Network analysis of frailty indicators in hospitalized elderly patients: unveiling the role of depression and hemoglobin as core factors

Ruxia Qiu,Yanhong Gu

DOI: https://doi.org/10.1007/s40520-023-02608-3

Abstract:Background: Frailty is a significant concern among hospitalized older adults, influenced by multiple factors. Understanding the complex interactions between these variables can be facilitated through a network perspective. Aim: This study aimed to identify the core factor and physiological indicator of frailty in hospitalized elderly patients and visualize their interactions within the network structure. Methods: Frailty was assessed using the Tilburg Frailty Indicators, with a score of 5 or higher indicating frailty. Additional variables related to sociodemographic, physical and clinical, psychological and cognitive aspects, as well as physiological indicators, were extracted from electronic health records. A partial correlation network analysis was conducted using an adaptive LASSO algorithm, based on univariate correlation and logistic regression, to examine the network structure and identify influential nodes. Results: The average age of participants was 70.74 ± 7.52 years, with 24.27% classified as frail. Frailty was associated with 38 of 145 initially included variables (P < 0.05). The network analysis revealed depression as the most central node, followed by drugs used, sleep disorders, loneliness, masticatory obstacles, drinking, and number of teeth missing. Hemoglobin emerged as the most central biochemical indicator in the network, based on network center index analysis (Strength = 4.858, Betweenness = 223, Closeness = 0.034). Conclusions: Frailty in hospitalized older adults is influenced by various social, physical, and psychological factors, with depression as the core factor of utmost importance. Changes in hemoglobin levels could serve as an essential indicator. This innovative network approach provides insights into the multidimensional structure and relationships in real-world settings.