The models presented less than a week ago have turned out to be very wrong – which is good for us.
On April 2 (just 6 days ago), the Corona Virus task force showed a model that predicted that there would be 100,000-240,000 deaths due to Covid 19. However, since that time, the predicted number of deaths have decreased steadily and significantly. Today’s estimate is that the number of deaths is now projected to be 60,415. This by the way, is the same number of people who died of the flu in 2018.
The decrease in predicted deaths is due to a reduction in the number of new cases, number of hospitalization and number of ICU patients in the past several days. This was especially true in ‘hot spots’ such as New York, Los Angeles and Italy. The projected fatality numbers will continue to drop if the trends continue.
The models are generated by the International Health Metrics and Evaluation (IHME) organization from the University of Washington. They are recognized as a reputable and respected group that does this kind of modeling all the time. However, models are only as good as the data that goes in and assumptions that the modelers make. In the vast majority of cases, the actual numbers used and specific assumptions made are either never disclosed or are only disclosed at later dates. In other words, we just have to take their word for it.
The IHME also predicted the number of hospital beds, numbers of ventilators etc. that each city might need. The IHME models for any states were off by a factor of 5. States that were predicted to have shortages are now predicted to have surplusses. On April 2, IHME predicted that Tennesse would have 3000 deaths. The current IMHE estimate is 600. Further on April 2, IMHE predicted that Tennessee would need 15,000 hospital beds. This caused great concern as Tennessee only has 7812 beds. However, it appears that they will actually only need 1232 beds. IMHE also predicted Tennessee would need 2000 ventilators and it appears that they will only need 208. Unfortunately, the IMHE greatly over estimated in many, many instances, including New York and California. The good news is there seems to be plenty of capacity rather than a shortage.
This is not really the fault of modeling – the limitations of modeling are well known to those familiar with the process and procedure. But these details are complicated and rarely communicated to public. The problem comes when decisive action is taken based on models that are not validated and changing almost hourly due to new data.
Great efforts were made to construct makeshift hospital beds, icus and settting up the manufacturing of ventilators. Fortunately, from the public health side this is all good news, but it could turn out to be a huge sum of money spent on equipment and facilities that are not needed – not mention the anxiety, fear and worry that was generated.
Models have their place, but we know the limitations of models. It can not accurately predict the track of a hurricane more than a few hours ahead, it can not predict which stocks will go up or down, or if the market will go up or down. In all these cases, we continue to use models because it is thought to be better than not having it. However, that’s not always true.