RACE, INCOME, AND OPERATIVE TIME Jeffrey H. Silber, M.D., Ph.D. Paul R. Rosenbaum, Ph.D. Xuemei Zhang, M.S. Orit Even-Shoshan, M.S. The Center for Outcomes Research, The Children’s Hospital of Philadelphia The Department of Pediatrics, Anesthesiology and Critical Care Medicine, PENN The Department of Statistics, The Wharton School, PENN The Leonard Davis Institute of Health Economics Measuring Operative Time • Until now, studies analyzing operative time in the Medicare population have been limited to single (or very few) institutions. • Operative time is usually defined as: – Surgical Time: Incision to closure – Anesthesia Time: Induction to Recovery Room Goals of this Talk • To examine the accuracy of using Anesthesia Time Units from Medicare bills to estimate operative time • To examine models that predict procedure length based on: – Patient comorbidities – Race – Income • To explore the relationship between race and operative time at individual hospitals • To discuss the implications of disparities in operative time Introduction • Operative time is a common measure of surgical quality – Longer procedures are associated with higher infection rates – Longer procedures are associated with physicians-in-training Measuring Operative Time in Medicare Patients • In 1994, billing for Anesthetist and anesthesiologist services changed to a “by the minute” system • Anesthetist bills (anesthesiologists, nurse anesthetists) from Medicare have never been utilized to examine procedure time on a large scale • Race and Income have not been studied with respect to procedure time Data • Medicare Claims on the 20 most common General Surgical procedures and the 20 most common Orthopedic procedures in Pennsylvania in 1995 and 1996, N = 77,638 • Chart Abstractions on a subset of 1931 Pennsylvania General Surgical and Orthopedic patients as part of the Surgical Outcomes Study 1 The Anesthesia Claim Time Abstraction • The Anesthesia Claim consists of two parts: – Base Units – Time Units • Time units are in 15 minute intervals. The first digit of the time unit is a decimal tenths digit – For example: “25” for units = 15 x 25/10 minutes or 37.5 minutes • We chose the single longest anesthesia bill associated with the same day as the principle procedure • We defined four specific times for abstraction on 1931 charts in the SOS study: – Start induction – Start incision – Closure – To recovery room or ICU (if going directly to ICU) Bill Time - Chart Time. N = 1931 patients from the Surgical Outcomes Study Chart Time as a Function of Claim Time, N = 1931 Variables Int. Slope P R2 Median Absolute Residual (minute) I: Unadjusted m-Estimation -1.21 0.97 .0001 .89 5.49 N/A II: Comorbidities, Procedures and Hospitals 2.47 0.98 .0001 .89 5.37 <.0001 P (Wald) Shapiro-Wilk Test P < 0.0001 The independent variable is anesthesia claim minutes, and the dependent variable is anesthesia chart minutes. N = 1931, ρ = 0.94. Estimated Anesthesia Time (Minutes) for Each General Surgical Principle Procedure N 25th %ile Median 75TH %ile 422 210 260 320 Anterior Rectal Resection 685 165 251 255 Left Hemicolectomy 1307 155 195 242 Part Lg Bowel Excis Nec 507 150 191 251 Sigmoidectomy 2922 140 176 221 Lg Bowel Stoma Closure 326 116 164 221 Part Sm Bowel Resection 1253 120 155 209 Right Hemicolectomy 34350 120 155 195 Unilat Thyroid Lobectomy 370 120 135 176 Cholecystectomy-Open 3883 105 135 176 Peritoneal Adhesiolysis 1268 94 125 173 Procedure Abd-Perineal Rect Resect These times include principle procedures that were also accompanied with other secondary procedures billed on the same day. 2 Estimated Anesthesia Time (minutes) for Each General Surgical Principle Procedure (when only one procedure performed) Estimated Anesthesia Time (Minutes) for Each Orthopedic Principle Procedure Procedure N 25th %ile Median 75TH %ile N 25th %ile Median 75TH %ile Abd-Perineal Rect Resect 84 206 240 285 Lumbar/Lum-sac Fus Post 273 236 299 380 Anterior Rectal Resection 222 146 176 218 Other Cervical Fus Ant 152 182 240 343 Left Hemicolectomy 262 143 180 221 Revise Hip Replacement 971 180 234 311 Part Lg Bowel Excis Nec 77 131 158 206 Spinal Canal Explor 3166 138 180 240 Sigmoidectomy 678 125 155 195 Revise Knee Replacement 1110 143 180 230 Lg Bowel Stoma Closure 159 101 125 180 Excision Intervert Disc 2565 131 165 221 Part Sm Bowel Resection 112 105 131 170 Total Hip Replacement 6215 140 165 197 Right Hemicolectomy 940 110 135 170 Total Knee Replacement 12718 131 150 180 Unilat Thyroid Lobectomy 286 108 135 167 Open Red-Int Fix Hum 690 107 142 191 Cholecystectomy-Open 978 90 116 146 Partial Hip Replacement 4735 105 129 155 Peritoneal Adhesiolysis 316 80 101 125 Shoulder Arthroplasty 909 105 124 158 These times only include principle procedures that did not have other secondary procedures billed on the same day. Procedure These times include principle procedures that were also accompanied with other secondary procedures billed on the same day. Estimating Anesthesia Time (Y = Mins. from bill) Estimated Anesthesia Time (minutes) for Each Orthopedic Principle Procedure (when only one procedure performed) Parameter Minutes P-value Multiple procedures 18.3 .0001 Admit from ER .10 .90 Procedure N 25th %ile Median 75TH %ile Lumbar/Lum-sac Fus Post 41 200 240 294 Transfer In 6.6 .0003 Other Cervical Fus Ant 30 155 195 245 Hx Coagulopathy 5.0 .0001 Revise Hip Replacement 639 180 225 300 Spinal Canal Explor 1478 131 170 218 Hx Diabetes 2.5 .0001 Revise Knee Replacement 792 140 179 227 Hx Paraplegia 4.7 .0004 Excision Intervert Disc 1529 122 155 196 Black Race Vs White 5.2 .0001 Total Hip Replacement 4770 140 165 195 Severity Score 1 1.5 .04 Total Knee Replacement 10551 131 150 180 Severity Score 2 .40 .59 Open Red-Int Fix Hum 459 101 131 170 Partial Hip Replacement 4102 105 125 150 Severity Score 3 & 4 2.2 .03 Shoulder Arthroplasty 129 101 131 176 Severity Score 0 Ref N/A N=77,638 patients, controlling for 39 (=n-1) surgical procedures and 182 (=n-1) hospitals and 34 patient covariates of which 8 are displayed above. The model was fit using m-estimation, rank R2 of 40%. These times only include principle procedures that did not have other secondary procedures billed on the same day. Race, Income, and Procedure Time Race, Income, and Procedure Time WL (n = 9,324) BM (n = 513) WM (n = 25,060) BH (n = 547) WH (n = 27,473) BL (n = 2002) BL adjusted by 183 hosp. 29 (<.0001) 9 (<.0001) 15 (<.0001) 3 (0.159) 30 (<.0001) 7 (<.0001) 15 (<.0001) 8 (<.0001) 27 (<.0001) 8 (<.0001) -2 (<.0001) WL (n = 9,324) WL adjusted by 183 hosp. --- -14 <.0001) -4 (.036) 1 (<.0001) 0 (.578) -14 (<.0001) -5 (.486) -2 (<.0001) 1 (.069) 0 (.8939) 12 (<.0001) BM (n = 513) BM adjusted by 183 hosp. 14 (<.0001) 4 (.036) 0 (.8939) 5 (.052) 12 (<.0001) 4 (.014) -15 (<.0001) -3 (<.0001) WM (n = 25,060) WM adjusted by 183 hosp. --- -15 (<.0001) 1 (.570) -3 (<.0001) 1 (.126) --- 12 (<.0001) BH (n = 547) BH adjusted by 183 hosp. --- 12 (<.0001) 0 (.848) WL (n = 9,324) BM (n = 513) WM (n = 25,060) BH (n = 547) WH (n = 27,473) BL (n = 2002) 29 (<.0001) 15 (<.0001) 30 (<.0001) 15 (<.0001) 27 (<.0001) WL (n = 9,324) --- -14 <.0001) 1 (<.0001) -14 (<.0001) 14 (<.0001) --- BM (n = 513) WM (n = 25,060) BH (n = 547) --- --- 3 Adjusted Differences Between Black and White Procedure Times in Pennsylvania Hospitals With the Largest Black Populations Hospital No. Black No. White Teaching Type Excess Mins. (b-w) Policy Implications P-Value 1 153 352 Major 3.3 .449 2 147 322 V Major 7.0 .115 3 130 326 V Major 0.8 .870 4 125 108 V Major 12.2 .039 5 118 1319 Major 14.3 .001 6 112 437 Major 12.1 .011 7 112 43 Major -4.6 .562 8 111 1006 Not Teaching 4.0 .366 9 109 71 V Major 6.9 .308 10 107 505 Major 9.3 .050 11 105 178 V Major 15.9 .004 12 100 1007 V Major -5.1 .274 The overall Wald test for equality was 0.029, suggesting there is variability in disparity across hospitals. General Surgical and Orthopedic Procedures Why might there be a disparity in operative time inside a hospital? Why do we care about 15 minutes? • Reasons consistent with fairness – Unobserved severity – Admission from ER – Emergency Surgery • Reasons based on injustice – Different Surgeon/Physician-in-Training for less empowered – Risk of Litigation – Racism (Conscious or Unconscious) Identifying the Provider: Who is holding the retractor? • In teaching hospitals it is not always clear from the chart or the claim who is performing surgery. • Is the resident doing the dissection or holding the retractor? • Differential times between the resident and the attending may help in identifying the two providers. For a typical procedure like Colectomy, which was described in Khuri et al. (Annals Surgery 2001), we can estimate the relative speed of the resident and attending, assuming the hospital location does not influence this time. VA Location N Complexit y % of Cases by Residents Hours of Surgery Mean Time Combined 3.085 Teaching 8,038 3.00 93.64 3.17 NonTeach 1,201 2.95 1.42 2.54 From this table we can solve for the time a resident takes to perform the colectomy (TR) and the time an attending takes to perform the surgery (TA). We have two equations and two unknowns: (0.9346)TR + (0.0636)TA = 3.17 (0.0142)TR + (0.9808)TA = 2.52 Solving, we get TA = 2.523 and TR = 3.213 hours. This is very close to those in table 8 above, as almost all the surgery in the teaching hospital was with residents and almost all the surgery in the non-teaching hospital was with attendings. 4 The Relative Risk of Getting a Resident in Black Versus White Patients TB = TR * PB,R + TA * PB,A From Regression Data [TB − TW ] PRB = + 1 PRW ( PRW ) [TR − TA ] TW = TR * PW,R + TA * PW,A Or, since PB,R + PB,A = 1and PW,R + PW,A = 1, we can write: TB = TR * PB,R + TA * (1 - PB,R ) and TW = TR * PW,R + TA * (1 - PW,R ). Hence, TB-TW = PB,R * [TR - TA] - PW,R*[TR - TA] and: [TB - TW]/[TR-TA] = PB,R - PW,R Or: From External Data or Matching Studies [TB - TW]/ [PW,R][TR-TA] = [PB,R / PW,R]-1 and finally: 1+ {[TB - TW] / {[PW,R][TR-TA]}} = [PB,R / PW,R] Assuming: Relative Risk of Obtaining a Resident if a Black Patient Versus a White Patient TR-TA = 30 mins TB – TW = 15 mins Relative Risk for Blacks 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Percent of Cases Performed by a Resident if Patient is White Advantages of using Procedure Time as a Measure of Disparity • Bills available for literally all procedures performed in the OR for Medicare patients • Procedure length is highly dependent on the surgeon, not patient compliance • Procedure length is less dependent on the past history of the patient—insensitive to unobserved severity • Time is not game-able, especially since it has not been used to measure disparity in the past Conclusions Acknowledgements • Procedure time can be well estimated with Medicare claims data • Procedure time is a function of race and income and the specific hospital, after adjustment for other patient factors • There is a significant procedure length racial disparity inside of hospitals • Mechanisms for this procedure length disparity need to be better understood • This work was funded through a grant from the Leonard Davis Institute of Health Economics at PENN and supported in part from AHRQ Grant Number HS-09460 5 Additional Slides for Potential Questions The End TEACHING N=480 Patients Sometimes NON-TEACHING N=96 Surgeons N=480 Patients Sometimes TEACHING Sometimes NON-TEACHING P = 0.214 109 minutes Sometimes 103 minutes P = 0.657 P = 0.002 P = 0.393 P = 0.0001 N=480 Patients ALWAYS N=480 Patients ALWAYS ALWAYS 121 minutes P = 0.002 ALWAYS 106 minutes 6