Patients with aneurysmal subarachnoid hemorrhage (aSAH) may have significant potentially harmful ionizing radiation exposure (PHIRE) from diagnostic tests and medical procedures (DTMP) during their initial hospitalization.

In this single-center, retrospective, observational study, we evaluated the incidence of PHIRE using all patients with radiographically proven aSAH who survived hospitalization over a 6-year period. Patient data were then used to fit a full logistic regression model, a reduced-variable logistic regression model with least absolute shrinkage and selection operator penalty, and a nonparametric tree-based model. Testing data were then used to calculate each predictive model's accuracy.

Of 192 patients included in this study, 69 (35.9%) met criteria for PHIRE. Patients with PHIRE were more likely to have a poor Hunt-Hess Score (40.6% vs. 12.2%, P < 0.0001), a poor modified Fischer Grading Scale score (30.4% vs. 16.3%, P = 0.03), ventriculostomy (91.3% vs. 47.2%, P < 0.0001), vasospasm (81.2% vs. 34.1%, P < 0.0001), and ventriculoperitoneal shunt (31.9% vs. 10.6%, P < 0.001). Parametric PHIRE prediction modeling with a full logistic regression model and reduced-logistic regression modeling with least absolute shrinkage and selection operator penalty demonstrated PHIRE prediction accuracy of 67% and 78% accuracy, respectively. Nonparametric tree-based PHIRE modeling demonstrated a prediction accuracy of 58%.

On the basis of our data, PHIRE occurs in approximately 35% of aSAH patients. The reduced-variable logistic regression model had the greatest predictive accuracy for PHIRE. Future studies should validate our findings and predictive models and, if our conclusions hold, further clarification of the risks of PHIRE and methods to reduce PHIRE should be investigated.