In-Hospital Mortality Risk Model of Gastric Cancer Surgery: Analysis of a Nationwide Institutional-Level Database With 94,277 Chinese Patients

Background: The objective of this study is to identify independent risks and protective factors and to construct a mortality prediction model for gastrectomy in the Chinese population. Study design: This is a population-based prospective cohort at an institutional level. Seventy-two participating hospitals reported their annual gastrectomy data between 2014 and 2016, while 44 variables covering the institution and surgical information were included in the analysis. We used R software to encode and complete data pre-processing. The first difference model was applied to build the risk model. Data from 2014 and 2015 were assigned to risk model development, while data from 2016 was used for validation. Results: In the included centers with 94,277 gastric cancer cases, the in-hospital mortality rate was 0.32%. The regression model revealed that provinces with low-middle GDP, hospitals with annual gastrectomy volume between 100 and 500, greater volume of urgent surgeries performed, larger proportion of males, and a higher proportion of liver metastasis were independent risk factors for mortality following gastric surgeries, while higher laparoscopic resection volume, greater volume of distal gastrectomy with B2 reconstruction, and larger proportion of palliative surgery were independent protective factors (p < 0.05, respectively). In the prediction test, the mean square error of the training set was 0.948, while that of the test set was 0.728, demonstrating the effectiveness of this model. Conclusions: We constructed the first mortality risk prediction model for gastric cancer surgery in the Chinese population. The identified risk factors will help with the therapy selection, while further informing Chinese medical policy decision-makers.