A nomogram for predicting cognitive decline in mild cognitive impairment based on visual working memory: prospective cohort study

A nomogram was developed to predict cognitive decline in individuals with mild cognitive impairment (MCI) using visual working memory performance and clinical features, aiding in early detection and prevention. Data from 76 MCI subjects were analyzed using logistic regression models incorporating demographic information, neuropsychological test results, and visual working memory performance. Nomogram-based models were developed and validated, with receiver operating characteristic (ROC) analysis, area under the curve (AUC), and decision curve analysis (DCA) used to evaluate efficacy. After one year, 44 individuals (57.9%) experienced cognitive decline, with 20.5% progressing to Alzheimer’s disease (AD). In the multivariate stepwise logistic regression model, four variables were retained: male gender ( OR (95% CI) : 2.62 (0.872–8.347), P = 0.091), lower baseline logic memory scores (0.734 (0.53–0.959), P = 0.038), lower baseline accuracy under high working memory load (1.068 (0.938–1.226), P = 0.328), and lower baseline capacity under high working memory load (0.285 (0.055–1.251), P = 0.109). Only logical memory test score was statistically significantly associated with cognitive decline, while other variables showed trend-level associations. The combination of visual working memory performance, male gender, and logic memory scores yielded an AUC of 0.774 for predicting cognitive decline in MCI. Only baseline logical memory score independently predicts cognitive decline. High-load VWM metrics show no statistical association after adjustment yet marginally enhanced model discrimination, requiring validation in larger cohorts. Chinese Clinical Trial Registry (ChiCTR), ChiCTR2000039550. Registered 31 October 2020.