1993 Understanding Base Rate Fallacy: Bringing Goldstein and Brown together with Tversky and Kahneman.

Dr. Roger Williams et al employed base rate to great success: beginning with first degree relatives of known carriers. In such a pool, a person with 310 total cholesterol has a 95% chance of carrying the mutation, whereas if we simply begin with the general population, the chance is under 4%. Note Williams’ comment that the “clinical clues are weak predictors of FH.” This is because these same “clinical clues” are shared with many other diseases besides FH. For example, high cholesterol is the dominating “clinical clue” both in FH and FCH scoring systems, while the only uniquely distinguishing characteristic for FH is the presence or absence of the LDL Receptor mutation. Barring genetic testing, if we tally up a score based mostly on cholesterol levels, we cannot overcome the fact that passing scores are shared among both on-target and off-target populations.  To bridge the divide between ideal-but-impractical genetic screening and the high error rate in the scoring systems, this third option was suggested by Dr. Roger Williams, employing the advantage of “a priori probabilities.” To recap the movement toward precision medicine, first there were “clinical clues” to distinguish FH from other diseases, such as FCH (e.g., exclude those with high triglyceride), then we advanced to genetic testing for the LDLR mutation, and then, due to the impracticality of testing entire populations, Dr. Williams employs a base rate strategy.  (By design, the numbers used in the games are not representative of the numbers relevant to FH prevalence. The sole purpose of this illustration to isolate the difference between the mathematical systems, not any given quantity within a given application. One system is biased: “those with characteristics shared with FH are therefore FH.” The other acknowledges base rate and takes advantage of a superior prior probability.)

Roger R. Williams base rate strategy for identifying familial hypercholesterolemia and avoiding misdiagnosing those with familial combined hyperlipidemia.