Published on Institute for Health Metrics and Evaluation (http://www.healthdata.org)

Home > Insights from IHME's latest COVID-19 model run

Insights from IHME's latest COVID-19 model run [1]

IHME Director and lead modeler, Dr. Christopher J.L. Murray, shares insights from the latest COVID-19 model run. Explore the latest forecasts: covid19.healthdata.org [2].

April 8, 2021

This transcript has been lightly edited for clarity 

In this week’s COVID-19 update from IHME, we see that the global epidemic is really not improving at all. New to our model is the expansion of estimates out to August 1 (they previously expanded to July 1). And like last week, there’s four key centers of ongoing transmission. 

First and foremost on the list of concerns is what’s unfolding in South Asia. There are rapid increases in cases Bangladesh, Pakistan and India. We’re seeing this in states such as Deli, where seroprevalence surveys suggest 65-70% have been infected. This makes us quite convinced that the outbreak in South Asia is driven by a so-called escape variant, which is where previous infection doesn’t necessarily protect you from the new variants (also known as the escape variants). And where vaccines are likely to be less effective, particularly the AstraZeneca vaccine. So the outbreak in India continues to increase very rapidly, and I think we can expect that to continue growing for quite some time.

(Read the COVID-19 policy briefing for Bangladesh [3], for India [4], and for Pakistan [5].)

The second area is the ongoing P1 epidemic in Brazil. New sequencing data was released from São Paulo this week that suggests P1 had spread into São Paulo much earlier than the data had previously expected, making the case that the general outbreak in the entire country is driven by the P1 variant. For those looking at the numbers, the seeming flattening of case numbers in the last few days around Easter is likely a data artifact. When there’s holidays, there’s usually lags in reporting.

Projections and Scenarios for Brazil:

  • In our reference scenario, which represents what we think is most likely to happen, our model projects 592,000 cumulative deaths on August 1. This represents 256,000 additional deaths from April 5 to August 1.
  • Daily deaths will peak at 3,480 on April 24, 2021.
  • We expect that 152.65 million will be vaccinated by August 1.
  • If universal mask coverage (95%) were attained in the next week, our model projects 60,000 fewer cumulative deaths compared to the reference scenario on August 1.
  • Under our worse scenario, our model projects 654,000 cumulative deaths on August 1, an additional 62,000 deaths compared to our reference scenario. 
  • By August 1, we project that 99,500 lives will be saved by the projected vaccine rollout.

(Read the COVID-19 policy briefing for Brazil [6], which includes detailed information about our projections.)

Third area of concern is in Europe, where cases are increasing slowly despite increasing social distancing mandates. Cases are increasing more rapidly in eastern countries where mask use is lower. The good news is that this holds out the prospect that further increases in vaccination in Europe, as long as mask use doesn’t drop too much, will really be the main strategy to remedy rising cases and deaths. 

We do see in the vaccine confidence data that in Eastern Europe particularly, vaccine confidence is really low. So there are countries where less than 30% of people are willing to be vaccinated. So there is a prospect for B.1.17 to continue expanding in those settings. Although, for the region overall, our reference scenario in our model projects that daily deaths will start to go down in May.

Projections and Scenarios for Europe: 

  • In our reference scenario, which represents what we think is most likely to happen, our model projects 1,566,000 cumulative deaths on August 1, 2021. This represents 303,000 additional deaths from April 5 to August 1.
  • Daily deaths will peak at about 5,620 in late April, and then start declining.
  • If universal mask coverage (95%) were attained in the next week, our model projects 55,000 fewer cumulative deaths compared to the reference scenario on August 1, 2021.
  • Under our worse scenario, in which mask use declines faster and mobility increases faster, our model projects 1,629,000 cumulative deaths on August 1, 2021, an additional 63,000 deaths compared to our reference scenario. Daily deaths remain above 1,000 on August 1 in this scenario.
  • By August 1, we project that 103,500 lives will be saved by the projected vaccine rollout. This does not include lives saved through vaccinations that have already been delivered.
  • Daily infections in the reference scenario drop below 100,000 in mid-June but remain above 200,000 through to August 1 in the worse scenario. 

(Read the COVID-19 policy briefing for Europe [7], which includes detailed information about our projections.)

The last area of focus or concern is in Canada and the United States. We’re seeing the largest outbreak in Michigan, that surge is really very impressive. Numbers are shooting up, and we’re finally seeing deaths creep up as well as cases and hospitalizations. Cases have gone up 500% in a month, and there’s no end in site so far. The challenging part for Michigan for us is that the increase there is not easily explained. There’s a lot of B.1.1.7 in Michigan, but mask use and mobility aren’t unusual in Michigan. And there’s more B.1.1.7, at least according to the data, in Maryland  – but a much bigger surge in Michigan. So they only way we can put all this together is to expect that the sequencing data is sort of out of sync in Michigan, and perhaps B.1.1.7 showed up there sooner, and there’s more transmission. If not, there’s some other factor that’s going on in Michigan.

(Read the COVID-19 policy briefing for Michigan [8], which includes detailed information about our projections.)

Across the border from Michigan, we’re seeing upticks in Ontario and Quebec [9] as well, and we’re starting to see bigger increases in transmission in adjacent states like Minnesota as well. Clearly, there’s a cluster of increasing transmission in that part of Canada and the US, and everybody is watching this very closely. The question is, is this a marker of what may happen in other parts of these countries?

Projections and Scenarios for the United States:

  • In our reference scenario, which represents what we think is most likely to happen, our model projects 619,000 cumulative deaths on August 1, 2021. This represents 64,000 additional deaths from April 5 to August 1. Daily deaths are expected to decline steadily until August 1.
  • If universal mask coverage (95%) were attained in the next week, our model projects 14,000 fewer cumulative deaths compared to the reference scenario on August 1, 2021.
  • Under our worse scenario, in which mask use declines more rapidly and mobility increases more quickly, our model projects 698,000 cumulative deaths on August 1, 2021, an additional 79,000 deaths compared to our reference scenario. In the worse scenario, daily deaths would increase until the beginning of June and then decline but remain over 750 a day on August 1.
  • By August 1, we project that 78,200 lives will be saved by the projected vaccine rollout.
  • At some point from April through August 1, nine states will have high or extreme stress on hospital beds. At some point from April through August 1, three states will have high or extreme stress on ICU capacity.

(Read the COVID-19 policy briefing for the United States [10], which includes detailed information about our projections.)

Overall, for the United States, it’s really this balance of the scale up vaccination, past infections, and how quickly do people re-open as to whether the variant spread of B.1.1.7 will tip us into a surge like Michigan elsewhere, or like what our reference scenario suggests, which is that things won’t be that bad and numbers will start to go down in May. Now it’s very easy, and our worse scenario demonstrates that, to see death numbers rising into June with only slightly lower mask use and slightly increase mobility. We’re on that knife edge of transmission where small changes can really shift us to an R-effective below 1 or R-effective over 1. This really puts enormous importance on people remaining cautious, not taking risks in terms of transmission, trying to get vaccinated as soon as possible and when they’re eligible, and hopefully we will win that race in terms of this spring surge. 

Additional Resources: 

  • COVID-19 policy briefings for over 200 world locations [11]
  • IHME publications related to COVID-19  [12]
  • World maps of COVID-19 mask use  [13]
  • COVID-19 model FAQs [14]

April 1, 2021

This transcript has been lightly edited for clarity 

In this week’s release of the IHME models for COVID-19, I think it’s important to look at four different areas of the world that are driving the global epidemic right now.

  1. First and most concerning is the P1-driven surge in Brazil and some neighboring states, such as Peru and Ecuador. In this week’s model, we have very substantially revised upwards our forecasts until July 1st, and this in part is due to much more in-depth analysis of what combinations of cross-variant immunity – between ancestral variants and P1 – and increased transmissibility of P1 can account for what we observed in Amazonas state in December and January. So when we put all that together, we find a much more alarming forecast for Brazil and neighboring countries.
     
  2.  The second great area of concern right now is the surge that we’re now seeing in South Asia, in Bangladesh, in Pakistan, and many states in India. We had been going through a very long period of declining cases and deaths, and now we are in a very rapid expansion in some places – Punjab is an example, and Bangladesh is having particularly sharp increases. Now when you look within India at a state like Delhi, where 65% of the population had been infected already – both from a seroprevalence survey and our own modeling – and now you’re seeing this big upsurge starting. It strongly suggests that the epidemic unfolding in South Asia is also related to one of the escape variants. Because sequencing is not as strong there, we’re not 100% sure, but there have been reports, for example, of mutations that look like escape variants that are accounting for this transmission.
     
  3. The third area of concern globally in terms of what’s happening is the continued expansion of daily cases in Europe, and now daily deaths going up at a much slower rate, but still going up. This is happening despite considerable vaccination underway in Europe and a very strong set of social distancing mandates in place. So the B.1.1.7 variant is driving increased transmission despite a lot of brakes on transmission that we would expect would be protecting them from continued expansion. In our models, despite this very concerning set of trends, we do think that the B.1.1.7-driven epidemic in Europe will peak sometime in late April and start to come down because, eventually, vaccination and declining seasonality will be enough to overwhelm an increased transmission from B.1.1.7. As we have forecasted for a number of weeks, we expect many countries – and we have now seen France do this – put tighter restrictions in place to help in putting the brakes on B.1.1.7.
     
  4. The last area of concern is in the United States and Canada, where we are seeing some places like Ontario, Michigan, Minnesota, and New Jersey with increasing case numbers. Some local communities, such as King County in Washington as well, where cases and hospitalizations are going up, and we expect deaths to follow soon as well despite vaccination. That is not as dramatic as these other three areas of concern, but certainly given what we’ve seen in Europe, this is enough of a concern that we should be monitoring this closely.

In all of these areas that we are speaking about, the three core strategies are still the same: 

Wherever you can, accelerate vaccination – of course that’s hard to say for places that don’t have access to vaccination – but for those that do, continue delivering vaccines as fast as possible. Next, maintaining mask use, even after vaccination. Finally, maintaining social distancing mandates and strengthening them when daily cases and hospitalizations begin to rise.

Additional Resources: 

  • COVID-19 policy briefings for over 200 world locations [11]
  • IHME publications related to COVID-19  [12]
  • World maps of COVID-19 mask use  [13]
  • COVID-19 model FAQs [14]

Links
[1] http://www.healthdata.org/covid/video/insights-ihmes-latest-covid-19-model-run
[2] https://covid19.healthdata.org/
[3] http://www.healthdata.org/sites/default/files/covid_briefs/161_briefing_Bangladesh.pdf
[4] http://www.healthdata.org/sites/default/files/files/Projects/COVID/2021/163_briefing_India_7.pdf
[5] http://www.healthdata.org/sites/default/files/covid_briefs/165_briefing_Pakistan.pdf
[6] http://www.healthdata.org/sites/default/files/covid_briefs/135_briefing_Brazil.pdf
[7] http://www.healthdata.org/sites/default/files/files/Projects/COVID/2021/44566_briefing_European_Region_12.pdf
[8] http://www.healthdata.org/sites/default/files/covid_briefs/545_briefing_Michigan.pdf
[9] http://www.healthdata.org/sites/default/files/covid_briefs/101_briefing_Canada.pdf
[10] http://www.healthdata.org/sites/default/files/files/Projects/COVID/2021/102_briefing_United_States_of_America_13.pdf
[11] http://www.healthdata.org/covid/updates
[12] http://www.healthdata.org/covid/publications
[13] https://www.healthdata.org/acting-data/covid-19-maps-mask-use
[14] http://www.healthdata.org/covid/faqs