> If you use deaths/(deaths+recoveries) you get a much higher figure You also have to take into account that mild cases are probably never detected and counted

> If you use deaths/(deaths+recoveries) you get a much higher figure You also have to take into account that mild cases are probably never detected and counted. So you are calculating the fraction of deaths among severe cases, not among all cases.