# Development and validation of a prognostic computed tomography scoring model for functional outcomes in patients with large hemispheric infarction following decompressive craniectomy

## TL;DR

This study created a new CT-based scoring model to predict poor recovery in patients with large brain strokes after surgery, showing strong accuracy and potential clinical use.

## Contribution

A novel CT-based prognostic scoring model for predicting functional outcomes after decompressive craniectomy in large hemispheric infarction patients.

## Key findings

- The model showed excellent discrimination with a C-index of 0.87 in the development cohort and 0.83 in the validation cohort.
- Each point in the score increased the risk of poor outcomes by 47.8%.
- The model's performance was better than ASPECTS and correlated with CPP and MAP.

## Abstract

There is no established prognostic scoring system developed for patients with large hemispheric infarction (LHI) following decompressive craniectomy (DC) based on imaging characteristics. The present study aimed to develop and validate a new computed tomography scoring model to assess the 6-month risk of poor functional outcomes (modified-Rankin scale [mRS] score of 4–6) in patients with LHI receiving DC.

This retrospective cohort study included patients at two tertiary stroke centers. A prediction model was developed based on a multivariable logistic regression. The final risk factors included the ASPECTS (Alberta Stroke Program Early Computed Tomography Score), longitudinal fissure cistern, Sylvian fissure cistern, and additional vascular territory involvement. 1,000 bootstrap resamples and temporal validation were implemented as validations for the scoring system.

Of the 100 individuals included in the development cohort, 71 had poor functional outcomes. The scoring model presented excellent discrimination and calibration with C-index = 0.87 for the development cohort, and C-index = 0.83 for the temporal validation cohort with non-significant Hosmer-Lemeshow goodness-of-fit test. The scoring model also showed an improved AUC compared to the ASPECTS. For each point in the score model, the adjusted risk of poor functional outcomes increase by 47.8% (OR = 1.48, p < 0.001). The scores were inversely correlated with MAP (mean arterial pressure, paired t-test, p = 0.0015) and CPP (cerebral perfusion pressure, rho = −0.17, p = 0.04).

In patients with LHI following DC, the score system is an excellent predictor of poor functional outcomes and is associated with CPP and MAP, which might be worth considering in clinical settings after further external validation.