Predictors of delirium not induced by alcohol or other psychoactive substances in older patients: a scoping review
Open Access
- 25 May 2026
- journal article
- Published by ECO-Vector LLC (Publications) in Neurology Bulletin
- Vol. LVIII (2) , 235-246
Abstract
The evolution of digital psychiatry is opening up new possibilities for predicting acute life-threatening psychiatric disorders, such as delirium (an acute state of mental confusion) not caused by alcohol or psychoactive substances (F05 according to ICD-10). A scoping review was conducted according to PRISMA-ScR recommendations in the eLibrary, CNKI, PubMed, and Google Scholar databases (March 2026). Original studies and reviews describing methods for diagnosing delirium, clinical conditions for enrollment, independent predictors, and statistical power of models were included. Among the 54 included publications, 37 models were based on prospective studies and 17 on retrospective studies; 36 models used logistic regression and 18 used machine learning methods. Three domains of delirium diagnosis have been identified: cognitive, complex thinking operations, and circadian. These domains are evident in all clinical diagnostic approaches used. The most valid clinical predictors include: age over 80 years, severity of somatic condition, reduced daily activity, premorbid cognitive decline/dementia, history of delirium, cardiovascular and cerebrovascular comorbidity, polypharmacy, use of anticholinergic drugs, abnormalities in vital signs, and visual and hearing impairments. Laboratory predictors of delirium are classified into three groups: markers of inflammation (interleukin-6, C-reactive protein, TNF-α, platelet indices), markers of neurodegeneration (neurofilament light chain, S100B protein, neuron-specific enolase), and markers of metabolic disorders (urea-to-creatinine ratio, electrolytes, lactate). Among electroencephalography techniques, spectral analysis of rhythm power, wavelet analysis, bispectral index, and patient status index are promising. Morphometric predictors include white matter hyperintensity, global brain atrophy, and diffusion tensor imaging abnormalities. Combining clinical and paraclinical data increases the statistical power of models (AUC up to 0.85–0.92). Modern models for predicting delirium in elderly patients are gradually switching from regression analysis to machine learning, which improves the accuracy of predictions. The integration of clinical predictors with biomarkers of inflammation and neurodegeneration, as well as quantitative electroencephalography indicators and morphometric data, represents the most promising direction for the development of validated risk assessment algorithms adapted to regional characteristics of clinical practice.Keywords
This publication has 66 references indexed in Scilit:
- EEG‐Derived Index Predicts Postoperative Delirium in Elderly Patients With Hip Fracture: A Prospective Study From a Tertiary Medical CenterBrain and Behavior, 2026
- Predicting delirium in older non-intensive care unit inpatients: development and validation of the DELIrium risK Tool (DELIKT)International Journal of Clinical Pharmacy, 2023
- Developing a machine learning model to identify delirium risk in geriatric internal medicine inpatientsEuropean Geriatric Medicine, 2021
- Association of Plasma Neurofilament Light with Postoperative DeliriumAnnals of Neurology, 2020
- Prediction of Incident Delirium Using a Random Forest classifierJournal of Medical Systems, 2018
- Performance of Electronic Prediction Rules for Prevalent Delirium at Hospital AdmissionJAMA Network Open, 2018
- Development and Validation of an Electronic Health Record–Based Machine Learning Model to Estimate Delirium Risk in Newly Hospitalized Patients Without Known Cognitive ImpairmentJAMA Network Open, 2018
- Incidence and Risk Factors for Postoperative Delirium in Patients After HepatectomyWorld Journal of Surgery, 2017
- A systematic review of neuroimaging in delirium: predictors, correlates and consequencesInternational Journal of Geriatric Psychiatry, 2017
- The AWOL tool: Derivation and validation of a delirium prediction ruleJournal of Hospital Medicine, 2013