5/31/2019 Acknowledgements 1 2 Study co-investigators Funding - PDF document

5/31/2019 Acknowledgements 1 2 Study co-investigators Funding source Jessica Lavery, MS Cancer Center Support Grant [P30 CA 008748] Peter Bach, MD, MAPP to Memorial Sloan Kettering Cancer Center Diane Li, BA Variation in 30-day mortality Ashley Russo, MD Disclosures: Vivian Strong, MD following cancer surgeries Katherine Panageas, DrPH None. across U.S. hospitals Allison Lipitz-Snyderman, PhD Assistant Attending, Center for Health Policy and Outcomes Department of Epidemiology and Biostatistics Memorial Sloan Kettering Cancer Center June 4, 2019 Quantitative performance reporting is common in Background 3 4 healthcare and other industries • Quality measurement and public reporting are common strategies to motivate improvement and provide comparative information to the public. • Cancer surgeries are largely left out of large-scale public reporting efforts. • Decades of research show a volume-outcome relationship. • Patients want comparative performance data on cancer surgical outcomes, but publicly available information across Source: CMS Hospital Compare the U.S. is limited. Table excerpt from: planecrashinfo.com/rates.htm Study objective Methods 5 6 To examine whether a measure of 30-day mortality after cancer • Data source: 100% Medicare Research Identifiable Files surgeries would be a candidate for large scale quality measurement. • Over 3,600 hospitals across the U.S. performing surgeries for 12 different cancer sites, 2011-2013. Evaluation criteria for comparative performance measurement: 1. Low measurement burden • Patients with Fee-for-Service Medicare coverage, ages 66 and 2. Significant variation in performance between hospitals over, assigned to the hospital performing surgery 3. High potential for public health benefit • Cancer sites: breast, colorectal, lung, prostate, kidney, other gynecologic, ovary, gastroesophageal, pancreas, liver, sarcoma, bones and joints, per prior validation work Allison Lipitz-Snyderman 1

5/31/2019 Number of surgeries and hospitals performing surgeries Analysis 7 8 for Fee-for-service Medicare patients, 2011-2013 • Performance metric: 30-day mortality after cancer surgery No. hospitals Median no. surgeries No. of surgeries performing surgery per hospital (IQR) • Linear mixed effects model to obtain an estimate of each hospitals’ risk adjusted mortality rate. Overall 340,489 3,776 34 (9, 108) Breast 119,217 3,537 17 (5, 44) Colorectal 85,857 3,471 14 (5, 34) • Adjusted for patients’ age, sex, race, Charlson comorbidities, Lung 33,513 1,844 9 (3, 23) emergent surgery; hospitals’ ownership, rural location, teaching Prostate 29,207 1,600 7 (<3, 21) Kidney 24,578 1,998 6 (<3, 15) status Other gynecologic 18,603 1,615 3 (<3, 12) Ovary 8,805 1,137 3 (<3, 10) • Examined the extent of variation between hospitals overall and Gastroesophageal 7,899 1,512 <3 (<3, 5) Pancreas 6,391 785 3 (<3, 8) by cancer site using Wald test of the random effect. Liver 3,562 556 3 (<3, 7) Sarcoma 1,984 506 <3 (<3, 4) Bones and Joints 873 302 <3 (<3, 3) Unpublished data Characteristics of surgeries for Fee-for-service Characteristics of hospitals performing surgeries for 9 10 Medicare patients, 2011-2013 Fee-for-service Medicare patients, 2011-2013 2 Cancer Site Examples Overall 2 Cancer Site Examples Column (%) Colorectal Gastroesophageal Overall Column (%) Colorectal Gastroesophageal No. of surgeries 340,489 85,857 7,899 No. of hospitals 3,776 3,471 1,512 Age (years) 66-69 22% 14% 20% Teaching (%) 93% 93% 85% 70-74 28% 22% 30% Ownership 75-79 22% 22% 25% Not-for-profit 63% 65% 70% 80-84 16% 21% 16% 85+ 12% 21% 10% Private 19% 19% 18% Female (%) 67% 54% 35% Government 18% 17% 12% White (%) 89% 88% 80% Rural (%) 36% 34% 12% No. comorbidities 0 45% 39% 33% 1 27% 26% 29% 2+ 29% 35% 39% Unpublished data Unpublished data Hospitals’ median risk-standardized 30-day mortality Potential limitations 11 12 ratio and test of between-hospital variation • May be unaccounted for factors that could contribute to Hospital median risk-standardized Test of variation differences between hospitals. 30-day mortality ratio (IQR) • Prior validation work on cancer stage. Overall 2.41 (2.31, 2.58) <.001 Overall estimated 600 Breast 0.23 (0.22, 0.24) <.001 deaths prevented if Colorectal 5.07 (4.85, 5.47) <.001 • Unknown generalizability outside the Fee-for-Service Medicare hospitals in the worse Gastroesophageal 5.81 (5.69, 5.88) 0.080 program. performing quintile improved to the Kidney 1.55 (1.50, 1.58) 0.036 median. Liver 4.19 (4.06, 4.26) 0.023 • Comprehensive evaluation of cancer surgical quality. Lung 4.76 (4.54, 5.24) <.001 Other gynecologic 1.41 (1.38, 1.41) 0.102 Ovary 3.43 (3.28, 3.48) 0.009 Pancreas 4.23 (4.05, 4.40) 0.030 *Bones and joints, prostate, and sarcoma are included in aggregate analysis but results by cancer site are not estimable. Unpublished data Allison Lipitz-Snyderman 2

5/31/2019 Conclusions and Implications 13 14 Members of the public and other stakeholders should not • Support for low burden measurement, variation between hospitals, and public health benefit. assume that all hospitals have equal outcomes of • There are some tradeoffs to consider. cancer surgery. • Low burden source, but time delay. • Low numbers of surgeries for many hospitals, but could impact the reliability of adjusted estimates • 30-day mortality after cancer surgeries is an untapped potential metric for large-scale comparative performance measurement. * Photographs by Allison Lipitz-Snyderman, New Jersey Transit train / train station 15 Questions / comments? Contact information: Allison Lipitz-Snyderman, PhD Center for Health Policy and Outcomes Department of Epidemiology and Biostatistics Memorial Sloan Kettering Cancer Center E-mail: snyderma@mskcc.org Allison Lipitz-Snyderman 3