Background:
Carotid endarterectomy (CEA) is the leading procedure for carotid stenosis, yet national data on trends in rates and outcomes are limited. We determined CEA rates among Medicare beneficiaries and evaluated mortality and readmission over 8 years.
Methods:
We used Medicare fee-for-service data to identify beneficiaries aged ≥65y who had their first CEA (ICD-9 38.12) from 2003-2010 and calculated annual rates per 100,000 person-years (PY). We fit mixed models to assess trends in patient-level outcomes, adjusting for demographics, comorbidities, and symptomatic status. We also evaluated hospital-level trends by calculating risk-standardized mortality (RSMR) and readmission (RSRR) rates. A spatial mixed model adjusted for age, sex, and race was fit to calculate county-specific risk-standardized CEA rates in 2003-2004.
Results:
There were 505,966 unique CEA hospitalizations. The annual number of CEA discharges decreased from 81,604 in 2003 to 47,597 in 2010 (42% decrease), though the patient characteristics remained largely similar. The national CEA rate was 283 per 100,000 PY in 2003, and there was considerable geographic variability (Figure A). This rate decreased each year to a low of 172 per 100,000 PY in 2010. The rate of stroke or death within 30 days decreased from 3.2 to 2.7%, with a significant adjusted annual reduction of 3% (Figure B). Annual reductions in other short- and long-term outcomes were similar, ranging from 2-3%. The median hospital-level 30-day RSMR decreased over time from 0.99 to 0.57%, while the variation between hospitals increased (interquartile range of 0.7-1.67 percentage points). The 30-day RSRR decreased from 11.0 to 10.1%, but there was more homogeneity across hospitals and years.
Conclusions:
CEA use among Medicare patients decreased dramatically from 2003-2010, while mortality and readmission outcomes improved. The relative importance of biological and sociological mechanisms for these trends merits further study.
Posterior predictive treatment assignment methods for causal inference in the context of time-varying treatments
