Outcomes of complex Gastrointestinal Procedures Performed in ...
WISCONSIN MEDICAL JOURNAL Outcomes of complex Gastrointestinal Procedures Performed in a community Hospital Matthew H. Guzzo, MD; Jeffrey Landercasper, MD; William C. Boyd, MD; Pamela J. Lambert, RNaBStract a Whipple procedure. LOS, morbidity, and mortalityBackground: Complex gastrointestinal (GI) procedures were less than or equivalent to published reports fromhave been defined as those that are associated with high-volume medical centers.higher morbidity and mortality, require a high level of Conclusion: Excellent outcomes for complex GI pro-technical expertise, and occur in less than 6000 patients cedures can be achieved at lower volume medical cen-per year in the United States. Prior studies suggest a ters. Regionalization strategies to improve patient caredirect volume-outcome relationship. should be based on outcome studies rather than volumeHypothesis: Complex GI procedures may be performed alone.with good outcomes in a lower volume hospital with acommitment to surgical residency training. intrOductiOn Numerous studies have described better patient out-Methods: Retrospective chart review of all patients un- comes at high-volume hospitals. The inverse relation-dergoing non-emergent operations that are considered ship between patient volume and mortality seems es-complex GI procedures (esophagectomy, total gastrec- pecially evident for complex procedures. Demonstratedtomy, major hepatic resection, pancreaticoduodenec- reductions in hospital length of stay (LOS), cost, mor-tomy, biliary tract anastomosis, and total abdominal bidity, and mortality have motivated coalitions of pub-proctocolectomy) from July 1989-June 1997 in a rural lic and private purchasers (e.g., the Leapfrog Group)referral medical center. to encourage patients to undergo specified complexResults: One hundred six consecutive patients under- procedures at high-volume hospitals.1,2 Medical cen-went complex GI procedures during a 7-year period ters and insurance companies have also created careending June 1997. Patients ranged from 19–90 years pathways to identify preferred treatment regimens and(mean 62). Forty-eight patients (45.3%) had 1 or more more appropriate locations of care.3 Studies reportingmajor comorbidities. Seventy-three patients (68.9%) correlations between patient outcome and hospital orhad operations for malignancies. Average length of stay surgeon volume have examined a wide range of surgi-(LOS) was 13.2 days (range 5-38). Major complica- cal procedures and drawn different conclusions.2,4-18tions occurred in 15 patients (14%). Two patients died Gordon et al evaluated statewide outcomes of 6 com-(mortality 1.9%), 1 after esophagectomy and 1 after plex high-risk gastrointestinal (GI) surgical procedures based on the average number of cases performed each year in Maryland.5 They found that high-volume hos-Doctor Guzzo is a staff surgeon in Sheboygan, Wis and past pitals (>201 procedures/year) had better outcomesChief Resident in Surgery at Gundersen Lutheran Medical Center than medium- (21-50 procedures/year) or low-volumein La Crosse, Wis. Doctor Landercasper is a staff surgeon andpast chair of the General and Vascular Surgery Department at centers (<10-20 procedures/year). The study reportedGundersen Lutheran Medical Center and current co-director of herein examines the outcomes of the same 6 complex,the Norma J. Vingers Center for Breast Care. Doctor Boyd is the high-risk GI surgical procedures that were performedchair of the General and Vascular Surgery Clinic at GundersenLutheran Medical Center. Ms Lambert is a registered nurse in the at a lower-volume community hospital.Research Department of Gundersen Lutheran Medical Foundationin La Crosse, Wis. Please address correspondence to: Jeffrey MEtHOdSLandercasper, MD, Dept of Surgery, Gundersen Lutheran, 1836South Ave, La Crosse, WI 54601; phone: 608.775.2894; fax The study population was based on 6 procedures608.775.4460; e-mail firstname.lastname@example.org. deemed to be complex, high-risk GI surgical proce-30 Wisconsin Medical Journal 2005 • Volume 104, No. 6
WISCONSIN MEDICAL JOURNALdures by Gordon et al.5 These procedures are associ- Table 1. Volume of Complex GI Proceduresated with higher morbidity and mortality rates, require Mean LOS a high level of technical expertise, and occur in less than Type of Operation N (%) (Days) Range6000 patients per year in the United States. The 6 pro- Total abdominal colectomy 32 (30.2) 12.4 5-38cedures and their International Classification of Disease Total gastrectomy 26 (21.5) 12.5 5-27(ICD-9 CM) codes were as follows: 42.40 to 42.42—ex- Biliary tract anastomosis 21 (19.8) 12.5 5-27cision of the esophagus; 43.91 to 43.99—total gastrec- Whipple procedure 15 (14.2) 17.1 7-30tomy; 45.8—total abdominal colectomy; 50.3—hepatic Esophagectomy 9 (8.5) 14.8 10-20 Hepatic lobectomy 3 (2.8) 10.3 10-11lobectomy; 51.31 to 51.39—biliary tract anastomosis; Total 106 (100) 13.2 5-38and 52.7—radical pancreaticoduodenectomy. Patientswho underwent these 6 procedures emergently were LOS=Length of stayexcluded. Age, race, sex, and admission status wererecorded. Each patient’s comorbidity score was as- Table 2. Major Complications in 106 Patientssigned using the Dartmouth-Manitoba variation of the Complication N %Charlson Co-morbidity index.19,20 Each comorbid con- Anastomotic leak 5 4.7dition was assigned 1 point and then added together. A ARDS 2 1.9disease counted as a comorbid condition if it was coded Wound dehiscence 1 0.9on the discharge summary as a secondary diagnosis of Small bowel perforation 1 0.9either the index admission or any prior discharge lists. Brachial embolus 1 0.9 All operations were performed at Gundersen Enterocutaneous fistula 1 0.9Lutheran Health System, La Crosse, Wisconsin, a 325- Pancreatic leak 1 0.9 Intraoperative esophageal perforation 1 0.9bed tertiary care hospital in rural western Wisconsin. Death 2 1.9Gundersen Lutheran is staffed by 386 physicians in La Total 15 14.2Crosse and outreach clinics that cover a referral areaof 19 counties in 3 states, with 500,000 residents. The ARDS=acute respiratory distress syndromeinstitution provides 2 fully accredited categorical gen-eral surgery positions in a 5-year residency program. years (mean age 62 years). Eighty-eight (83%) patientsAll 6 complex procedures were performed with at least had either 0 or 1 comorbid disease. Fifteen patients1 resident and 1 staff physician. The surgical team that (14.2%) had 2, and 3 patients (2.8%) had >3 comorbidi-performed the complex operation monitored and was ties. Seventy-three (68.9%) patients underwent opera-in charge of the patients’ postoperative care. Minimal tions for malignancy. Table 1 lists the number of eachcross-coverage by residents and staff who were un- procedure performed.familiar with patients occurred. Night and weekend A total of 33 complications occurred in 22 patients, in-coverage was almost always provided by the attending cluding 15 major (45.5%) and 18 minor (54.5%) compli-surgeon and surgical resident who performed the pro- cations. Two perioperative deaths occurred, for an overallcedure. The surgical team remained primary physicians mortality rate of 1.9%. One patient died after esophago-for patients admitted to an open intensive care unit. gastrectomy and another after radical pancreaticoduode- Primary outcomes recorded for each patient included nectomy. Each of the two patients had 2 comorbid con-length of stay, perioperative death, and complications. A ditions. Table 2 lists each major complication.perioperative death was defined as death within 30 daysof the operation or the patient dying as a consequence diScuSSiOnof having surgery. A major complication was defined as Luft et al suggested a possible benefit of regionalizationany postoperative event or condition that was directly of medical resources in 1979 based on surgical volume.6related to the operative procedure and required medical Subsequently, numerous studies have been conductedor surgical intervention. to study the hypothesis that increased volume is associ- ated with better results.4,5-11,13-16,18 In evaluating outcomerESultS after esophagectomy, at least 3 reports revealed bet-The study population included 106 patients treated from ter outcomes in higher-volume centers.12 Dimick et alJuly 1989 to June 1997 at Gundersen Lutheran Hospital and Kuo et al found reductions in mortality rates for(average 13.25 cases/year). Patients were all Caucasians, esophagectomies in statewide studies in Maryland andwith a male majority (59.4%). Age ranged from 19 to 90 Massachusetts.21,22 Dimick et al also reported a shorter Wisconsin Medical Journal 2005 • Volume 104, No. 6 31
WISCONSIN MEDICAL JOURNALhospital stay and a lower cost for esophagectomies per- ogy practice with surgeons who were trained at “high-formed at high-volume hospitals in Maryland.21 Swisher volume” fellowship programs.8 Schwarz et al concludedet al compared outcomes from 13 national cancer in- that “quality can be independent of quantity.”8stitutions and 88 community hospitals.23 Mortality was In a report of esophagectomy performed at 2 low-reduced from 12.2% to 3.0% (P=0.05) in hospitals per- volume tertiary care centers (43 total patients duringforming more than 5 esophagectomies per year. 5 years), the 30-day mortality rate was 4.7% and the Choti et al studied outcomes after hepatic resections anastomotic leak rate was 11.6%.15 Padmanabhan et alin relation to hospital volume.13 Mortality in high-vol- concluded that elective esophagectomy could be safelyume groups was 1.5%, compared to 7.9% in low-vol- performed at low-volume centers and urged that re-ume hospitals (P<0.01, RR=5.2). They concluded that gionalization/referral patterns for high-risk proceduresboth major and minor hepatic surgery and resections should be guided by local outcomes and not by the totalfor metastatic disease could be performed more safely number of procedures performed at a specific center.15at higher-volume referral centers. Dimick et al detailed In their published report of surgical outcomes for 6the postoperative complication and mortality rate of all complex high-risk GI surgical procedures, Gordon etpatients undergoing hepatic resection in Maryland from al found that hospitals with higher procedure volumes1994 to 1998. High-volume hospitals (>60 cases over 5 in Maryland had a shorter LOS and a lower in-hospitalyears) had better outcomes.4 mortality rate.5 In their publicly available statewide da- Many high-volume centers have reported their 30- tabase, they compared low- and medium-volume hos-day mortality rates and postoperative mortality rates pitals to their own high-volume institution. Our studyafter pancreaticoduodenectomy.9,10 Begg et al and evaluated the same 6 high-risk procedures performedBirkmeyer et al reported mortality rates of 3.0% and at our institution during the same 8-year period end-5.8%, respectively.9,10 They concluded that their “high- ing in 1997. The 6 procedures were performed an aver-volume” centers had lower mortality rates than lower- age of 13.25 times per year. This would categorize ourvolume hospitals. Glasgow et al discovered similar find- hospital as a low-volume group as defined by Gordonings in their statewide evaluation of patient outcomes et al.5 Gordon et al concluded that higher-volume cen-after pancreatic resection in California.14 They found ters were more experienced with complex proceduresbetter outcomes and lower hospital charges in higher- and better able to decrease morbidity, leading to shortervolume centers and concluded that regionalizing such duration of hospital stay.5 Our study population had anhigh-risk procedures would better serve patients with average hospital stay of 13.2 days. This is lower thanpancreatic cancer.14 the 16.4 and 19.4 day average LOS for the high- and However, not all studies report better outcomes middle-volume providers reported in Maryland. Ourin higher-volume centers. The Veterans Health 30-day perioperative mortality rate was 1.9%, which isAdministration (VHA), in an effort to consolidate cost lower than the mortality rates reported in the statewidewithout compromising care, created the VHA National surgical registry of Maryland. In Maryland, Gordon etSurgical Quality Improvement Program. Data were col- al found an 8.4% mortality rate for medium-volumelected on 68,631 operations. Khuri and Daley found no hospitals and a 2.9% rate for higher-volume providersstatistically significant association between procedure (Table 3). Mortality was recorded in our series not justor specialty volume and 30-day mortality rate or 30-day for patients who died during the 30 days of the peri-stroke rate for carotid endarterectomies.7 The authors operative period, but also included those patients whoconcluded that the assumptions that better surgical out- were discharged home but expired thereafter from acomes are achieved in high-volume hospitals cannot be surgical complication. Evaluating only hospital mortal-proven and the “volume of surgery should not be used ity rates, as reported in the Maryland registry, wouldas a surrogate for quality of surgical care.”7 exclude those patients who did well initially only to Schwarz et al examined the outcome of a complex sur- expire from complications after being discharged fromgical procedure performed in a lower-volume practice.8 the hospital. The mean age, gender, elective admissionThey documented outcomes for 54 patients undergo- status, and percentage of patients with malignancy wereing pancreaticoduodenectomy for periampullary cancer all similar when comparing our study population to thefrom 1987-1998 (4.9 cases/year). The authors had no 30- Maryland registry.day perioperative deaths and their LOS improved from The science of how to assess and improve patient16.5 days to 12 days after 1995. They believed that their outcomes in a region or nation is complex and fraughtexcellent outcomes were related to an exclusive oncol- with many confounding variables.2,24,25 Investigators have32 Wisconsin Medical Journal 2005 • Volume 104, No. 6
WISCONSIN MEDICAL JOURNALattempted to correlate patient outcome with hospital Table 3. Mortality of High-Risk Procedures Based on Volumevolume, surgeon volume, American Board of Surgery Avg. # certification, surgical subspecialty certification, site of Volume Group* Cases/Year Morbidity Mortality P value†residency training (university versus non-university), Minimal 4.3 NR 14.2% .0003years of surgical experience, presence of a surgical resi- Low 13.4 NR 12.7% .0008dency program, presence of dedicated “intensivist” avail- Medium 23.8 NR 8.4% .0156ability 24 hours per day, closed versus open intensive care High 213.9 NR 2.9% .5342units, quality of nursing and anesthesia care, and quan- Gundersen Lutheran 13.5 15/106 2/106 — (14.2%) (1.9%)†tity of ICU nurses.4,16-18,26-28 Many investigations focuson only 1 of the aforementioned variables, but all may * Defined by Gordon et al.5be important. Factors that may confound these studies † Comparison of volume group in Maryland statewide registry5 to Gundersen Lutheran using Pearson’s Chi squareand render conclusions difficult to interpret are detailedin Table 4. Publication bias may also occur. Institutionsare more likely to submit and have their results acceptedfor publication if they are generally positive. There is an Table 4. Confounding Variables in Outcome Assessmentobvious disincentive to publish poor outcomes. Retrospective, inadequate and incorrect coding during data The relationship of a hospital’s teaching status to pa- retrievaltient outcome has also been studied. Flood et al stated Data collection from administrative and billing codes ratherthat a hospital’s teaching status had no effect on patient than medical databasesoutcomes.11 This study classified a hospital as a teach- Differing definitions of procedures, complications, and postop-ing hospital if both the American Medical Association erative death(AMA) and the American Hospital Association reported Case mix difference between hospitals and patientsthem to be conducting 1 or more residency training Comorbiditiesprograms and if the AMA source reported there to be Stage of diseaseat least 2 resident physicians actually present in the hos- Cultural variations (e.g., length of stay practice, incidence ofpital.11 These criteria do not address the type of residen- obesity, malnutrition, other)cies or the type of residents present in the hospital. In a Publication biassurgical outcomes assessment, the presence of a surgical Only best outcomes submitted and acceptedresidency may be more influential in determining out- Disincentive to publish poor outcomescome than other types of training programs. Continuity Difference in postoperative care, hospital support, anesthesia,of care and involvement of a surgical resident in-house quantity of ICU nurses, quality of nursing support, ICU support,may have contributed to the good outcomes demon- 24-hour intensiviststrated in this study. Teaching versus non-teaching hospital with 24-hour resident coveragecOncluSiOn Hawthorne effect (better outcomes occur when investigatorsRegionalization of surgical procedures is controversial. realize they are under observation)The outcomes study reported herein demonstrates goodresults in performing complex, high-risk GI proceduresin a low-volume center with a commitment to surgical acknOWlEdGMEntSeducation. Postoperative outcomes are dependent on Presented in part at the Wisconsin Surgical Society,multiple factors, and hospital volume by itself is not an Madison, Wisconsin, October, 2001. The authors grate-adequate measure of quality of care. Therefore, region- fully acknowledge the financial support of Gundersenalization of surgical procedures should not be based on Lutheran Medical Foundation and the technical assis-volume alone. tance of Angela Kuhn in manuscript preparation. We do not dispute the statistical association of theusual reported direct positive relationship between hos- rEFErEncES 1. Birkmeyer JD, Finlayson EVA, Birkmeyer CM. Volume stan-pital volume and outcome. However, the ultimate goal dards for high-risk surgery: potential benefits of the Leapfrogof best patient care is not simply achieved by regional- initiative. Surgery. 2001;130:415-422.ization of patients to a few medical centers but, rather, 2. Finlayson EVA, Birkmeyer JD. Effects of hospital volume on life expectancy after selected cancer operations in olderby directing patients to centers that demonstrate good adults: a decision analysis. J Am Coll Surg. 2003;196:410-outcomes regardless of volume. 417. Wisconsin Medical Journal 2005 • Volume 104, No. 6 33
WISCONSIN MEDICAL JOURNAL3. Lilienfeld D, Engin MS. Health policy research: information 17. Prystowsky, JB, Bordage G, Feinglass JM. Patient outcomes guiding clinical decision making at the population level. In: for segmental colon resection according to surgeon’s train- Gordon TA, Cameron JL, Eds. Evidence-Based Surgery. ing, certification, and experience. Surgery. 2002;132:663- Hamilton, London: B.C. Decker Inc; 2000;81-88. 672.4. Dimick JB, Pronovost PJ, Cowan JA, Lipsett, PA. 18. Cowan JA, Dimick JB, Thompson BG, et al. Surgeon volume Postoperative complication rates after hepatic resection in as an indicator of outcomes after carotid endarterectomy: an Maryland hospitals. Arch Surg. 2003;138:41-46. effect independent of specialty practice and hospital volume.5. Gordon TB, Bowman HM, Bass EB, et al. Complex gastroin- J Am Coll Surg. 2002;195:814-821. testinal surgery: impact of provider experience on clinical and 19. Romano PS, Roos LL, Jollis JG. Adapting a clinical comor- economic outcomes. J Am Coll Surg. 1999;189:46-56. bidity index for use with ICD-9-CM administrative data: differ-6. Luft HS, Bunker JP, Enthoven AC. Should operations be ing perspectives. J Clin Epidemiol. 1993;46:1075-1079. regionalized? the empirical relation between surgical volume 20. Charlson ME, Pompei P, Ales KL, et al. A new method of and mortality. N Engl J Med. 1979;301:1364-1369. classifying prognostic comorbidity in longitudinal studies: de-7. Khuri SF, Daley J. Relation of surgical volume to outcome in velopment and validation. J Chron Dis. 1987;40:373-383. eight common operations. Ann Surg. 1999;230:414-432. 21. Dimick JB, Cattaneo SM, Lipsett PA, et al. Hospital volume is8. Schwarz RE, Keny H, Ellenhorn JD. A mortality-free decade related to clinical and economic outcomes of esophageal re- of pancreatoduodenectomy: is quality independent of quan- section in Maryland. Presented at the 37th Annual Society of tity? Am Surg. 1999;10:949-954. Thoracic Surgeons meeting, January 29, 2001, New Orleans,9. Begg CB, Cramer LD, Hoskin WJ. Impact of hospital vol- LA. ume on operative mortality for major cancer surgery. JAMA. 22. Kuo EY, Chang Y, Wright CD. Impact of hospital volume 1998;280:1747-1751. on early clinical outcomes after esophagectomy for cancer.10. Birkmeyer JD, Finlayson SR. Effect of hospital volume on Presented at the 37th Annual Society of Thoracic Surgeons in-hospital mortality with pancreaticoduodenectomy. Surgery. meeting, January 29, 2001, New Orleans, LA. 1999;125:250-256. 23. Swisher SG, Deford L, Merriman KW, et al. Effect of opera-11. Flood AB, Scott WR, Ewy W. Does practice make perfect? tive volume on morbidity, mortality and hospital use after part II: the relation between volume and outcomes and other esophagectomy for cancer. J Thorac Cardiovasc Surg. hospital characteristics. Med Care. 1984;22:115-125. 2000;119:1126-1134.12. Mathisen DJ. What’s new in general thoracic surgery. Am 24. Russell TR. Invited commentary: volume standards for high- Coll Surg. 2001;192:737-749. risk operations: an American College of Surgeons’ view.13. Choti MA, Bowman HM, Pitt HA, et al. Should hepatic re- Surgery. 2001;130:423-424. sections be performed at high-volume referral centers? J 25. Daly JM. Clinical volume and patient outcome: how much Gastrointest Surg. 1998;2:11-20. cause and effect? Cont Surg. 2003;59:106-108.14. Glasgow RE, Mulvihill SJ. Hospital volume influences out- 26. Dimick JB, Pronovost PJ, Lipsett PA. The effect of ICU physi- come in patients undergoing pancreatic resection for cancer. cian staffing and hospital volume on outcomes after hepatic West J Med. 1996;165:294-300. resection. J Int Care Med. 2002:17:41-47.15. Padmanabhan RS, Byrnes MC, Helmer SD, et al. Should 27. Pronovost PJ, Jenckes MW, Dorman T, et al. Organizational esophagectomy be performed in a low-volume center? Am characteristics of intensive care units related to outcomes of Surg. 2002;68:348-352. abdominal aortic surgery. JAMA. 1999;281:1310-1317.16. Ko CY, Chang JT, Chaudhry S, et al. Are high-volume 28. Dimick JB, Swoboda SM, Pronovost PJ, et al. Effect of nurse surgeons and hospitals the most important predictors of to patient ratio in the intensive care unit on pulmonary com- in-hospital outcome for colon cancer resection? Surgery. plications and resource use after hepatectomy. Am J Crit 2002;132:268-273. Care. 2001;10:376-382.34 Wisconsin Medical Journal 2005 • Volume 104, No. 6