Traditional Machine Learning versus Deep Learning Approaches for Multi-Outcome Perioperative Risk Prediction
- 1Stony Brook University Department of Surgery, Stony Brook University Department of Biomedical Informatics
Abstract
Whether newer modeling approaches improve perioperative risk prediction is unclear. Holding 73 prespecified preoperative variables fixed -- 69 observed and used as tabular inputs, all 73 serialized to text -- we benchmarked six tabular model families and three pretrained clinical text-encoder embedding classifiers for four 30-day outcomes, training on 4,995,670 ACS NSQIP cases (2018-2022) and testing temporally on 963,565 cases (2024). The FT-Transformer led or tied on AUROC for all four outcomes (0.755-0.955), exceeding logistic regression by only 0.009-0.030; the three encoders fell 0.005-0.009 short of the best tabular model and spanned 0.003 or less. On identical cases the ACS NSQIP Surgical Risk Calculator matched the best model for mortality but trailed for any complication. Rankings changed under AUPRC and did not hold in a 3,490-case single-institution deployment check, not independent of the PUF. Once the predictor set is fixed, architecture and encoder choice move discrimination little.
Related
- Dynamic machine-learning model for early prediction of intraoperative hypotension using 5-minute perioperative time-series data: a multicenter retrospective study
- Development and validation of machine learning models for preoperative prediction of perioperative outcomes in patients with thymic epithelial tumors
- Artificial intelligence and machine learning for prediction of postoperative complications: A systematic review focused on anesthesiology, perioperative risk stratification, and clinical applicability.
- Optimized Feature Engineering and Machine Learning Approaches for Early Prediction of Renal Cell Carcinoma
- Can GPT Be Used as an Alternative Prediction Model to Traditional Machine Learning and Neural Networks on Low-Volume Clinical Data?
BibTeX
@misc{ding2026traditional,
title = {{Traditional Machine Learning versus Deep Learning Approaches for Multi-Outcome Perioperative Risk Prediction}},
author = {Z. Ding and G. Han and T. Ma and F. Wang and J. Hajagos and T. Kurc and A. R. Kumar},
howpublished = {medRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.02.26361824},
url = {https://www.medrxiv.org/content/10.64898/2026.09.02.26361824v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Traditional Machine Learning versus Deep Learning Approaches for Multi-Outcome Perioperative Risk Prediction
AU - Z. Ding
AU - G. Han
AU - T. Ma
AU - F. Wang
AU - J. Hajagos
AU - T. Kurc
AU - A. R. Kumar
PY - 2026
DA - 2026/09/02
PB - medRxiv
DO - 10.64898/2026.09.02.26361824
UR - https://www.medrxiv.org/content/10.64898/2026.09.02.26361824v1
AB - Whether newer modeling approaches improve perioperative risk prediction is unclear. Holding 73 prespecified preoperative variables fixed -- 69 observed and used as tabular inputs, all 73 serialized to text -- we benchmarked six tabular model families and three pretrained clinical text-encoder embedding classifiers for four 30-day outcomes, training on 4,995,670 ACS NSQIP cases (2018-2022) and testing temporally on 963,565 cases (2024). The FT-Transformer led or tied on AUROC for all four outcomes (0.755-0.955), exceeding logistic regression by only 0.009-0.030; the three encoders fell 0.005-0.009 short of the best tabular model and spanned 0.003 or less. On identical cases the ACS NSQIP Surgical Risk Calculator matched the best model for mortality but trailed for any complication. Rankings changed under AUPRC and did not hold in a 3,490-case single-institution deployment check, not independent of the PUF. Once the predictor set is fixed, architecture and encoder choice move discrimination little.
ER -