Stevens–Johnson syndrome seems to occur as a response to several different drugs, but apparently only to people with a specific HLA genotype. Part of the solution here is that currently we only approve a drug if it is safe for nearly 100% of the population, even if it is useful for far fewer (see statins) and the path forward needs to give more consideration of the differences in people, and having suitable genotyping a prerequisite for some medications.
Agreed on the general principle, but note that the incidence of SJS varies by drug. For example, lamotrigine causes SJS in 10-30% of people taking it, depending on how fast the dosage is increased.
There they don't address it by trying to predict who will show SJS symptoms and who won't; they just put people in it slowly (which reduces incidence) and take people off the drug (or hold at a given dose until the body accepts the drug) if they show symptoms.
