Survival bias is a type of sampling error or selection bias that occurs when the selection process of a trial favours certain individuals who made it past a certain obstacle or point in time and ignores the individuals who did not (and are generally less visible).
A classic example comes from a study of bombers during world war two. Planes were studied after returning from bombing runs, and it was recommended that the areas of the planes with the most damage should be reinforced. The problem with this solution was that they were only looking at the planes that survived battle. In fact, the areas of the planes that looked like they sustained the least damage were the most important to reinforce, because those were the areas that, if hit, would cause the plane to be lost (and therefore not included in the sample).
Another classic (and somewhat tragic) example comes in the study of cats falling from high rise buildings. Observational data indicated that cats falling from higher than 6 stories actually had fewer injuries than cats falling from less than 6 stories. A number of different theories have been suggested to explain this observation, but the most likely explanation is that cats that are obviously dead are not brought to veterinarians and therefore are not captured in the data. Cats falling from higher are more likely to die, and also more likely to have a greater number of injuries, but are excluded from the data, resulting in a biased sample.
In emergency medicine, trials often only start when the patient is stable enough to be consented for research. Unfortunately, this might result in a biased sample of patients, because it excludes everyone who dies very early in their resuscitation, or who is too sick to consent.
Mangel, Marc; Samaniego, Francisco (June 1984). “Abraham Wald’s work on aircraft survivability”. Journal of the American Statistical Association. 79 (386): 259–267. doi: 10.2307/2288257
Whitney, WO; Mehlhaff, CJ (1987). “High-rise syndrome in cats”. Journal of the American Veterinary Medical Association. 191 (11): 1399–403. PMID 3692980