Abstract: Objective To construct a nomogram risk prediction model for delayed neurocognitive recovery (dNCR) in elderly patients with Parkinson’s disease (PD) after subthalamic nucleus (STN) deep brain stimulation (DBS). Methods A retrospective analysis was conducted on 291 elderly PD patients who underwent STN-DBS treatment at Nanjing Brain Hospital from 2016 to 2024. The characteristic variables were selected based on the data of patients with or without dNCR after STN-DBS. Fifteen characteristics, including prior repetitive transcranial magnetic stimulation(rTMS), disease duration, years of education, levodopa-equivalent daily dose (LEDD), scores of Hamilton Anxiety Scale(HAMA), Hamilton Depression Scale(HAMD), Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment(MoCA), total cholesterol, and histories of hypertension, diabetes, cerebral infarction, and brain atrophy were enrolled in multivariable logistic regression analysis to construct a prediction nomogram model, and a calibration curve was drawn to evaluate the accuracy of the model. Results There were 89(30.58%) patients presenting with dNCR after STN-DBS. Multivariable logistic regression model identified LEDD, baseline MMSE score, years of education, rMST history, and brain atrophy as the influencing factors for dNCR after STN-DBS (all P<0.05). The model demonstrated good discriminative ability and calibration upon bootstrap validation (500 replicates), and showed consistency between the actual probability and the predicted probability. Conclusions Higher MMSE score, longer education duration, and previous rTMS treatment independently reduce the risk of post-operative dNCR, whereas elevated LEDD and brain atrophy are the risk factors. The constructed nomogram model showed robust discrimination and calibration, providing a clinically useful tool for individualized risk assessment before STN-DBS in elderly PD patients.   

Key words: deep brain stimulation, delayed neurocognitive recovery, Parkinson’s disease, prediction model, aged