Participants were recruited from the Psychiatric Hospital of Wenzhou between January 2024 and January 2025. All participants were diagnosed with Alzheimer’s disease according to the criteria of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) (17). The diagnostic process included comprehensive clinical evaluations conducted by two licensed psychiatrists, with at least one holding a senior professional title, to ensure diagnostic accuracy and reliability. To minimize diagnostic bias, a consensus diagnosis was required in cases of initial disagreement between clinicians. Inclusion criteria were as follows: (1) age 60 years or older; (2) confirmed diagnosis of Alzheimer’s disease based on DSM-5 criteria; (3) availability of complete Clinical Dementia Rating (CDR) assessments; and (4) provision of written informed consent by participants or their legally authorized representatives. Exclusion criteria included: (1)history of stroke with focal neurological signs and imaging findings consistent with cerebral small vessel disease (Fazekas score ≥2); (2) mental or intellectual developmental disorders; (3) other diseases known to cause cognitive impairment; (4) comorbid conditions that would prevent cooperation with cognitive assessments; (5) refusal to provide informed consent; (6) insufficient clinical documentation in the case report form; (7) focal neurological symptoms and signs consistent with stroke (including hemiplegia, central facial paralysis, Babinski’s sign, sensory disturbances, dysarthria); (8) CT/MRI evidence of multiple macrovascular infarctions, lacunar infarctions, extensive periventricular white matter lesions, or strategically located single infarcts; and (9) clinical diagnosis of vascular dementia. A total of 134 participants met the inclusion criteria and were included in the final analysis.

All participants underwent individual, face-to-face clinical assessments conducted in the outpatient clinic of the hospital. Each evaluation was performed one-on-one by a physician holding a valid psychiatric practice registration certificate. The assessment protocol included structured clinical interviews, during which demographic information, medical history, and cognitive assessments were systematically collected. The Clinical Dementia Rating (CDR) scale was administered by trained clinicians, who rated participants across six functional domains: memory, orientation, judgment and problem solving, community affairs, home and hobbies, and personal care. Each assessment session lasted approximately 45–60 minutes to ensure a thorough evaluation of all domains. Participants were subsequently classified into two groups based on their CDR global scores: the mAD group, with CDR scores of 0.5–1.0, and the Mod-sAD group, with CDR scores of 2.0–3.0 (18).

A total of 134 participants diagnosed with Alzheimer’s disease were included in the final analysis, comprising 37 patients with mAD and 97 patients with Mod-sAD. The demographic characteristics of the participants were presented in Table 1. The two groups were well-matched across most demographic variables, with no significant differences observed in gender distribution (p = 0.11), educational levels (p = 0.75), smoking status (p = 0.36), or drinking habits (p = 0.22). However, patients in the Mod-sAD group were significantly older than those in the mAD group (p = 0.009).

Significant differences were observed across all six domains of the CDR between the mAD and mod-sAD groups (Figure 1). The Mod-sAD group showed significantly higher impairment scores, compared to the mAD group, across all cognitive and functional domains (all p < 0.001). Cohen’s d values indicated large effect sizes across all domains, with Home and Hobbies showing the largest effect size (Cohen’s d = 2.71), followed by Memory (Cohen’s d = 2.67), Community Affairs (Cohen’s d = 2.61), and Judgment and Problem Solving (Cohen’s d = 2.58). Orientation (Cohen’s d = 2.12) and Personal Care (Cohen’s d = 1.83). Correlation analysis revealed significant relationship between demographic variables and CDR domain scores (presented in Table 2). Gender was negatively correlated with memory impairment (r = –0.18, p < 0.05). Age was positively correlated with all functional domains, including memory (r = 0.22, p < 0.05), orientation (r = 0.19, p < 0.05), judgment and problem solving (r = 0.18, p < 0.05), community affairs (r = 0.22, p < 0.01), home and hobbies (r = 0.33, p < 0.001), and personal care (r = 0.33, p < 0.001).

Centrality analyses (Figure 2) further highlights shifts in node importance, particularly in terms of strength. In the mAD network (Figure 3), judgment and problem-solving exhibited the highest strength (strength = 1.65). By contrast, in the Mod-sAD network, memory emerged as the node with the highest strength (strength = 1.62), while the strength of judgment and problem-solving showed a sharp decline (strength = −0.39). Moreover, the Mod-sAD network demonstrated increased strength values for orientation (strength = 0.40) relative to the mAD group (strength = −1.32). In addition, memory showed a notable increase in both closeness (closeness = 1.08) and betweenness (betweenness = 1.63), compared to the mAD group (closeness = −0.04; betweenness = 0.78). Considering that all edge weights were positive, expected influence values were identical to strength values and thus were not reported separately.

To assess potential redundancy among the CDR domains and to rule out clustering coefficient inflation that might bias centrality estimates, clustering coefficients were calculated for each node. Across all nodes, clustering coefficients were relatively low, ranging from 0.11 to 0.18, with an overall mean clustering coefficient of 0.16 (SD = 0.02). The node with the highest clustering coefficient was judgment and problem solving (Clustering coefficient = 0.18). As shown in Figure 4, there was no significant correlation between strength and clustering coefficient in the symptom network model(|r| < 0.30, p > 0.56).

One of the most salient findings of this study was the significantly greater network density and global strength observed in the moderate-to-severe Alzheimer’s disease (Mod-sAD) group compared to the mild AD (mAD) group. This pattern suggests that individuals with more severe AD tend to show more interconnected cognitive and functional impairments, reflecting a more integrated system of deficits. These results are consistent with the cascade hypothesis of neurodegeneration, which posits that pathological processes in AD propagate through interconnected brain networks, resulting in progressive deterioration across multiple cognitive domains (21, 22). The increased connectivity observed in advanced stages may be indicative of more pervasive and interdependent functional impairments. Although our data, based solely on CDR domains, cannot directly address underlying neuropathological mechanisms, this pattern is broadly in line with neuroimaging evidence showing altered connectivity in Alzheimer’s disease (23, 24). For example, Brier et al. (2014) reported that disruptions in default mode network connectivity are associated with greater cognitive impairment severity, suggesting that network-level alterations may contribute to the clinical manifestations of advanced AD (25).

The centrality analyses revealed a fundamental reorganization of core symptoms between disease stages, with memory becoming the node with the highest strength and betweenness in moderate to severe AD. This finding provides important insights into the evolving role of memory dysfunction throughout AD progression. In the early stages, memory impairments represented a relatively isolated deficit that has not yet extensively permeated other cognitive domains. Across disease stages, memory dysfunction appears increasingly central within the overall symptom profile, suggesting that memory-related deficits may be linked to a broader pattern of functional impairments. This shift in centrality aligns with the well-established understanding of AD as primarily a disorder of memory systems, particularly involving the hippocampus and associated medial temporal lobe structures (26, 27). The increased centrality of memory in moderate-to-severe stages reflects the progressive involvement of memory-related neural networks, which become increasingly critical for maintaining other cognitive functions. Greater impairment of memory systems may coincide with widespread dysfunction across interconnected cognitive domains. Conversely, judgment and problem-solving demonstrated the highest centrality in mild AD but showed a dramatic decline as the disease progressed to moderate-to-severe stages.

The substantial increase in orientation centrality from mild to moderate-to-severe AD represents another significant finding warranting discussion. Orientation, encompassing awareness of time, place, and person, showed markedly increased connectivity and influence within the symptom network as AD progressed. This finding may reflect the fundamental role of orientation in organizing and coordinating other cognitive functions (28). As orientation becomes increasingly impaired in advanced AD, it may serve as a critical bridge connecting various cognitive deficits, potentially explaining why disorientation is such a prominent and distressing feature of moderate-to-severe dementia. Gennaro et al. further highlighted that the conversion from normal aging to AD may be traced through allocentric distance-based deficits, a core component of spatial orientation capacity, underscoring the pivotal role of orientation impairments in early disease detection and progression monitoring (28). From a clinical perspective, the emergence of orientation as a central node in advanced AD has important implications for assessment and intervention strategies.

Beyond the centrality symptoms shifts, several new edge connections emerged as AD progressed from mild to moderate-to-severe stages. Specifically, Memory-Personal Care, Orientation-Judgment and Problem Solving, and Orientation-Home and Hobbies—provides insights into the dynamic reorganization of cognitive-functional networks during disease progression. This finding aligns with previous network studies suggesting that AD progression involves not merely the loss of connections, but also the formation of new pathological or compensatory pathways (29). The Memory-Personal Care connection particularly supports the cascade model of functional decline, where cognitive impairments progressively impact instrumental and basic activities of daily living (30, 31). This coupling reflect the increasing reliance of self-care abilities on intact memory systems, consistent with studies showing that memory deficits predict functional deterioration in moderate-stage AD (32). The emergence of Orientation-Judgment and Problem-Solving connections corroborate findings that executive dysfunction and disorientation become increasingly interrelated as AD advances (33, 34). Similarly, the Orientation-Home and Hobbies connection reflects the documented relationship between spatial disorientation and the abandonment of complex leisure activities (35). These emerging connections represent either compensatory recruitment of cognitive resources, as suggested by neuroimaging studies showing hyperactivation in early AD stages (36), or pathological coupling reflecting shared vulnerability to neurodegeneration (37). Future longitudinal network studies are needed to distinguish between these mechanisms and their implications for intervention strategies.