When did the NY infection start and how bad would it be without mitigation?
Whether the lock-down in New York has been effective and what the unmitigated death rate would have been are interesting questions. Experts have published a paper for review suggesting that the source of the epidemic was from Europe as evidenced by genome testing and it was introduced in mid February. This would imply that the virus was spreading silently for some time before cases and then deaths were seen.
Utilizing our Monte Carlo simulation tools, we can easily test this theory out by seeding the simulation with a very small number of infected individuals. The simulation space was set to allow no mitigations and play out through the population using the same parameters as our NY case study.
Initially a seed of one individual per million was used, and because of the random characteristics assigned to the individual and victims, the infection did not reliably spread in the simulation. When one infected individual per million was seeded in four successive days, there was a fairly consistent start to the infection. The onset of the death curve was fit to the actual data for New York. This is a reasonable marker for correlation as deaths that occurred early on were from infections that began long before any mitigation steps were taken.
The start of the spread of the infection, based on the calibration of the seed date is approximately February 24th, 2020. The results of four runs are show with an overlay of actual deaths and the New York case study curve (which matches current death trends). A more exhaustive number of runs were not performed to establish an error range, but the distribution is likely to fairly tight. The four person seed runs were unmitigated and result more than twice as many deaths. From this we can conclude the actual infection did indeed start in mid to late February and that lock-down was still saved a significant number of people.
If we take a look at the number of deaths per day for the same case, the infection peak still runs the same general shape. The unmitigated runs have a similar shape and timing but again peak out at more than twice what the NY Case Study model predicts.
If we scale the Case Study Model to fit on top of the unmitigated curves we find that there is a minor degradation in the back porch, due to the mitigations effectiveness cutting into the back side of the infection curve in mid March.
What can we conclude from this? Several things. First, the infection likely started spreading sometime around February 24th. Second, we can also conclude that the mitigation steps, although late, did save lives.
For reference, the timing of the mitigation steps in the NY Case Study model, as a reduction in transmission factor, is shown below.