Nick,
Nice examples. Maricianah makes a good point that it is extremely helpful to evaluate face and content validity of new measures of subtle constructs, such as repro autonomy. The Cronbach's alpha is a measure of reliability, but it is very desirable to have test-retest. Validity assessment is often much harder though. I don't know this field well enough to propose additional validation assessments. There's no gold standard presumably, so most evaluations will depend on construct validity.
The Moaddab et al paper is very surprising. First, it seems like they "tricked" CDC wonder into delivering small cell sizes that the database developers intended to obscure. I'm surprised this isn't against the data use agreement for Wonder, and it certainly seems questionable. Second, their application of a factor analysis to completely disparate constructs is puzzling as a data reduction approach. Usually you would do factor analysis if you actually thought there was a coherent underlying construct which all of the items measured. That is not the case here, they are just trying to select among a large set of covariates with a relatively modest sized set of observations. Factor analysis doesn't seem appropriate. Third, the interpretation that trends are due to racial disparities is puzzling given that the trends occur in whites and blacks.
Maria