OUP user menu

Quality of gastric cancer care in designated cancer care hospitals in Japan

Takahiro Higashi , Fumiaki Nakamura , Yasuhiro Shimada , Tetsu Shinkai , Toru Muranaka , Wataru Kamiike , Eiji Mekata , Ken Kondo , Yuichi Wada , Hironori Sakai , Mikinobu Ohtani , Takashi Yamaguchi , Nobuyuki Sugiura , Shunichi Higashide , Yoshio Haga , Akitoshi Kinoshita , Tetsuo Yamamoto , Takahiro Ezaki , Shuichi Hanada , Fujio Makita , Tomotaka Sobue , Takeshi Okamura
DOI: http://dx.doi.org/10.1093/intqhc/mzt041 418-428 First published online: 4 June 2013


Objective To develop a set of process-of-care quality indicators (QIs) that would cover a wide range of gastric cancer care modalities and to examine the current state of the quality of care provided by designated cancer care hospitals in Japan.

Design A retrospective medical record review.

Settings Eighteen designated cancer care hospitals throughout Japan.

Participants A total of 1685 patients diagnosed with gastric cancer in 2007.

Main Outcome Measures Provision of care to eligible patients as described in the 29 QIs, which were developed using an adaptation of the RAND/UCLA (University of California, Los Angeles) appropriateness method by a panel of nationally recognized experts in Japan.

Results Overall, the patients received 68.3% of the care processes recommended by the QIs. While ‘deep venous thrombosis prophylaxis before major surgery’ was performed for 99% of the cases, ‘documentation before endoscopic resection’ was completed for only 12% of the cases. The chemotherapy care was less likely to meet the QI standards (61%) than pre-therapeutic care (76%), surgical treatment (66%) and endoscopic resection (71%; overall difference: P < 0.001). A comparison based on the types of care showed that documentation and patient explanation were performed less frequently (60 and 53%, respectively) than were diagnostic and therapeutic processes as recommended in the QIs (85%; overall P < 0.001).

Conclusions Although many required care processes were provided, some areas with room for improvement were revealed, especially with respect to chemotherapy, documentation and patient explanation. Continuous efforts to improve the quality and develop a system to monitor this progress would be beneficial in Japan.

  • cancer control in Japan
  • cancer
  • gastric cancer
  • quality measurements
  • quality indicators


Gastric cancer is the second leading cause of cancer death in the world [1]. Although its incidence has been decreasing globally, gastric cancer is still common, especially in East Asia. In Japan, the age-adjusted annual incidence is estimated to be five times higher than in the USA, although the survival is generally higher in East Asia than in North America and Europe among patients with similar stages of the disease [24].

The processes of care for gastric cancer have largely remained unexamined. Several reports that have aimed to examine its quality in western countries have almost exclusively targeted surgical therapy, focusing on the number of lymph nodes pathologically examined and the achievement of cancer-free surgical margins [510]. These studies revealed inadequate surgical management in several settings [7, 9, 10], and surveys of surgeons [5] and pathologists [8] showed that many of them were not aware of these quality criteria. The practices in East Asia have not even been studied. These situations underscore the importance of quality management in this area.

The management of gastric cancer includes pre-therapeutic evaluation and multiple treatment modalities [1113]. In particular, endoscopic resection and chemotherapy have been gaining importance due to the rapid advancement of technology and medical knowledge. The indication for endoscopic resection is expanding with the development of endoscopic submucosal dissection [14, 15], and adjuvant chemotherapy with S-1 increases the survival of patients with extensive (D2) lymphadenectomy [16, 17]. Given the expanding treatment choices, the coordination of care with appropriate documentation and patient-centered decision-making with sufficient patient explanation are more important than ever.

In this study, we developed process quality indicators (QIs) based on clinical practice guidelines for gastric cancer and evaluated the quality of care provided to patients who received care in 18 hospitals across Japan.


Development of QIs

To measure the quality of gastric cancer care, we developed explicit process-of-care QIs. Explicit process QIs describe the standards of care with specific definitions of patients and the care processes that such patients should receive [18]. The degree to which the actual care provided is concordant with the described care standards represents the quality of care. To examine the various aspects of care, we developed multiple QIs.

In the development of the QIs, we adopted the RAND/UCLA (University of California, Los Angeles) appropriateness method [19]. This method is used extensively in various fields of medicine [2025], and prior studies have demonstrated predictive validity of the QIs developed by this method [26]. The details of the development processes have been reported previously [27]. Briefly, we first compiled a set of candidate QIs based on the Japanese Clinical Practice Guidelines for Gastric Cancer Treatment, second edition [28]. Then, we had these candidate QIs evaluated by a multidisciplinary panel of 11 clinical experts (four surgeons, five medical oncologists and two endoscopists). The panel members rated each QI candidate for its validity on a scale of 1–9, with 1 being the least valid and 9 being the most valid. The rating was assigned twice, first before the discussion meeting and then right after the discussion. During the discussion meeting, the distribution of the first ratings among the members was presented without identifying the raters to highlight the agreement/disagreement for each QI. The panel was asked to suggest modifications to the QIs whenever necessary. The discussion session was moderated by a health services researcher who has the experience in moderating similar discussions to develop QIs. The QIs that received a second rating of 7 or higher by half or more experts and 3 or lower by two or fewer experts were considered valid, as has been judged in prior research [2025]. In total, 31 QIs were accepted as the final set. In this study, we present the results on 29 QIs, excluding two QIs related to perioperative antibiotic use due to the unreliability of chart abstraction for these QIs, which was identified after the implementation.

Implementation of QIs

Patient sample

The clinical information for the patients diagnosed with gastric cancer in 2007 was obtained from 18 cancer care hospitals. These hospitals agreed to participate voluntarily. All of them were cancer care hospitals designated by the government to function as the center of cancer care system in the community. They operated a hospital-based cancer registry that recorded all the cancer patients who had been encountered. We used their registry to identify the patient sample for the study. The patients were eligible if they received either endoscopic/surgical resection or chemotherapy. The patients who received only radiation therapy or only supportive care were not included because there were no QIs that targeted such patients. Patients were also excluded if they had another cancer concurrently that affected the management of gastric cancer or if they had atypical histology, such as sarcoma or lymphoma.

The staff cancer registrars, who manage the cancer registry in the participating hospitals, reviewed the medical record of study patients from the initial visit related to the cancer diagnosis to the most recent visit at the time of abstraction in 2009. An electronic data collection form created for this project guided the data collection and ensured the consistency of the data using the built-in algorithms. The study protocol was approved by the institutional review board at the National Cancer Center of Japan.

Calculation of quality scores

We calculated the score for each QI by the proportion of patients who received the described care among those eligible for the QI (i.e. who met the criteria for the target patients, the denominator). The patients who met the eligibility criteria but did not remain under the care of the same hospital for the time limits in the QI (e.g. an endoscopic follow-up was indicated within 6 months, but the patient transferred to another hospital at 3 months) were excluded from the denominator. For the QIs that listed multiple events per patient, such as a QI that required an explanation of risks and benefits before a new chemotherapy regimen started, we allowed a ‘partial credit’ for the calculation of the scores. For example, if a patient received three chemotherapy regimens and received an appropriate explanation for only the first two regimens, the patient was regarded as having received ‘two-thirds’ of the indicated care. The quality scores for the QIs were calculated both overall and at the hospital level. The hospital scores were analyzed for the distributions of the quality scores using a box and whiskers plot.

We also conducted sub-analyses according to the types of QI. For the QIs that required multiple factors to be documented (e.g. a QI that required the documentation of the size of the lesion, its location and so on after surgery), we examined each factor that constituted the QI individually and the complete documentation so that we could identify the factors that needed particular attention for improvement. For those QIs with multiple events for which partial credit per patient was allowed in the main scoring, we analyzed the changes in the scores by testing two alternative methods of calculating the QI scores: (i) requiring a ‘pass’ for 100% of the events for each patient (100% threshold) and (ii) requiring a ‘pass’ for more than half of the eligible events (50% threshold).

The QIs could be classified by the modality (i.e. pre-therapeutic care, surgical treatment, endoscopic resection and chemotherapy) and the types of care processes indicated (i.e. diagnostic/therapeutic processes, appropriate documentation of clinical information and patient explanation). We compared the average quality scores across the modality of care and the types of care. The difference in the overall proportions of care provision meeting the QI recommendation were statistically tested using the non-parametric Kruskal–Wallis tests.


Characteristics of the patients

The medical records of 1685 patients from 18 hospitals were abstracted. The mean age was 68.1, and 27.5% were female (Table 1). The majority (64.5%) of the patients underwent surgical resection, and 20.7% were treated with endoscopic resection. Approximately one-third of the patients received chemotherapy at least once during the information collection period.

View this table:
Table 1

Sample characteristics (n = 1685)

Quality of care delivered

On average, the patients were eligible for 9.9 QIs and passed the QI 68.3% of the time.

The quality scores varied across the QIs (Table 2). The QIs with the lowest scores were ‘documentation before endoscopic resection’ (QI4: 12%) and ‘pre-surgical explanation including rate of complications and surgical mortality’ (QI7: 26%), the QIs with the highest scores were ‘deep venous thrombosis prophylaxis during surgery’ (QI9: 99%) and ‘dietary guidance after surgery’ (QI12: 98%). The comparison across the modalities revealed that pre-therapeutic care was provided 76% of the time, surgical and endoscopic treatment was provided 66 and 71% of the time, respectively, and chemotherapy was provided 61% of the time (overall test for difference P < 0.001). The patient explanation and documentation were performed as recommended 53 and 60% of the time, respectively, while the recommended diagnostic and therapeutic processes were performed 85% of the time (overall P < 0.001). The variation across hospitals is presented in a box and whiskers plot (Figs 1 and 2). A large variation was observed for many QIs.

View this table:
Table 2

Quality scores for gastric cancer care

Figure 1

Average quality score by the modality targeted.

Figure 2

Boxplot of quality scores across hospitals (n = 18) in each QI. Note: the QI# corresponds to the numbers in Table 2.

Sub-analysis of influential factors and scoring methods

By analyzing the individual items separately in the QIs that constituted the appropriate documentation of the endoscopic, surgical and pathological information, we explored the influential factors in the quality scores for these QIs (Table 3). In many QIs, most of the factors were usually documented, while several particular factors were frequently missing. In the QI that targeted the macroscopic findings before surgical resection, for example, the location of the tumor and the depth of the tumor invasion were documented in 50 and 61% of the applicable cases, respectively, while other factors were documented for >90% of the cases.

View this table:
Table 3

Documentation and explanation of respective clinical factors in the multi-factor QIs

The modification of the scoring methods influenced the scores for the QIs for which multiple events were possible (Table 4). The scores used in the main analysis (use of partial credit per patient) produced scores between the 100 and 50% threshold.

View this table:
Table 4

Changes in the quality scores by scoring method


This study examined the quality of the gastric cancer care provided in 18 cancer hospitals in Japan. To our knowledge, this is the first large-scale multicenter study to examine the quality of gastric cancer care that covered the various modalities and types of care. Quality measurements using QIs developed by a formal, standardized process that aggregated the clinical expertise of multidisciplinary experts and the collection of detailed clinical information derived from medical records provided us with a starting point to further discuss the quality of cancer care.

For some QIs, scores approached 100%, while other QIs showed variations in the scores across hospitals. The largest variations occurred in the documentation practices, but the performance of the treatments, such as the extent of lymph node dissection and adjuvant chemotherapy, was also different. For adjuvant chemotherapy, however, its efficacy in Japan was reported in 2007 [16], the year of cancer diagnosis for the current study sample. Therefore, our data may be showing some effects of the transition process of gastric cancer practices adapting to new evidence. Future studies should follow changes to the adherence of these new guidelines over time.

The quality scores for the QIs targeting patient explanation were lower than the other types of care, which may be attributed to the lack of documentation of patient explanation rather than a lack of explanation itself. If we had asked patients about the explanations they received, the scores may have been higher. However, in the process of developing the QIs, we specified that we would evaluate the care based on the medical records and asked the expert panel to judge a QI as valid if a lack of documentation itself represented a suboptimal quality of care.

The comparison across modalities showed that scores on chemotherapy QIs tended to be lower than those targeting other modalities. This difference may have occurred because many believe that surgical or endoscopic resection is the mainstay of treatment for gastric cancer care. Although the current Japanese clinical practice guidelines recommend adjuvant chemotherapy with S-1 for Stage II or III cases [11], the previous edition of the Japanese clinical practice guideline, published in 2002 (with minor revisions in 2004), recommended chemotherapy only for unresectable cases and accepted adjuvant chemotherapy only as an experimental clinical trial. Such a view of chemotherapy by the doctors as having a supplementary role in gastric cancer treatment may have led to the finding that written informed consent was obtained in less than two-thirds of the cases. Because chemotherapy can cause serious adverse effects along with its benefit, more attention must be paid to the quality of the chemotherapy to improve future practices.

The chemotherapy QIs were applied to multiple events for each patient, such as periodic check-ups and explanations before the start of each regimen. There are multiple options regarding how to score these QIs. One possibility, which we chose to use, is to give partial credit when a patient received the recommended care for some proportion of the eligible events. Another way is to count only the patients who received perfect care, i.e. received the recommended care 100% of the eligible times. The latter approach is similar to that proposed by Nolan and Berwick [29], who justified this approach by asserting that the low scores resulting from raising the threshold would capture people's attention and thereby foster systematic improvements to achieve high-quality care. However, this all-or-none approach would affect the chemotherapy QIs more severely than the surgical or endoscopic QIs. We wanted to avoid the low scores for chemotherapy being attributed to the scoring method. Determining the optimal scoring methods will require further discussion.

Many QIs explicitly allowed the consideration of exceptional cases by qualifying the documentation of reasons for omitting the indicated care process as ‘passing’ the QI. This explicit allowance was strongly emphasized by the expert panel in the discussion for the QI development to avoid unfairly penalizing the appropriate handling of exceptional cases. Prior research on the quality measurements for clinical areas other than cancer reported that ∼3–5% of the cases were considered exceptions to standard recommendations [30, 31]. The reasons a case could be listed as exceptional in the data collection include a patient's refusal of recommended care and a patient's general condition. Sometimes, the clinical judgment of the patient's condition may be subtle, and the validity of the judgment may be difficult to determine from the medical records. Additionally, the variation in such clinical judgment may contain opportunities for quality improvement [32]. Future research should examine more closely the reasons given for exceptions.

We must also note the limitations of this study. First, we used an expert panel process to select QIs. Different panelists from other specialties or disciplines may have had different opinions, which may have led to a different set of QIs. Nonetheless, the quality measurement activities are at the starting stage in the area of gastric cancer care in Japan. Our QI set should be considered as the initial model, rather than the final set. Continuous improvement of both clinical practice and quality measurement are necessary. Secondly, we relied on medical records as the information source. The records may not have contained all the information about the care provided or the decision-making process. As mentioned above, verifying the data with respect to the patient explanation with the patients themselves may be desirable. When we needed to consider the exceptional treatment for particular cases, the reasons may not be always documented. To some extent, the quality scores may have reflected the quality of the documentation [33]. Although we believe that documentation is an important part of clinical care quality, the significance of the quality deficits in this respect may be viewed differently. Thirdly, our QIs focused on the processes of care rather than the outcomes. For a process measure to be valid, it must be related to the outcome [18, 34]. Although we selected the processes that the expert panel believed would ultimately improve patient outcomes, the observed variation in score quality suggests that the physicians practicing at the frontline did not agree with the validity of the QIs. The process-outcome link remains to be determined by future research. Fourthly, the medical records were abstracted by tumor registrars. We believe that this method has made the review more objective, preventing the possibility that physicians give credit to suboptimal care using implicit justification. However, the possibility remains that some care labeled as poor may actually have reasonable justification in the eyes of specialized experts. We should regard our quality measurement as a start for the improvement of care rather than a judgment about the quality. Fifthly, the hospitals participating in this study were designated cancer care hospitals by the Japanese government and were thus motivated to participate in the quality measurement project. Because they play a central role in cancer care in their community, their quality may be higher than that of average hospitals in Japan. Furthermore, the hospitals agreed voluntarily to be assessed for their quality of care. Our findings may have limited generalizability, and more serious quality problems may exist in non-specialized hospitals. Nevertheless, we still detect room for improvement. These areas of improvement may apply to other hospitals. Sixthly, several QIs only had a few eligible patients. We must use caution if we are to interpret the scores on them (e.g. comparison across hospitals) because the calculated scores are unstable. Still, such QIs can represent important clinical standards. Rather than simply dismissing these QIs, a larger target population with longer study periods or multiple hospitals should be used if we focus on these QIs in future. Finally, the QIs were primarily based on Japanese clinical practice guidelines. Due to the differences in patient populations and clinical outcomes [4], the standards of care agreed upon in this environment may differ from other situations [11]. However, we believe that the care that should be performed in regular practice is by and large similar, which is in contrast to the differences often highlighted in the medical literature. Although our QIs may not all be applied to gastric cancer care in its current form, they will provide useful information to experts, even in different environments, who aim to develop their own quality criteria to improve the outcomes of their patients.

To our knowledge, our study is the first to comprehensively examine the quality of gastric cancer care, covering not only surgical therapy but also other treatment modalities. While many parts of this care agreed with the standards, suboptimal care and variations across hospitals also existed. In particular, we found considerable room for improvement in the chemotherapy practice. This quality information provides an important opportunity to take measures to improve the care for the patients.


This project was funded as a grant-in-aid by the Japanese Ministry of Health, Labor and Welfare and the National Hospital Organization.


The authors thank Drs Narikazu Boku (St. Marianna University School of Medicine Hospital), Kazumasa Fujitani (National Hospital Organization National Osaka Hospital), Tetsuya Hamaguchi (National Cancer Center Hospital), Seiji Ito (Aichi Cancer Center Hospital), Ken Kato (National Cancer Center Hospital), Kei Muro (Aichi Cancer Center Hospital), Ichiro Oda (National Cancer Center Hospital), Hiroyuki Ono (Shizuoka Cancer Center), Atsushi Sato (Showa University Hospital) and Takaki Yoshikawa (Kanagawa Cancer Center) for their contributions in the expert panel to develop the QIs and the tumor registrars for their efforts in collecting the data from the medical records.


View Abstract