derived assumptions for R0 and generation time, which were not properly calibrated to UK data and may have underweighted other models. Comparing UK and EU COVID death counts (Priority: 4/5): The episode critiques viral graphics claiming the UK had more daily deaths than the whole EU, emphasizing that reported deaths on a given day are not the same as deaths that occurred that day and that country reporting rules differ substantially. Spain’s death-reporting footnote problem (Priority: 4/5): A detailed example shows Spain temporarily reporting zero deaths because only deaths registered the day before were counted, illustrating how footnotes and registration delays can severely distort international comparisons. Antibody tests and the misuse of '100% accurate' claims (Priority: 5/5): The programme dissects media and company claims that Roche and Abbott antibody tests were 100% accurate, showing that these claims depended on tiny subgroups and did not reflect overall test performance. Government testing targets and definition changes (Priority: 4/5): The episode returns to UK testing statistics, explaining how antibody tests and tests posted out were included in headline counts, allowing the government to claim 200,000 tests a day without publishing equivalent completed-test figures.

Key Arguments: The UK was not four weeks behind Italy in March 2020; the evidence suggests it was closer to two weeks, and perhaps faster than the government believed even by mid-March. The government and SAGE appeared to rely on a doubling time of about five to seven days, but UK data available by March 14 already suggested a much shorter, roughly three-day doubling period. The Imperial College model used early Wuhan estimates for R0 and generation time that were too low/long respectively, producing an inflated doubling time and underestimating UK growth. Model consensus can fail when some modelling groups dominate and plausible alternative estimates do not get passed up the chain to SAGE and ministers. Headlines comparing same-day reported deaths across countries are misleading because reporting delays, backdating, and national counting rules differ widely. Spain’s temporary zero-death reports were an artifact of a strict daily-registration rule, not a true absence of deaths. Claims that antibody tests are 100% accurate were based on cherry-picked subgroups, small sample sizes, and time-sliced analyses rather than the full study populations. A positive antibody result is usually reliable, but negative results can miss infections, especially depending on when the test is taken after symptom onset. The UK government’s testing totals were boosted by counting tests sent out and by including antibody tests, not just completed swab tests. Without transparent published data on completed tests and returns, headline testing figures are of limited value for assessing real capacity or performance.

Data Points: UK coronavirus deaths vs press conference timing: 1,000 deaths announced 16 days later; actually passed on 24 March, 12 days after 12 March press conference - Used to show the outbreak was farther along than officials said Italy comparison claim: about 4 weeks behind Italy - Sir Patrick Vallance’s statement at the 12 March press conference Observed UK rise in deaths: 10 deaths on 12 March to 81 deaths by 16 March - Evidence cited for much faster-than-assumed epidemic growth Doubling time used by government: 5 to 6 days - Boris Johnson’s 16 March statement based on SAGE advice Alternative estimated doubling time: around 3 days - Kit Yates and calibrated modelling using available UK data Imperial College R0 estimate: about 2.4 - Based on early Wuhan data in the 16 March paper Generation time used in Imperial model: 6.5 days - Combined with R0 to imply a slower doubling time Government daily deaths comparison graphic: UK 359 vs EU 314 - Twitter image based on a Newsnight graphic that lacked important caveats Spain reported deaths on two days: 0 new deaths on 1 and 2 June - Result of restrictive counting based on the day before report Spain reporting rule: Only deaths where the date of death is the day before the report are added daily - Footnote explained by FT’s John Byrne Murdoch UK and EU 7-day average deaths: UK about 25 deaths behind EU in latest days discussed - A smoother comparison method that avoids daily reporting noise Roche study COVID-positive samples: 93 samples; 15 missed - Shows the test was not universally 100% sensitive Roche late-symptom subgroup: 8 of 8 positive over 40 days after symptoms - Basis for a 100% accuracy headline Roche leaflet study subgroup: 29 samples - Key result in instructions-for-use leaflet used to support accuracy claims Abbott late-testing threshold: 17 days after symptoms - Company data slicing produced no false negatives in that narrower subgroup Government antibody testing volume: around 40,000 a day - By early June, enough to push total tests over 200,000 on some days Government posted tests: about 90,000 on some days - Included as tests ‘sent out’ rather than completed Care home deadline: 6 June - Postal tests surged before the target to send tests to every care home

Pivotal Quotes: "On the curve, we're maybe four weeks or so behind in terms of the scale of the outbreak." — Sir Patrick Vallance: 12 March coronavirus press conference, later challenged by UK data "According to Sage, it looks as though we're now approaching the fast growth part of the upward curve, and without drastic action, cases could double every five or six days." — Boris Johnson: 16 March statement reflecting the government’s official understanding of epidemic growth "The idea that these will always come out with 100% accuracy is not tenable." — John Deeks: Critique of media/company claims about antibody test performance

Implications: The episode shows how bad assumptions, selective data, and sloppy metrics can shape policy and public understanding. For future crises, transparent modelling, calibrated local data, and careful statistical definitions are essential to avoid misleading decisions and headlines.