Predicting Future Elective Colon Resection for Diverticulitis Using Patterns of Health Care Utilization

Area of Science:

• Colorectal surgery
• Health informatics
• Machine learning in medicine

Background:

• Recurrent diverticulitis is a common indication for elective colon surgery, despite recommendations against early resection.
• Rising rates of surgery for diverticulitis highlight the need to identify patients likely to elect this procedure.
• Shared decision-making can reduce low-value surgical care, but predicting patient preference for surgery remains challenging.

Purpose of the Study:

• To develop and validate Machine Learning (ML) algorithms for predicting future elective colon surgery in patients diagnosed with diverticulitis.
• To assess the accuracy of ML models in forecasting surgical intervention based on health care utilization (HCU) data.
• To identify patients at higher risk of electing surgery for diverticulitis to facilitate shared decision-making.

Main Methods:

• Utilized MarketScan® database claims data (diagnosis, procedure, drug codes) from 2009-2012 for patients with new diverticulitis diagnoses.
• Developed and trained three ML algorithms to predict elective colon resection occurring 52-104 weeks post-diagnosis.
• Evaluated model performance using Area Under the Curve (AUC) with varying lengths of historical HCU data.

Main Results:

• The study included 82,231 patients with incident diverticulitis; 1.2% underwent elective colon resection.
• A Gradient Boosting Machine model, trained on 152 weeks of claims data, achieved an AUC of 75% for predicting surgery.
• Models trained on shorter data periods showed reduced predictive accuracy (AUCs of 68% and 57%).
• The majority (85%) of identified resections were classified as low-value care.

Conclusions:

• Machine learning applied to HCU data can predict elective surgery for diverticulitis with moderate accuracy weeks to months in advance.
• Identifying patients likely to opt for surgery enables targeted shared decision-making initiatives.
• This approach offers an opportunity to reduce the utilization of costly, low-value surgical care for diverticulitis.