Drivers of Inter-Hospital CT Dose Variation in Saudi Arabian Tertiary Brain Imaging: Protocol-Level Analysis from a Five-Centre Dose Survey

Abstract

Purpose: To identify and quantify the specific protocol, level determinants of inter hospital variation in CT dose indices for adult non-contrast brain CT at five Saudi tertiary hospitals, and to model the dose reduction expected from targeted protocol optimization.
Materials and Methods: The study performed a retrospective cross-sectional analysis of 488 adult non-contrast brain CT examinations at five tertiary hospitals. Primary dose metrics were CTDIvol (mGy) and DLP (mGy.cm). Protocol parameters collected during the study were reconstruction algorithm (iterative reconstruction versus filtered back projection), tube voltage (120 versus 130 kVp), and scan length at all five sites. The Kruskal-Wallis test with Dunn post-hoc comparisons and Bonferroni correction tested inter-hospital differences. Mann-Whitney U tests compared dose distributions by protocol subgroup. Epsilon-squared quantified effect sizes. Intra-hospital protocol-stratified analysis (within Hospital D, 120 kVp versus 130 kVp subgroups) isolated the contribution of individual protocol variables. Calculated for all hospitals were coefficients of variation to characterize within-hospital dose consistency.
Results: Inter-hospital differences were statistically significant (Kruskal-Wallis H = 47.3, p < 0.001, epsilon-squared = 0.18). Three independent protocol factors drove the variation. Tube voltage was the dominant variable: patients scanned at 130 kVp had a median CTDIvol 9.6 mGy higher than those at 120 kVp (56.7 versus 47.1 mGy, epsilon-squared = 0.21). Reconstruction algorithm accounted for an estimated 6 to 10 mGy difference between the filtered back projection only hospital and the iterative reconstruction-equipped hospitals at comparable kVp settings. Scan length added approximately 6.5 per cent to DLP at the above-DRL hospital. Coefficients of variation for CTDIvol were near identical across all five hospitals (17.8 to 18.2 per cent), demonstrating that within-hospital relative variability reflects AEC physiology rather than protocol inconsistency, and confirming that protocol-level changes will reduce the entire dose distribution proportionally.
Conclusion: Tube voltage selection is the single most impactful modifiable protocol determinant of brain CT dose variation between Saudi tertiary hospitals, with an effect size exceeding that of hospital membership itself. The absence of iterative reconstruction software at Hospital D accounts for the majority of its above-DRL performance and is remediable through a software upgrade rather than capital equipment replacement. The combination of 120 kVp standardization, iterative reconstruction activation, and scan range restriction is projected to reduce Hospital D’s facility median from 56.0 mGy to approximately 43 to 47 mGy without capital expenditure.