Surgeon Specialty and Operative Mortality With Lung Resection PP Goodney, FL Lucas, TS Stukel, JD Birkmeyer VA Outcomes Group, White River Junction, VT Dartmouth-Hitchcock Medical Center, Lebanon, NH
Background Several studies have reported variation in outcomes according to surgeon specialty –colorectal resection for cancer Wigmore et al, Ann Surg 1999 –carotid endarterectomy Hannan et al, Stroke 2001 Not all studies have confirmed this finding Cowan et al, JACS 2002
Lung resection Resection for lung cancer –approximately 25,000 cases per year in the U.S. –Performed by: General surgeons Cardiothoracic surgeons Non-cardiac thoracic surgeons
Research question Does surgeon specialty affect operative mortality in lung resection for lung cancer?
Subjects and databases Study population –All Medicare beneficiaries –Age Patient selection –Procedure code for lung resection (pneumonectomy or lobectomy) –Diagnosis code for lung cancer –Unique physician identifier number (UPIN) present on discharge abstract
Unique physicians in lung resection inpatient file n=4793 Surgeon Specialty Assignment
Unique physicians in lung resection inpatient file n=4793 Yes n=2179 Surgeon Specialty Assignment American Board of Thoracic Surgery member? No n=2614
Unique physicians in lung resection inpatient file n=4793 Yes n=2179 Cardiothoracic surgeon n=1516 Non-cardiac thoracic surgeon n=663 Surgeon Specialty Assignment American Board of Thoracic Surgery member? Perform CABG? Yes No n=2614
Unique physicians in lung resection inpatient file n=4793 Yes n=2179 Cardiothoracic surgeon n=1516 Non-cardiac thoracic surgeon n=663 General surgeon n=2614 Surgeon Specialty Assignment American Board of Thoracic Surgery member? Perform CABG? Yes No n=2614
Analysis Unit of analysis: patient Main exposure: surgeon specialty –General, cardiothoracic, non-cardiac thoracic Main outcome measure: operative mortality –Combination of death before discharge or within thirty days of the index procedure
Analysis Using multiple logistic regression models, adjusted for the following: Patient variables: Age, sex, race Comorbidity score Admission acuity Extent of resection Surgeon variables: Surgeon volume Clustering Hospital variables: Hospital volume Bed size Teaching status Medical school affiliation ACS-approved cancer center
RESULTS
Patient characteristics General surgeons Cardiothoracic surgeons Non-cardiac thoracic surgeons No. patients (n) 9,2639,7926,490 Age >75 (%) Female (%) African American (%) Charlson score >3 (%) Pneumonectomy (%)
Hospital characteristics General surgeons Cardiothoracic surgeons Non- cardiac thoracic surgeons Bed size Teaching status (% with residents) Medical school affiliation (%) ACS cancer program (%) >45 lung resections per year (%)
Surgeon characteristics General surgeons Cardiothoracic surgeons Non- cardiac thoracic surgeons No. surgeons (n)2,6141, Surgeon age (years) Years in practice (years) Average cases/yr, lung resection >20 lung resections per year (n, %)71 (2.7%)65 (4.3%)87 (13%)
Adjusted operative mortality, by surgeon subspecialty General Cardiothoracic Non-cardiac thoracic p <0.001 between all groups
Adjusted operative mortality, by extent of resection General Cardio. NCTSGeneral Cardiothoracic NCTS p <0.001 between all groups
Adjusted operative mortality, with high-volume surgeons General Cardiothoracic Non-cardiac thoracic p <0.01 between non-cardiac thoracic surgeons and others
Adjusted operative mortality, in high-volume hospitals General Cardiothoracic Non-cardiac thoracic p <0.01 between non-cardiac thoracic surgeons and others
Odds ratios of operative mortality General surgeons Cardio- thoracic surgeons Non-cardiac thoracic surgeons Crude Adjusted for patient characteristics Patient and hospital characteristics Patient and hospital characteristics, and hospital volume Patient and hospital characteristics, hospital and surgeon volume
Summary Operative mortality with lung resection varies by surgeon specialty Risks were lowest for non-cardiac thoracic surgeons Hospital and surgeon volume account for some, but not all of this effect
Limitations Administrative data for risk adjustment Error in assignment of surgeon specialty –Bias would tend towards the null
Why does performance differ across specialty? Additional training Structural differences across specialty –Larger hospitals –Medical school affiliations –ACS-approved cancer programs Another possibility – –Differences in processes of care Intensivist-managed ICUs, epidural catheters, pulmonary protocols Many of these processes are unmeasured in current quality improvement initiatives such as the STS database
Do our findings matter? Although these differences are statistically significant, are they clinically important? Differences are small (~1%) for lobectomy, but larger (~5%) for pneumonectomy How much is enough? –Only patients can decide
Conclusion Surgeon specialty impacts operative mortality with lung resection Some, but not all of this difference can be explained by volume Further study of these differences may hold potential for improvement
Acknowledgement Scottie Siewers VA Outcomes Group