The mathematical model of postoperative mortality in cases of duodenal injuries: A single center review of a 14-year experience

Objective: Traumatic injuries of the duodenum are quite rare, often accompanied by local and systemic postoperative complications, as well as high mortality. Therefore, identifying predictors of early death in this category of patients is an urgent problem. Materials and methods: A retrospective study included 72 patients older than 18 years old with traumatic injuries of the duodenum >= II degree, according to the criteria of the American Association of Surgery for Trauma, who were operated from 2006 to 2020. Patients were assigned into two groups: the 1st group (survivors, n=58) and the 2nd group (non-survivors, n=14). Different methods of statistical analysis were used to construction the mathematical model. Results: Multiple injuries of intra-abdominal and anatomical structures of the retroperitoneal space were in 72.2% patients. Mean RTS, ISS, and APACHE II scores in the deceased and the surviving patients were and 6.89 +/- 1.0 and 3.93 +/- 0.7; and 12.36 +/- 3.9 and 20.6 +/- 4.5; 14.05 +/- 3.2 and 26.67 +/- 4.1, respectively (p=0.000). All indicators were identified whose changes were most different in surviving and deceased patients on admission to the hospital: blood hemoglobin, serum albumin and lactate and the level of systolic blood pressure (area under curve amounted 0.831, 0.934, 0.956, 0.816 respectively, p=0.000). The analysis of the factor structure made it possible to determine the contribution of each indicator to discrimination and to divide patients into groups. It was found that the indicator 'serum lactate' played most major role in discrimination (-0.9723) and the following were 'serum albumin' (0.5934), 'systolic blood pressure' (0.4636) and 'blood hemoglobin' (0.4437). Based on the data obtained, the mathematical model was developed for predicting of mortality with sensitivity 86.7% and specificity 100%. Conclusions: These research results have shown that the proposed mathematical forecasting model is most suitable for predicting treatment outcomes in the early stages of patients with duodenal injuries with ensures high accuracy. At the same time, the forecasting problem can be solved using the APACHE II and RTS severity scores for these patients, although the forecasting accuracy in this case may be lower.