RT Journal
A1 Briesacher BA, Madden JM, Soumerai SB
T1 CAution in generalizing part d results to medicare population
JF Archives of Internal Medicine
JO Archives of Internal Medicine
YR 2011
FD February 28
VO 171
IS 4
SP 365
OP 369
DO 10.1001/archinternmed.2011.9
UL http://dx.doi.org/10.1001/archinternmed.2011.9
AB 
There may be another reason for the anomalous findings. The investigators excluded a large proportion (approximately 60%) of the original 2005 elderly and Medicare-eligible MEPS sample in order to use a longitudinal study design. Such a large exclusion can significantly and systematically diminish the generalizability of the results to the reference population, in this case Medicare beneficiaries. As a result, other findings in the article by Millet et al1 are also inconsistent with previous research. For instance, Table 1 of their article1 suggests that 44% of the Medicare population did not have drug coverage in 2005 and 19% remained without coverage in 2006, when others put those estimate closer to 24% and 10%, respectively.6 These are not trivial differences, since the magnitude of the effect of Medicare Part D is directly related to levels of drug coverage before and after implementation of the program.