Identification and Validation of a Novel Model: Predicting Short-Term Complications After Local Flap Surgery for Skin Tumor Removal

Abstract

BACKGROUND The aim was to analyze the risk factors for the occurrence of complications after local flap transfer and to construct a simple prediction model to help surgeons in the perioperative screening of high-risk patients. MATERIAL AND METHODS Short-term complications were defined as any postoperative infection, dehiscence, bleeding, subcutaneous effusion, fat liquefaction, arteriovenous crisis, and tissue necrosis that required medical consultation or intervention. To explore 16 factors influencing short-term complications after local flap transfer, least absolute shrinkage and selection operator (LASSO) logistic regression was used to reduce the dimensionality of the data and to screen for predictors. Independent risk factors affecting the development of complications after local flap transfer were analyzed using logistic multiple regression models. The consistency (C-)index, receiver operating characteristic (ROC) curves, and calibration curves were used to check the model's discrimination and calibration. Decision curve analysis (DCA) curves were used to evaluate the clinical applicability of this model, and internal validation was assessed using bootstrap validation. RESULTS The C-index of the nomogram model to predict short-term complications after local flap transfer was 0.763 (95% CI: 0.702-0.824), the area under the ROC curve was 0.763, and the internal validation C-index was 0.747. The calibration curve showed good agreement between observed and predicted values, and the DCA showed the model can benefit patients. CONCLUSIONS The model identified the relevant factors influencing short-term complications after local flap transfer, facilitating the identification and targeted intervention of patients at high risk of flap complications after surgery.

Figure 1

Predictor selection using a LASSO logistic regression model. (A) Penalty profile of scalar coefficients of 16 possible influencing factors. The 5-fold cross-validation of the minimum criterion was used to identify the optimal penalty coefficient in the LASSO model. (B) A dotted vertical line is drawn at the best value using the minimum standard and 1 standard error of the minimum standard (1-SE standard). Six features with non-zero coefficients resulted in the optimal lambda. Figure created with R software (version 4.2.0).

Figure 2

The nomogram for complications of the local flap is shown. The complications of the local flap risk nomogram were developed by incorporating the following characteristics: secondary operation, hypoproteinemia, smoking history, infringement of bone, body mass index, and defect area. Figure created with R software (version 4.2.0).

Figure 3

Internal validation of the nomogram. (A) Calibration diagram of the line diagram. The Y-axis is the actual incidence of complications diagnosis. The X-axis is the complication’s predicted risk. The dashed diagonal line represents the perfect prediction of the ideal model. The solid line shows the bias correction performance of the nomogram, whereas the dotted line closer to the diagonal shows a better prediction. (B) Area under the receiver operating characteristic curve analysis was used to determine the accuracy of the identification model for complications. Figure created with R software (version 4.2.0).

Figure 4

Decision curve analysis for the nomogram. Decision curve analysis demonstrates the clinical usefulness of the nomogram. The Y-axis measures the net benefit. The blue line is the nomogram predicting the risk of complications. The solid gray line assumes that complications will occur in all patients. The thin solid black line assumes that no patient will develop complications. In this analysis, the decision curve provided greater net benefits across the range of 4% to 69%. Figure created with R software (version 4.2.0).