Trends and Variation in the Utilization and Diagnostic Yield of Chest Imaging for Medicare Patients With Suspected Pulmonary Embolism in the Emergency Department

OBJECTIVE. The purpose of this study is to assess trends and variation in chest CT utilization in the emergency department (ED) and its diagnostic yield for suspected pulmonary embolism (PE) among a national sample of Medicare beneficiaries. The relationship between hospital and provider characteristics is also discussed. MATERIALS AND METHODS. We conducted an observational analysis of Medicare beneficiaries evaluated in the ED for suspected PE from 2000 to 2009. Standard Medicare analytic files representing a 20% sample of fee-for-service beneficiaries were linked to the American Hospital Association Annual Survey of Hospitals, American Medical Association Physician Masterfile, Medicare Physician Identification and Eligibility Registry, and Dartmouth Atlas Project to calculate geographic-and physician-level chest CT utilization (i.e., the proportion of ED visits involving chest CT examination for suspected PE) and diagnostic yield (i.e., the proportion of chest CT examinations with a positive PE diagnosis). RESULTS. Of 2.5 million ED visits, 2.5% (n = 164,274) included chest CT for suspected PE; 6.2% visits (n = 10,121) resulted in positive findings for PE. Between 2000 and 2009, chest CT utilization increased fivefold. Geographic variation in CT utilization (median, 2.38%; interquartile range [IQR], 1.91-2.92%) and diagnostic yield (median, 6.31%; IQR, 5.11-7.66%) was observed between 306 hospital referral regions. Physician use of imaging was explained by greater experience (lower utilization and higher yield) and emergency medicine board certification (lower utilization and equivalent yield). CONCLUSION. CT utilization in the ED for suspected PE has steadily risen, whereas diagnostic yields have declined over time. Wide variation in practice is observed at the physician and geographic levels and is explained by several physician and hospital characteristics. Taken together, our findings suggest a substantial inefficiency of chest CT use and substantial opportunities for improvement.