Old updates for March-April
Update 30/4
Back on track. Yesterday’s uptick in new cases has been revised away. New cases in Andalucia only 44. Six regions below 10, two on the mainland. Can anybody explain why the government seems to be planning a uniform national relaxation, with the same rules in Madrid and Catalonia as here, though their infection rates are ten times as high? And why not wait another week or so to get to zero?
Update 29/4
Upticks, a big one in deaths. Two steps forward, one step back. Andalucia still below 100 a day, but getting to zero and staying there is hard.
Update 28/4
Decline continues. Steeper for new cases (-53% in the 5 days since the change of basis) than for new deaths (-18%). Murcia has a case again: complete suppression is very difficult.
Update 27/4
Small upticks, not significant unless they continue. Murcia back on zero new cases. Andalucia stays low. Good explanation by Simon Wren-Lewis of the costly British dithering.
Update 26/4
Not quite Hallelujah but unalloyed good news. New cases in Spain 1,729, the lowest since March 16 (second day of the lockdown) and a drop of 41% in one day. Andalucia at last below 100, joining 15 other regions. Only three are above 100 – Madrid, Catalonia, and Castile. New deaths at 288, the lowest since March 17, are 24% down in one day. Hospitalisations track reported cases closely (see pdf here, Figura 2), so hospital overload is almost certainly past. Unalloyed did I say? I take that back, total reported deaths are 23,190: certainly an undercount (as elsewhere) from deaths in care homes etc. that went untested, especially early on. The headline total will probably now end up under 25,000. I doubt if Sanchez can spin this total as a success, any more than Macron, Conte and Johnson can for theirs. Jacinda Ardern is unfortunately not available to replace any of them.
Update 25/4
ISCIII have not only changed the reporting basis on cases, but corrected all the old data as well, in a large and messy csv file. I’m not going to back-correct my table, so I’ve just marked the series as discontinuous, and dropped the now meaningless 7-day smoothed averages. This means we will need a few days more to see the trend re-emerge. Extremadura and Murcia report zero new cases, along with Ceuta. The spike in Andalucia has mysteriously vanished.
Update 24/4
ISCIII have changed the reporting basis for new cases to “PCR+”. No technical note supplied, but I assume this means tests only, not informed clinical guesses. The number of new cases in Spain has dropped 40%, but tripled in Andalucia. Effectively this means both are new series – I’ll try and fix the charts tomorrow to show this. Assuming no change in the reporting method for new deaths, these show an encouraging 17% one-day fall to a record low. Wait for a second day before celebrating.
Update 23/4
Highest total of new cases in Andalucia for a week. It’s still low, but should be going in the opposite direction.
Update 22/4
Two so-so days. The indicators seem stuck on another plateau. I don’t understand it. When R>1, cases and later deaths increase rapidly; when R<1, and it must be because we are below the peak, cases and deaths should drop equally fast. Possibilities include: changes in testing which are getting reporting closer to the true levels, making up for substantial undercounts at the peak; a gradual relaxation of lockdown compliance, restoring R to very closer to 1. If it’s the former, the fall should become steeper very soon. If it’s the latter, we are in for a long wait – or even a resurgence of the ofcoal has togo. It pandemic.
Update 20/4
The headline total of cases has passed 200,000, but as I’ve noted before, there is mounting evidence that all such totals are massive undercounts of the true incidence, including many asymptomatic carriers. More worrying is the slow pace of the decline. Since 13 regions out of 19 are now below 100 new cases a day, and our own Andalucia hovers just above, you wonder if there hasn’t been massive non-compliance with the lockdown in Madrid (+2079) and Catalonia (+1076).
New Zealand’s useful four-level scheme for exit.
Update 19/4
Total deaths now over 20,000, but new deaths (410) under half the peak (950). My new guess at total deaths is 23,000.
Update 18/4
The reverse of yesterday: nationally deaths up, cases flat. (ISCIII corrected yesterday’s numbers). The regional map gets odder. Andalucia now just above 100 new cases a day, about to join 10 other regions out of 19 below this benchmark. Micro-region Melilla had zero new cases. Those in Madrid sharply down at 953 but in Catalonia sharply up at 1,627.
Update 17/4
Cases flat, deaths sharply down. One of these is bad data, from noise or testing bias. I’m betting it’s the cases, but then I would say that. I was however wrong to suggest earlier that reported deaths are likely to be free from trend bias. That would not hold if you have significant numbers of untested deaths (at home or DOA at hospital). If test kits are in short supply they won’t be used on dead bodies; when they are plentiful, and medics are less overloaded, it’s worthwhile to get an accurate diagnosis of the cause(s) of death.
Nine Spanish regions out of 19 reported less than 100 new cases in the day. Madrid had 1,299.
Update 16/4
A disappointing day, with upticks all round. It looks as if we are stuck on a plateau. The regional disparities remain very wide. The low incidence in the warmer and drier south, from Valencia to Extremadura, could be explained by the climate: but not why Asturias is so much lower than Castilla/Leon.
Update 15/4
ISCIII have corrected yesterday’s numbers and the Andalucian case rate for the last two days is back to the normal range. Upticks in new cases, deaths declining on trend. 48% of the national cumulative cases are in just two regions, Madrid and Catalonia, with just 33% of the national population. Maybe arguing about politics or football in crowded cafés and meeting halls is not good for you?
Update 14/4
New reported cases in Andalucia have crashed to only 5! Too good to be true IMHO. Extrapolating the average decline over the last 5 days, you would expect about 150. This is still good progress. Wait and see. For Spain: trend decline in new cases, uptick in deaths.
Update 13/4
Blips over, all three indicators back on a nice trend decline. New cases in Andalucia are back below 200, and this time I think it’s for good. The UK has taken over from Spain and Italy as the European hotspot.
Update 12/4 – Easter Sunday
I had of course hoped for good news today, but it’s the usual mixed bag. New cases in Spain (whatever that means) are down 14% and now stand at 45% of the peak. But new deaths, and new cases in Andalucia, have both ticked up. General conclusion: the slow decline continues.
{sermonette} Suppose new deaths had been 400, as they will be someday later this week. I would have said, and will say: good news. But each of those 400 represents a human tragedy, each its own little black bubble of misery and loss. The thing is, both viewpoints can be true. Most of us are still untouched by personal loss and grief, though we may know people who are, and can offer them our sympathy and support. For most, the task is simply to keep fighting and accept the small burden of confinement in good spirits. This includes arming ourselves with solid information about the progress of the war. I write this as a small contribution to that fight, as well as to maintain my own morale. So I don’t apologise for my data-driven good cheer.{/sermonette}
Quarantine in Spain has been extended to April 26, under current rules, except for a relaxation of the work suspension for non-essential workers. Sanchez warns that a further 15-day extension will probably be needed, under conditions to be worked out. This is being rather pessimistic. If you take a peak of March 31 (new cases was one day earlier, new deaths one day later), and assume the downslope is a mirror image of the upslope, that brings us back to 5 March. On that day total reported cases stood at 198 and the total deaths at 1. It should be possible to switch at that point to an Asian strategy of Test/Track/Targeted isolation, the Taiwanese three Ts, along with moderate social distancing (say 1 metre spacing, no big sporting or entertainment events). The big decision is when to reopen schools. Of course, this all depends on the speed of the fall. There’s no a priori reason why it should be slow. This virus needs human hosts: in the open, it inactivates after a few hours or days. This survey of airborne microfauna gives the typical half-life of a robust virus in an aerosol as one hour. It’s that easy to break the transmission chain if you try.
Update 10/4
After a few poor days, the good news resumes. New deaths are 35% below the peak. New cases remain concentrated in Madrid and surrounding regions. “Again, in spite of that, we call this Friday good.” T.S. Eliot, East Coker.
Update 8/4
Uptick continues for a second day, and is more worrying. Total deaths have passed 14,500 – so much for the optimistic projection of 13,000 I made on 30 March, which I later revised to a still optimistic 17,000. There is no reason to change the call that the peak has passed, but the decline is slower than we all hoped. The gap between regions has surprisingly widened (see map and table on the ISCIII page): new cases in Andalucia (new Prussian motto: Ordnung Muẞ Sein) are a third of those in tiny Rioja. [My mistake, I looked at the 14-day number thinkng it was the 1-day. Sorry. The point holds that Andalucia is low.] It does not take many dickheads to block progress.
Update 7/4
Uptick in new cases in Spain, but for the reasons I’ve given we should not worry much about this. Also, and a bit more worrying, in new deaths, but it is a noisier series. The decline in new cases in Andalucia is large, for the second day running. New cases are the lowest they have been for a fortnight.
Update 6/4
The decline continues – striking in new cases, for both Andalucia and Spain; less so for new deaths.
We have to be cautious about relying on the numbers of cases – deaths are certain, and the number of misattributions of cause will be low and random. One problem is the lack of a time lag between the peaks in new cases (March 30) and new deaths (March 31). It takes 1-2 weeks between serious symptoms and death, so you would expect a lag. The peak in new deaths was 16 days after the start of the lockdown. This fits the expected reduction in R at that point, and in real new cases a week later. Much of the recent growth in reported cases may well be an artefact of better testing. This adds to the well-known biases from asymptomatic and mild cases who never get tested, and the 25% - 30% false negative rate in the tests themselves. (Luckily false positives are rare, well under 1% - we know this from Iceland, which tested a large random sample of its small population and got a 1% positive rate, including most of the true cases.) The sharpness of today’s fall in new cases in Spain would be nice if true, but we can’t be sure.
Update 5/4
It’s definite: the peak is behind us! For three successive days, new cases and new deaths in Spain have fallen; yesterday by substantial percentages (14% and 17%). For Andalucia it’s only three out of the last four days, but the pattern is clear.
Could this be reversed? Possible but very unlikely. 130,000 men, women and children in Spain have had Covid-19 with symptoms serious enough to go to a doctor or hospital, and 12,000 have died. That’s a lot of human beings. To a statistician, it’s also a huge sample and a lot of data points. The downwards trend is not noise. It could be reversed if people stopped distancing. That’s a real risk, but not just yet IMHO. How long before the lockdown can be safely relaxed? New cases passed 100 on 3 March. So IF THE CURVE IS SYMMETRICAL it’s the end of April at earliest before we pass this benchmark on the way down. If the government can set up Korean-style testing/tracing, it could be safe to relax a bit before. Meanwhile, stay vigilant.
Again assuming symmetry, and a peak on March 30-31, total cases will end up around 200,000 and total deaths around 17,000. I hope the deaths will come out lower – possible as ICU overload and the ventilator shortage will end soon, and the exhausted doctors and nurses (clap) must be learning how to treat better. Italy has passed the peak too at about 15,000 cumulative deaths, so may end up with 30,000. Spain is doing somewhat better – pro rata with Italy, deaths would be 21,000.
Update 4/4
Looking pretty good today. New cases and new deaths down by significant numbers in Spain. New cases in Andalucia are up, but still below the peak. Note on timing: ISCIII say they update as of 20h on the reference day. I think that means that’s when they close the reporting window. The data are published on day + 1 around noon. That’s when I update this tracker.
Update 3/4
Plateau in Spain continues with downticks if you look hard. Bigger downtick in Andalucia. Half of today’s new cases are in just two highly urbanised regions, Madrid and Catalonia. The lockdown does seem to have prevented a massive expansion in other regions, including ours. I guess relative timing rather than mobility of carriers. Regions that locked down with lower cases stayed that way.
Update 1/4
Oof, the spike yesterday in total cases has not been repeated and numbers have fallen back to the plateau. I was premature in calling a peak – I’ll wait for 3 consecutive days of falls before venturing this again.
The ISCIII map shows that Castilla-La Mancha is among the worst-hit regions. It consists of Madrid commuter towns and a vast and very low-density farming plain. Are the cases all in the former? If not, it’s an argument against thinking that modern rural life offers protection.
Update 31/3
A mixed bag. The headline total of new cases has hit a new record, and Spain will pass 100,000 cases tomorrow. But the uptick in Andalucia yesterday now indeed looks as if it was just a blip, and the trend has been flat for a week. New cases may be biased by better testing – the data on deaths is the better indicator, and it’s still flat.
Update 30/3
Slow decline continues in new cases in Spain and has started in new deaths. If 28 March was the peak in these, and the curve is perfectly symmetrical, the total deaths would come out at 13,000: much better than Italy. The UK target is keeping under 20,000, very similar adjusted for population. But there’s a nasty uptick in new cases in Andalucia. Let’s hope it’s just a blip.
Update 29/3
Unambiguous good news! New cases in Spain fell by 20% in one day, in Andalucia by 16%. We have definitely turned the corner IMHO. The timing suggests that this is the result of the lockdown imposed a fortnight ago. More surprisingly, new deaths seem to have gone flat – I would have expected them to keep rising for another week. Hats off to Spanish hospital staff and GPs– they are getting better at treatment, in spite of horrendous overload.
Update 28/3
I’ve added new cases and new deaths to the table and charts. This is better for spotting turning points. It definitely looks as if new cases in Spain have peaked. If the downslope is similar to the upslope, this suggests a total caseload of around 140,000. The apparent death rate is very high, perhaps because only hospitalized cases may be reported - 20% if you set new deaths against new cases 1 week back. Total deaths would then be of the order of 30,000. That’s guesswork of course. But it looks as if Spain won’t face a hecatomb of hundreds of thousands.
Update 27/3
Have I been doing this all wrong? The Spanish government announced on 21 March that it had bought 640,000 Covid testing kits. This plainly reflects a decision to expand testing beyond hospital settings. This is going to catch an increasing number of mild and asymptomatic cases among key workers and traced contacts. It’s the right policy to cut transmission, but it makes recently reported cases a less reliable indicator of the underlying trends. The bias is in one direction: the trend is better than the indicator suggests.
There is nothing I can do about this, but I have added a chart of deaths. This is the hardest indicator of all. It does however lag. Lancet article: “Recently, WHO reported that the time between symptom onset and death ranged from about 2 weeks to 8 weeks”. Add this to the 5-14 day incubation period, and the lockdown won’t show up in death rates before April.
But the slowdown continues! The 1-day growth rate in reported cases in Spain is down to 14%, and 11% in Andalucia. New daily cases are DOWN from 8,758 to 7,871: we may have seen the first peak. (But I Am Not An Epidemiologist). The 7-day GR gets rid of the noise, and it’s inching steadily down.
Update 25/3 There is no longer any difference in growth rate between Spain as a whole and Andalucia. The Spanish has converged on the lower Andalucian one. Both are now just below 20% in the 7-day average. The 1-day GR is the same, which is disappointing.
Update 23/3 Slowdown continues. 7-day growth rate now <30% for Spain, <20% for Andalucia. Notes/sources on this page restored. ISCIII have added new data to their page on hospitalisations and new cases.
Update 21/3 Definite signs of a limited slowdown in the 7-day rates. This cannot be due to the lockdown starting 15 March, but must come from voluntary community action before.