Abstract
Introduction Rapidly progressive dementia (RPD) is a clinical challenge due to the relatively high prevalence of potentially reversible causes, some of which have specific treatments. Consequently, the diagnostic approach is broader than in slowly progressive dementias. Nevertheless, neurodegenerative (ND) diseases remain a leading etiology in most RPD cohorts. Underreporting of prior cognitive decline contributes to extensive testing, higher costs, longer diagnostic timelines, and invasive procedures, often without identifying causes beyond ND pathology. Objective This study aims to evaluate the predictive utility of factors such as age of onset, symptom duration, absence of neurological signs, and absence of crisis for identifying ND etiologies in patients with RPD. Therefore, we aim to identify a subgroup of patients for whom the diagnostic and therapeutic approach could be optimized, facilitating an accurate diagnosis and improving the prognosis and treatment of these cases. Materials and methods A retrospective, observational, cross-sectional study was conducted of 135 diagnoses, with RPD patients treated at a public hospital in Buenos Aires (2013-2024). Data included demographics, comorbidities, clinical features, diagnostic delay, laboratory and imaging findings, final diagnosis, follow-up, and outcomes. Quantitative data were summarized using means or medians, while categorical data were expressed as percentages. R and RStudio software (version 4.3.2, October 31, 2023) (Posit Software, Boston, MA, US) were used for variable analysis. Results A comprehensive analysis was performed on 135 patients diagnosed with RPD, aiming to develop an algorithm to identify ND diseases as the underlying etiology. Of these, 30 patients were excluded after the initial diagnostic evaluation revealed a definite cause. Among the remaining 105 patients, an additional 35 were excluded because imaging findings suggested a non-ND etiology (such as temporal, thalamic, insular, frontal, and cortical hyperintensities on diffusion-weighted imaging (DWI), T2, and fluid-attenuated inversion recovery (FLAIR) or contrast enhancement). In the remaining 70 patients, the predictive value of selected variables, such as age, symptom duration, absence of neurological signs, and crisis, was evaluated. Among these, age and symptom duration showed significant predictive value for ND etiology (sensitivity (Se) 55%, specificity (Sp) 75%, positive predictive value (PPV) 72%, negative predictive value (NPV) 58%; p = 0.02). Logistic regression, including age, disease progression, and absence of neurological signs, identified age as an independent predictor (p = 0.004). Model performance was moderate (area under the curve (AUC) 0.74; 95% confidence interval (CI) 0.63-0.86). Finally, when applying the model to the cohort, 11 patients were identified, all with ND etiology. Conclusion Considering these factors, ND disorders-historically viewed as diagnoses of exclusion-may become a central component of the diagnostic approach to RPD, alongside autoimmune encephalitis (AE). We propose the development of an algorithm based on a predictive model that allows for the identification of a subgroup of patients in whom the diagnostic approach can be optimized, avoiding unnecessary studies and prioritizing the early identification of a possible ND etiology.