Normal Pressure Hydrocephalus is defined as an imbalance between production and
reabsorption of cerebral spinal fluid which results in enlarged ventricles placing pressure on
surrounding brain structures. Intervention requires insertion of a ventriculoperitoneal (VP) shunt
and early interventions has been shown to improve cognition (Vibha & Tripathi, 2021). Diagnostic
workup includes repeated neuropsychological assessment (NPA) as part of an CSF drain trial.
When cognitive improvement is seen on NPA post drain trial, then this is viewed as a predictor of
a cognitive benefit and supports surgical intervention. Use of NPA post drain trial as a predictor
of outcome has resulted in mixed findings, possibly due to prior studies not accounting for practice
effects of repeated exposure to neurocognitive tests or failing to simultaneously factor in
comorbidities that are found to influence cognition. In one of the few studies (El Ahamedieh et al.,
2020) that included a large sample, neuroimaging, and reliable change indices, their NPA was
limited (episodic memory, global cognition). Thus, the current study aimed to predict shunt
placement via statistically accounting for practice effects and by refining the prediction model
while accounting for vascular comorbidity.

In this retrospective study, 58 (M_age = 73.3; 57.9% male; 78.9% Caucasian, M_edu = 15 years) patients who have undergone neuropsychological assessment as part of drain trials were included in the analytic sample. Reliable Change Index (RCI) was used to evaluate meaningful change in pre- and post- CSF drain trial on Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) Index scores and Trail Making Test- B (TMT-B). Comorbidities were evaluated in vascular, physical, psychological, substance use, and imaging domains which were dichotomized (yes/no) and summed to create a comorbidity index (CMI). Two logistic regressions were used to predict shunt placement (yes/no), first using RCIs only (model 1), and then, CMI was added to investigate if it provided significant improvement to model fit (model 2).

Forty-three (74%) patients went on to receive shunt intervention. Both logistic models
had good fit (Model 1: ꭓ² (7) = 15.9, p = 0.02, Nagelkerke R² = 0.21; Model 2: ꭓ² (1) = 16.7, p =
0.02, Nagelkerke R² = 0.22), though no significant difference was observed between them
suggesting that adding comorbid factors did not improve predictive power (Δꭓ² (7) = 1.03, p =
0.31). In both models, it was observed that those who did not proceed to shunt placement surgery were
more likely to see a decrease in RBANS Language Index (β = -0.71691, p < .04) while an
improvement in TMT-B was associated with a greater likelihood of shunt placed (β = 0.34294, p
< 0.03).

The current study demonstrate that shunt placement was predicted by improvements in divided attention via TMT-B, which aligns with the extant literature suggesting that removing excess CSF will relieve pressure on frontal subcortical pathways resulting in improved executive functioning. Comorbidities did not impact outcome.