Symptom network analysis in elderly patients with chronic gastritis at two time points: a repeated cross-sectional study

Abstract

Objective

Based on the biopsychosocial model and the Comprehensive Geriatric Assessment framework, this study employed network models to compare symptom association patterns at two time points (hospital admission and 6-month follow-up) and to examine differences in symptom network structure between these time points.

Methods

A longitudinal design with convenience sampling was used to recruit 360 hospitalized chronic gastritis patients aged ≥ 60 years. Assessments were conducted at admission (T1) and 6 months after discharge (T2). Data were collected using a general information questionnaire, the Comprehensive Geriatric Assessment, and the Gastrointestinal Symptom Rating Scale. Symptom networks were estimated using R, and centrality indices were calculated to identify high-centrality symptoms and structural changes.

Results

Network analysis of physical and psychological symptoms in elderly patients with chronic gastritis revealed that anxiety, depression, and cognitive function (MD9, MD10, MD11) constituted high-centrality correlational nodes at T1, whereas frailty, self-care ability, and fall risk (MD8, MD6, MD7) emerged as the most strongly interconnected nodes at T2. It is important to note that these designations solely reflect partial correlation strength and do not imply causal relationships or establish clinical priority. The centrality of gastrointestinal symptoms (abdominal pain, diarrhea, constipation) and activities of daily living (Barthel Index) significantly decreased at T2 (strength C = − 0.176 to − 0.824, P < 0.05), whereas psychosocial factors (anxiety, depression, well-being) remained stable (P > 0.05). Global network strength increased from 4.75 (T1) to 5.95 (T2) (P < 0.001).

Conclusion

The symptom network demonstrated temporal shifts in its statistical connectivity structure, characterized by evolving correlational hubs. While psychological indicators (anxiety, depression, cognitive function) emerged as the most strongly connected nodes at T1, somatic functional status indicators (frailty, self-care capacity, fall risk) exhibited heightened connectivity at T2. Notably, psychosocial factors persisted as stable correlational hubs within the bidirectional mind-body network across both time points, albeit not necessarily representing definitive primary clinical targets. These findings offer descriptive empirical support and potential clinical reference points for designing whole-cycle, stratified, integrated traditional Chinese and Western medicine interventions for older adults with chronic gastritis.

As China enters a moderately aging society, health challenges among older adults have become increasingly prominent [1]. Chronic gastritis (CG) is characterized by recurrent gastrointestinal symptoms, including abdominal pain, bloating, nausea, and constipation. The disease is typically long-standing and prone to relapse, and is frequently accompanied by psychological comorbidities such as anxiety and depression, substantially impairing quality of life. Its prevalence and severity generally increase with advancing age [2]. In older populations, the onset and progression of CG are closely associated with age-related physiological decline, multimorbidity, and transitions in social roles. Current conventional treatment models predominantly focus on localized gastric mucosal inflammation, often overlooking the multidimensional health needs of older patients, which may contribute to prolonged disease courses or therapeutic resistance [3].

Comprehensive Geriatric Assessment (CGA), a cornerstone of geriatric medicine, employs a multidisciplinary approach to systematically evaluate physical function, psychological status, social environment, and quality of life in older adults. Based on these assessments, individualized treatment and care plans are developed to maintain and improve functional capacity and overall well-being, emphasizing holistic evaluation of psychosomatic symptoms [4]. However, most existing studies focus merely on static analyses of physical and psychological symptoms, and fail to conduct in-depth research on change patterns of multifactorial network structures across two time points. Based on the biopsychosocial medical model, this study integrated symptom network analysis with CAG and conducted a multi-time-point longitudinal analysis involving two observation time points, covering biological dimensions (gastrointestinal symptoms, physical function, frailty, fall risk, cognition, and pathological changes), psychological dimensions (anxiety, depression, well-being), and social dimensions (social participation, interpersonal relationships, social support, and demographic characteristics). The symptom association structures at the two time points were compared, and the temporal shifts in indicator interrelationships were characterized based on cross-sectional comparisons at two time points.

Given that integrated traditional Chinese and Western medicine intervention provides holistic conditioning advantages over Western medicine alone, this study incorporated traditional Chinese medicine (TCM)-related observation indicators. Existing research has demonstrated that after 6 months of standardized integrated TCM-Western medicine intervention, elderly patients with CG may show varying degrees of improvement in gastrointestinal symptoms, emotional status, and cognitive indicators [5]. Factors such as physiological compensation, medication adjustments, and lifestyle modifications can all contribute to statistically significant changes in scale scores. Moreover, Montreal Cognitive Assessment (MoCA) scores are responsive to short-term intervention, and the relief of gastrointestinal discomfort can mitigate subjective cognitive complaints [5]. These findings provide clinical and theoretical basis for the between-group differences in indicators at the two time points and can also serve as a reference for developing individualized, full-cycle, mind-body integrated interventions combining TCM and Western medicine for elderly patients with CG.

Materials and methods

Study design and participants

This longitudinal cohort study adopted convenience sampling to recruit elderly inpatients diagnosed with CG from the Departments of Geriatrics and Spleen-Stomach Diseases at a Binhai County Hospital of Traditional Chinese Medicine, Jiangsu Province, China, from November 2024 to June 2025.

Eligible participants met the following criteria: (1) Age ≥ 60 years, any gender. (2) Diagnosis of CG confirmed by endoscopy and histopathology according to the Chinese Guidelines for the Diagnosis and Treatment of CG (2022, Shanghai) [6, 7]. Endoscopic and pathological findings were independently reviewed by at least two senior gastroenterologists (associate chief physician level or above). TCM syndrome differentiation was independently performed by two senior TCM or integrative gastroenterology physicians. In cases of diagnostic discrepancy, a third senior specialist adjudicated until consensus was achieved. (3) Enrollment from an integrated Chinese and Western medicine hospital and receipt of routine treatment based primarily on integrated Chinese and Western medicine approaches. (4) Provision of written informed consent and the ability to comply with study procedures.

Exclusion criteria included: (1) Severe primary cardiac, cerebrovascular, hepatic, or renal disease as the principal admission diagnosis. (2) Individuals with severe anxiety disorders, depressive disorders, other mental disorders, or cognitive impairments (including those requiring prescription psychiatric medications for these conditions) as assessed by a licensed physician shall be excluded. (3) incomplete clinical data.

Sample Size Estimation: A total of 24 independent variables were considered for baseline description, with the initial sample size estimated based on the rule of ten participants per variable [8], yielding a preliminary estimate of 240 participants. After incorporating an anticipated 20% proportion of invalid questionnaires, the minimum required sample size was calculated to be 288 participants. Ultimately, 360 patients were recruited, exceeding the adjusted target and enhancing statistical power and the stability of network parameter estimation. Importantly, recruitment procedures and eligibility criteria remained unchanged throughout the study period; thus, the larger sample reflects the natural availability of eligible patients and does not introduce additional selection bias beyond that inherent to convenience sampling.

A total of 24 independent variables—covering demographic characteristics, disease-related and TCM-related indicators, gastrointestinal symptoms, and CGA items—were used for baseline description, confounding adjustment, and subgroup analyses. By contrast, only 14 core clinical symptom nodes were selected to construct the final symptom network model, as the two sets of variables served different research purposes. The obtained sample size was sufficient for network analysis, achieving satisfactory statistical performance and model stability (e.g., correlation stability coefficient > 0.5). This reflects a rational sample design and supports the credibility of the findings.

This study was approved by the Ethics Committee of Binhai County Hospital of Traditional Chinese Medicine, Jiangsu Province (Approval No. [2024]2) and was performed in accordance with the Declaration of Helsinki. The purpose of the study was explained to all participants before the survey was conducted, and written informed consent was obtained. All supplementary materials have been anonymized; no personally identifiable information is included. The ethics certificate provided in the Appendix is specific to this CG study.

Measures

In this study, the Cronbach’s alpha coefficients of some scales were relatively high. This can be attributed to the high homogeneity of symptoms and strong inter-item associations in the elderly population, as well as the large sample size and complete data. This pattern is consistent with the reliability distribution characteristics observed in geriatric inpatient populations, indicating that the scales have good internal consistency and yield stable, reliable results.

Demographic and clinical characteristics

A self-designed structured questionnaire collected demographic information (age, sex, marital status, education level, monthly income, and medical insurance type) and clinical data (disease duration, family history of gastrointestinal malignancy, smoking and alcohol use, diagnosis, TCM syndrome classification, ¹³C urea breath test results, and gastroscopy findings).

Gastrointestinal symptoms

Gastrointestinal symptom severity was assessed using the Chinese version of the Gastrointestinal Symptom Rating Scale (GSRS) [9]. The GSRS comprises 15 items across five domains: abdominal pain, reflux, dyspepsia, diarrhea, and constipation. Items are rated on a 4-point Likert scale, yielding total scores ranging from 15 to 60, with higher scores indicating greater symptom severity [10]. In this study, internal consistency was acceptable (Cronbach’s α = 0.833).

CGA domains

A multidimensional CGA framework was applied to systematically evaluate physical, psychological, social, and functional domains [4]. The specific scales are shown as follows:

1. Physical function

Activities of Daily Living (ADL) was assessed using the Barthel Index (BI) [11], which evaluates 10 basic daily activities (total score 0–100). Higher scores reflect greater independence. Cronbach’s α = 0.920.

Fall risk was measured using the Falls Risk Assessment Scale for Older People (FRASE) [12], consisting of 35 items across eight domains (score range: 0–53). Higher scores indicate elevated fall risk. Cronbach’s α = 0.759.

• (2). Psychological status

Symptoms of anxiety were assessed using the Self-Rating Anxiety Scale (SAS) [13], which consists of 20 items rated on a 4-point scale. Raw scores are multiplied by 1.25 to generate standardized scores (range 25–100). Cronbach’s α = 0.991.

Depressive symptoms were assessed using the Self-Rating Depression Scale (SDS) [14], which also comprises 20 items rated on a 4-point scale, with standardized scoring procedures identical to the SAS. Cronbach’s α = 0.990.

Global cognitive performance was evaluated using the MoCA [15]. The MoCA assesses eight cognitive domains, with total scores ranging from 0 to 30. One additional point was added for participants with ≤ 12 years of formal education. Scores < 26 indicated cognitive impairment. Cronbach’s α = 0.980.

• (3). Quality of life

Subjective well-being was measured using the Memorial University of Newfoundland Scale of Happiness (MUNSH) [16], comprising 24 items across four domains: (positive affect, negative affect, positive experience, and negative experience). Total scores range from 0 to 48, with higher scores reflecting greater happiness. Cronbach’s α = 0.990.

• (4). Social function

Social participation was assessed according to the industry standard issued by the Ministry of Civil Affairs, the Assessment of Ability of Older Adults (MZ/T001-2013) [17]. Five domains were rated from 0 to 4 (total 0–20), with higher scores indicating greater impairment. Cronbach’s α = 0.998.

Role adaptation (perception, satisfaction, and adaptation to one’s own roles) was evaluated using the Interpersonal Functioning Scale [18], a 16-item measure (score range: 16–48), with lower scores indicating poorer interpersonal functioning. Cronbach’s α = 0.962.

Frailty was screened using the FRAIL scale [19], which evaluates fatigue, resistance, ambulation, comorbidity burden, and weight loss. Scores range from 0 to 5: 0 = robust, 1–2 = pre-frail, ≥ 3 = frail. Cronbach’s α = 0.569.

Data collection procedures

All participants were assessed at hospital admission (T1) and the 6-month post-discharge follow-up (T2). All subjects completed the follow-up without any loss to follow-up or missing data. Trained research staff administered all questionnaires in a standardized, quiet setting using combined paper-based and electronic formats. Given the potential fatigue and reduced endurance of older participants, assessments were permitted to be completed in multiple sessions. All data were anonymized and coded prior to analysis. Double data entry was independently performed by two investigators, followed by cross-verification to ensure accuracy and minimize transcription errors.

A strict 6-month longitudinal follow-up was conducted in this study. The full sample was tracked through inpatient registration, admission education, telephone follow-up, and community collaboration. All 360 patients successfully underwent both the T1 (baseline, at admission) and T2 (6 months after discharge) evaluations, with no loss to follow-up and missing data. This high retention rate was achieved through proactive tracking and support; nevertheless, the absence of attrition may introduce a volunteer bias, which is acknowledged in the Discussion.

Statistical analysis

Statistical analyses were conducted using SPSS version 27.0 and R version 4.3.3. Continuous variables were presented as mean ± standard deviation (SD), and categorical variables as frequencies and percentages. Normality of continuous variables was checked using the Shapiro–Wilk test, which indicated approximately normality for most variables. For the few variables with minor deviations, the large sample size justified the use of paired t-tests based on the central limit theorem. As the T1 and T2 data were obtained from the same subjects in a paired design, a paired t-test was employed to compare scale scores across the two time points. Since no distinct subject subgroups were included in this study, independent-samples t-tests and one-way ANOVA were not performed.

Standardized effect sizes (Cohen’s d) were calculated for all paired comparisons, including both the original scale scores and the changes in node strength derived from the network analysis (see below). Effect sizes were interpreted according to Cohen’s conventional criteria: 0.2 = small, 0.5 = medium, and 0.8 = large.

A weighted symptom network was constructed using the qgraph package in R. The network was estimated as a regularized partial correlation network based on the Gaussian graphical model, with extended Bayesian information criterion (EBIC) penalization for regularization. The EBIC criterion automatically identified non-zero edges while eliminating irrelevant associations. Only the 14 core clinical symptom nodes (all continuous variables, each representing a scale score) were included in the network model; the remaining demographic and other independent variables were not entered. Model assumptions included multivariate normality of the variables and sparsity of partial correlations among them, these assumptions were approximately satisfied given the sample size and the regularization procedure.

Because T1 and T2 data were paired dependent samples from the same subjects, the NetworkComparisonTest (NCT) was employed in its dependent (paired) sample mode to compare the networks across the two time points, in line with the two-wave repeated-measures design. Bootstrap sampling (n = 1000) was performed to examine the stability of the overall network and centrality indices. The correlation stability (CS) coefficient was calculated, with a value > 0.5 considered as the criterion for a stable and reliable network. Node strength and closeness centrality were prioritized for interpretation; betweenness centrality was calculated only for descriptive reference due to its poor reliability.

Reverse scoring was applied to the Barthel Index, MoCA, Interpersonal Functioning Scale, and MUNSH to unify the direction of network analysis, so that higher scores consistently indicated greater symptom severity or functional impairment. The reverse-scored value was calculated as: Reverse Score = Theoretical Full Score − Original Score. There were no missing data in this study; therefore, no imputation was performed.

Results

General characteristics of elderly patients with CG

The sample of 360 elderly patients with CG consisted of 61.67% males, with a mean age of 71.53 ± 8.26 years; 60.00% had an education level of primary school or lower; 78.61% reported a monthly income under 2000 yuan; and 82.22% were covered by resident medical insurance. Chronic atrophic gastritis (CAG) accounted for 77.22% of cases. In terms of TCM syndrome differentiation, spleen-stomach deficiency syndrome was the most common (33.33%). The positive rate of Helicobacter pylori infection was 12.50%. Detailed demographic and clinical characteristics are presented in Table 1.

Table 1.

General demographic data of elderly patients with chronic gastritis (n = 360)

Variables	Groups	n	%
Gender	Male	222	61.67
	Female	138	38.33
Age Group	60～70	164	45.56
	71～80	130	36.11
	> 80	66	18.33
Marital Status	Married	358	99.44
	Unmarried	2	0.56
Education Level	Primary school or below	216	60.00
	Junior high school	118	32.78
	Senior high school or higher	26	7.22
Monthly Income (RMB)	< 2000	283	78.61
	≥ 2000	77	21.39
Type of health insurance	Employee medical insurance	58	16.11
	Resident medical insurance	296	82.22
	Commercial health insurance	6	1.67
Disease Duration	< 2 years	129	35.83
	2～5 years	148	41.11
	≥ 6 years	83	23.06
Family History of Digestive Tract Cancer	Yes	0	0.00
	No	360	100.00
Smoking History	Yes	27	7.50
	No	333	92.50
Drinking History	Yes	28	7.78
	No	332	92.22
Diagnosis	Chronic non-atrophic gastritis	82	22.78
	Chronic atrophic gastritis	278	77.22
TCM Diagnosis	Liver-stomach disharmony syndrome	87	24.17
	Spleen-stomach damp-heat syndrome	68	18.89
	Spleen-stomach weakness syndrome	120	33.33
	Stomach yin deficiency syndrome	51	14.17
	Stomach collateral blood stasis syndrome	34	9.44
13C (Carbon-13) Urea Breath Test	Positive	45	12.50
	Negative	315	87.50
Gastroscopy Result	Other	29	8.06
	Superficial gastritis	87	24.17
	Atrophic gastritis	244	67.78

Descriptive statistics of symptom scores at T1 and T2

All 360 patients completed the 6-month follow-up without loss or missing data at either T1 or T2. All patients received standardized integrated TCM-Western medicine treatment and home-based health management throughout the study. Intervention factors, including medication, diet, and rehabilitation, were holistically managed under a unified protocol. Therefore, the overall intervention context during the 6-month follow-up can reasonably account for the observed changes in scale scores from a descriptive perspective. However, because individual-level treatment data were not collected, no causal attribution can be made regarding which specific components contributed to the changes.

Compared with T1, most indicators showed a trend toward improvement at T2. The syndrome scores of abdominal pain (d = 0.473), reflux (d = 0.753), dyspepsia (d = 0.504), and diarrhea (d = 0.577) all decreased significantly (all P < 0.001).

The negative Barthel Index score decreased (d = 0.126, P = 0.017), as did the frailty screening score (d = 0.169, P = 0.001), suggesting that although the improvements in self-care ability and overall frailty status were statistically significant, the effect sizes were relatively small. In the psychological domain, anxiety (d = 0.781) and depression (d = 0.983) scores decreased significantly, and the negative MoCA score also declined (d = 0.331, P < 0.001). The MoCA assessment was administered face-to-face by uniformly trained investigators. The study predominantly involved elderly patients with limited education and some degree of cognitive decline, but patients with severe cognitive or psychiatric disorders were excluded. All scales were administered with one-on-one assistance from investigators, controlling for cognitive bias. Detailed results are shown in Table 2.

Table 2.

Symptom scores of elderly patients with chronic gastritis at T1 and T2 (n = 360)

Variables	T1 stage	T2 stage	t value	P value	Cohen’s d(95%CI)
Abdominal Pain Syndrome	5.42 ± 1.29	5.24 ± 1.25	8.977	< 0.001	0.473 (0.365–0.581)
Reflux Syndrome	5.79 ± 2.61	5.38 ± 2.55	14.277	< 0.001	0.753 (0.640–0.866)
Dyspepsia Syndrome	5.40 ± 2.12	5.20 ± 2.10	9.556	< 0.001	0.504 (0.396–0.612)
Diarrhea Syndrome	3.61 ± 1.47	3.36 ± 1.36	10.939	< 0.001	0.577 (0.468–0.686)
Constipation Syndrome	8.14 ± 2.13	8.37 ± 3.83	-1.131	0.259	0.060 (-0.046–0.166)
Barthel Index (Negative)	21.01 ± 16.72	17.92 ± 16.59	2.391	0.017	0.126 (0.022–0.230)
Fall Risk	5.04 ± 3.35	5.96 ± 6.84	-2.313	0.021	0.122 (0.018–0.226)
Frailty Screening	2.19 ± 1.00	1.81 ± 1.93	3.206	0.001	0.169 (0.065–0.273)
Anxiety	68.63 ± 16.99	54.58 ± 16.87	14.812	< 0.001	0.781 (0.667–0.895)
Depression	68.91 ± 16.35	54.58 ± 16.23	18.63	< 0.001	0.983 (0.865–1.101)
Montreal Cognitive Assessment (Negative)	13.88 ± 13.53	9.43 ± 8.57	6.274	< 0.001	0.331 (0.225–0.437)
Social Participation	8.74 ± 9.26	9.13 ± 8.20	-0.596	0.551	0.031 (-0.075–0.137)
Interpersonal Relationship (Negative)	33.44 ± 12.85	32.54 ± 9.35	1.075	0.283	0.057 (-0.049–0.163)
Well-being (Negative)	26.83 ± 16.20	25.90 ± 14.22	0.809	0.419	0.043 (-0.063–0.149)

The Barthel Index, MoCA, interpersonal relationship, and happiness scores shown are negatively transformed scores, calculated as: Negative score = Full score–Original score. This transformation is applied solely to unify the direction for network analysis. The T2 mean scores for anxiety and depression are identical (54.58) due to rounding; their raw means differ trivially. Cohen’s d values were interpreted according to conventional criteria: 0.2 = small, 0.5 = medium, 0.8 = large

Symptom network analysis in elderly patients with CG

Network analysis was conducted to explore the structural characteristics of symptom clusters at T1 and T2. At T1, cognitive function (MD11), anxiety (MD9), and depression (MD10) exhibited the highest node strength and closeness centrality, forming the predominant correlational hubs upon admission; however, this statistical pattern does not imply psychological factors are the root clinical cause of all symptoms. By T2, centrality metrics for the Barthel Index (MD6), fall risk (MD7), and frailty screening (MD8) increased significantly, indicating stronger inter-variable correlations among somatic functional indicators six months post-discharge. Crucially, this shift reflects changes in partial correlation magnitude rather than confirming somatic function as the dominant causal driver, as psychosocial nodes maintained stable connectivity across both time points. (See Fig. 1.)

Fig. 1.

Cross-sectional network of symptom clusters in T1 and T2 stages among older adults with chronic gastritis

Betweenness centrality was excluded from the core interpretation owing to poor stability. By T2, the centrality metrics of the Barthel Index, fall risk (MD7), and frailty (MD8) had risen markedly, suggesting that physical function progressively became a critical hub linking the network. Centrality metrics are presented in Fig. 2.

Fig. 2.

Centrality measures of symptom nodes in T1 and T2 stages

Strength difference tests demonstrated that, at T1, diarrhea syndrome (MD4) and constipation syndrome (MD5) exhibited the lowest strength values and were therefore more likely to differ significantly from other symptoms. By T2, test results showed that significant differences in node strength were more prevalent. The distribution of black and white cells indicated that, with disease progression, the strength differences among symptom nodes intensified significantly, reflecting a heightened differentiation in the association strength of indicators across time points in the network structure. (Fig. 3)

Fig. 3.

Difference test of symptom node strength between T1 and T2 stages

Bootstrap analysis of edge weights (95% confidence intervals) showed differences in estimation precision between T1 and T2. At T1, confidence intervals narrowed substantially when edge weights exceeded 0.25, indicating accurate estimation of stronger connections. At T2, this threshold increased to 0.40, suggesting that only stronger associations achieved high estimation stability in the later stage. This pattern reflects strengthened symptom interconnections over time and provides a basis for subsequent centrality stability testing. (Fig. 4)

Fig. 4.

Bootstrap 95% confidence intervals showing edge weights of the network in T1 and T2 stages

A subsample bootstrap procedure (n = 1,000) was used to assess the stability of the centrality indices. At T1, the CS coefficients were 0.75 (strength), 0.361 (closeness), and 0 (betweenness). At T2, the CS coefficients were 0.75 (strength), 0.75 (closeness), and 0.283 (betweenness). Therefore, strength centrality was sufficiently stable at both time points. Closeness centrality was stable at T2 but not at T1 (CS = 0.361 < 0.5). Betweenness centrality was unstable at both time points (CS < 0.5). See Fig. 5.

Fig. 5.

Results of the subsample Bootstrap test for the symptom network in T1 and T2 stages

Comparative network analysis between T1 and T2

Global network strength increased significantly from 4.75 (T1) to 5.95 (T2) (P < 0.001), indicating tighter interconnections among symptoms and functional indicators over time.

Node centrality analysis demonstrated that the centrality of abdominal pain syndrome (MD1; strength C = − 0.176, P = 0.004, d = 0.42), diarrhea syndrome (MD4; C = − 0.666, P = 0.001, d = 0.88), constipation syndrome (MD5; C = − 0.551, P = 0.001, d = 0.75), and the Barthel Index representing functional capacity (MD6; C = − 0.824, P = 0.001, d = 1.12) significantly decreased. These findings suggest that these variables served as key nodes in the network at T1, whereas their relative importance diminished at T2. This pattern indicates an alleviation of gastrointestinal symptoms and improvement in functional limitations, thereby reducing their correlational influence within the overall health network.

In contrast, all psychosocial functioning nodes (anxiety, depression, social participation, subjective well-being) showed no statistically significant temporal changes in centrality metrics (P > 0.05, |d| < 0.10). This stability confirms that psychosocial factors remained consistent correlational hubs at T2, even as gastrointestinal and basic daily function nodes declined in connectivity. Although somatic functional indicators formed the most strongly correlated cluster at T2, psychosocial dimensions retained persistent, stable interconnections across all symptom domains, maintaining their network-wide associational influence. See Table 3.

Table 3.

Difference test of centrality measures of symptom nodes between T1 and T2 stages (n = 360)

Nodes	Variables	Strength C	Strength P value
MD1	Abdominal Pain Syndrome	-0.176	0.004
MD2	Reflux Syndrome	-0.113	0.071
MD3	Dyspepsia Syndrome	-0.075	0.097
MD4	Diarrhea Syndrome	-0.666	0.001
MD5	Constipation Syndrome	-0.551	0.001
MD6	Barthel Index (Negative)	-0.824	0.001
MD7	Fall Risk	-0.202	0.079
MD8	Frailty Screening	0.110	0.363
MD9	Anxiety	0.061	0.539
MD10	Depression	0.011	0.907
MD11	Montreal Cognitive Assessment (Negative)	-0.029	0.710
MD12	Social Participation	-0.019	0.811
MD13	Interpersonal Relationship (Negative)	0.052	0.464
MD14	Well-being (Negative)	0.037	0.636

Discussion

Predominant spleen–stomach deficiency and elevated socioeconomic risk in elderly cag: implications for comprehensive intervention

In accordance with the 2024 National Report on the Development of Elderly Care Affairs, participants aged ≥ 60 years including the 60~64-year-old subgroup, were enrolled [20]. Among the 360 enrolled patients, the mean age was 71.53 ± 8.26 years, and chronic atrophic gastritis (CAG) accounted for 77.22% of all cases, consistent with the age-dependent increasing prevalence of atrophic gastritis reported by Dilaghi et al. [21], which reported an increasing prevalence of atrophic gastritis with advancing age. In terms of TCM syndrome differentiation, spleen-stomach deficiency syndrome accounted for 33.33%, aligning with age-related decline in spleen and stomach function in older adults. Accordingly, treatment should follow the principle of strengthening the spleen and harmonizing the stomach [22].

Most participants enrolled in this study were from disadvantaged socioeconomic backgrounds characterized by low educational attainment and limited household income. Such populations typically suffer from constrained access to health-related information, insufficient health literacy, and unsatisfactory treatment adherence, which may exert adverse effect long-term disease management. From a clinical perspective, these findings highlight the necessity of targeted health education and refined clinical management protocols to boost treatment engagement and medication adherence among this patient cohort. At T2, fall risk increased. Rodrigues et al. [23] demonstrated that elevated fall risk in older adults is associated with sarcopenia, functional decline, reduced resistance training, and poor quality of life. Therefore, clinical management should adopt a comprehensive approach addressing both gastrointestinal symptoms and psychological status. Regular assessments of physical function, nutritional status, and fall risk are warranted, alongside individualized dietary guidance, medication safety education, and appropriate physical exercise, with the aim of improving quality of life and ensuring patient safety. The proportion of married individuals in our sample was 99.44%, a figure that may restrict the external validity of the findings. This observation may be attributed to the fact that hospitalized patients receiving spousal care demonstrate better follow-up adherence, which leads to the underrepresentation of unmarried individuals in the sample. Existing evidence has confirmed that marital support affects health behaviors, treatment adherence and disease prognosis [24]. We collected demographic information including income and insurance type to provide a more complete baseline profile. These variables can influence treatment-related behaviors and were originally intended for use in later subgroup stratification; however, because this study prioritized symptom network analysis, they were not incorporated into the main analytical model.

Evolution of the symptom network from psychological dominance to physical–functional core

Network analysis showed that at T1, anxiety (MD9), depression (MD10), and cognitive function (MD11) exhibited elevated node strength and closeness centrality. It is important to note that centrality indices represent statistical associations, not causal relationships or clinical priority; drawing on existing brain–gut axis literature [25] can theoretically inform exploration of association mechanisms, but the cross-sectional correlation network cannot confirm causality. The elevated centrality of the Barthel Index, fall risk, and frailty screening nodes at T2 merely reflects a statistical change in the distribution of variable associations between the two time points.

Although all participants received standardized integrated TCM-Western medicine treatment during hospitalization, detailed individual data on medication type, dosage, treatment duration, and medication adherence were not collected. Therefore, the decreased associations among gastrointestinal indicators and the shift in network node centrality cannot be attributed to the effects of the integrated intervention. Based on the association data, the centrality of the anxiety and depression nodes showed no statistically significant change after the improvement of somatic symptoms, suggesting an asynchronous nature of the mind–body connection. Psychological distress is influenced by multiple factors, including long-term adaptation to chronic illness, age-related psychological traits, and changes in social roles; therefore, only statistical associations exist between somatic and psychological domains, and no single deterministic effect can be established. The cognitive function node (MD11) occupied a prominent position in the T1 network and was closely associated with psychological, somatic, and gastrointestinal symptoms. This may be closely associated with the fact that elderly patients, due to long-term gastrointestinal discomfort, experience reduced social activity, decline in physical function, and cognitive impairment, which in turn lead to misinterpretation of disease information, decreased treatment adherence, and somatic amplification. This finding is consistent with the study by Liang X et al. [26], which showed that specific gut microbial signatures influence both hippocampal volume and cognitive function.

At T2, somatic function variables formed the cluster with the highest node strength, marking a quantifiable shift in the pairwise exhibiting the strongest statistical associations. Notably, this did not diminish the importance of psychological factors; anxiety, depression, and cognition remained stably interconnected across all symptom domains, serving as psychosocial factors that persisted as stable correlational hubs alongside the newly dominant somatic function cluster. This temporal pattern is consistent with the findings of Zhu Y et al. [27] on the correlations among sarcopenia, depression, cognitive function, and frailty. Their cross-sectional study also confirmed that decline in physical function and increased frailty in older adults are important core features of the geriatric syndrome complex [26, 27]. This is consistent with the structural pattern in which somatic function nodes were dominant in the T2 network of this study, and also with the typical symptomatic characteristics of the geriatric frailty syndrome.

The study by Sahin UK et al. [28] showed that dependence in activities of daily living is a core risk factor for sarcopenia, while frailty continuously impairs the daily living ability of older adults. These two factors—dependence and frailty—interact and mutually exacerbate each other, forming a vicious cycle of persistent functional decline. This bidirectional association is also reflected in the present study, particularly in the somatic-function-dominated symptom network at T2. Comparing the cross-sectional networks at T1 and T2, we found that the strength centrality of the gastrointestinal symptom cluster decreased significantly at T2 (with a negative Strength C value), indicating that the connectivity and predominant role of gastrointestinal symptoms in the overall symptom network were markedly attenuated compared with T1. In contrast, the T2 network was characterized by elevated centrality of somatic function nodes (e.g., Barthel Index, fall risk, frailty screening), which aligns with the bidirectional cycle between functional dependence and frailty reported by Sahin et al. [28]. Based on existing evidence, integrated TCM-Western medical treatment [29], TCM appropriate techniques, and lifestyle interventions can effectively relieve gastrointestinal symptoms in elderly patients. This symptom improvement is consistent with the observed temporal structural shift, characterized by a reduced influence of gastrointestinal symptoms within the network.

The structural temporal differences between the T1 and T2 symptom networks suggest that clinical intervention strategies can be optimized according to the structural features of the symptom network, shifting from a model targeting gastrointestinal symptoms alone to a comprehensive management approach centered on frailty prevention/control and somatic function maintenance.

The symptom and function network structure of elderly patients with CG evolved marked from T1 to T2, as reflected by a significant increase in global network strength (T2 = 5.95 vs. T1 = 4.75, P < 0.001). Compared with T1, the interconnections among symptoms and somatic function indicators were stronger at T2, indicating a more intricate overall symptom–function association system. This temporal evolution was driven by changes in the composition of symptom profiles between the two time points. Specifically, the T1 network was centered primarily on gastrointestinal symptom associations, whereas the T2 network additionally incorporated associations with somatic function, psychological, and social indicators, forming a complex, multi-dimensional, multi-system correlation network. Cross-sectional comparison showed that node strengths of the gastrointestinal symptom cluster (abdominal pain, diarrhea, and constipation) and the Barthel Index decreased significantly at T2 (change in strength ranged from − 0.824 to − 0.176, P < 0.01), indicating that their core driving weight in the overall symptom network was significantly diminished compared with T1. From the perspective of symptom dimensions, gastrointestinal symptoms and Barthel Index scores (reverse-scored, so lower scores indicate better function) improved significantly from T1 to T2, whereas psychosocial indicators such as anxiety, depression, and social participation showed no significant temporal changes. This differential characteristic indicates that the psychosocial status of elderly patients with CG does not improve in parallel with somatic symptom relief. Possible explanations include long-term adaptation to chronic disease, weakened social roles, inherent psychological traits in older adults, and inadequate clinical attention to psychosocial dimensions.

From the perspective of core structure evolution, psychosocial nodes (anxiety, depression, and social adjustment) ascended as the central associative hubs at T2, while gastrointestinal symptom nodes lost their previously dominant roles. Global network strength also increased significantly (P < 0.001), confirming that the association pattern among symptoms and somatic function exhibits temporal heterogeneity, jointly influenced by physiological changes, clinical interventions, and psychosocial status. Consistent with the brain-gut axis framework [30], the observed evolution suggests that somatic and psycho-emotional pathways form stable, bidirectional reinforcing associations, manifesting as distinct mind-body coupling patterns over time. Together with anxiety, depression, cognitive function, and social adjustment, these pathways form highly interconnected psychosomatic symptom clusters, such that fluctuations in a single symptom can have more widespread effects on the overall health status of patients. It is worth noting that anxiety and depression already showed high closeness centrality in the T1 network, indicating that early negative emotions play a prominent role in the gastrointestinal symptom association system and serve as key contributors to the formation of these associations. Overall, the temporal evolution demonstrates that psychosocial factors emerge as a central dimension associated with the health status of elderly CG patients, further corroborating the regulatory role of the brain-gut axis in the psychosomatic symptom network. Clinically, this implies a shift from solely controlling gastrointestinal symptoms to holistic interventions that strengthen psychological support and promote social participation, thereby improving long-term outcomes.

The findings indicate that the symptom network structures of elderly patients with CG at different time points are jointly shaped by biopsychosocial factors, rather than by gastric pathology. Specifically, the biological dimension encompasses gastrointestinal injury, somatic function, and frailty-related indicators; the psychological dimension includes anxiety and depression, linked via the brain-gut axis and somatic symptoms; and the social dimension is shaped by education level, economic income, social participation, and family support. These factors are interrelated. Therefore, integrated, multimodal interventions centered on whole-person health are recommended. Analysis of the structural changes in the two-time-point cross-sectional network revealed a gradual transition of network hubs from gastrointestinal indicators to psychological, social, and somatic function indicators. The network distribution characteristics suggest that clinical management should extend beyond the symptomatic treatment of gastrointestinal symptoms to also take into account psychological status, social participation ability, and daily self-care ability. Leveraging the differential structural features, stratified intervention strategies can be formulated. At T1, priority should be given to emotional intervention and brain-gut axis-related health management, with routine screening for anxiety, depression, and cognitive indicators. At T2, the emphasis should shift to somatic function maintenance, including sarcopenia prevention and control, nutritional support, and social participation promotion, with concurrent monitoring of the Barthel Index, fall risk, and frailty screening. This integrated management approach aims to maintain somatic function, delay the progression of frailty, embody the concept of “active aging,” and comprehensively optimize the health status of elderly patients with CG.

Several limitations should be acknowledged. First, the sample was derived from a single center, which may limit the generalizability of the findings to broader populations. Second, while the 100% follow-up rate and absence of missing data minimized attrition and information bias, the lack of individual-level treatment data precludes causal inference regarding specific intervention components. Third, the network analyses are cross-sectional at each time point, and the observed temporal changes do not imply causality. Fourth, the FRAIL scale had low internal consistency in this cohort (Cronbach’s α = 0.569), which may reduce the precision of frailty-related centrality and edge-weight estimates. Accordingly, the elevated centrality of the frailty node (MD8) at T2 should be interpreted cautiously, and its network strength should not be overemphasized clinically. Future studies using more robust frailty tools (e.g., the Fried Frailty Phenotype) are needed to confirm these findings. Additionally, the stability of betweenness centrality was poor (CS < 0.5), and closeness centrality was unstable at T1, warranting cautious interpretation. Future multicenter studies with larger samples, validated high-reliability frailty assessment tools, and longitudinal network modeling are needed to validate and extend these findings.

Conclusions

Based on cross-sectional data at admission (T1) and six months post-discharge (T2), this study investigated the evolution of symptom networks in elderly patients with CG using CGA. The symptom network exhibited temporal shifts in highly connected hubs: psychological-emotional indicators (anxiety, depression, cognition) formed the dominant cluster at admission (T1), whereas somatic-functional indicators (self-care ability, fall risk, frailty) emerged as the strongest correlational cluster at the 6-month follow-up (T2). Critically, psychosocial variables persisted as stable, cross-system hubs at both time points, without diminishing in mind-body linkage strength. Collectively, these findings reveal a persistent, psychosocial-mediated mind-body interaction pattern underlying dynamic symptom correlations in older adults with chronic gastritis. This network analysis advances the theoretical framework of psychosomatic comorbidity and supports a whole-cycle, stratified management strategy: prioritizing gastrointestinal symptom relief and psychological-cognitive screening during acute hospitalization and shifting to targeted interventions focused on physical function preservation, frailty prevention, and sustained psychosocial support during long-term follow-up.

Authors' contributions

Y.Xin and X.G.Zuo contributed to the conception and design of the study, and were major contributors in writing the manuscript.Z.Y.Zhu and W.Wei were responsible for data collection.D.J.Zhang made substantial revisions to the manuscript.All authors read and approved the final manuscript.

Funding

Henan Province Higher Education Teaching Reform Research and Practice Project (Graduate Education Category) (2023SJGLX235Y).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of Binhai County Hospital of Traditional Chinese Medicine, Jiangsu Province (Approval No. [2024]2) and complied with the Declaration of Helsinki. The purpose of the study was explained to all participants before the survey was conducted and informed consent was obtained.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.