Research on prognosis prediction and nursing decision system for plastic rhinoplasty patients based on patient-reported outcomes
QIAO Haixia
MA Na
MA Xianjie
JIA Weina
YANG Minli
Abstract：Objective To address the incongruence between high postoperative complications and patient satisfaction in plastic rhinoplasty,we developed a prognostic prediction model and clinical decision system based on Automated Machine Learning(AutoML).Methods Retrospectively,256 rhinoplasty patients from Xijing Hospital(March 2019 to January 2024)were included.Multidimensional data encompassing clinical baselines,surgical procedures,perioperative management,and postoperative behaviors were collected.Prog-nostic indicators included 1-month adverse events and 1-year Rhinoplasty Outcome Evaluation(ROE)scores.An Improved Catch Fish Optimization Algorithm(ICFOA)constructed the AutoML framework,validated and optimized via CEC2022 benchmark functions.SHAP analysis quantified key variables,culminating in a MATLAB-based visual Clinical Decision Support System(CDSS).Results The IC-FOA-AutoML model outperformed others in the test set:AUC for 1-month complications reached 0.9158,and R2 for 1-year ROE score re-gression was 0.8783.SHAP(SHapley Additive exPlanations)identified core predictors-short-term:postoperative nasal collisions,preoperative enlarged nostril pores,postoperative nasal folliculitis;long-term:preoperative ROE scores,education level,and suboptimal prosthesis carving.Decision curve analysis confirmed enhanced net benefits over traditional methods.Conclusion This study established an AutoML-driven prognostic management framework for rhinoplasty patients,bridging the gap between surgical practice and patient-reported outcomes through dynamic risk stratification and explainable artificial intelligence,offering a novel paradigm for aesthetic surgical decision-making.
Keywords：RhinoplastyAutomated machine learningPatient-reported outcomesPrognosis predictionExplainable artificial intelligenceClinical decision support system
Publication Date：2026-01-15
Online Publishing Date：2026-03-25(First online date of this platform, not the publication date of the document)
Pages：7( 43-48,后插5 )