What people are saying about the breathalyser is partly correct, but what has been said about forensic back calculations is wrong.
I have dealt with many drink driving cases in which suspects have attempted to obstruct justice by deliberately drinking alcohol after the collision but before police arrive.
Step 1. Identify the container from which they have been drinking and seize it, so that the alcohol content and amount consumed can be measured. Take account of the evidence offered by the suspect and others as to the post incident drinking.
Step 2. Obtain blood and urine samples, the concentration of alcohol in which is a function of what has been drunk, and crucially when. Clearly breath alcohol accessible to a breathalyser can be affected by alcohol in the mouth from something drunk in the previous few minutes, which is why suspects are asked about when they last had a drink. By the time a suspect reaches an evidential breath test machine in a police station this mouth alcohol will have cleared. What is breathed out then via the lungs is a function of blood alcohol concentration, which trails behind consumption as the digestive system has to absorb it. Urine concentration trails further because of the need for further processing by the kidneys.
Step 3 Carefully assess all of this information against the huge volume of experimental data which exists regarding alcohol concentrations and consumption which comes from both medical experiments and real world data where suspects alcohol consumption has been well documented.
Step 4. Prosecute when the back calculation points to a figure significantly in excess of legal limits to allow a margin for error.
Simples.
(Actually it is far from simple, but you know, science. This process has also been rigorously tested in court by, you know, the law).