← Back to List

Program Number: B009
Session Name: Poster Session

Neural Network Machine Learning for Determining Surgical Appropriateness in Head and Neck Subspecialty Referrals

Angeline A Truong¹; John Paul Arios¹; Kevin Xin²; Max L Jiam¹; Nicole T Jiam¹; Patrick K Ha¹; Katherine C Wai¹; ¹University of California, San Francisco; ²University of California, Irvine

Background: Complex and inaccurate referral workflows lead to delays in patient management and have severe consequences for patient outcomes. A machine learning algorithm that could predict whether or not a patient should receive head and neck surgery based on clinical characteristics could reduce time to diagnosis and management. We aim to define the ability of a neural network machine learning algorithm to predict referrals to head and neck cancer care and the algorithm’s ability in determining whether patients require surgical intervention.

Methods: A neural network model was developed to predict indications for surgery. The neural network was trained for 20 epochs on 105,000 patient data points from the National Cancer Database (NCDB) and 44,487 patient data points from the Surveillance, Epidemiology, and End Results (SEER) database. At the end of each training epoch, the neural network was tested on an internal validation set consisting of 57,100 patient data points from NCDB and 19,066 patient data points from SEER. Independent variables included age at diagnosis, cancer location, sex, stage, pathologic TNM, and grade. The dependent variable was whether or not the patient was an appropriate surgical candidate. Thirty-nine patients who were referred for specialized head and neck cancer care from 2020-2023 were randomly selected. Data about patients’ head and neck cancer location and pathology were obtained from the electronic medical record. The accuracy of the machine learning model was then compared against the treatment plan recommended in the attending physician’s clinical note, which was obtained through retrospective chart review.

Results: The model had a 79% accuracy in predicting whether patients underwent surgery for curative intent following referral to head and neck cancer care. Sensitivity was 85%, specificity was 50%, positive predictive value was 90%, and negative predictive value was 38% (Figure 1). A subgroup analysis was performed on all patients who underwent surgery to determine clinical differences between patients who were correctly or incorrectly predicted by the machine learning model. Of the 33 patients who underwent surgery for curative intent as part of their treatment plan following referral, 28 (85%) were correctly predicted by the model to undergo surgery. There were no significant differences in age, sex, or cancer location between correctly and incorrectly predicted surgical patients. There were no significant differences in the proportion of patients from counties defined as a poverty area between correctly and incorrectly predicted surgical patients. There was a trend towards differences in cancer stage between correctly and incorrectly predicted surgical patients with improved accuracy for lower stage cancer, though this did not reach significance (p=0.054) (Table 1).

Conclusions: We demonstrate that our machine learning model accurately predicted head and neck cancer surgery recommendations based on age at referral, cancer location, histology ICD code, sex, stage, AJCC 8th edition pathologic T, N, M, and grade. Overall, this machine learning algorithm shows promise in predicting appropriate referrals to specialized head and neck surgical cancer care, which may expedite the referral process and minimize delays to care.

 

 

← Back to List