Impact of a board certification system and implementation of clinical practice guidelines for pancreatic cancer on mortality of pancreaticoduodenectomy

Purposes The aim of this study was to clarify the impact of a board certification system and the implementation of clinical practice guidelines for pancreatic cancer (PC) on the mortality of pancreaticoduodenectomy in Japan. Methods By a web questionnaire survey via the National Clinical Database (NCD) for departments participating in the NCD, quality indicators (QIs) related to the treatment for PC, namely the board certification systems of various societies and the adherence to clinical practice guidelines for PC, were investigated between October 2014 and January 2015. A multivariable logistic regression analysis was performed to evaluate the relationship between the QIs and mortality of pancreaticoduodenectomy. Results Of 1415 departments that registered at least 1 pancreaticoduodenectomy between 2013 and 2014 in NCD, 631 departments (44.6%), which performed pancreaticoduodenectomy for a total of 11,684 cases, answered the questionnaire. The mortality of pancreaticoduodenectomy was positively affected by the board certification systems of the Japanese Society of Gastroenterological Surgery, Japanese Society of Hepato-Biliary-Pancreatic Surgery, Japanese Society of Gastroenterology, and Japanese Society of Medical Oncology as well as by institutions that used magnetic resonance imaging of ≥ 3 T for the diagnosis of PC in principle. Conclusions The measurement of the appropriate QIs is suggested to help improve the mortality in pancreaticoduodenectomy. Masamichi Mizuma and Hiroyuki Yamamoto equally contributed


Introduction
Quality indicators (QIs) are utilized to measure the quality of care, which can be defined as "the degree to which health services for individuals and populations increase the likelihood of desired health outcomes and are consistent with current professional knowledge" [1]. The quality of care has been reported to be evaluated from three aspects: "structure", "process" and "outcome" [2]. For example, board-certified experts are an indicator of structure. In addition, diagnostic methods or treatments recommended in a clinical practice guideline correspond to process indicators. The evaluation and improvement of the quality of care in each institution ultimately lead to the uniform accessibility of medical care nationwide. Therefore, assessing the quality of care using QIs is very important.
The registration of surgical cases in the National Clinical Database (NCD), which is linked to the board certification system of some surgical societies, began in 2011. Most surgical cases (90%-95%) performed in Japan are included in the NCD [3]. Approximately 10,000 cases of pancreaticoduodenectomy (PD), classified as having a high degree of difficulty in the surgical difficulty category defined by the Japanese Society of Gastroenterological Surgery (JSGS), are registered per year on NCD [4]. Risk models of the eight main procedures, including PD, were created using NCD data [5][6][7][8][9][10][11][12] and are used in the risk calculator on the NCD web site, which is available in clinical settings. PD is still a high-risk procedure, and the operative mortality and morbidity need improvement. The evaluation of QIs related to PD is thought to contribute to the improvement of the surgical outcome.
A questionnaire survey of the board certification system and the implementation of clinical practice guidelines for cancers of the esophagus, stomach, colorectum, liver, pancreas, biliary tract, lung and breast was conducted using the NCD to investigate their impact on the surgical mortality by a study group for "the utilization of high-accuracy organ cancer registration in the clinical practice guidelines and medical specialist training" and was supported by a grant from the Ministry of Health, Labour and Welfare of Japan. The results concerning esophageal and colon cancers have been recently reported [13,14].
The present study aimed to elucidate the impact of the board certification system and the adherence to the clinical practice guideline for pancreatic cancer on mortality of PD.

Web questionnaire using the NCD registration system
The questionnaire form was created with the NCD registration system. The questionnaire survey of the QIs related to the treatment for pancreatic cancer was performed via the NCD web page between October 1, 2014, and January 31, 2015. The QIs of the questionnaire, which were chosen by discussion among experts on pancreatic diseases (MM, MU, TS and MG), are shown in Table 1. Q1-16, mainly asking whether or not there is a board-certified expert in each society related to the treatment of pancreatic cancer, were created as structure indicators. Q3 regarding boardcertified institutions of the Japanese Society of Hepato-Biliary-Pancreatic Surgery (JSHBPS) was answered separately for Training Institutions A and B. In the application for the board certification system of JSHBPS, Training Institutions A and B need to perform 50 and 30 high-level hepato-biliary-pancreatic surgeries annually, as defined by the JSHBPS, respectively [15]. Board-certified experts of the Japanese Society of Gastroenterology (JSGE) or Japanese Society of Medical Oncology (JSMO), who are not necessarily surgeons, may participate in preoperative care for PD. Thus, the board certification systems of the JSGE and JSMO were considered for the questionnaire because they may affect the outcomes of PD. Q17-22 were selected as process indicators from Clinical Questions (CQs) of Clinical Practice Guidelines for Pancreatic Cancer Based on Evidence-Based Medicine 2013 [16]. The subjects of the questionnaire were a total of 1415 departments that performed at least 1 case of PD between 2013 and 2014, including a total of 20,183 PD cases in this study (Fig. 1).

Operative mortality for each QI
Responses to the QI questionnaire were obtained from 631 departments (44.6%), which performed 11,684 pancreaticoduodenectomies (57.9%) during the study period. These PD cases were analyzed using the NCD database. Operative death was observed in 292 cases (2.5%) (Fig. 1). Operative death was defined as any death within the index hospitalization period up to 90 days after surgery or any death after discharge within 30 days after surgery. The operative mortality was analyzed for each QI of the questionnaire.

The multivariable logistic regression analysis
The relationship between each QI of the questionnaires and operative death was analyzed by multivariable logistic regression models fitted with a generalized estimating equation, considering the clustering of patients by the hospital level. According to a previous report on the risk model using the NCD [5], the following variables were used to adjust risk factors by the patient background: age, respiratory distress (any), the activity of daily life (ADL) within 30 days before surgery (any assistance), angina, weight loss > 10%, American Society of Anesthesiologists (ASA) performance status grade ≥ Class 3, Brinkman index > 400, body mass index (BMI) > 25 kg/ m 2 , serum creatinine > 3 mg/dl, platelet count < 120,000/ μl, prothrombin time-international normalized ratio (PT-INR) > 1.1, white blood cell (WBC) count > 11,000/μl and activated partial thromboplastin time (APTT) > 40 s. Q1 and Q17 were excluded from the multivariable analysis because very few departments answered "no" and "Not performed in principle", respectively.

Statistical analyses
The STATA 15 software program (STATA Corp., College Station, TX, USA) was used for all statistical analyses. The significance of categorical variates was calculated using the chi-square test or Fisher's exact test. The risk-adjusted odds ratio (AOR) and 95% confidence interval (CI) were calculated in multivariable logistic regression analyses. P < 0.05 was considered statistically significant.
This specific project was approved by the Ethics Committee of Fukushima Medical University (No. 1057).  Are contrast media used in CT or MRI to diagnose pancreatic cancer? Q18 Is MRI of 3 T or more performed to diagnose pancreatic cancer? Q19 Is radical resection performed for cases with Stage 0-IVa* pancreatic cancer without invasion of SMA or CA? Or does your institution refer them to other institutions for radical resection? *General Rules for the Study of Pancreatic Cancer the 6th Edition (the 3rd English Edition) by Japan Pancreas Society Q20 Is S-1 monotherapy performed as the first choice in adjuvant chemotherapy for pancreatic cancer Q21 Is chemoradiotherapy or chemotherapy performed as the first-line therapy for locally advanced unresectable pancreatic cancer? Q22 Is either gemcitabine monotherapy, gemcitabine plus erlotinib combination therapy, or S-1 monotherapy performed as the first-line chemotherapy for locally advanced unresectable or metastatic pancreatic cancer?

Patient demographics and crude operative mortality
The crude operative mortality was investigated for each risk factor according to the previous report of the risk model for PD (Table 2) [5]. All risk model variables except for a Brinkman index > 400 were significantly correlated with operative death. Tables 3 and 4 indicate the response distribution and crude operative mortality rate in each QI for the structure and process indicators, respectively. As shown in Qs2-5, the board-certified institutions of the JSGS, JSHBPS, JSGE, and JSMO showed a significantly lower mortality rate than the non-certified institutions (p < 0.001). Regarding QIs related to board-certified experts or instructors, institutions having an expert surgeon and instructor board-certified by the JSHBPS (p < 0.001), a gastroenterologist and instructor board-certified by the JSGE (p = 0.004 and p < 0.001), and an oncologist and instructor board-certified by the JSMO (p = 0.001 and p = 0.004) showed significantly lower operative mortality than others (Table 3).

The response distribution and crude operative mortality for each QI
However, regarding QIs for process indicators, departments that used magnetic resonance imaging (MRI) of ≥ 3 T for the diagnosis of pancreatic cancer in principle (p < 0.001), performed radical resection for pancreatic cancer or referred the case to other institutions for radical resection in principle (p = 0.018) and performed S-1 adjuvant therapy for pancreatic cancer in principle (p = 0.001) showed a significantly lower operative mortality rate than others (Table 4). In Q18, 285 departments (45.2%) responded with "Performed in principle" concerning the use of MRI of ≥ 3 T for the diagnosis of pancreatic cancer.     [0.37-0.66], p < 0.001). In contrast, there was no significant difference between the JSHBPS board-certified training institution B and the institutions that were not boardcertified (Fig. 2b). Although institutions accredited by the JSGE showed a significantly lower AOR (0.50 [0.34-0.74], p < 0.001) than those not certified by or related to the JSGE, related institutions showed no significant difference (Fig. 2c). Institutions with an expert surgeon or instructor board-certified by the JSHBPS (Fig. 2i, j), a gastroenterologist or instructor of gastroenterology board-certified by the JSGE (Fig. 2k, l), and an oncologist or instructor of oncology board-certified by the JSMO (Fig. 2m, n) showed significantly lower AOR values than those without them.

Results of the multivariable logistic regression analysis
Regarding the use of MRI of ≥ 3 T for the diagnosis of pancreatic cancer (Q18), both "Not performed in principle" and "Doctor's discretion" were significantly poor risk factors (p < 0.001 and p = 0.01) (Fig. 3a). Regarding radical resection (Q19) and S-1 adjuvant chemotherapy (Q20), "Doctor's discretion" showed a significantly higher AOR than "Performed in principle". "Not performed in principle" showed no significant difference in Q19 and Q20 (Fig. 3b, c).

Discussion
The present study revealed the following three points using questionnaires and the data of the NCD: (1) Mortality of PD was positively affected by the institution certification systems of the JSGS, JSHBPS, JSGE and JSMO. (2) Institutions with an expert or instructor board-certified by the JSHBPS, JSGE or JSMO showed a low PD mortality. (3) The mortality of PD was low in institutions that used MRI of ≥ 3 T for the diagnosis of pancreatic cancer in principle. These findings suggest to be useful as a QI for PD in Japan.
According to the NCD, the operative mortality of PD between 2011 and 2012 was reported to be lowest in training institution A (board-certified by the JSHBPS) followed by institution B and non-certified institutions. In addition, the participation of an expert surgeon or instructor who was board-certified by the JSHBPS in PD resulted in a lower operative mortality compared to that with no such participation [17]. The current study, which analyzed NCD data collected between 2013 and 2014, also showed a similar impact of the board certification system of the JSHBPS on the operative mortality of PD. The board certification system of the JSHBPS is suggested to be a good QI in PD for pancreatic cancer.
In contrast, regarding the board certification system of the JSGS, the present study indicated no marked correlation between the operative mortality and the presence of a board-certified expert surgeon or instructor. A previous report showed that the number of expert surgeons boardcertified by the JSGS was a surrogate marker of the operative mortality in eight main procedures, including PD [18]. The present study's lack of an investigation of the number of expert surgeons board-certified by the JSGS might have been associated with the absence of a correlation with the operative mortality. In our study, a favorable outcome of PD was observed in institutions board-certified by the JSGE or JSMO. Furthermore, institutions with experts board-certified by the JSGE or JSMO who did not necessarily participate directly in PD still showed a significantly lower operative mortality for PD than in those without. These results suggest that institutions that specialize in gastroenterology or oncology have more favorable outcomes from surgery due to an indirect effect, as gastroenterologists and oncologists are involved in preoperative care, including oncological judgement, chemotherapy and nutritional management, for patients scheduled for PD. Therefore, these results imply that systematic multidisciplinary approach for preoperative care improves the safety of PD. There are no reports on the relationship between the operative mortality of PD and the board certification systems of the JSGE or JSMO. These are novel findings as factors related to the operative mortality of PD.
To our knowledge, there have been no reports concerning the implementation of clinical practice guidelines for pancreatic cancer, including associations with the mortality of PD. In the present study, institutions using MRI of ≥ 3 T in principle for the diagnosis of pancreatic cancer had a significantly lower mortality rate of PD than those who did not or did so only at the doctor's discretion (Q18). Although adherence to Q18 was low compared with other QIs, this might be due to the considerable number of institutions unable to perform MRI of ≥ 3 T. Since possession of an MRI machine of ≥ 3 T depends on a hospital's financial standing, the results of Q18 may reflect the effects of the hospital volume. Interestingly, the present study showed that QIs in radical resection (Q19) and S-1 adjuvant chemotherapy (Q20) had higher AORs for "Doctor's discretion" than for "Performed in principle". In a previous study in this project concerning esophageal cancer, similar results were found in some QIs [13]. These findings suggest the importance of organizational compliance with clinical practice guidelines for pancreatic cancer.
Despite patient selection bias due to old age, which may be considered a factor of non-operative indication, especially in elderly patients with comorbidity, this study demonstrated that age was a significant risk factor for mortality in PD, as previously reported [5]. Mortality following PD for elderly patients with pancreatic cancer has been reported to be affected by specific comorbidities (chronic obstructive pulmonary disease, chronic kidney disease, dementia and sepsis) as patient factors [19]. The present study was conducted with risk adjustment for various patient factors, including the age, as described in the Methods section. However, as a structure indicator, a previous report showed that non-teaching hospitals have a higher risk of PD mortality for elderly patients with pancreatic cancer than teaching hospitals [20]. The present study clarified the correlation between the mortality of PD and board certification systems of various academic societies as structure indicators. Thus, the assessment of structural indicators is crucial for reducing the mortality of PD.
The utilization of administrative claims data in Japan for the wide-scale measurement of QIs in the treatment of various cancers, namely colorectal, lung, stomach, liver, breast, prostate and cervical cancer, has been reported [21]. When comparing NCD data with administrative claims data, the advantage is that the impact of QIs on surgical outcomes can be analyzed, as in the current study project [13]. At clinical settings in Japan, the NCD Breast Cancer Registry is used to assess the QIs recommended by the clinical practice guidelines. Registered NCD users can compare the implementation rates of the QIs in their institutes with those of the national average on the NCD web site, which helps eliminate cancer care disparity. Thus, the NCD is a useful tool for evaluating QIs related to each type of cancer.
The limitations of this study are similar to those previously described [13]. First, we cannot exclude respondents' bias in the questionnaire surveys. The respondents were users registered in the NCD and not necessarily representative of the department. In other words, the answers may not necessarily reflect the policies of the department. Second, we received no answer from more than half of the institutions. There may be differences in the implementation rate of QIs or the mortality of PD between respondents and nonrespondents. Third, there may have been selection bias for the QIs, which were selected by discussion among experts of pancreatic diseases, as mentioned above. Finally, PD cases with diseases other than pancreatic cancer were included in this study.
In conclusion, the mortality of PD was positively impacted by the institutional certification systems of the JSGS, JSHBPS, JSGE and JSMO. Institutions with an expert or instructor who was board-certified by the JSHBPS, JSGE or JSMO showed a lower mortality rate of PD than those without such a staff member. Furthermore, institutions performing MRI of ≥ 3 T for the diagnosis of pancreatic cancer showed a lower mortality from PD than others. The NCD is a useful tool for evaluating the quality of cancer care, especially for analyzing the impact of QIs on surgical outcomes.