They use the same logic for Kern County, California, and New York. Their logic is, that if you test N people and find M positive cases, that in the population at large you'll have an infection rate of M/N. This would be true if the N tests were performed on random members of the population. But because the tests are only performed when there's reason to suspect infection, M/N is a higher infection rate than in the population at large. Their inflated number for total infected persons makes the death rate appear smaller.