Neural predisposing factors of postoperative delirium in elderly patients with femoral neck fracture

Abstract

Elderly adults are more likely to develop delirium after major surgery, but there is limited knowledge of the vulnerability for postoperative delirium. In this study, we aimed to identify neural predisposing factors for postoperative delirium and develop a prediction model for estimating an individual's probability of postoperative delirium. Among 57 elderly participants with femoral neck fracture, 25 patients developed postoperative delirium and 32 patients did not. We preoperatively obtained data for clinical assessments, anatomical MRI, and resting-state functional MRI. Then we evaluated gray matter (GM) density, fractional anisotropy, and the amplitude of low-frequency fluctuation (ALFF), and conducted a group-level inference. The prediction models were developed to estimate an individual's probability using logistic regression. The group-level analysis revealed that neuroticism score, ALFF in the dorsolateral prefrontal cortex, and GM density in the caudate/suprachiasmatic nucleus were predisposing factors. The prediction model with these factors showed a correct classification rate of 86% using a leave-one-out cross-validation. The predicted probability computed from the logistic model was significantly correlated with delirium severity. These results suggest that the three components are the most important predisposing factors for postoperative delirium, and our prediction model may reflect the core pathophysiology in estimating the probability of postoperative delirium.

Figure 1

Flowchart for participant enrollment.

Figure 2

Mapping brain regions playing essential roles in prediction models. Gray matter (GM) density in the left caudate and suprachiasmatic nucleus (SCN) was significantly d00000000000ecreased in the delirium (DEL) group relative to the non-delirium (No-DEL) group (A), whereas the amplitude of low-frequency fluctuation (ALFF) in the right dorsolateral prefrontal cortex (R. DLPFC) was significantly increased in the DEL group relative to the No-DEL group (B).

Figure 3

Small-world network properties of the structural (A) and functional (B) networks. All measures were normalized by that of 1,000 random networks. Abbreviation: DEL, delirium; No-DEL, non-delirium.

Figure 4

Flowchart of the prediction model for postoperative delirium from variable selection to validity test. Abbreviations: ALFF; amplitude of low-frequency fluctuation; BFI, Big Five Inventory; BFI-C, conscientiousness of the BFI; BFI-N, neuroticism of the BFI; CAU, caudate; DLPFC, dorsolateral prefrontal cortex; DMPFC, dorsomedial prefrontal cortex; FA, factional anisotropy; FN, functional network; GM, gray matter; ITG, inferior temporal gyrus; MMSE, Mini-Mental State Examination; MTG, middle temporal gyrus; PCL, paracentral lobule; SCN, suprachiasmatic nucleus; and SN, structural network.

Figure 5

Performance of the delirium prediction models and correlations between the predicted probability and severity of postoperative delirium: Analysis of the receiver operating characteristics for the clinical (A), biological (B), and combined (C) model; leave-k-out cross validation analysis for each model (D); and Pearson’s correlation analysis between the Korean version of Delirium Rating Scale (KDRS) score and the predicted probability of developing postoperative delirium obtained from the biological model (E) and combined model (F), respectively.