Sometimes it's better to have a consistent yardstick than an accurate one. Noise vs bias. This definition has the significant merit of having a lot lower noise

Sometimes it's better to have a consistent yardstick than an accurate one. Noise vs bias. This definition has the significant merit of having a lot lower noise than any clinical definition of a Covid death, where you'd have wildly different estimations from doctor to doctor, let alone from country to country. It does have bias, true, but you can deal with bias. You do pinpoint studies to measure it and use them to correct bigger trends. You compare with the overall mortality change. And once you have an estimation of the bias, you can apply it back to the overall mortality numbers and get a much better result.