Predicting postoperative complications of head and neck squamous cell carcinoma in elderly patients using random...

Area of Science:

• Oncology
• Medical Informatics
• Data Science

Background:

• Head and Neck Squamous Cell Carcinoma (HNSCC) disproportionately affects elderly patients.
• Postoperative complications in HNSCC are challenging to predict early, impacting treatment.
• Developing predictive models for these complications is crucial for improving patient outcomes.

Purpose of the Study:

• To develop and validate a predictive model for postoperative complications in elderly HNSCC patients.
• To identify key variables associated with postoperative complications.
• To evaluate the performance of data mining algorithms for this prediction task.

Main Methods:

• Utilized data from 525 elderly HNSCC patients (2006-2011).
• Employed five data mining algorithms to construct predictive models on a training set (n=513).
• Validated the best model using cross-validation and an external testing set (n=12).

Main Results:

• The Random Forest algorithm model, using 44 selected variables, achieved 89.08% accuracy and 0.949 AUC in the training set.
• The model demonstrated 83.33% accuracy and 0.781 AUC in the external testing set.
• Key demographic, disease, and treatment variables were identified as significant predictors (P<0.05).

Conclusions:

• Data mining, particularly the Random Forest algorithm, offers a promising approach for predicting postoperative complications in elderly HNSCC patients.
• Computational prediction models can aid in the early identification of at-risk patients.
• This predictive capability can potentially optimize treatment and improve outcomes for this vulnerable population.