Modeling preoperative risk factors for potentially lethal morbidities using a nationwide Japanese web-based database of patients undergoing distal gastrectomy for gastric cancer

Most risk models for mortality and morbidity after distal gastrectomy have been created based on relatively small retrospective studies, and a model originating from nationwide database has been lacking. This study aimed to identify preoperative risk factors that predict the postoperative morbidities closely associated with mortality in gastric cancer patients undergoing distal gastrectomy, using data from the National Clinical Database (NCD), a nationwide web-based database in Japan. We analyzed records from the NCD for 65,906 patients who underwent distal gastrectomy in 1,986 hospitals during 2011 and 2012. Using 80 % of these patients, we identified independent preoperative predictors for morbidities closely related with mortality. These risk factors were then validated using the remaining 20 % of patients from 2011 to 2012 and the further 35,575 records from 2013. Operative mortality was 1.07 % and overall morbidity was 14.2 % for the development data set. We selected eight morbidities that were closely associated with postoperative mortality, and then identified between 13 and 25 independent preoperative risk factors for each of the eight morbidities. Of these, old age, female gender, and poor ADL were the factors most frequently associated with the morbidities. The C-indices for each morbidity from the 2011 to 2012 validation data set were favorable as follows: unplanned intubation (0.797), pneumonia (0.784), systemic sepsis (0.748), renal failure (0.832), cardiac events (0.728), large blood transfusions (0.700), central nervous system (CNS) events (0.779), and anastomotic leakage (0.658). Similar C-indices were obtained for each mortality using the 2013 validation data set. In this study, we successfully constructed an acceptable risk model using preoperative risk factors to predict eight postoperative morbidities highly associated with mortality in gastric cancer patients. This risk model could help to tailor perioperative management and improve clinical outcomes for patients who undergo distal gastrectomy.