We Need Better Data!!

Any decision process is better if more of the most important data is available and is taken into account in the process. Today, we are making big decisions on responses to the Coronavirus quite blindly. Dr. Deborah Birx pointed out at the Tuesday, March 17, 2020, news conference that models that are leading to wildly different predictions, some of which are frightening, are based on legally unexamined and unchallenged assumptions. These predictions are being published or broadcast by mass media, but we really do not know enough about the validity of many modeling assumptions to give credence to these predictions.

Let's look at some of the big unknowns:

• How contagious is the disease?

We make a reasonable assumption that the disease is most contagious during its asymptomatic incubation period. What we do not know is the duration of that incubation period. Public health professionals are using 14 days as a guideline for quarantining of people exposed to the virus, because it is at the outer edge of the normal range of incubation periods, but the larger data studies on incubation periods are much less precise.

The latest study with the largest available retrospective data comes from the University of Massachusetts at Amherst. It finds that the median incubation period is a little over 5 days and that 97.5% of those infected will have noticeable symptoms within 11.5 days. The question of whether people will self-quarantine once they know or suspect they are infected is an important one in forecasting the probable spread of the disease. Yet we do not know the answer to this question with any degree of precision. The length of the incubation period affects the degree to which people expose others to the virus before they are aware of their contagiousness.

https://www.sciencedaily.com/releases/2020/03/200317175438.htm

Israel is enacting legislation that authorizes the government to use cell phone data for contact tracing. While that is something we may consider, the privacy implications of this kind of tactic to learn how many people an infected person has exposed to the virus are such that we are highly unlikely to do that here.

• What percentage of the population is infected with a serious form of the Coronavirus?

We do not know what percentage of the population is infected because our testing is deliberately skewed toward people with noticeable symptoms. There are estimates of between 10x-20x the known cases of people with the virus that have yet to be tested, most of whom may never be tested.

The reason that is relevant is that if we knew how many people had the virus and had the same number of deaths we have today, we might find that the true death rate could be as little as marginally more than a seasonal flu death rate or as large as the 1.7% we currently see in the United States. It makes a big difference to our planning if we are dealing with a death rate comparable to the SARS virus, which was significantly higher than a seasonal flu or with a seasonal flu death rate, which is relatively manageable without extraordinary measures.

Professor Neeraj Sood of the University of Southern California has recommended that the US do a nationwide random sample testing based on the demographics of the entire population, not just those presenting themselves with symptoms. I agree. That sample should be large enough to have a variability rate of no more than a fraction of a percent.

• We should know far more about who is dying or progressing to a serious level of the illness.

We believe that the virus is most deadly to old people with compromised immune systems, but we do not have granular enough data to understand which chronic or complex disease patterns pose the highest risks. It would seem logical that anyone with a chronic disease that involves severely compromised lung capacity would be at very high risk. Accordingly, it would make most sense that smokers with emphysema should be targeted for testing and monitored more closely than people diagnosed with hypertension who are taking their medications. However, we do not have enough data to prioritize the large older chronic disease population to do a true risk assessment. Dr. Birx also discussed that at the March 17, 2020, news conference.

• The question of whether the virus "caused" someone's death is important in terms of how "cause" is defined.

My 98-year-old father-in-law died seven weeks ago in his sleep. The cause of death was listed as "pneumonia." Even before he had what, to any of us, would have been a relatively minor bronchial infection, his life expectancy could have been measured in months or even weeks. The virus he had (not Coronavirus, but that is not relevant for this point) hastened his death by a few weeks or, at most, a few months. Legally and from a reporting standpoint, the cause of death designation was accurate, but very misleading.

However, the question of "cause of death" in any of these cases is extremely important. Are we going to take extreme measures to isolate 300 million people, destroy their ability to earn a living, separate them from their elderly loved ones, and have all the other horrible consequences of this "social isolation" strategy if we find that the virus, on average, shortened the lives of those who died by a few weeks or a small number of months?

We are focusing heavily on "contagiousness," as opposed to whether the virus is lethal for the vast majority of people who come into contact with those who are infected. That is the wrong focus. If we suddenly discovered an "epidemic" level of people with the common cold or mild cases of the flu, would we take the same actions we are taking today? That is the way we need to think about the problem.

My hypothesis, absent far better data, is that there is an exceptionally vulnerable part of our population that will be devastated if it comes into contact with the Coronavirus, not because the Coronavirus is exceptionally lethal, but because the combination of the virus and the horrifically compromised immune systems of that part of the population is lethal. What none of us know, including me, is whether my hypothesis is right, wrong, or needs to be refined. We need to be figuring this out as quickly as possible.

The Italian public health and clinical authorities are having to make horrible choices about who lives or dies. They are making decisions on the spot that some patients will die in a few weeks or months of some other illness and are making implicit decisions not to deploy scarce healthcare resources to keep them alive for a few weeks or even months. They are focusing on people with longer-term life expectancy. Their scarce resource is not money. It is the combination of intensive care facilities and healthcare resources that is scarce.

To use a comparison to another kind of health challenge, we debate whether someone with Stage 4 cancer should receive a specialty treatment costing several hundred thousand dollars that has a 10% chance of extending life by one month. In these cases, we urge patients and their families not to choose life extending treatments for which the rest of us have to pay. In this case, we are reporting on every death and looking at our ability to treat every patient, regardless of whether we are extending life by one month or 20 years.

As a society, we may explicitly decide we want to do that, but if we do, then we have to weigh all the other negative health consequences of social distancing: stress, adverse effects on mental health from isolation, enforced physical inactivity, and severely impaired financial wellbeing. Very little of that is happening now. We have made the single-minded decision to prevent every infection, even if the odds of that infection having serious health consequences are infinitesimally small, and even if the negative health impacts of devoting disproportionate societal resources to this far outweigh the effects of this illness.

As I have said in repeated postings, we cannot make decisions of this magnitude with the skimpy and flawed data sets we have today.

• We need more complete individual health records. In the Netflix series, Diagnosis, hosted by Dr. Lisa Sanders, the 3rd episode was about a man who had serious memory loss problems. By crowdsourcing his symptoms, she and he were able to figure out that the underlying problem was a condition arising from his service in the 1st Gulf War. The condition was identified at the time as happening with many military personnel stationed in Kuwait, but not captured in a medical record 25 years later. He had no lifelong health record. That is not accidental or aberrational. It is the predictable result of a deliberate effort by healthcare providers, insurance companies, and healthcare software providers like Epic to keep patient data fragmented inside their systems. Shame on them! Every patient should be able to present to every physician a complete lifelong health history that does not depend on the patient's or the caregiver's memory.
• We need more complete population level health data. We also need to have more ability for public health authorities to pull diagnostic and treatment data out of records, not to identify specific individual health conditions, but to do broad population-level analytics in situations like this. We depend too heavily on upward reporting by individual providers or healthcare systems. That is too clumsy a way to make decisions.
• We need an infrastructure that enables better digital health. If we were serious about more continuous biometric vital signs monitoring and video telehealth consultations, we would need every home to have Internet access and virtually all homes to have high-speed fiber optic broadband access. South Korea, Japan, Singapore, and Sweden, to name a few countries, have this. Less than half of our residences have broadband, according to data from Microsoft, which has a far better handle on this than the FCC, since Microsoft is continually downloading software into our home computers and phones at our homes and knows download speeds. In her book Fiber, Susan Crawford notes that getting 100% penetration of electricity into every American home took about 50 years and only happened at the end because the Franklin Roosevelt administration understood the importance of electricity to a good quality of life. Broadband is the same way. We have too much of our system controlled by telecoms and cable companies that, for short-term profit reasons, have chosen not to invest and deploy broadband.

This is too serious a crisis to be managed at all levels of government and by public health authorities with as little data as we currently have. At a minimum, the people who do statistical modeling need to be far more rigorous in examining what they assume in the data they receive from others or the data they collect at one point in time that changes later.