Validity and significance of 30‐day mortality rate as a quality indicator for gastrointestinal cancer surgeries

Abstract Background and Aim Benchmarking has proven beneficial in improving the quality of surgery. Mortality rate is an objective indicator, of which the 30‐day mortality rate is the most widely used. However, as a result of recent advances in medical care, the 30‐day mortality rate may not cover overall surgery‐related mortalities. We examined the significance and validity of the 30‐day mortality rate as a quality indicator. Methods The present study was conducted on cancer surgeries of esophagectomy, total gastrectomy, distal gastrectomy, right hemicolectomy, low anterior resection, hepatectomy, and pancreaticoduodenectomy that were registered in the first halves of 2012, 2013 and 2014 in a Japanese nationwide large‐scale database. This study examined the mortality curve for each surgical procedure, “sensitivity of surgery‐related death” (capture ratio) at each time point between days 30‐180, and the association between mortality within 30 days, mortality after 31 days, and preoperative, perioperative, and postoperative factors. Results Surgery‐related mortality rates of each surgical procedure were 0.6%‐3.0%. Regarding 30‐day mortality rates, only 38.7% (esophagectomy) to 53.3% (right hemicolectomy) of surgery‐related mortalities were captured. The capture ratio of surgery‐related deaths reached 90% or higher for 120‐day to 150‐day mortality rates. Factors associated with mortality rate within 30 days/after the 31st day were different, depending on the type of surgical procedure. Conclusion Thirty‐day mortality rate is useful as a quality indicator, but is not necessarily sufficient for all surgical procedures. Quality of surgery may require evaluation by combining 30‐day mortality rates with other indicators, depending on the surgical procedure.


| INTRODUCTION
Benchmarking and evaluations of surgical quality have proven beneficial and indispensable for improving surgical quality. 1,2 A technique that focuses on the "structure," "process," "results," and "outcomes" of medical care has been proposed as a method for evaluating the quality of medical care, and a variety of other parameters are also being used as indicators in the assessment of surgical quality. 3 For example, number of surgeries has been used as a "structural" indicator, rate of laparoscopic surgeries and length of hospital stay have been used as "process" indicators, and surgery-related mortality rate, as well as the rate of complications, has been used as "outcome" indicators. In recent years, interest has been increasingly shown in, as well as expectations from, outcome-oriented evaluations and measurements of medical care quality.
Mortality is an objective indicator, and assessment methods using overall mortality, surgery-related mortality, in-hospital mortality, 30day mortality, and 90-day mortality are commonly used. Historically, the 30-day mortality rate has been used to measure performance across a wide range of surgical disciplines, and the American College of Surgeons National Surgical Quality Improvement Program (ACSNSQIP) database, which is widely considered as the gold standard database for surgical quality improvements and future pay-for-performance programs, records 30-day complication and mortality outcomes after surgery. [4][5][6] The 30-day mortality rate, as an objective indicator, may not entirely encompass the overall surgery-related mortality rate. 7,8 As a result of recent progress in medical care, especially advances in anaesthesia 9 and intensive care, 10  However, this does not necessarily mean that all patients who avoid death during the early postoperative period will continue to survive.
Medical circumstances, such as length of hospital stay after surgery, are different in each country, and patients particularly in Europe and the USA are discharged from hospitals during the early postoperative period. As a result, surgery-related complications that occur after discharge might be impossible to determine on the basis of 30-day mortality rates or in-hospital mortality rates. There is an absence of available data pertaining to how much of the surgery-related mortality rate can be determined based on the 30-day mortality rate, although this observation varies depending on situations in each country.
Herein, our study examines the significance and validity of the 30-day mortality rate as a surgery-related quality indicator (QI).  11 The NCD, a Web-based data management system, continuously involves individuals who approve data, those in charge of annual case reports from various departments, and data entry personnel, thereby assuring data traceability. Among gastrointestinal surgeries registered in the NCD, data on items similar to those of the ACSNSQIP have been collected in regard to esophagectomy, total gastrectomy, distal gastrectomy, right hemicolectomy, low anterior resection, hepatectomy, pancreaticoduodenectomy, and acute diffuse peritonitis. Hepatectomy included only extended lobectomy, lobectomy and segmentectomy other than lateral segmentectomy for primary liver cancer or gallbladder cancer.
NCD records between January 1, 2012 and December 31, 2014 were analyzed for this study. The study was conducted on cases of gastrointestinal cancers, which were treated with esophagectomy, total gastrectomy, distal gastrectomy, right hemicolectomy, low anterior resection, hepatectomy, and pancreaticoduodenectomy. To examine the significance and validity of the 30-day mortality rate as a surgical QI, the study was carried out without including acute diffuse peritonitis, in which the patient's condition before surgery may have a great influence on the mortality rate. In addition, cases of non-curative resection were excluded to exclude cancer-related deaths. In order to increase the quality of the endpoints, a longer duration was given to the postoperative observation period, and data on surgical cases operated between January and June, which allowed for acquisition of follow-up data from 30 days to 180 days after surgery, were analyzed. Records from patients who refused use of their data were excluded from this analysis. Records with missing data for age, gender, or status at postoperative day 30 were also excluded. Because non-curative resection cases had been excluded, postoperative mortality was equivalent to surgery-related mortality and, therefore, was considered as an endpoint.

| Mortality curve for each operative procedure
Mortality curves were drawn using death within 30 days or postoperative death as events to visualize when mortality events occurred in the postoperative course.

| Sensitivity of surgery-related mortality (capture ratio)
In order to assess the mortality at various time points, ranging from 30 days to 180 days, which covers the surgery-related mortality rate, we calculated the sensitivity of mortality at each point (capture ratio) on the basis of the 210-day mortality, and we also calculated respective 95% confidence intervals.
2.4 | Association between mortality within 30 days, mortality after 31 days, and preoperative, perioperative, and postoperative factors Preoperative, perioperative, and postoperative factors, extracted during the creation of surgery-related mortality models, were used for examining association of mortality within 30 days and after 31 days.
We investigated the distribution of patients who died within 30 days as well as that of patients who died after the 31st day, and the difference between the two was tested using Fisher's exact test. In both categories, a two-sided P-value of <.05 was considered statistically significant. Statistical analyses were carried out using STATA14 (Stata Corp., College Station, TX, USA).
The present study followed the ethical guidelines of human subjects based on the Helsinki Declaration. Review and approval by the ethics committee was not carried out because existing unlinkable, anonymized data were used in the present study.

| Mortality curves for each surgical procedure
When survival or death at day 210 was used as an endpoint, the surgery-related mortality rate was 3.0% for patients with hepatectomies, 2.5% for patients with pancreaticoduodenectomies, 2.3% for patients with esophagectomies, 1.4% for patients with total gastrectomies, 1.1% for patients with right hemicolectomies, 0.8% for patients with distal gastrectomies, and 0.6% for patients with low anterior resections, respectively, in descending order. All types of surgical procedures showed mortality rates that increased over time.
For right hemicolectomies, low anterior resections, and hepatectomies, the mortality rate nearly reached a plateau between days 90 and 150. For esophagectomies, total gastrectomies, gastrectomies, and pancreaticoduodenectomies, the mortality rate continuously increased until day 210 ( Figure 1).
3.3 | Sensitivity of the 30-to-180-day mortality rate (capture ratio for surgery-related mortalities) in comparison with the 210-day mortality rate (surgery-related mortalities) The 30-day mortality rate captured only 38.7% (esophagectomies) to 53.3% (right hemicolectomies) of surgery-related mortalities. The capture ratio for surgery-related mortalities reached 90% or greater when the 120-day mortality rate was taken into consideration for right hemicolectomy, low anterior resection, hepatectomy, and pancreaticoduodenectomy, as well as when the 150-day mortality rate was taken into account for esophagectomy, total gastrectomy, and distal gastrectomy. The capture ratio for surgery-related mortalities reached 90% or greater when the 120-day mortality rate was taken into consideration for right hemicolectomy, low anterior resection, hepatectomy, and pancreaticoduodenectomy, and when the 150-day mortality rate was taken into consideration for esophagectomy, total gastrectomy and distal gastrectomy. The capture ratio of 90-day mortality rate was 80.3% in esophagectomies, 80.3% in total gastrectomies, 81.6% in distal gastrectomies, 89.0% in right hemicolectomies, 88.5% in low anterior resections, 85.2% in hepatectomies and 81.2% in pancreaticoduodenectomies (Table 2).     hemicolectomy), central nervous system disorder (right hemicolectomy), sepsis (right hemicolectomy), and septic shock (total gastrectomy, low anterior resection) were also extracted. In contrast, our results also indicated that factors highly associated with the type of surgical procedure, such as reoperation within 30 days and anastomotic leakage, had significantly lower percentages of deaths occurring within 30 days after surgery. Our data show that factors linked to surgery-related deaths differ depending on the type of surgical procedure, and that when the 30-day mortality rate was used as a QI for an evaluation of surgery outcome or quality, the capture ratio for the determination of deaths associated with surgical technique-related complications may be low or otherwise useless, depending on the respective surgical procedure. Furthermore, our data suggest that when the rate of surgical site infections (except those as a result of anastomotic leakage) is high in surgeries other than low anterior resection, the surgery-related mortality rate is likely to be high even when the 30-day mortality rate is low. Also, when the rate of anastomotic leakage is high in esophagectomy, total gastrectomy, and distal gastrectomy, the surgery-related mortality rate is likely to be high, even when the 30-day mortality rate is low. Similarly, in esophagectomy, right hemicolectomy, pancreaticoduodenectomy, the surgery-related mortality rate is likely to be high when the rate of reoperation within 30 days is high, even when the 30-day mortality rate is low, and in hepatectomies with large amounts of bile leakage, surgery-related mortality rate is likely to be high, even when the 30-day mortality rate is low and, as a result, surgery-related mortality may be difficult to evaluate
Meanwhile, in low anterior resections, surgery-related mortalities can be evaluated properly through evaluation of the 30-day mortality rate. Thus, in order to evaluate the outcomes or qualities of surgery, the 30-day mortality rate and other indicators, such as complications, will need to be assessed in combination with outcomes or qualities of surgery. For example, the incidence of anastomotic leakage in total gastrectomy and distal gastrectomy or bile leakage in hepatectomy may be useful. And to use these indicators could fit the actual clinical feelings of surgeons. A comparison with worldwide outcomes or qualities of surgery, or benchmarking, will be necessary to improve the outcomes or qualities of surgery in Japan. 16 However, data regarding Japan, in which surgery-related mortality is well determined, were based on the Japanese health-care system, and it remains unknown whether these observations are unique to Japan or are universal and shared worldwide. Our results, which were derived from data collected in Japan, could be used to guide an evaluation of their association with medical circumstances in all countries worldwide, through an international endeavor.

| CONCLUSION
The 30-day mortality rate is definitely useful as a QI for the evalua- but it is not necessarily sufficient to cover all types of surgical procedures. Depending on the type of surgical procedure, evaluations of surgical outcome (or qualities) may need to be carried out in combination with the use of the 30-day mortality rate and other indicators, such as complications.