OBJECTIVES: The cost-effectiveness of CAH screening (a state-level policy) was assessed by Yoo and Grosse (2009) using national incidence (NATL). Incidence of rare genetic diseases often varies geographically, however, it is likely that some state variation is due to sampling error. This research assesses the extent of the sampling error explanation through a series of simulations.

METHODS: We obtained actual state-level incidence (ASI) for the 50 US states plus DC for 2006. For the simulation we assumed that state equaled national incidence and constructed a Beta distribution with alpha parameter equaling predicted state cases, and Beta parameter equaling state births minus predicted cases.  We then ran a Monte Carlo simulation of 1000 iterations and calculated the proportion of iterations for which the incidence draw was more extreme than the ASI (i.e. draw > ASI if ASI > NATL or draw < ASI if ASI < NATL). Small numbers of iterations more extreme than the ASI are consistent with the ASI not equaling NATL. Extreme was defined as less than 5% of draws in the simulation less than the ASI (if ASI< NATL) or greater than the ASI (if ASI > NATL).

RESULTS: ASI per 1000 births ranged from 0 to .343 (median, .038).  There were 21 states with ASI < NATL (including 13 values of 0) and 25 states with ASI > NATL.  Of the 46 states reporting ASI,