The devil’s in the details: Variation in estimates of late-life activity limitations across national cohort studies