More information about text formats
A non-vaccinator is not the culprit for the existence of immunosuppressed persons. Hence, we must use overall and not conditionals probabilities here and pass to the population level. Let’s assume that the prevalence of immunosuppressed persons is 4/10.000 and that measles attack rate is 2/10.000 for the population. Then the probability to observe an immunosuppressed person that gets infected by measles is very low, say 4/10.000 * 2/10.000 = 8 * 10^(-8). The probability to get an encephalitis from measles vaccine is 1 * 10^(-6) and hence more probable.
Risk vs Risk is always problematic and should be avoided.