Original article
Impact of hospital volume on hospital mortality, length of stay and total costs after pancreaticoduodenectomy R. Yoshioka1 , H. Yasunaga2 , K. Hasegawa1 , H. Horiguchi3 , K. Fushimi4 , T. Aoki1 , Y. Sakamoto1 , Y. Sugawara1 and N. Kokudo1 1
Hepato-Biliary-Pancreatic Surgery Division, Department of Surgery, Graduate School of Medicine, and 2 Department of Health Economics and Epidemiology Research, School of Public Health, University of Tokyo, 3 Department of Clinical Data Management and Research, Clinical Research Centre, National Hospital Organization Headquarters, and 4 Department of Health Care Informatics, Tokyo Medical and Dental University, Tokyo, Japan Correspondence to: Dr N. Kokudo, Hepato-Biliary-Pancreatic Surgery Division, Department of Surgery, Graduate School of Medicine, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8655, Japan (e-mail: [email protected])
Background: High morbidity and mortality rates after pancreaticoduodenectomy (PD) have led to
concentration of this surgery in high-volume centres, with improved outcomes. The extent to which better outcomes might be apparent in a healthcare system where the mortality rate is already low is unclear. Methods: The Japanese Diagnosis Procedure Combination database was used to identify patients undergoing PD between 2007 and 2010. Patient data included age, sex, co-morbidities at admission, type of hospital, type of PD, and the year in which the patient was treated. Hospital volume was defined as the number of PDs performed annually at each hospital, and categorized into quintiles: very low-, low-, medium-, high- and very high-volume groups. The Charlson co-morbidity index was calculated using the International Classification of Diseases, tenth revision, codes of co-morbidities. Results: A total of 10 652 patients who underwent PD in 848 hospitals were identified. The overall in-hospital mortality rate after PD was 3·3 per cent (350 of 10 652), and for the groups ranged from 5·0 per cent for the very low-volume group to 1·4 per cent for the very high-volume group (P < 0·001). Multivariable analysis revealed a significant linear relationship between higher hospital volume and shorter postoperative length of stay compared with the very low-volume group, and between increasing hospital volume and lower total costs. Conclusion: A significant relationship exists between increasing hospital volume, lower in-hospital mortality, shorter length of stay and lower costs for patients undergoing PD in Japan. Centralization of PD in this healthcare system is therefore justified. Paper accepted 10 December 2013 Published online 24 February 2014 in Wiley Online Library (www.bjs.co.uk). DOI: 10.1002/bjs.9420
Introduction
Pancreaticoduodenectomy (PD) is a high-risk procedure. Although a relatively low postoperative mortality rate of 0·8–2 per cent has been reported1 – 3 for high-volume institutions, nationwide surveys4 – 6 suggest much higher mortality rates of 6·4–8·4 per cent. The inverse relationship between hospital volume and postoperative mortality after PD is well established4 – 11 . This volume–outcome relationship has also been demonstrated for length of stay and total inpatient costs12 – 14 . Longterm survival after resection for pancreatic cancer has also been related to higher hospital volume15,16 . A recent 2014 BJS Society Ltd Published by John Wiley & Sons Ltd
meta-analysis17 confirmed significant associations between hospital volume and both postoperative mortality and long-term survival after PD in studies adjusted for case mix. The benefits of centralization of PD on postoperative mortality have been advocated in Europe and the USA4,6,7,9 – 11 . In Japan, a similar relationship between hospital volume and short-term operative outcome has also been reported for several surgical procedures18,19 . The nationwide survey20 by the Japanese Society of Gastroenterological Surgery (JSGS) identified the relationship between hospital volume and mortality rates for major surgical procedures in gastrointestinal surgery, including PD. The postoperative BJS 2014; 101: 523–529
R. Yoshioka, H. Yasunaga, K. Hasegawa, H. Horiguchi, K. Fushimi, T. Aoki et al.
524
mortality rate of 2·7 per cent after PD (including highand low-volume centres) was lower than that reported for other countries. Although the data suggested an inverse relationship between hospital volume and postoperative mortality, the analysis lacked the risk adjustment that most would consider essential21 . The present study used a nationwide administrative database with risk-adjusted analyses to assess the relationship between hospital volume, postoperative mortality, length of stay and total costs after PD. The objective of the present study was to see whether the volume–outcome relationship was still present with the very low level of mortality identified in Japan. By 2010, there were 952 participating hospitals in the database, which included 3·2 million patients, representing approximately 45 per cent of all inpatient admissions to acute-care hospitals in Japan.
is conducted between July and December each year to collect administrative claims data and clinical details for all inpatients. The clinical data include: unique identifiers of hospitals; patient age and sex; diagnoses and co-morbidities recorded with text data in Japanese and the International Classification of Diseases 10th revision (ICD-10) codes; drugs administered; length of stay; discharge status; and total costs. Total costs are estimated using the item-by-item reference costs for surgical, pharmaceutical, laboratory and other inpatient services according to the Japanese national fee schedule. To optimize the accuracy of their diagnoses, attending physicians are obliged to record diagnoses with reference to medical charts. Given the anonymous nature of the data, informed consent was not required for this study, which was approved by the Institutional Review Board at the University of Tokyo.
Patient selection and data Methods
Diagnosis Procedure Combination database Details of the Diagnosis Procedure Combination (DPC) database have been described elsewhere19 . This case-mix patient classification system is linked with a lump-sum payment procedure. All 82 university hospitals in Japan are obliged to use this system, but adoption by community hospitals is voluntary. A survey of participating hospitals Table 1
Patients who had a PD between July and December in each year from 2007 to 2010 were identified. Because the data from January to June of each year were not available, the number of PDs at each hospital in each full year was calculated by doubling the number of PDs performed during the 6 months from July to December. Patient data included age, sex, co-morbidities at admission, type of hospital, type of PD (standard, PD with lymph node dissection, PD with resection of adjacent organ,
Patient demographics Annual hospital volume (no. of PDs/year)
No. of hospitals No. of patients Mean(s.d.) age (years) Sex ratio (F : M) Charlson co-morbidity index 0–2 3 4 ≥5 Patients in academic hospital Type of PD Standard PD with lymph node dissection PD with resection of adjacent organ PD with vascular resection and reconstruction Hepatopancreaticoduodenectomy Study period 2007 2008 2009 2010
Total
< 8 (very low)
8–11 (low)
12–17 (medium)
18–28 (high)
> 28 (very high)
P*
848 10 652 67·3(10·2) 4051 : 6601
497 2242 68·1(9·5) 809 : 1433
159 2031 68·1(9·7) 783 : 1248
101 2216 67·4(10·5) 890 : 1326
60 2123 66·9(10·3) 820 : 1303
31 2040 65·8(10·7) 749 : 1291
4589 (43·1) 3384 (31·8) 1515 (14·2) 1164 (10·9) 3451 (32·4)
972 (43·4) 724 (32·3) 304 (13·6) 242 (10·8) 90 (4·0)
967 (47·6) 593 (29·2) 246 (12·1) 225 (11·1) 235 (11·6)
953 (43·0) 701 (31·6) 348 (15·7) 214 (9·7) 720 (32·5)
921 (43·4) 676 (31·8) 290 (13·7) 236 (11·1) 943 (44·4)
776 (38·0) 690 (33·8) 327 (16·0) 247 (12·1) 1463 (71·7)
2114 (19·8) 5523 (51·8) 1429 (13·4) 1500 (14·1) 86 (0·8)
638 (28·5) 1021 (45·5) 414 (18·5) 159 (7·1) 10 (0·4)
471 (23·2) 994 (48·9) 320 (15·8) 234 (11·5) 12 (0·6)
470 (21·2) 1156 (52·2) 280 (12·6) 298 (13·4) 12 (0·5)
301 (14·2) 1229 (57·9) 223 (10·5) 353 (16·6) 17 (0·8)
234 (11·5) 1123 (55·0) 192 (9·4) 456 (22·4) 35 (1·7)
< 0·001
2287 (21·5) 2673 (25·1) 2478 (23·3) 3214 (30·2)
491 (21·9) 528 (23·6) 596 (26·6) 627 (28·0)
433 (21·3) 479 (23·6) 518 (25·5) 601 (29·6)
444 (20·0) 549 (24·8) 521 (23·5) 702 (31·7)
486 (22·9) 552 (26·0) 426 (20·1) 659 (31·0)
433 (21·2) 565 (27·7) 417 (20·4) 625 (30·6)
< 0·001
< 0·001† < 0·001 < 0·001
< 0·001
2
Values in parentheses are percentages unless indicated otherwise. PD, pancreaticoduodenectomy. *χ test, except †ANOVA.
2014 BJS Society Ltd Published by John Wiley & Sons Ltd
www.bjs.co.uk
BJS 2014; 101: 523–529
Outcome–volume relationship for pancreaticoduodenectomy
PD with vascular resection and reconstruction, and hepatopancreaticoduodenectomy), and the year in which the patient was treated. Hospital volume was defined as the number of PDs performed annually at each hospital, and was categorized into quintiles (very low, low, medium, high and very high volume), with approximately equal numbers of patients assigned to each group. The Charlson co-morbidity index (CCI) was calculated using ICD-10 codes of co-morbidities based on the protocol of Quan and colleagues22,23 , in which a weight (1–6) is assigned to each of 17 co-morbid disease categories, and the score is the sum of those weights. This method has been well validated, and has produced consistent results in studies using a large administrative database22,23 .
Outcome measurements The primary endpoint was in-hospital mortality, defined as death at any time before hospital discharge. Secondary Table 2
525
endpoints were postoperative length of stay and total costs during the hospital stay.
Statistical analysis Descriptive statistics of the patient population included mean(s.d.) values to describe continuous variables and proportions to describe categorical variables. χ2 tests were used to compare categorical data, and analyses of variance to compare continuous variables among the five hospital volume categories. Multivariable logistic regression analysis was performed for in-hospital mortality, and multivariable linear regression analyses for length of stay and total costs (given in euros, based on US $1 = ¤0·80) to examine associations between hospital volume and outcomes with adjustment for patient backgrounds. Only variables with P ≤ 0·100 in univariable analysis were entered into multivariable analyses. Because the multicentre data were structured hierarchically into two
Bivariable analyses and multivariable logistic regression for in-hospital mortality
Sex M F Age (years) ≤ 59 60–69 70–79 ≥ 80 Charlson co-morbidity index 0–2 3 4 ≥5 Type of hospital Academic Non-academic Hospital volume (no. of PDs/year) < 8 (very low) 8–11 (low) 12–17 (medium) 18–28 (high) > 28 (very high) Type of PD Standard PD with lymph node dissection PD with resection of adjacent organ PD with vascular resection and reconstruction Hepatopancreaticoduodenectomy Study period 2007 2008 2009 2010
No. of patients (n = 10 652)
In-hospital mortality*
P‡
Adjusted odds ratio†
P
6601 4051
264 (4·0) 86 (2·1)
< 0·001
Reference 0·48 (0·38, 0·62)
< 0·001
2085 3640 4073 854
31 (1·5) 108 (3·0) 162 (4·0) 49 (5·7)
< 0·001
Reference 1·94 (1·30, 2·87) 2·75 (1·86, 4·01) 4·15 (2·61, 6·58)
0·001 < 0·001 < 0·001
4589 3384 1515 1164
127 (2·8) 99 (2·9) 41 (2·7) 83 (7·1)
< 0·001
Reference 1·01 (0·76, 1·34) 0·90 (0·62, 1·31) 2·55 (1·85, 3·50)
0·968 0·587 < 0·001
3451 7201
82 (2·4) 268 (3·7)
< 0·001
Reference 1·03 (0·72, 1·47)
0·869
2242 2031 2216 2123 2040
113 (5·0) 82 (4·0) 68 (3·1) 59 (2·8) 28 (1·4)
< 0·001
Reference 0·78 (0·56, 1·08) 0·61 (0·43, 0·86) 0·53 (0·37, 0·76) 0·25 (0·14, 0·43)
0·131 0·006 0·001 < 0·001
2114 5523 1429 1500 86
72 (3·4) 158 (2·9) 55 (3·8) 61 (4·1) 4 (5)
0·090
Reference 0·95 (0·71, 1·28) 1·11 (0·77, 1·59) 1·60 (1·11, 2·29) 2·00 (0·67, 5·98)
0·750 0·575 0·012 0·214
2287 2673 2478 3214
79 (3·5) 76 (2·8) 83 (3·3) 112 (3·5)
0·516
Values in parentheses are *percentages and †95 per cent confidence intervals. PD, pancreaticoduodenectomy. ‡χ2 test.
2014 BJS Society Ltd Published by John Wiley & Sons Ltd
www.bjs.co.uk
BJS 2014; 101: 523–529
R. Yoshioka, H. Yasunaga, K. Hasegawa, H. Horiguchi, K. Fushimi, T. Aoki et al.
levels (hospitals and patients), clustering of outcomes within hospitals was accounted for by using generalized estimating equations24 . The threshold for significance was P < 0·050. All statistical analyses were conducted using SPSS version 19 (IBM, Armonk, New York, USA).
60
Hospitals Patients
50 40 %
526
30 20
Results 10
Table 3
0
Low
Very low
Medium
High
Very high
Annual hospital volume
Proportion of hospitals in each volume group, and the proportion of patients who underwent pancreaticoduodenectomy at these hospitals
Fig. 1
60
50 In-hospital mortality (%)
A total of 10 652 patients who underwent PD in 848 hospitals during the survey period were identified. There were 2242, 2031, 2216, 2123 and 2040 patients in the very low (fewer than 8 PDs per year), low (8–11 PDs per year), medium (12–17 PDs per year), high (18–28 PDs per year) and very high (more than 28 PDs per year) volume groups respectively. Table 1 shows the demographics of the patients in these five hospital-volume groups. Patients in the very high-volume group were younger than those in the very low-volume group (65·8(10·7) versus 68·1(9·5) years respectively; P < 0·001). There was a higher proportion of patients with a CCI of 5 or more in the very highvolume group (12·1 per cent versus 10·8 per cent in the very low-volume group; P < 0·001). The proportion with combined vascular resection was higher in the very highvolume group compared with that in the very low-volume group (22·4 versus 7·1 per cent respectively; P < 0·001). The proportion of hospitals performing fewer than eight PDs annually was 58·6 per cent (Fig. 1). The overall in-hospital mortality after PD was 3·3 per cent (350 of 10 652 patients), and for the groups ranged from 5·0 per cent for very low-volume to 1·4 per cent for very high-volume groups (P < 0·001, χ2 test) (Table 2). Logistic regression analysis indicated an adjusted odds ratio (OR) of 0·25 (95 per cent confidence interval (c.i.) 0·14 to 0·43; P < 0·001), for the very high-volume group compared with the very low-volume group. In χ2 analysis, academic hospitals had significantly lower inhospital mortality than non-academic hospitals (P < 0·001), but no significant difference was found in multivariable logistic regression analysis (adjusted OR 1·03, 0·72 to 1·47; P = 0·869) (Table 2). Regarding type of PD, although no
40
30
20
10
0
10
20 40 30 Hospital volume (PDs per year)
50
Scatter plot of hospitals according to annual hospital volume of pancreaticoduodenectomies (PDs) and in-hospital mortality rate
Fig. 2
significant difference in in-hospital mortality was found on χ2 testing (P = 0·090), logistic regression analysis showed a significant difference between standard PD and PD with vascular resection and reconstruction (adjusted OR
Multivariable linear regression for postoperative length of stay and total costs Postop. length of stay (days)
Hospital volume (no. of PDs/year) < 8 (very low) 8–11 (low) 12–17 (medium) 18–28 (high) > 28 (very high)
Total costs (¤)
Mean(s.d.)
Coefficient
P
Mean(s.d.)
Coefficient
P
43·9(24·3) 40·5(22·9) 37·2(21·7) 36·9(21·9) 32·8(19·2)
Reference −3·4 (−5·4, −1·3) −6·6 (−8·6, −4·4) −7·0 (−9·5, −4·5) −11·3 (−14·6, −8·0)
0·001 < 0·001 < 0·001 < 0·001
40 294(17 150) 38 532(17 188) 36 336(17 038) 37 416(26 650) 33 719(16 894)
Reference −2244 (−3776, −714) −5206 (−6824, −3589) −4813 (−6605, −3022) −10 076 (−12 507, −7645)
0·004 < 0·001 < 0·001 < 0·001
Values in parentheses are 95 per cent confidence intervals unless indicated otherwise. PD, pancreaticoduodenectomy.
2014 BJS Society Ltd Published by John Wiley & Sons Ltd
www.bjs.co.uk
BJS 2014; 101: 523–529
Outcome–volume relationship for pancreaticoduodenectomy
1·60, 1·11 to 2·29; P = 0·012). In-hospital mortality for all hospitals is shown in Fig. 2. In the multivariable analysis, a significant linear relationship was found between higher hospital volume and shorter postoperative length of stay, and between increasing hospital volume and lower total costs (Table 3). Discussion
The present study found a significant inverse relationship between hospital volume and in-hospital mortality for patients undergoing PD. A higher hospital volume was also related to shorter postoperative length of stay and lower total costs. Although the overall mortality rate of 3·3 per cent in this study was low compared with that in nationwide surveys in Western countries, with reported rates between 6·4 and 7·5 per cent4 – 6,9,25 , the same relationship between hospital volume and in-hospital mortality was found. This provides supportive evidence for the validity of centralization of PD in Japan and further concentration of resources in other countries where this process has been started. The results are consistent with those of a study by the JSGS20 , although the present investigation is the first from Japan to record an inverse relationship following PD between hospital volume and in-hospital mortality using risk-adjustment analysis. Although many studies have reported on the volume–outcome relationship, factors that influence this have not been elucidated completely. Schmidt and colleagues1 reported surgeon experience to be an important dominant of postoperative morbidity. In the present study, some low-volume hospitals had in-hospital mortality rates as low as those in some very high-volume hospitals (Fig. 2), implying that surgeon volume might be important. One study26 attributed 51 per cent of the hospital volume effect to surgeon volume, whereas another27 found that system clinical resources were more influential. These factors could not be evaluated by the DPC database. In the present study, higher hospital volume was associated with shorter length of stay and lower total costs. Several reports have examined the relationship between hospital volume and costs of hospitalization for PD. Although significantly lower costs, attributed to better postoperative outcomes, were identified in one tertiary referral hospital compared with community hospitals in Maryland, USA12 , another US study found no association between higher hospital volume and lower costs, but did find an inverse relationship between surgeon volume and costs14 . More resources in high-volume centres might result in better postoperative outcomes and lower costs. Prolonged hospital stays and increased costs 2014 BJS Society Ltd Published by John Wiley & Sons Ltd
527
may reflect postoperative complications, late detection or inappropriate management of complications. The volume–outcome relationship is likely, therefore, to reflect the quality of the operation and postoperative management. Even in very high-volume hospitals, the mean postoperative length of stay in the present study was 32·8 days, much longer than that seen in the West at around 10 days2,3 . According to Organization for Economic Cooperation and Development (OECD) Health Data 201228 , the national mean length of hospital stay in Japan was 18·2 days, compared with 4·9 days in the USA and 7·7 days in the UK. This probably reflects differences in the healthcare system between Japan and Western countries, as well as cultural attitudes. In Japan, most hospitals provide both early postoperative care and subsequent rehabilitation in a single hospitalization29 . Readmission rates within 30 days of PD may be around 15 per cent in the West30 , so that some events occurring after discharge in Western countries would be included during the postoperative stay in Japan. In the course of promoting centralization, a major concern is preserving patient accessibility to optimal care, where the capacity of high-volume hospitals, potential delays and distance the patient must travel are important31,32 . More than 20 per cent of PD operations in Japan are carried out in very low-volume centres; nearly 60 per cent of the hospitals in the present study were in this category (Fig. 1). Although the present study has the advantage of a very large sample size collected over a short, recent period, there are several limitations. As adoption of the DPC system by community hospitals is voluntary, the database may not be representative of all hospitals. In particular, the participation rate of very low-volume hospitals may have been underestimated, so that overall in-hospital mortality may be higher in Japan as a whole. Details of postoperative complications, notably pancreatic fistula, delayed gastric emptying and postoperative haemorrhage that are important after pancreatic surgery33 – 35 , were not included in the DPC database. Length of stay and total costs were used as surrogate markers of the postoperative course following PD. Surgeon volume was not included in the analysis owing to lack of data and, finally, because the DPC database is administrative, information on postdischarge and long-term outcomes were not available. Despite these limitations, the strong linear relationship between higher hospital volume and decreased in-hospital mortality, shorter length of stay and reduced total costs, supports the validity of centralization of PD. www.bjs.co.uk
BJS 2014; 101: 523–529
528
R. Yoshioka, H. Yasunaga, K. Hasegawa, H. Horiguchi, K. Fushimi, T. Aoki et al.
Acknowledgements
This study was funded by a Grant-in-Aid for Research on Policy Planning and Evaluation from the Ministry of Health, Labour and Welfare, Japan (grant no. H22Policy-031), by a Grant-in-Aid for Scientific Research B (no. 22390131) from the Ministry of Education, Culture, Sports, Science and Technology, and by the Funding Programme for World-Leading Innovative R&D on Science and Technology (FIRST programme) from the Council for Science and Technology Policy, Japan (grant no. 0301002001001). Disclosure: The authors declare no conflict of interest. References 1 Schmidt CM, Turrini O, Parikh P, House MG, Zyromski NJ, Nakeeb A et al. Effect of hospital volume, surgeon experience, and surgeon volume on patient outcomes after pancreaticoduodenectomy: a single-institution experience. Arch Surg 2010; 145: 634–640. 2 Fern´andez-del Castillo C, Morales-Oyarvide V, McGrath D, Wargo JA, Ferrone CR, Thayer SP et al. Evolution of the Whipple procedure at the Massachusetts General Hospital. Surgery 2012; 152: S56–S63. 3 Cameron JL, Riall TS, Coleman J, Belcher KA. One thousand consecutive pancreaticoduodenectomies. Ann Surg 2006; 244: 10–15. 4 Topal B, Van de Sande S, Fieuws S, Penninckx F. Effect of centralization of pancreaticoduodenectomy on nationwide hospital mortality and length of stay. Br J Surg 2007; 94: 1377–1381. 5 Balzano G, Zerbi A, Capretti G, Rocchetti S, Capitanio V, Di Carlo V. Effect of hospital volume on outcome of pancreaticoduodenectomy in Italy. Br J Surg 2008; 95: 357–362. 6 de Wilde RF, Besselink MG, van der Tweel I, de Hingh IH, van Eijck CH, Dejong CH et al. Impact of nationwide centralization of pancreaticoduodenectomy on hospital mortality. Br J Surg 2012; 99: 404–410. 7 Learn PA, Bach PB. A decade of mortality reductions in major oncologic surgery: the impact of centralization and quality improvement. Med Care 2010; 48: 1041–1049. 8 van Heek NT, Kuhlmann KF, Scholten RJ, de Castro SM, Busch OR, van Gulik TM et al. Hospital volume and mortality after pancreatic resection: a systematic review and an evaluation of intervention in the Netherlands. Ann Surg 2005; 242: 781–788. 9 McPhee JT, Hill JS, Whalen GF, Zayaruzny M, Litwin DE, Sullivan ME et al. Perioperative mortality for pancreatectomy: a national perspective. Ann Surg 2007; 246: 246–253. 10 Gordon TA, Bowman HM, Tielsch JM, Bass EB, Burleyson GP, Cameron JL. Statewide regionalization of pancreaticoduodenectomy and its effect on in-hospital mortality. Ann Surg 1998; 228: 71–78. 2014 BJS Society Ltd Published by John Wiley & Sons Ltd
11 Finks JF, Osborne NH, Birkmeyer JD. Trends in hospital volume and operative mortality for high-risk surgery. N Engl J Med 2011; 364: 2128–2137. 12 Gordon TA, Burleyson GP, Tielsch JM, Cameron JL. The effects of regionalization on cost and outcome for one general high-risk surgical procedure. Ann Surg 1995; 221: 43–49. 13 Rosemurgy A, Cowgill S, Coe B, Thomas A, Al-Saadi S, Goldin S et al. Frequency with which surgeons undertake pancreaticoduodenectomy continues to determine length of stay, hospital charges, and in-hospital mortality. J Gastrointest Surg 2008; 12: 442–449. 14 Ho V, Aloia T. Hospital volume, surgeon volume, and patient costs for cancer surgery. Med Care 2008; 46: 718–725. 15 Bilimoria KY, Talamonti MS, Sener SF, Bilimoria MM, Stewart AK, Winchester DP et al. Effect of hospital volume on margin status after pancreaticoduodenectomy for cancer. J Am Coll Surg 2008; 207: 510–519. 16 Birkmeyer JD, Sun Y, Wong SL, Stukel TA. Hospital volume and late survival after cancer surgery. Ann Surg 2007; 245: 777–783. 17 Gooiker GA, van Gijn W, Wouters MW, Post PN, van de Velde CJ, Tollenaar RA; Signalling Committee Cancer of the Dutch Cancer Society. Systematic review and meta-analysis of the volume–outcome relationship in pancreatic surgery. Br J Surg 2011; 98: 485–494. 18 Otake H, Yasunaga H, Horiguchi H, Matsutani N, Matsuda S, Ohe K. Impact of hospital volume on chest tube duration, length of stay, and mortality after lobectomy. Ann Thorac Surg 2011; 92: 1069–1074. 19 Yasunaga H, Horiguchi H, Matsuda S, Fushimi K, Hashimoto H, Ohe K et al. Relationship between hospital volume and operative mortality for liver resection: data from the Japanese Diagnosis Procedure Combination database. Hepatol Res 2012; 42: 1073–1080. 20 Suzuki H, Gotoh M, Sugihara K, Kitagawa Y, Kimura W, Kondo S et al. Nationwide survey and establishment of a clinical database for gastrointestinal surgery in Japan: targeting integration of a cancer registration system and improving the outcome of cancer treatment. Cancer Sci 2011; 102: 226–230. 21 Khuri SF, Daley J, Henderson W, Hur K, Gibbs JO, Barbour G et al. Risk adjustment of the postoperative mortality rate for the comparative assessment of the quality of surgical care: results of the National Veterans Affairs Surgical Risk Study. J Am Coll Surg 1997; 185: 315–327. 22 Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005; 43: 1130–1139. 23 Quan H, Li B, Couris CM, Fushimi K, Graham P, Hider P et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol 2011; 173: 676–682.
www.bjs.co.uk
BJS 2014; 101: 523–529
Outcome–volume relationship for pancreaticoduodenectomy
529
24 Hubbard AE, Ahern J, Fleischer NL, Van der Laan M, Lippman SA, Jewell N et al. To GEE or not to GEE: comparing population average and mixed models for estimating the associations between neighborhood risk factors and health. Epidemiology 2010; 21: 467–474. 25 Teh SH, Diggs BS, Deveney CW, Sheppard BC. Patient and hospital characteristics on the variance of perioperative outcomes for pancreatic resection in the United States: a plea for outcome-based and not volume-based referral guidelines. Arch Surg 2009; 144: 713–721. 26 Nathan H, Cameron JL, Choti MA, Schulick RD, Pawlik TM. The volume–outcomes effect in hepato-pancreato-biliary surgery: hospital versus surgeon contributions and specificity of the relationship. J Am Coll Surg 2009; 208: 528–538. 27 Joseph B, Morton JM, Hernandez-Boussard T, Rubinfeld I, Faraj C, Velanovich V. Relationship between hospital volume, system clinical resources, and mortality in pancreatic Resection. J Am Coll Surg 2009; 208: 520–527. 28 Organization for Economic Cooperation and Development (OECD). OECD Health Data 2012. http://www.oecd.org/ health/healthpoliciesanddata/oecdhealthdata2012.htm [accessed 7 January 2013]. 29 Hashimoto H, Ikegami N, Shibuya K, Izumida N, Noguchi H, Yasunaga H et al. Cost containment and quality of care in Japan: is there a trade-off? Lancet 2011; 378: 1174–1182.
30 Ahmad SA, Edwards MJ, Sutton JM, Grewal SS, Hanseman DJ, Maithel SK et al. Factors influencing readmission after pancreaticoduodenectomy: a multiinstitutional study of 1302 patients. Ann Surg 2012; 256: 529–537. 31 Forshaw MJ, Gossage JA, Stephens J, Strauss D, Botha AJ, Atkinson S et al. Centralisation of oesophagogastric cancer services: can specialist units deliver? Ann R Coll Surg Engl 2006; 88: 566–570. 32 Stitzenberg KB, Sigurdson ER, Egleston BL, Starkey RB, Meropol NJ. Centralization of cancer surgery: implications for patient access to optimal care. J Clin Oncol 2009; 27: 4671–4678. 33 Wente MN, Bassi C, Dervenis C, Fingerhut A, Gouma DJ, Izbicki JR et al. Delayed gastric emptying (DGE) after pancreatic surgery: a suggested definition by the International Study Group of Pancreatic Surgery (ISGPS). Surgery 2007; 142: 761–768. 34 Bassi C, Dervenis C, Butturini G, Fingerhut A, Yeo C, Izbicki J et al. Postoperative pancreatic fistula: an international study group (ISGPF) definition. Surgery 2005; 138: 8–13. 35 Wente MN, Veit JA, Bassi C, Dervenis C, Fingerhut A, Gouma DJ et al. Postpancreatectomy hemorrhage (PPH): an International Study Group of Pancreatic Surgery (ISGPS) definition. Surgery 2007; 142: 20–25.
2014 BJS Society Ltd Published by John Wiley & Sons Ltd
www.bjs.co.uk
BJS 2014; 101: 523–529
Impact of hospital volume on hospital mortality, length of stay and total costs after pancreaticoduodenectomy R. Yoshioka1 , H. Yasunaga2 , K. Hasegawa1 , H. Horiguchi3 , K. Fushimi4 , T. Aoki1 , Y. Sakamoto1 , Y. Sugawara1 and N. Kokudo1 1
Hepato-Biliary-Pancreatic Surgery Division, Department of Surgery, Graduate School of Medicine, and 2 Department of Health Economics and Epidemiology Research, School of Public Health, University of Tokyo, 3 Department of Clinical Data Management and Research, Clinical Research Centre, National Hospital Organization Headquarters, and 4 Department of Health Care Informatics, Tokyo Medical and Dental University, Tokyo, Japan Correspondence to: Dr N. Kokudo, Hepato-Biliary-Pancreatic Surgery Division, Department of Surgery, Graduate School of Medicine, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8655, Japan (e-mail: [email protected])
Background: High morbidity and mortality rates after pancreaticoduodenectomy (PD) have led to
concentration of this surgery in high-volume centres, with improved outcomes. The extent to which better outcomes might be apparent in a healthcare system where the mortality rate is already low is unclear. Methods: The Japanese Diagnosis Procedure Combination database was used to identify patients undergoing PD between 2007 and 2010. Patient data included age, sex, co-morbidities at admission, type of hospital, type of PD, and the year in which the patient was treated. Hospital volume was defined as the number of PDs performed annually at each hospital, and categorized into quintiles: very low-, low-, medium-, high- and very high-volume groups. The Charlson co-morbidity index was calculated using the International Classification of Diseases, tenth revision, codes of co-morbidities. Results: A total of 10 652 patients who underwent PD in 848 hospitals were identified. The overall in-hospital mortality rate after PD was 3·3 per cent (350 of 10 652), and for the groups ranged from 5·0 per cent for the very low-volume group to 1·4 per cent for the very high-volume group (P < 0·001). Multivariable analysis revealed a significant linear relationship between higher hospital volume and shorter postoperative length of stay compared with the very low-volume group, and between increasing hospital volume and lower total costs. Conclusion: A significant relationship exists between increasing hospital volume, lower in-hospital mortality, shorter length of stay and lower costs for patients undergoing PD in Japan. Centralization of PD in this healthcare system is therefore justified. Paper accepted 10 December 2013 Published online 24 February 2014 in Wiley Online Library (www.bjs.co.uk). DOI: 10.1002/bjs.9420
Introduction
Pancreaticoduodenectomy (PD) is a high-risk procedure. Although a relatively low postoperative mortality rate of 0·8–2 per cent has been reported1 – 3 for high-volume institutions, nationwide surveys4 – 6 suggest much higher mortality rates of 6·4–8·4 per cent. The inverse relationship between hospital volume and postoperative mortality after PD is well established4 – 11 . This volume–outcome relationship has also been demonstrated for length of stay and total inpatient costs12 – 14 . Longterm survival after resection for pancreatic cancer has also been related to higher hospital volume15,16 . A recent 2014 BJS Society Ltd Published by John Wiley & Sons Ltd
meta-analysis17 confirmed significant associations between hospital volume and both postoperative mortality and long-term survival after PD in studies adjusted for case mix. The benefits of centralization of PD on postoperative mortality have been advocated in Europe and the USA4,6,7,9 – 11 . In Japan, a similar relationship between hospital volume and short-term operative outcome has also been reported for several surgical procedures18,19 . The nationwide survey20 by the Japanese Society of Gastroenterological Surgery (JSGS) identified the relationship between hospital volume and mortality rates for major surgical procedures in gastrointestinal surgery, including PD. The postoperative BJS 2014; 101: 523–529
R. Yoshioka, H. Yasunaga, K. Hasegawa, H. Horiguchi, K. Fushimi, T. Aoki et al.
524
mortality rate of 2·7 per cent after PD (including highand low-volume centres) was lower than that reported for other countries. Although the data suggested an inverse relationship between hospital volume and postoperative mortality, the analysis lacked the risk adjustment that most would consider essential21 . The present study used a nationwide administrative database with risk-adjusted analyses to assess the relationship between hospital volume, postoperative mortality, length of stay and total costs after PD. The objective of the present study was to see whether the volume–outcome relationship was still present with the very low level of mortality identified in Japan. By 2010, there were 952 participating hospitals in the database, which included 3·2 million patients, representing approximately 45 per cent of all inpatient admissions to acute-care hospitals in Japan.
is conducted between July and December each year to collect administrative claims data and clinical details for all inpatients. The clinical data include: unique identifiers of hospitals; patient age and sex; diagnoses and co-morbidities recorded with text data in Japanese and the International Classification of Diseases 10th revision (ICD-10) codes; drugs administered; length of stay; discharge status; and total costs. Total costs are estimated using the item-by-item reference costs for surgical, pharmaceutical, laboratory and other inpatient services according to the Japanese national fee schedule. To optimize the accuracy of their diagnoses, attending physicians are obliged to record diagnoses with reference to medical charts. Given the anonymous nature of the data, informed consent was not required for this study, which was approved by the Institutional Review Board at the University of Tokyo.
Patient selection and data Methods
Diagnosis Procedure Combination database Details of the Diagnosis Procedure Combination (DPC) database have been described elsewhere19 . This case-mix patient classification system is linked with a lump-sum payment procedure. All 82 university hospitals in Japan are obliged to use this system, but adoption by community hospitals is voluntary. A survey of participating hospitals Table 1
Patients who had a PD between July and December in each year from 2007 to 2010 were identified. Because the data from January to June of each year were not available, the number of PDs at each hospital in each full year was calculated by doubling the number of PDs performed during the 6 months from July to December. Patient data included age, sex, co-morbidities at admission, type of hospital, type of PD (standard, PD with lymph node dissection, PD with resection of adjacent organ,
Patient demographics Annual hospital volume (no. of PDs/year)
No. of hospitals No. of patients Mean(s.d.) age (years) Sex ratio (F : M) Charlson co-morbidity index 0–2 3 4 ≥5 Patients in academic hospital Type of PD Standard PD with lymph node dissection PD with resection of adjacent organ PD with vascular resection and reconstruction Hepatopancreaticoduodenectomy Study period 2007 2008 2009 2010
Total
< 8 (very low)
8–11 (low)
12–17 (medium)
18–28 (high)
> 28 (very high)
P*
848 10 652 67·3(10·2) 4051 : 6601
497 2242 68·1(9·5) 809 : 1433
159 2031 68·1(9·7) 783 : 1248
101 2216 67·4(10·5) 890 : 1326
60 2123 66·9(10·3) 820 : 1303
31 2040 65·8(10·7) 749 : 1291
4589 (43·1) 3384 (31·8) 1515 (14·2) 1164 (10·9) 3451 (32·4)
972 (43·4) 724 (32·3) 304 (13·6) 242 (10·8) 90 (4·0)
967 (47·6) 593 (29·2) 246 (12·1) 225 (11·1) 235 (11·6)
953 (43·0) 701 (31·6) 348 (15·7) 214 (9·7) 720 (32·5)
921 (43·4) 676 (31·8) 290 (13·7) 236 (11·1) 943 (44·4)
776 (38·0) 690 (33·8) 327 (16·0) 247 (12·1) 1463 (71·7)
2114 (19·8) 5523 (51·8) 1429 (13·4) 1500 (14·1) 86 (0·8)
638 (28·5) 1021 (45·5) 414 (18·5) 159 (7·1) 10 (0·4)
471 (23·2) 994 (48·9) 320 (15·8) 234 (11·5) 12 (0·6)
470 (21·2) 1156 (52·2) 280 (12·6) 298 (13·4) 12 (0·5)
301 (14·2) 1229 (57·9) 223 (10·5) 353 (16·6) 17 (0·8)
234 (11·5) 1123 (55·0) 192 (9·4) 456 (22·4) 35 (1·7)
< 0·001
2287 (21·5) 2673 (25·1) 2478 (23·3) 3214 (30·2)
491 (21·9) 528 (23·6) 596 (26·6) 627 (28·0)
433 (21·3) 479 (23·6) 518 (25·5) 601 (29·6)
444 (20·0) 549 (24·8) 521 (23·5) 702 (31·7)
486 (22·9) 552 (26·0) 426 (20·1) 659 (31·0)
433 (21·2) 565 (27·7) 417 (20·4) 625 (30·6)
< 0·001
< 0·001† < 0·001 < 0·001
< 0·001
2
Values in parentheses are percentages unless indicated otherwise. PD, pancreaticoduodenectomy. *χ test, except †ANOVA.
2014 BJS Society Ltd Published by John Wiley & Sons Ltd
www.bjs.co.uk
BJS 2014; 101: 523–529
Outcome–volume relationship for pancreaticoduodenectomy
PD with vascular resection and reconstruction, and hepatopancreaticoduodenectomy), and the year in which the patient was treated. Hospital volume was defined as the number of PDs performed annually at each hospital, and was categorized into quintiles (very low, low, medium, high and very high volume), with approximately equal numbers of patients assigned to each group. The Charlson co-morbidity index (CCI) was calculated using ICD-10 codes of co-morbidities based on the protocol of Quan and colleagues22,23 , in which a weight (1–6) is assigned to each of 17 co-morbid disease categories, and the score is the sum of those weights. This method has been well validated, and has produced consistent results in studies using a large administrative database22,23 .
Outcome measurements The primary endpoint was in-hospital mortality, defined as death at any time before hospital discharge. Secondary Table 2
525
endpoints were postoperative length of stay and total costs during the hospital stay.
Statistical analysis Descriptive statistics of the patient population included mean(s.d.) values to describe continuous variables and proportions to describe categorical variables. χ2 tests were used to compare categorical data, and analyses of variance to compare continuous variables among the five hospital volume categories. Multivariable logistic regression analysis was performed for in-hospital mortality, and multivariable linear regression analyses for length of stay and total costs (given in euros, based on US $1 = ¤0·80) to examine associations between hospital volume and outcomes with adjustment for patient backgrounds. Only variables with P ≤ 0·100 in univariable analysis were entered into multivariable analyses. Because the multicentre data were structured hierarchically into two
Bivariable analyses and multivariable logistic regression for in-hospital mortality
Sex M F Age (years) ≤ 59 60–69 70–79 ≥ 80 Charlson co-morbidity index 0–2 3 4 ≥5 Type of hospital Academic Non-academic Hospital volume (no. of PDs/year) < 8 (very low) 8–11 (low) 12–17 (medium) 18–28 (high) > 28 (very high) Type of PD Standard PD with lymph node dissection PD with resection of adjacent organ PD with vascular resection and reconstruction Hepatopancreaticoduodenectomy Study period 2007 2008 2009 2010
No. of patients (n = 10 652)
In-hospital mortality*
P‡
Adjusted odds ratio†
P
6601 4051
264 (4·0) 86 (2·1)
< 0·001
Reference 0·48 (0·38, 0·62)
< 0·001
2085 3640 4073 854
31 (1·5) 108 (3·0) 162 (4·0) 49 (5·7)
< 0·001
Reference 1·94 (1·30, 2·87) 2·75 (1·86, 4·01) 4·15 (2·61, 6·58)
0·001 < 0·001 < 0·001
4589 3384 1515 1164
127 (2·8) 99 (2·9) 41 (2·7) 83 (7·1)
< 0·001
Reference 1·01 (0·76, 1·34) 0·90 (0·62, 1·31) 2·55 (1·85, 3·50)
0·968 0·587 < 0·001
3451 7201
82 (2·4) 268 (3·7)
< 0·001
Reference 1·03 (0·72, 1·47)
0·869
2242 2031 2216 2123 2040
113 (5·0) 82 (4·0) 68 (3·1) 59 (2·8) 28 (1·4)
< 0·001
Reference 0·78 (0·56, 1·08) 0·61 (0·43, 0·86) 0·53 (0·37, 0·76) 0·25 (0·14, 0·43)
0·131 0·006 0·001 < 0·001
2114 5523 1429 1500 86
72 (3·4) 158 (2·9) 55 (3·8) 61 (4·1) 4 (5)
0·090
Reference 0·95 (0·71, 1·28) 1·11 (0·77, 1·59) 1·60 (1·11, 2·29) 2·00 (0·67, 5·98)
0·750 0·575 0·012 0·214
2287 2673 2478 3214
79 (3·5) 76 (2·8) 83 (3·3) 112 (3·5)
0·516
Values in parentheses are *percentages and †95 per cent confidence intervals. PD, pancreaticoduodenectomy. ‡χ2 test.
2014 BJS Society Ltd Published by John Wiley & Sons Ltd
www.bjs.co.uk
BJS 2014; 101: 523–529
R. Yoshioka, H. Yasunaga, K. Hasegawa, H. Horiguchi, K. Fushimi, T. Aoki et al.
levels (hospitals and patients), clustering of outcomes within hospitals was accounted for by using generalized estimating equations24 . The threshold for significance was P < 0·050. All statistical analyses were conducted using SPSS version 19 (IBM, Armonk, New York, USA).
60
Hospitals Patients
50 40 %
526
30 20
Results 10
Table 3
0
Low
Very low
Medium
High
Very high
Annual hospital volume
Proportion of hospitals in each volume group, and the proportion of patients who underwent pancreaticoduodenectomy at these hospitals
Fig. 1
60
50 In-hospital mortality (%)
A total of 10 652 patients who underwent PD in 848 hospitals during the survey period were identified. There were 2242, 2031, 2216, 2123 and 2040 patients in the very low (fewer than 8 PDs per year), low (8–11 PDs per year), medium (12–17 PDs per year), high (18–28 PDs per year) and very high (more than 28 PDs per year) volume groups respectively. Table 1 shows the demographics of the patients in these five hospital-volume groups. Patients in the very high-volume group were younger than those in the very low-volume group (65·8(10·7) versus 68·1(9·5) years respectively; P < 0·001). There was a higher proportion of patients with a CCI of 5 or more in the very highvolume group (12·1 per cent versus 10·8 per cent in the very low-volume group; P < 0·001). The proportion with combined vascular resection was higher in the very highvolume group compared with that in the very low-volume group (22·4 versus 7·1 per cent respectively; P < 0·001). The proportion of hospitals performing fewer than eight PDs annually was 58·6 per cent (Fig. 1). The overall in-hospital mortality after PD was 3·3 per cent (350 of 10 652 patients), and for the groups ranged from 5·0 per cent for very low-volume to 1·4 per cent for very high-volume groups (P < 0·001, χ2 test) (Table 2). Logistic regression analysis indicated an adjusted odds ratio (OR) of 0·25 (95 per cent confidence interval (c.i.) 0·14 to 0·43; P < 0·001), for the very high-volume group compared with the very low-volume group. In χ2 analysis, academic hospitals had significantly lower inhospital mortality than non-academic hospitals (P < 0·001), but no significant difference was found in multivariable logistic regression analysis (adjusted OR 1·03, 0·72 to 1·47; P = 0·869) (Table 2). Regarding type of PD, although no
40
30
20
10
0
10
20 40 30 Hospital volume (PDs per year)
50
Scatter plot of hospitals according to annual hospital volume of pancreaticoduodenectomies (PDs) and in-hospital mortality rate
Fig. 2
significant difference in in-hospital mortality was found on χ2 testing (P = 0·090), logistic regression analysis showed a significant difference between standard PD and PD with vascular resection and reconstruction (adjusted OR
Multivariable linear regression for postoperative length of stay and total costs Postop. length of stay (days)
Hospital volume (no. of PDs/year) < 8 (very low) 8–11 (low) 12–17 (medium) 18–28 (high) > 28 (very high)
Total costs (¤)
Mean(s.d.)
Coefficient
P
Mean(s.d.)
Coefficient
P
43·9(24·3) 40·5(22·9) 37·2(21·7) 36·9(21·9) 32·8(19·2)
Reference −3·4 (−5·4, −1·3) −6·6 (−8·6, −4·4) −7·0 (−9·5, −4·5) −11·3 (−14·6, −8·0)
0·001 < 0·001 < 0·001 < 0·001
40 294(17 150) 38 532(17 188) 36 336(17 038) 37 416(26 650) 33 719(16 894)
Reference −2244 (−3776, −714) −5206 (−6824, −3589) −4813 (−6605, −3022) −10 076 (−12 507, −7645)
0·004 < 0·001 < 0·001 < 0·001
Values in parentheses are 95 per cent confidence intervals unless indicated otherwise. PD, pancreaticoduodenectomy.
2014 BJS Society Ltd Published by John Wiley & Sons Ltd
www.bjs.co.uk
BJS 2014; 101: 523–529
Outcome–volume relationship for pancreaticoduodenectomy
1·60, 1·11 to 2·29; P = 0·012). In-hospital mortality for all hospitals is shown in Fig. 2. In the multivariable analysis, a significant linear relationship was found between higher hospital volume and shorter postoperative length of stay, and between increasing hospital volume and lower total costs (Table 3). Discussion
The present study found a significant inverse relationship between hospital volume and in-hospital mortality for patients undergoing PD. A higher hospital volume was also related to shorter postoperative length of stay and lower total costs. Although the overall mortality rate of 3·3 per cent in this study was low compared with that in nationwide surveys in Western countries, with reported rates between 6·4 and 7·5 per cent4 – 6,9,25 , the same relationship between hospital volume and in-hospital mortality was found. This provides supportive evidence for the validity of centralization of PD in Japan and further concentration of resources in other countries where this process has been started. The results are consistent with those of a study by the JSGS20 , although the present investigation is the first from Japan to record an inverse relationship following PD between hospital volume and in-hospital mortality using risk-adjustment analysis. Although many studies have reported on the volume–outcome relationship, factors that influence this have not been elucidated completely. Schmidt and colleagues1 reported surgeon experience to be an important dominant of postoperative morbidity. In the present study, some low-volume hospitals had in-hospital mortality rates as low as those in some very high-volume hospitals (Fig. 2), implying that surgeon volume might be important. One study26 attributed 51 per cent of the hospital volume effect to surgeon volume, whereas another27 found that system clinical resources were more influential. These factors could not be evaluated by the DPC database. In the present study, higher hospital volume was associated with shorter length of stay and lower total costs. Several reports have examined the relationship between hospital volume and costs of hospitalization for PD. Although significantly lower costs, attributed to better postoperative outcomes, were identified in one tertiary referral hospital compared with community hospitals in Maryland, USA12 , another US study found no association between higher hospital volume and lower costs, but did find an inverse relationship between surgeon volume and costs14 . More resources in high-volume centres might result in better postoperative outcomes and lower costs. Prolonged hospital stays and increased costs 2014 BJS Society Ltd Published by John Wiley & Sons Ltd
527
may reflect postoperative complications, late detection or inappropriate management of complications. The volume–outcome relationship is likely, therefore, to reflect the quality of the operation and postoperative management. Even in very high-volume hospitals, the mean postoperative length of stay in the present study was 32·8 days, much longer than that seen in the West at around 10 days2,3 . According to Organization for Economic Cooperation and Development (OECD) Health Data 201228 , the national mean length of hospital stay in Japan was 18·2 days, compared with 4·9 days in the USA and 7·7 days in the UK. This probably reflects differences in the healthcare system between Japan and Western countries, as well as cultural attitudes. In Japan, most hospitals provide both early postoperative care and subsequent rehabilitation in a single hospitalization29 . Readmission rates within 30 days of PD may be around 15 per cent in the West30 , so that some events occurring after discharge in Western countries would be included during the postoperative stay in Japan. In the course of promoting centralization, a major concern is preserving patient accessibility to optimal care, where the capacity of high-volume hospitals, potential delays and distance the patient must travel are important31,32 . More than 20 per cent of PD operations in Japan are carried out in very low-volume centres; nearly 60 per cent of the hospitals in the present study were in this category (Fig. 1). Although the present study has the advantage of a very large sample size collected over a short, recent period, there are several limitations. As adoption of the DPC system by community hospitals is voluntary, the database may not be representative of all hospitals. In particular, the participation rate of very low-volume hospitals may have been underestimated, so that overall in-hospital mortality may be higher in Japan as a whole. Details of postoperative complications, notably pancreatic fistula, delayed gastric emptying and postoperative haemorrhage that are important after pancreatic surgery33 – 35 , were not included in the DPC database. Length of stay and total costs were used as surrogate markers of the postoperative course following PD. Surgeon volume was not included in the analysis owing to lack of data and, finally, because the DPC database is administrative, information on postdischarge and long-term outcomes were not available. Despite these limitations, the strong linear relationship between higher hospital volume and decreased in-hospital mortality, shorter length of stay and reduced total costs, supports the validity of centralization of PD. www.bjs.co.uk
BJS 2014; 101: 523–529
528
R. Yoshioka, H. Yasunaga, K. Hasegawa, H. Horiguchi, K. Fushimi, T. Aoki et al.
Acknowledgements
This study was funded by a Grant-in-Aid for Research on Policy Planning and Evaluation from the Ministry of Health, Labour and Welfare, Japan (grant no. H22Policy-031), by a Grant-in-Aid for Scientific Research B (no. 22390131) from the Ministry of Education, Culture, Sports, Science and Technology, and by the Funding Programme for World-Leading Innovative R&D on Science and Technology (FIRST programme) from the Council for Science and Technology Policy, Japan (grant no. 0301002001001). Disclosure: The authors declare no conflict of interest. References 1 Schmidt CM, Turrini O, Parikh P, House MG, Zyromski NJ, Nakeeb A et al. Effect of hospital volume, surgeon experience, and surgeon volume on patient outcomes after pancreaticoduodenectomy: a single-institution experience. Arch Surg 2010; 145: 634–640. 2 Fern´andez-del Castillo C, Morales-Oyarvide V, McGrath D, Wargo JA, Ferrone CR, Thayer SP et al. Evolution of the Whipple procedure at the Massachusetts General Hospital. Surgery 2012; 152: S56–S63. 3 Cameron JL, Riall TS, Coleman J, Belcher KA. One thousand consecutive pancreaticoduodenectomies. Ann Surg 2006; 244: 10–15. 4 Topal B, Van de Sande S, Fieuws S, Penninckx F. Effect of centralization of pancreaticoduodenectomy on nationwide hospital mortality and length of stay. Br J Surg 2007; 94: 1377–1381. 5 Balzano G, Zerbi A, Capretti G, Rocchetti S, Capitanio V, Di Carlo V. Effect of hospital volume on outcome of pancreaticoduodenectomy in Italy. Br J Surg 2008; 95: 357–362. 6 de Wilde RF, Besselink MG, van der Tweel I, de Hingh IH, van Eijck CH, Dejong CH et al. Impact of nationwide centralization of pancreaticoduodenectomy on hospital mortality. Br J Surg 2012; 99: 404–410. 7 Learn PA, Bach PB. A decade of mortality reductions in major oncologic surgery: the impact of centralization and quality improvement. Med Care 2010; 48: 1041–1049. 8 van Heek NT, Kuhlmann KF, Scholten RJ, de Castro SM, Busch OR, van Gulik TM et al. Hospital volume and mortality after pancreatic resection: a systematic review and an evaluation of intervention in the Netherlands. Ann Surg 2005; 242: 781–788. 9 McPhee JT, Hill JS, Whalen GF, Zayaruzny M, Litwin DE, Sullivan ME et al. Perioperative mortality for pancreatectomy: a national perspective. Ann Surg 2007; 246: 246–253. 10 Gordon TA, Bowman HM, Tielsch JM, Bass EB, Burleyson GP, Cameron JL. Statewide regionalization of pancreaticoduodenectomy and its effect on in-hospital mortality. Ann Surg 1998; 228: 71–78. 2014 BJS Society Ltd Published by John Wiley & Sons Ltd
11 Finks JF, Osborne NH, Birkmeyer JD. Trends in hospital volume and operative mortality for high-risk surgery. N Engl J Med 2011; 364: 2128–2137. 12 Gordon TA, Burleyson GP, Tielsch JM, Cameron JL. The effects of regionalization on cost and outcome for one general high-risk surgical procedure. Ann Surg 1995; 221: 43–49. 13 Rosemurgy A, Cowgill S, Coe B, Thomas A, Al-Saadi S, Goldin S et al. Frequency with which surgeons undertake pancreaticoduodenectomy continues to determine length of stay, hospital charges, and in-hospital mortality. J Gastrointest Surg 2008; 12: 442–449. 14 Ho V, Aloia T. Hospital volume, surgeon volume, and patient costs for cancer surgery. Med Care 2008; 46: 718–725. 15 Bilimoria KY, Talamonti MS, Sener SF, Bilimoria MM, Stewart AK, Winchester DP et al. Effect of hospital volume on margin status after pancreaticoduodenectomy for cancer. J Am Coll Surg 2008; 207: 510–519. 16 Birkmeyer JD, Sun Y, Wong SL, Stukel TA. Hospital volume and late survival after cancer surgery. Ann Surg 2007; 245: 777–783. 17 Gooiker GA, van Gijn W, Wouters MW, Post PN, van de Velde CJ, Tollenaar RA; Signalling Committee Cancer of the Dutch Cancer Society. Systematic review and meta-analysis of the volume–outcome relationship in pancreatic surgery. Br J Surg 2011; 98: 485–494. 18 Otake H, Yasunaga H, Horiguchi H, Matsutani N, Matsuda S, Ohe K. Impact of hospital volume on chest tube duration, length of stay, and mortality after lobectomy. Ann Thorac Surg 2011; 92: 1069–1074. 19 Yasunaga H, Horiguchi H, Matsuda S, Fushimi K, Hashimoto H, Ohe K et al. Relationship between hospital volume and operative mortality for liver resection: data from the Japanese Diagnosis Procedure Combination database. Hepatol Res 2012; 42: 1073–1080. 20 Suzuki H, Gotoh M, Sugihara K, Kitagawa Y, Kimura W, Kondo S et al. Nationwide survey and establishment of a clinical database for gastrointestinal surgery in Japan: targeting integration of a cancer registration system and improving the outcome of cancer treatment. Cancer Sci 2011; 102: 226–230. 21 Khuri SF, Daley J, Henderson W, Hur K, Gibbs JO, Barbour G et al. Risk adjustment of the postoperative mortality rate for the comparative assessment of the quality of surgical care: results of the National Veterans Affairs Surgical Risk Study. J Am Coll Surg 1997; 185: 315–327. 22 Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005; 43: 1130–1139. 23 Quan H, Li B, Couris CM, Fushimi K, Graham P, Hider P et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol 2011; 173: 676–682.
www.bjs.co.uk
BJS 2014; 101: 523–529
Outcome–volume relationship for pancreaticoduodenectomy
529
24 Hubbard AE, Ahern J, Fleischer NL, Van der Laan M, Lippman SA, Jewell N et al. To GEE or not to GEE: comparing population average and mixed models for estimating the associations between neighborhood risk factors and health. Epidemiology 2010; 21: 467–474. 25 Teh SH, Diggs BS, Deveney CW, Sheppard BC. Patient and hospital characteristics on the variance of perioperative outcomes for pancreatic resection in the United States: a plea for outcome-based and not volume-based referral guidelines. Arch Surg 2009; 144: 713–721. 26 Nathan H, Cameron JL, Choti MA, Schulick RD, Pawlik TM. The volume–outcomes effect in hepato-pancreato-biliary surgery: hospital versus surgeon contributions and specificity of the relationship. J Am Coll Surg 2009; 208: 528–538. 27 Joseph B, Morton JM, Hernandez-Boussard T, Rubinfeld I, Faraj C, Velanovich V. Relationship between hospital volume, system clinical resources, and mortality in pancreatic Resection. J Am Coll Surg 2009; 208: 520–527. 28 Organization for Economic Cooperation and Development (OECD). OECD Health Data 2012. http://www.oecd.org/ health/healthpoliciesanddata/oecdhealthdata2012.htm [accessed 7 January 2013]. 29 Hashimoto H, Ikegami N, Shibuya K, Izumida N, Noguchi H, Yasunaga H et al. Cost containment and quality of care in Japan: is there a trade-off? Lancet 2011; 378: 1174–1182.
30 Ahmad SA, Edwards MJ, Sutton JM, Grewal SS, Hanseman DJ, Maithel SK et al. Factors influencing readmission after pancreaticoduodenectomy: a multiinstitutional study of 1302 patients. Ann Surg 2012; 256: 529–537. 31 Forshaw MJ, Gossage JA, Stephens J, Strauss D, Botha AJ, Atkinson S et al. Centralisation of oesophagogastric cancer services: can specialist units deliver? Ann R Coll Surg Engl 2006; 88: 566–570. 32 Stitzenberg KB, Sigurdson ER, Egleston BL, Starkey RB, Meropol NJ. Centralization of cancer surgery: implications for patient access to optimal care. J Clin Oncol 2009; 27: 4671–4678. 33 Wente MN, Bassi C, Dervenis C, Fingerhut A, Gouma DJ, Izbicki JR et al. Delayed gastric emptying (DGE) after pancreatic surgery: a suggested definition by the International Study Group of Pancreatic Surgery (ISGPS). Surgery 2007; 142: 761–768. 34 Bassi C, Dervenis C, Butturini G, Fingerhut A, Yeo C, Izbicki J et al. Postoperative pancreatic fistula: an international study group (ISGPF) definition. Surgery 2005; 138: 8–13. 35 Wente MN, Veit JA, Bassi C, Dervenis C, Fingerhut A, Gouma DJ et al. Postpancreatectomy hemorrhage (PPH): an International Study Group of Pancreatic Surgery (ISGPS) definition. Surgery 2007; 142: 20–25.
2014 BJS Society Ltd Published by John Wiley & Sons Ltd
www.bjs.co.uk
BJS 2014; 101: 523–529
