I recently came across this Venn diagram on a blog that is new to me, Blueberry Town, and thought it captured my own thoughts well. The author also had a separate post entitled, Tip-Toeing In A Dark House, which expressed similar ideas, though better stated and more comprehensive, to my recent post, I Don't Know And You Don't Either. Below is a lengthy excerpt and I urge you to read the entire post.----------------------------------------------------------------------------------------------
Imagine you wake up in a house you have never been in before, and it is pitch dark, and there is no electricity. You need to find your way to the bathroom, and then the kitchen, and eventually to the outside. Your only option is to tip toe – there might be a Lego block, or you might fall down the stairs – and grope the walls and furniture, and eventually make your way as you gather more tactile information in little bits, with still no understanding of the bigger picture.
Everyone in the world right now, from the most sage public health expert to the wisest head of state to the anxious citizen shopping for groceries, is tip-toeing through that dark unfamiliar house.
Among the things that we do not know to any level of certainty:
- Prevalence. We not only do not know how many current cases there have been in any given population, we don’t know how many there have been. Sure, certain experienced countries and homogeneous micro-states probably have better information than most other countries, but even when we have done extensive antibody testing, such as in California’s Santa Clara County, the extrapolations for the broader population have a huge margin of error.
- Transmission. We know that an infected person in contact or close proximity to another person can transmit Covid-19. Other than that, there is a great deal we do not know, including whether the amount of virus at exposure makes a big difference, or whether strolling on a beach in high wind poses any risk at all.
- R0, pronounced “R naught.” See here for the concept. R0 is the “basic reproduction rate” of the virus, which can be thought of as “the expected number of cases directly generated by one case in a population where all individuals are susceptible to infection.” The estimated R0 for Covid-19 has a massive proposed range, between 1.4 and 5.7, per the obsessively updated Wikipedia page. If you play with any of the epidemiological models, you quickly learn that there is tremendous leverage, and therefore uncertainty, in this number.
- Rt. Rt is the actual transmission rate per case. In a nutshell, if “t” is above 1 then we are suffering an increasing rate of infection, and if it is below 1 we are gaining against the rate of infection. Click around this site for more, including purported Rt for each American state. The problem, of course, is that we can only measure Rt for cases we have confirmed, and we have very little idea how many cases there are that haven’t been confirmed.
- The mortality rate from Covid-19. Notwithstanding evidence of excess mortality and the problem of people who die suspiciously without being tested, we probably know the rough number of Covid-19 as well as we know anything. What we do not know is the percentage of infected people who die, and under what circumstances, because we really have no clue how many people are, or have been, infected.
- The mechanism and circumstances by which Covid-19 kills people. This is an excellent article on what we do and do not know about Covid-19’s mechanism of action, but only read it if you have a robust emotional state. There remains lots of mystery.
- Who is at risk of dying, and why? We know that old age and various co-morbidities, such as diabetes, hypertension, compromised lungs, and autoimmune diseases are correlated with much higher mortality rates than for young and healthy people. We don’t know why some otherwise healthy people crash and become very ill.
- Treatments. As the world’s doctors treat more patients and share their knowledge, we are learning more every day about how to respond to patients sick with Covid-19. In addition to practical interventional medicine, there are around 800 clinical trials around the world studying the safety and efficacy of old and new therapies. All this progress is great for people getting sick later rather than sooner, but we still have no clue whether we will have a robust cure in the near future.
- Immunity, antibodies, and the testing therefor. We do not know the extent to which survivors of Covid-19 develop antibodies, how long they persist, whether they usually confer substantial immunity or only do so sometimes, how long that immunity persists — weeks, months, years? — and whether a test for antibodies, even if it works to do that much, is therefore a test for immunity.
- Vaccines. There is a global scramble to develop a vaccine and test it sufficiently to green-light mass production and administration. Vaccines are moving along the developmental and regulatory process more quickly than under normal circumstances, probably in part because we are taking risks to accelerate their development. A vaccine would be like a light coming on in the dark house, showing us exactly the path to the front door. If we knew that the mortality rate from Covid-19 were very high, it might be wise to take even more risks to get a vaccine. But what if the mortality rate isn’t very high because the disease is far more prevalent than we now know? Do we want to take the risk of vaccinating the entire population with a short-cut vaccine if the mortality rate from Covid-19 is only, say, twice that of the seasonal flu?
- Oh. And the models. There has been a lot of controversy about the various epidemiological models and their predictions under various social-distancing measures. “Listen to the experts!” being both the ultimate admonition and excuse for our times, models are necessary — no, they are not crackpot conspiracies — so that people in authority have some rationale for making decisions that carry the force of law. However, because those models need to embed assumptions about all of the variables in the foregoing bullets that we know little or nothing about, they inherently have extremely wide margins of error. In other words, they barely improve the likelihood that those decisions will ultimately prove to be the optimal ones, even if models help us choose a direction. As we pick our way through the dark house, the models are, perhaps, a flash of light, maybe a single distant lightning strike, that plays through the windows for a fraction of a second. We might see the top of the staircase, but we won’t see a toy left on the third step from the top, or the bad guy lurking with his cudgel.
Many of the measurements we are using today to give us a handle on where we are headed in this pandemic are almost useless at this point. Confirmed cases are a terrible metric because they are so dependent on testing availability and strategies. I live in Arizona and the state has just announced it is loosening the criteria for testing and making it more readily available so we can expect the number of confirmed cases to increase but we will have no idea if that indicates the number of infected people is increasing or decreasing.
Deaths are a better metric since they are more certain of being counted but is a badly trailing indicator since the fatalities represent people who may have become infected a month earlier.
Two measurements that would give us a better feel for trends but are only sporadically available are hospitalizations and infections and deaths in long-term care facilities.
Hospitalizations (of both confirmed and suspected Covid cases) are also a lagging indicator but by 10-14 days from date of infection rather than a month as with deaths. It would also tell us more accurately about trends than positive test results. In Maricopa County (Phoenix metro) we are running at about 15 hospitalizations a day in a population of 4.5 million, which is slightly less than the peak from March 22 to April 9.
What is happening in our long-term facilities would give us a better handle on who and how we need to protect vulnerable populations and more precise information on relative risk of those under such care versus the over 65 group not in such facilities. Maricopa County also tracks this metric. Overall 60% of fatalities have been in long-term care and over the last week it is running closer to 75%. As an area resident, and someone who is over 65, it has been helpful to have the county data on hospitalizations and long-term care. The New York City Department of Health also is doing a fine job on detailed and useful metrics as are some other jurisdictions.
And we still don't know with great certainty what works to contain Covid and what doesn't. We know New York has had a rough time of it (123 deaths per 100,000 population), yet the other three big states, California, Texas, and Florida have done much to date despite charting very different courses. California (5 deaths per 100,000) went into lockdown mode very early, yet Florida (6 deaths per 100,000) went into it very late, and never to the extent of California. A month ago I thought Florida was headed for disaster between delaying social distancing and stricter measures; the largest elderly population of any state; letting spring break go ahead; and New Yorkers fleeing their state for Florida. Yet it did not happen, at least so far. And Texas (3 deaths per 100,000) is probably in the middle of the spectrum on state action between California and Florida. And, maybe in a month, the situation will have drastically deteriorated in one or more of those states.
Something else detracting from our ability to try to make sense of what is going on, or to more easily accept we don't know what is going on, is the obsession of many across the political spectrum from using their "priors" and triumphantly grabbing on to any bit of data reinforcing their political preferences without assessing the validity of such data by asking the most basic questions. The result is a lot of people end up being made stupider by such discussions.
No comments:
Post a Comment