Development and Validation of a Prognostic Classification Model Predicting Postoperative Adverse Outcomes in Older...

Area of Science:

• Geriatric Surgery
• Health Informatics
• Machine Learning in Medicine

Background:

• Older adults face higher risks of postoperative complications.
• Existing risk stratification tools are often resource-intensive.

Purpose of the Study:

• Develop and validate an open-source, patient-level predictive model for adverse outcomes in older general surgery patients.
• Utilize data from the Observational Health Data Sciences and Informatics (OHDSI) framework.

Main Methods:

• Employed the Observational Medical Outcomes Partnership (OMOP) common data model and machine learning algorithms.
• Trained and tested the model using data from Seoul National University Bundang Hospital (SNUBH) and Seoul National University Hospital (SNUH).
• Validated model performance using the area under the receiver operating characteristic curve (AUC).

Main Results:

• The least absolute shrinkage and selection operator (LASSO) logistic regression model demonstrated strong predictive accuracy for 90-day mortality and emergency department visits (internal AUC 0.723, external AUC 0.703).
• The model also showed good performance for predicting postoperative delirium, prolonged postoperative stay, and prolonged hospital stay.
• LASSO outperformed traditional methods like age and Charlson comorbidity index in prediction.

Conclusions:

• The developed model provides a valuable tool for clinicians and patients to understand individualized surgical risks and benefits.
• This open-source model enhances the assessment of postoperative adverse events in older adults.