Medical Student Albert Einstein College of Medicine
Introduction: Neuroimaging characteristics of subdural hematoma (SDH), particularly in patients with dementia, are under-reported. This preliminary analysis aimed to estimate the average marginal effect of these characteristics on 1-year mortality following a diagnosis of SDH.
Methods: Medical records from Dec 2021 to Jan 2023 were reviewed retrospectively with a 1:2 sampling rate (dementia vs. non-dementia patients). An optimal cut-off for predicting 1-year mortality using neuroimaging characteristics was calculated via 1000 bootstrap samples. Full matching on the propensity score was employed with a propensity score estimated using logistic regression of the binary neuroimaging variable with the optimal cut-off value on the covariates including age, sex, and dementia status. The relative risk (RR) with 95% confidence intervals (CI) was reported from logistic regression after propensity score matching.
Results: 54 adults (mean age 77±15) were analyzed, with 50% (27/54) being male, 39% (21/54) having dementia, and 7.4% (4/54) having chronic SDH. An optimal cut-off of 7 mm for hematoma width predicting 1-year mortality (AUC=0.76) was better than a 13 mm cut-off for midline shift (AUC=0.60), as such hematoma width as a binary variable was chosen for propensity score matching. Propensity score standardized mean differences were < 0.3 for covariates, indicating good balance for preliminary analysis. Covariates could not be entered into the model due to an event rate of < 10 (13%, 7/54) and no 1-year mortality occurred in those with a hematoma size of < 7 mm, as such actual hematoma width was used as the sole predictor. The effect of hematoma width on 1-year mortality: RR=1.06 [1.02, 1.1], p=.005, N=54.
Conclusion : Risk of 1-year mortality appears to rise significantly by 6% on average for each 1 mm increase in hematoma width. A sample size of 231 is needed to include covariates with a propensity score standardized mean difference > 0.1.
.jpg)
