# Predictors of Recurrence of Chronic Subdural Hematoma Among Adult Filipinos After Surgery: Developing a Preliminary Model for Prognosis

## TL;DR

This study developed a model to predict the recurrence of chronic subdural hematoma in adult Filipinos after surgery, using factors like bilateral surgery and hospital stay length.

## Contribution

The study introduces a preliminary predictive model for CSDH recurrence tailored to adult Filipinos post-surgery.

## Key findings

- Bilateral surgery and intraoperative tranexamic acid were associated with lower CSDH recurrence.
- Longer hospital stays increased the risk of recurrence.
- The model showed strong predictive accuracy with an AUC of 0.911.

## Abstract

Objective: The objective of this study is to develop a preliminary model for prediction of chronic subdural hematoma (CSDH) recurrence for adult Filipinos who underwent CSDH surgery.

Methods: This study addresses a retrospective cohort of 71 adult patients who underwent CSDH surgery at a tertiary medical center in the Philippines. The outcome of interest was recurrence of CSDH detected on post-operative scans after discharge. Candidate predictors were chosen a priori based on our institutions experience and available literature. Multivariate logistic regression analysis was performed to determine the final model guided by Akaike Information Criteria (AIC) and receiver operating characteristic (ROC) curve criteria.

Results: A total of 71 adult patients underwent CSDH surgery at our institution. Seven (9.9%) had recurrence of subdural hematomas, where four patients underwent repeat CSDH surgery, and three were managed conservatively. The analysis revealed that bilateral surgery (odds ratio (OR) = 0.06; SE 1.35; 95% confidence interval (CI) of 0.01, 0.78; p = .032), intraoperative tranexamic acid (OR = 0.03; SE 1.68; 95% CI of 0.01, 0.87; p = .042), and length of hospital stay (OR = 1.18; SE 0.08; 95% CI of 1.01, 1.41; p = .032) were independently associated with CSDH recurrence. The final model was significantly better at predicting recurrence compared to baseline probability (Wald χ² (6, N= 71) = 18.427, p = .005). The model had good sensitivity (SE) and specificity in predicting recurrence (area under the curve (AUC) = 0.911; SE 0.054; 95% CI of 0.806 to 1.000; p < .001). A Youden’s index value (Jmax = .716) corresponds to a predicted probability of 0.171 (SE = 85.7%, Specificity = 85.9%, positive predictive value (PPV) = 40.0%, negative predictive value (NPV) = 98.2%).

Conclusion: The authors have a developed a preliminary model predicting recurrence after surgery for St. Luke’s Medical Center, one of the largest tertiary hospital systems in the Philippines.