In a retrospective single-center cohort of 284 surgically treated CSDH patients, smoking history, laterality, hematoma density, greatest hematoma thickness, midline shift ≥10 mm, and postoperative pneumocephalus were significant predictors of recurrence, and an online dynamic nomogram built from them showed good training discrimination (AUC 0.89) but more modest internal validation performance (AUC 0.74), warranting external multicenter validation.
Key Findings
Results
The overall postoperative recurrence rate of chronic subdural hematoma in this cohort was 19.01%.
54 of 284 surgically treated patients experienced recurrence.
Patients were treated between January 2018 and May 2025 at a single center (retrospective design).
Recurrence was defined as reaccumulation or enlargement of hematoma on postoperative CT within 6 months, accompanied by recurrent or worsening neurological symptoms, regardless of whether repeat surgery was performed.
33 recurrence events occurred in the training cohort and 21 in the validation cohort.
Results
Six variables were statistically significant in multivariable logistic regression for postoperative CSDH recurrence.
The significant variables were smoking history, laterality, hematoma density, greatest hematoma thickness, cerebral midline shift ≥10 mm, and postoperative pneumocephalus.
Univariate analysis, multivariable logistic regression, predictor selection, and nomogram construction were performed exclusively in the training cohort (n = 198).
The cohort was split 7:3 into training (n = 198) and validation (n = 86) cohorts before model development.
Results
The nomogram showed good discrimination in the training cohort.
AUC was 0.89 (95% CI, 0.82-0.95).
Sensitivity was 0.82 and specificity was 0.85 in the training cohort.
Results
Discrimination was lower in the validation cohort, with reduced sensitivity.
AUC was 0.74 (95% CI, 0.59-0.88).
Sensitivity was 0.52 and specificity was 0.85 in the validation cohort.
The validation cohort had 86 patients with 21 recurrence events, and the confidence interval is wide.
Results
Bootstrap optimism correction indicated modest overfitting of the model.
Apparent AUC was 0.892.
Mean optimism was 0.021.
Optimism-corrected AUC was 0.871.
Results
The model showed no statistically significant evidence of poor calibration in the training cohort.
An online dynamic nomogram was built to provide individualized recurrence risk estimation, with external multicenter validation still needed.
The authors state the online nomogram provides a convenient framework for individualized recurrence risk estimation.
Only internal validation was performed in this single-center retrospective study.
The authors state that further external multicenter validation is warranted to assess generalizability and clinical applicability.
What This Means
This study looked at 284 patients who had surgery for chronic subdural hematoma, a collection of old blood between the brain and its outer covering that often follows head injury, especially in older adults. About 1 in 5 patients (19%) had the hematoma come back within 6 months. The researchers searched for features that were linked to recurrence and used them to build an online calculator (a nomogram) that estimates an individual patient's risk.
Six factors were linked to recurrence: smoking history, which side the hematoma was on, hematoma density on CT, the greatest hematoma thickness, midline shift of the brain of 10 mm or more, and air inside the skull after surgery (pneumocephalus). The tool performed well on the patients used to build it (AUC 0.89), but less well on a separate group of 86 patients held back for testing (AUC 0.74), where it caught only about half of the recurrences. A statistical correction for overfitting gave an AUC of about 0.87.
This research suggests that a tool using routine clinical and imaging information could help estimate recurrence risk. However, the study came from a single hospital, was retrospective, and had a small validation group. The authors say the tool needs testing in other hospitals before it can be considered reliable for wider clinical use.
Zhang X, Huang Z, Li S, Wu B, Li M, Wang H, et al.. (2026). Risk factors for postoperative recurrence of chronic subdural hematoma and construction and validation of an online dynamic nomogram prediction model.. Frontiers in neurology. https://doi.org/10.3389/fneur.2026.1877761