Evaluating Risk Factors That Affect Chronic Wound Healing with Logistic Regression and Neural Network Models

Chronic wound impose immense burdens and suffering on patients, and the healing is influenced by a complex interplay of physiological, social, and psychological factors. This study aims to explore risk factors affecting chronic wound healing. We collected demographic variables data from 232 chronic wound patients. The rank-sum test was used to analyze the risk factors that affect chronic wound prognosis. Statistically significant parameters were included in binary logistic regression and multilayer perceptron neural network models to identify independent risk factors for chronic wound prognosis and assess the diagnostic performance of related risk factors in predicting chronic wound outcomes. We found significant differences in education level, number of wounds, size of wounds, intervention methods, non-steroidal drug use, pain, hemoglobin, and albumin affecting chronic wound outcomes (P = 0.016, P = 0.018, P = 0.042, P < 0.001, P = 0.009, P = 0.005, P = 0.004, P = 0.002). In the binary logistic regression model, education level, number of wounds, size of wounds, and intervention methods emerged as independent risk factors (P = 0.048, P = 0.046, P = 0.048, P < 0.001). In the multilayer perceptron neural network model, the accuracy of training set and test set were at 70.6% and 72.2%, and the ROC AUC value was 0.772, with normalized importance of 100% and 83.5% for wound size and albumin, respectively. In conclusion, wound healing is influenced by multiple factors, among which wound size and albumin are the main factors affecting wound healing.