A series of 100 patients with thoracic esophageal cancer who underwent subtotal esophagectomy through a right thoracotomy between 1986 and 1989, were statistically analyzed to assess the risk factors predicting hospital mortality from complications. Hospital mortality was termed as "complication death", and the analyzed factors were age, pulmonary function [% vital capacity (%VC) or % forced expiratory volume1.0 (%FEV1.0)], cardiac function [ECG and Master test], renal function [creatinine clearance (CCR)], hepatic function [15' indocyanine green test (R15.ICG)], diabetes mellitus [75 g oral glucose tolerance test (75OGTT)], depth of tumor invasion [T-factor], and the type of operative procedure [operation]. Each patient was scored according to risk severity on a scale from 0-3, with the higher numbers representing higher risk. Patients not succumbing to complication death had less than 8 points in the total score, while those who suffered a complication death had 8 or more points. Through stepwise logistic regression analysis, we produced a prediction formula. In cases where eight or more points are scored by the semi-quantitative analysis, or 0 or more, by the prediction formula, the operative procedure should be changed to a less radical one for improved prognosis. The introduction of this semi-quantitative analysis for postoperative risk reduced the incidence of complication death from 6% to 3%, and of hospital mortality from 13% to 3%.