Because DES were introduced in 2003, we calculated average annual payer-perspective costs among CAD patients during 2002–2006 (including 2002 costs as a baseline), in each U.S. Hospital Referral Region (HRR)6
using a 5% random sample of fee-for-service Medicare beneficiaries, excluding patients younger than 66 and older than 85 (DES use declines markedly at older ages).5
Calculations were separately performed on each of three CAD sub-cohorts categorized annually by clinical events: patients with acute myocardial infarction (AMI), patients with acute coronary syndrome (ACS) but no AMI, and patients without ACS. We did not assume DES-associated healthcare cost growth was confined solely to DES recipients, thus cohorts included all CAD patients regardless of treatments received. Costs included all facility and provider Medicare payments, including non-cardiovascular costs, inflated to 2006 dollars using the consumer price index. This design captured costs downstream of major cardiovascular procedures and events, as patients were retained in the cohort through 12/31/2006 or until death. Annual DES rates within each HRR and subcohort were also calculated.
Substantial geographic variation in DES use across HRRs enabled measurement of the relationship between higher DES use and higher healthcare costs. Multivariable regression models were estimated predicting annual HRR-level healthcare costs among CAD patients as a function of the local DES rate, HRR “fixed effects” that controlled for time-invariant differences in costs across HRRs, and time varying controls such as an annual HRR-specific medical cost index (controlling for geographic variability in healthcare inflation), patients’ average DxCG Risk Score (predicting comorbidity-associated costs), and general time trend controls. Models were estimated separately for each subcohort.
To fully describe the national expenditure implications of the per-patient DES cost increases estimated by regression models, we computed the total change in national expenditures attributable to DES by multiplying the total number of Medicare beneficiaries nationwide in each CAD subgroup by the per-patient 2002–2006 cost increase predicted by the models.