Network analysis of patient-reported complaints in climacteric syndrome: a single-center study

Abstract

Background

This study analyzed medical records of patients with climacteric syndrome at a Korean medicine hospital and investigated the interrelationships among their symptoms using network analysis.

Methods

A retrospective chart review was conducted on 81 women (aged 45–60 years) who visited a women’s health clinic between 1 January 2020 and 31 December 2021. Network analysis was performed on unstructured text from first-visit medical records to identify symptom patterns and compute centrality indices.

Results

The most common symptoms were ‘pain’, ‘hot flush’, ‘facial blushing’, ‘fatigue’, and ‘sleep disorder’. Among them, pain and hot flush showed high node strength, betweenness centrality, and closeness centrality, which indicated their central role in symptom co-occurrence. Although there were no statistically significant differences in the frequency of symptoms between the premenopausal and postmenopausal groups, variations in symptom centrality and modular analysis results before and after menopause were noted, emphasizing the importance of detailed symptom analysis in clinical settings.

Conclusions

By analyzing medical records of patients with climacteric syndrome, this study clarified symptom patterns in menopausal women, providing data for targeted and stage-specific treatments.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00404-026-08386-3.

What does this study add to the clinical work?

Pain and hot flush were the most central symptoms in the climacteric symptom network, with distinct symptom clusters observed across menopausal stages; these findings emphasize the need for stage-specific assessment and management strategies in clinical practice

Background

The menopausal transition, which occurs during midlife in women, is characterized by a high prevalence of symptomatic experiences, as evidenced by research on middle-aged women [1]. Previous studies on the prevalence and severity of symptoms in premenopausal and postmenopausal stages have primarily relied on surveys and interviews with healthy middle-aged women [2, 3] However, these methods are limited, as there can be a gap between the symptoms observed in healthy middle-aged women and those in women with climacteric syndrome. To address this gap and enhance clinical understanding, this study focuses specifically on patients who actively seek medical attention, examining their reported symptoms and the interrelationships therein. This approach is based on an analysis of medical records from the first visit of the patients who presented at a hospital for the treatment of menopausal symptoms during the transitional period.

While extant literature [4–6] has explored the typology, severity, and prevalence of climacteric symptoms, as well as patients’ quality of life, these studies predominantly highlight correlations between symptoms and demographic factors such as time elapsed since menopause, hormone therapy utilization, and age. Although a few prior studies have investigated the network structure of menopausal symptoms [7, 8], to our knowledge, no study has applied network analysis to examine the interrelationships among symptoms reported by patients seeking clinical treatment.

To explain the nature of complaints expressed by patients presenting with climacteric symptoms in Korean clinical settings, we conducted an analysis of medical records from women who sought treatment for climacteric syndrome at a women’s health clinic within a Korean Medicine hospital. Our methodology incorporated text mining and network analysis techniques to examine unstructured text from first-visit medical records. This investigation into the distribution and typology of primary complaints and clinical symptoms associated with climacteric syndrome is intended to provide foundational data for the enhancement of medical care in this domain.

Methods

Study design and setting

This retrospective chart review was conducted to identify patterns of clinical symptoms from the medical records of patients with climacteric syndrome. This study was approved by the Institutional Review Board (IRB) of Dongguk University Ilsan Korean Medicine Hospital (approval number: DUIOH 2022-03-003). The IRB waived the requirement for obtaining informed consent from the patients owing to the retrospective nature of this study. This study also followed the guidelines of the Declaration of Helsinki and Tokyo for humans. Throughout the study process, the personal information of subjects was treated as strictly confidential and used only by the researchers.

This study examined all the 45- to 60-year-old female patients who were diagnosed with the climacteric syndrome at the department of obstetrics and gynecology of Dongguk University Ilsan Korean Medicine Hospital between 1 January 2020 and 31 December 2021. Only those diseases that were limited to the following International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes were considered: N95.1 for menopausal and female climacteric states; N95.2 for postmenopausal atrophic vaginitis; N95.8 for other specified menopausal and perimenopausal disorders; and N95.9 for menopausal and perimenopausal disorders, unspecified [9]. Based on their menopausal status, we divided the patients into two groups to compare their premenopausal and postmenopausal symptom patterns. Menopausal state was defined as the 12 consecutive months without menstruation and surgical menopause after bilateral oophorectomy according to World Health Organization criteria [10].

Data collection and preprocessing

At the time of their initial visit, patients’ demographic information, ICD-10 codes, and symptoms were extracted from their medical records. Symptom extraction and standardization were performed by clinician-researchers with clinical experience in Korean Medicine, who routinely manage patients with climacteric symptoms. Since the symptoms were written in various narrative forms, they were standardized to calculate the frequency. Some recorded symptoms included specific details, such as subtypes or the sites of occurrence, in parentheses. Two symptom-counting strategies were parallelly implemented, as follows: the first strategy, termed specific symptom counting (SSC), differentiated symptoms based on their specifics. For instance, if a patient had ‘pain’, ‘pain (pelvis)’, and ‘pain (shoulders)’, each aspect was counted once as a separate symptom. The second strategy, general symptom counting (GSC), grouped all instances of a symptom together regardless of their specifics. Accordingly, ‘pain’, ‘pain (pelvis)’, and ‘pain (shoulders)’ were integrated and counted as a single instance of ‘pain’. After symptom standardization (SSC and GSC), we constructed patient-by-symptom binary matrices (presence/absence) for subsequent analyses.

Network analysis and visualizations

To gain an intuitive understanding of symptom interrelationships, networks were visualized using symptom co-occurrence matrices, in which each element represented the frequency of simultaneous occurrence of two symptoms in a patient. In these networks, nodes represented symptoms, and an edge was formed between two interconnected nodes, with their co-occurrence frequency being its weight.

Subsequently, two network analyses were conducted. First, an analysis of the full dataset was performed using both SSC and GSC approaches. To ensure validity and exclude random co-occurrences, a threshold of >3 co-occurrences was applied. Second, comparative networks for the premenopausal and postmenopausal groups were established using the GSC approach. For these subgroup analyses, the threshold was adjusted to >2 to account for the reduced sample sizes.

Three centrality indices (Node strength, Betweenness centrality, Closeness Centrality) were computed in each network. Node strength is the sum of the weights of all the edges connected to an interested node, indicating how often the symptom co-occurs overall [11]. Betweenness centrality measures the fraction of all the shortest paths in a network passing through a specific node [11]. A symptom’s high betweenness centrality can be considered a significant mediation between two other symptoms. Closeness is the inverse of the average length of the shortest paths to all other nodes; it indicates the number of distinct symptoms to which a symptom is connected [11]. These metrics collectively assess each symptom’s importance within the network.

We evaluated the robustness of the networks using bootstrap procedures [12]. First, we conducted a nonparametric bootstrap (1000 iterations; resampling patients with replacement) to assess edge-weight accuracy. For each bootstrap sample, we recomputed co-occurrence weights and derived 95% confidence intervals for each edge. In addition, we calculated the selection probability for each edge, defined as the proportion of bootstrap samples in which the edge met the original cooccurrence threshold. Second, we performed a case-dropping bootstrap to examine the stability of centrality indices. For a range of case-dropping proportions (10%–70%), centrality indices were recalculated and compared with those from the full sample using Spearman’s rank correlation. We summarized stability using the correlation stability (CS) coefficient, defined as the maximum proportion of cases that can be dropped while retaining a high correlation (Spearman’s ρ > 0.7) with the original centrality rankings. Because cooccurrence counts scale with sample size, we proportionally adjusted the co-occurrence threshold in the case-dropping procedure to reflect the retained sample fraction. A CS coefficient of at least 0.25 is recommended, and values above 0.50 indicate high stability [12].

In addition to bootstrap-based robustness assessment, we conducted a threshold-perturbation sensitivity analysis to evaluate the stability of centrality results under alternative co-occurrence cutoffs. For each network, we reconstructed comparative networks using thresholds within ±2 of the original cutoff, excluding the original threshold. For each pair of networks, we restricted comparisons to the intersection of active symptoms (node strength > 0 in both networks) to avoid artifacts driven by isolated nodes. Centrality indices (strength, betweenness, and closeness) were then compared between the original and perturbed networks using Spearman’s rank correlation [12]. This analysis confirmed that the centrality rankings were robust to moderate (±1) variations in the co-occurrence threshold (see Online Resource 3).

To examine the symptoms’ modularity in the network, we used the Louvain method for community detection [13]. This algorithm is an efficient community detection method to maximize modularity, which changes depending on the community to which each node belongs. It divides the symptoms of climacteric syndrome into groups within each network and facilitates the observation of patterns and hierarchies in symptom manifestation. In addition, during the premenopausal and postmenopausal symptom network analysis, symptoms were categorized into clinical classifications based on ICD-10, the Kupperman menopausal index, and the Menopausal Rating Scale (MRS) [14–16], which were compared with the communities identified by the Louvain algorithm. This classification process required the researchers’ input since some symptoms could not be easily classified into a single category. All network visualizations were created using Cytoscape 3.9.1 software (https://cytoscape.org/) using the prefuse force directed layout [17].

Statistical analysis

This study is mainly descriptive and results are primarily presented as counts, means, standard deviations, and percentages. To evaluate the statistical significance of variations in symptom occurrence between premenopausal and postmenopausal patients, two-proportion z-tests and Fisher’s exact tests were conducted. Further, p-values were adjusted using the Benjamini–Hochberg method to account for multiple comparisons [18]. All statistical analyses were conducted using Statsmodels in Python 3.9 [19].

Results

Examination of patient records

This study considered 81 first-visit patients’ medical records. The patients’ mean age was 52.3±3.8 years. Among the patients considered, 49 (60.5%) were premenopausal and 32 (39.5%) were postmenopausal. Table 1 summarizes the study population’s characteristics.

Table 1.

Characteristics of the study population

	N = 81 n (%)/Mean ± SD
Age (years)	52.3 ± 3.8
Body height (cm)	158.2 ± 6.0
Body weight (kg)	57.0 ± 8.1
BMI (kg/m2)	22.8 ± 3.3
Menopausal status
Premenopausal	49 (60.5)
Postmenopausal	32 (39.5)
Diagnosis (ICD-10 codes)
Menopausal and female climacteric states (N95.1)	58 (71.6)
Postmenopausal atrophic vaginitis (N95.2)	1 (1.2)
Other specified menopausal and perimenopausal disorders (N95.8)	6 (7.4)
Menopausal and perimenopausal disorder, unspecified (N95.9)	15 (18.5)

There were three and four missing values for body height and weight, respectively. Mean imputation was used for this purpose

BMI, body mass index, ICD-10 International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, SD standard deviation

According to the clinical diagnosis based on KCD codes, 58 patients, that is, the majority of the population (>70%) were diagnosed with N95.1. N95.8 and N95.9 were the second and third common diagnoses, with 15 and 6 patients, respectively. N31.8 and N95.2 were coded for one patient each, which were adopted for patients with atrophic vaginitis and an overactive bladder, respectively.

Symptomatic profiles and comparisons

In this study, symptoms were extracted from the medical records of 81 patients using the SSC and GSC approaches. Using the SSC approach, 186 distinct symptoms were observed, each occurring in 1 to 36 patients. In contrast, the GSC approach identified 126 symptoms, with occurrences ranging from 1 to 45 patients.

To identify the predominant symptoms of the climacteric syndrome, symptoms were arranged based on the number of patients experiencing them, and the top 10% were classified as predominant (Fig. 1). The SSC approach indicated that hot flush, facial blushing, fatigue, pain (shoulders), and irregular menstruation were the most frequent symptoms (Fig. 1A). For GSC, the most common symptoms were pain, hot flush, sleep disorder, facial blushing, and fatigue (Fig. 1B). Notably, hot flush, which 36 (44.4%) patients experienced, was ranked as either the primary or secondary most prevalent symptom for both SSC and GSC. Further, dyspepsia, facial blushing, fatigue, headache, irregular menstruation, pain, palpitation, sleep disorder, sweating, and dryness were consistently identified as the predominant symptoms for both SSC and GSC approaches, falling within the top 10% of symptom frequency in both cases. This highlights the convergence of certain predominant symptoms, demonstrating their consistent presence across both approaches.

Fig. 1.

Predominant symptoms in patients with climacteric syndrome. Predominant symptoms were identified as the top 10% of symptoms by frequency using A the specific symptom counting (SSC) approach and B the general symptom counting (GSC) approach. The dotted line indicates the average number of occurrences of all symptoms

It is important to clarify the trends associated with specific details, rather than simply calculating general frequency, to gain deeper insights into patients’ experiences. Therefore, the predominant symptoms identified by both SSC and GSC methods were selected, and their frequencies in relation to specific details were examined (Table 2). The symptoms of pain, sleep disorder, and dryness met the criteria for examination, with their types of specifics being numbered 20, 4, and 6, respectively. For pain, the prevalent specifics (i.e., the sites of pain) were shoulders, generalized, and low back. For sleep disorder, the specifics were subtypes. In patients with climacteric syndrome, fragmented sleep was the primary sleep disorder subtype. This was followed by sleep onset insomnia and shallow sleep. For dryness, the common specifics observed in patients were vagina, skin, and throat. These findings offer a detailed perspective on symptom manifestations, which would be associated with the syndrome’s pathological mechanisms.

Table 2.

Symptom frequencies based on specifics in patients with the climacteric syndrome

Symptom	Total count n (%)
Pain	85
Specified	85
Shoulders	22 (25.9)
Generalized	12 (14.1)
Low back	11 (12.9)
Knee joints	8 (9.4)
Pelvis	7 (8.2)
Neck	4 (4.7)
Back	3 (3.5)
Feet	3 (3.5)
Wrists	3 (3.5)
Othersª	12 (14.1)
Unspecified	0
Sleep disorder	38
Specified	25
Fragmented sleep	12 (48.0)
Sleep onset insomnia	6 (24.0)
Shallow sleep	4 (16.0)
Sleep maintenance insomnia	3 (12.0)
Othersª	0 (0)
Unspecified	13
Dryness	32
Specified	28
Vagina	10 (35.7)
Skin	6 (21.4)
Throat	6 (21.4)
Eyes	4 (14.3)
Othersª	2 (7.2)
Unspecified	4

Percentages were calculated for specified symptoms alone

ªOthers includes the specifics observed less than three times.

To explore differences in symptoms across menopausal stages, we compared the predominant symptoms of premenopausal and postmenopausal patients, which were identified using the GSC approach (Fig. 2). Whereas a total of 93 symptoms were observed in premenopausal patients with each symptom occurring in 1 to 24 patients, 89 distinct symptoms were observed in postmenopausal patients, with each symptom affecting 1 to 21 individuals. Pain, hot flush, facial blushing, fatigue, sleep disorder, dryness, and sweating were included in the top 10% of symptoms by frequency, and identified as predominant symptoms in both the premenopausal and postmenopausal groups. On the other hand, irregular menstruation and headache were listed only in the predominant symptoms of the premenopausal group, while cold sensation and depression were in those of the postmenopausal group alone. None of the symptoms showed statistical significance for the difference in frequency between the premenopausal and postmenopausal groups.

Fig. 2.

Comparison of the top 10% most frequent symptoms between premenopausal and postmenopausal groups. The bar plot compares the occurrence frequency of the symptoms identified as the top 10% most frequent in both premenopausal (gray bars) and postmenopausal (black bars) groups

Network analysis

Examining symptom networks helps determine the systemic patterns in symptom occurrence, which may not be apparent at the individual symptom level. This perspective led to the establishment of symptom co-occurrence networks for all patients with climacteric syndrome, which included 26 symptoms and 69 edges in the SSC approach (Fig. 3A) and 22 symptoms and 78 edges in the GSC approach (Fig. 3B).

Fig. 3.

Symptom networks of patients with the climacteric syndrome. Two symptom networks were built, one for each approach: A SSC and B GSC. The color of the nodes indicates the module to which each node belongs. Node sizes reflect their degrees, while edge thicknesses represent their weights

The centrality indices for each node in the two networks are depicted in Fig. 4. In the symptom network using the SSC approach, hot flush had the highest node strength (121), followed by facial blushing (71) and pain (shoulders) (70). Likewise, hot flush also showed the highest closeness centrality (0.833), with pain (shoulders) (0.676) and facial blushing (0.641) following. As for betweenness centrality, hot flush (0.490), headache (0.163), and pain (shoulders) (0.119) exhibited the highest values. In the symptom network through GSC, the highest node strength and closeness centrality were shown by pain (176 and 0.913, respectively), hot flush (121, 0.778), and sleep disorder (97, 0.75). As for betweenness centrality, pain (0.283), hot flush (0.175), and facial blushing (0.120) had the highest values. In both SSC and GSC networks, hot flush and pain (including shoulder pain) consistently showed the highest values in all indices, indicating that they are central to the co-occurrence of symptoms in climacteric syndrome.

Fig. 4.

Centrality indices in the symptom networks of patients with the climacteric syndrome. The symptoms included in each network are represented using distinct lines

Case-dropping bootstrap analyses indicated that the centrality estimates were stable in both networks, with all three indices meeting the stability criterion in the SSC network (CS coefficient = 0.50 for strength, closeness, and betweenness) and in the GSC network (CS coefficient = 0.70 for strength and closeness and 0.50 for betweenness) (Online Resource 1) [12].

Subsequently, to compare and clarify the differences in symptom patterns between premenopausal and postmenopausal women, symptom co-occurrence networks from premenopausal and postmenopausal groups were constructed (Fig. 5). The premenopausal network included 19 symptoms with 49 edges, whereas the postmenopausal network had 16 symptoms with 53 edges. The values of the centrality indices for the network nodes are depicted in Fig. 6. In the premenopausal symptom network, the symptoms with the highest node strength, betweenness centrality, and closeness centrality were pain (81, 0.408, 0.857, respectively), hot flush (57, 0.240, 0.75), and facial blushing (40, 0.130, 0.667). In contrast, in the postmenopausal network, pain (79), hot flush (57), and sweating (47) had the highest node strength. Betweenness centrality was dominated by pain (0.320), dryness (0.152), and sleep disorder (0.066), and closeness centrality by pain (0.882), hot flush (0.789), and sweating (0.789). Pain and hot flush emerged as central in both networks, reaffirming their role in symptom co-occurrence across menopausal stages. Edge-level bootstrap results further supported this core structure, with symptom pairs involving pain and/or hot flush consistently appearing among the most stable edges in both the premenopausal and postmenopausal networks, as also observed in the SSC and GSC networks (Online Resource 2). At the same time, distinct symptoms showed prominence depending on the stage. Facial blushing was more central in the premenopausal network, whereas sweating, sleep disorder, and dryness played a greater role in the postmenopausal group. These findings suggest a shift and the dynamic nature of symptom co-occurrence across menopause.

Fig. 5.

Symptom networks for the premenopausal and postmenopausal groups. A, C depict the symptom networks for the premenopausal group, while B, D represent the symptom networks for the postmenopausal group. In all networks, symptoms are connected if they co-occur with other symptoms more than twice. Nodes with the same color indicate the same module in A, B or the same clinical classification in C, D

Fig. 6.

Centrality indices in the premenopausal and postmenopausal symptom networks. The symptoms included in each network are represented using distinct lines

Consistent with these findings, case-dropping bootstrap analyses showed that all three centrality indices met the stability criterion in both menopausal-stage networks (CS coefficient = 0.70 for strength and closeness and 0.50 for betweenness in both groups) (Online Resource 1) [12].

To explore the modular structures among symptoms in the symptom co-occurrence network, we conducted community detection using the Louvain method across the SSC and GSC approaches, as well as the premenopausal and postmenopausal networks (Fig. 5A, B). Some identical symptom communities were consistently found in both the SSC and GSC networks, including [hot flush, facial blushing, cold sensation, dyspepsia, urinary frequency], [amenorrhea, fatigue, irregular menstruation], and [headache, dizziness, palpitation]. For the analysis using the premenopausal and postmenopausal networks, pain and numbness were consistently found to belong to the same community across menopausal stages, as was the case in the GSC network. Further, [amenorrhea, fatigue, irregular menstruation], [hot flush, facial blushing, urinary frequency], and [headache, dizziness, palpitation] were consistently grouped into distinct communities, remaining separate from one another in the premenopausal network, as well as in the SSC and GSC networks. In contrast, the postmenopausal symptom network did not show any clear modularity. When the symptoms were categorized according to the clinical classification depicted in Table 3 (Fig. 5C, D), some symptom groups showed similarities to the communities identified by the Louvain method. Interestingly, symptoms in the pain & numbness-related group, such as pain, arthralgia, and numbness, belonged to a distinct community in both the premenopausal and postmenopausal symptom networks, which was particularly separate from the symptoms in the dermatological group such as dryness and pruritus. Also, urologic symptoms, including urinary frequency and nocturia, were consistently separate from the neurologic & psychiatric group symptoms. These findings not only indicate the presence of consistent symptom co-occurrence patterns represented as communities or clinical classifications in the symptom networks but also highlight the influence of pathophysiological and clinical backgrounds to these patterns.

Table 3.

Clinical classification of the symptoms in the premenopausal and postmenopausal networks

Dermatological	Gastrointestinal	Menstrual	Neurological and psychiatric	Pain and numbness-Related	Urologic	Vasomotor
Dryness Pruritus	Dyspepsia	Amenorrhea Irregular menstruation	Depression Dizziness Fatigue Headache Sleep disorder	Arthralgia Numbness Pain	Nocturia Urinary frequency	Facial blushing Chest tightness Cold sensation Cold sweats Heat sensation Hot flush Palpitation Sweating

Discussion

In women, midlife years, 45–60 years, coincide with the period of menopausal transition, and studies on middle-aged women suggest that they are highly symptomatic. Significant research efforts have been expended in identifying the symptoms related to the hormonal changes underlying menopause and those that are caused by aging, some other health states, psychosocial factors, and lifestyle. Some important symptoms are vasomotor symptoms (hot flush and perspiration) and atrophic vaginitis (dyspareunia and blood-stained vaginal discharge).

Despite this extensive research, the pattern of climacteric syndrome remains inconsistent across individuals. The frequency of symptoms varies from one individual to another and is slightly influenced by the presence of diseases; further, symptoms are preferentially associated with the syndrome [20]. Given the complexity of these symptoms and their interrelatedness, we employed network analysis to understand the relationships between them.

The network analysis of the symptoms of a disease or syndrome is performed for a diverse range of diseases, from psychiatric disorders, such as depressive disorder, schizophrenia, and posttraumatic stress disorder, to broader diseases, such as cancer, multiple sclerosis, and the recent coronavirus disease 2019 pandemic [21–26]. The network analysis approach offers insights into the potential etiological properties of, and the patterns and hierarchies among symptoms [27, 28]. The findings of the study can not only be utilized in patient treatment strategies but also provide a foundation for further investigative research [29]. Therefore, this study’s findings offer meaningful clinical insights and could be considered in the treatment and management of patients with climacteric syndrome.

Pain, hot flush, sleep disorder, and fatigue were the most common symptoms in both premenopausal and postmenopausal women. Conversely, irregular menstruation and headache were prevalent only in the premenopausal group, while cold sensation and depression were specific to the postmenopausal group. Irregular menstruation is a natural physiological symptom occurring before menopause, and previous studies have noted an increase in tension-type headaches during the perimenopausal stage in middle-aged women [30].

According to the centrality analysis, symptoms’ strength centrality values vary in premenopausal and postmenopausal patients. For instance, pain and hot flush have the highest node strength for both premenopausal and postmenopausal patients, indicating that these symptoms occur frequently or are intense and strongly linked to other symptoms across menopausal stages. However, facial blushing has a high node strength for premenopausal patients, whereas sweating, dryness, and sleep disorder display high strength values for postmenopausal women. Similarly, betweenness centrality quantifies how often a symptom acts as a bridge within the shortest paths between other symptom pairs in a network. The symptoms with high betweenness centrality are important because they mediate the flow of symptom manifestation or information. For example, pain and hot flush consistently appear as central intermediaries in both premenopausal and postmenopausal phases, suggesting their significant role in connecting various other symptoms. Further, closeness centrality reflects how ‘close’ a symptom is to all other symptoms in the network and indicates its ability to quickly interact with or affect other symptoms. In Fig. 6, symptoms such as sweating and sleep disorder indicate an increase in closeness centrality in postmenopausal patients. Accordingly, these symptoms play a central role in symptom interactions, overall symptom manifestations, and symptom management in the postmenopausal phase.

Modular analysis using the Louvain method showed that symptom patterns were clear in the premenopausal group but less defined in the postmenopausal group. However, when symptoms were grouped according to clinical classification, they were categorized in a way that aligned with the modular analysis results for both premenopausal and postmenopausal stages. In the symptom networks for the premenopausal and postmenopausal groups, vasomotor symptoms and physical discomfort were more prominent before menopause, whereas chronic symptoms and psychological changes became more pronounced after menopause. A comprehensive understanding of these changes will help healthcare providers develop treatment and management strategies appropriate to each stage. For instance, approaches to alleviate vasomotor symptoms are essential before menopause, whereas managing chronic symptoms and providing psychological support are crucial considerations after menopause.

According to the modular analysis result, symptoms were categorized as [amenorrhea, fatigue, irregular menstruation], [hot flush, facial blushing, urinary frequency], and [headache, dizziness, palpitation] in the premenopausal network, and [pain, hot flush, urinary frequency], [sweating, cold sensation, depression, facial blushing], and [sleep disorder, fatigue, dryness] in the postmenopausal network. These differences likely reflect hormonal and physiological response changes due to menopause and aging. When comparing the two networks, the premenopausal network revealed a cluster of vasomotor symptoms, including [headache, dizziness, palpitation]. Over time, as aging progresses, accompanied by sustained low estrogen levels and a decline in metabolic rate, symptom clusters such as [sleep disorder, dryness, fatigue] become more prominent in the postmenopausal stage [31].

Furthermore, based on modular analysis, the symptoms that coalesce and manifest distinctly were provisionally categorized. This categorization facilitates the investigation and identification of additional symptoms that, although not explicitly articulated by the patient, may be latent. For instance, sleep disturbances subsequent to menopause can exacerbate climacteric symptoms; thus, when a patient presents with symptoms of climacteric syndrome, these can be concurrently assessed and managed effectively. This result aligns with findings from previous studies, which reported that the majority of patients with senile dryness who also complained of sleep disorders [32] and reported the relationship between sleep quality, sweating, hot flash, anxiety and depression in peri/postmenopausal women [33].

In contrast to previous studies [7, 8], the symptom network derived from patients who sought care for climacteric complaints in this study exhibited a distinct pattern of central symptoms. Notably, pain emerged as a highly central symptom, along with hot flushes, fatigue, and sleep disorders. This finding differs from prior network analyses of menopausal symptoms, which employed structured questionnaires such as the Menopause-specific quality of life symptoms (MENQOL), the Menopause Rating Scale (MRS), and Patient Health Questionnaire (PHQ) [31, 32]. In those studies, vasomotor and psychological symptoms were typically identified as central, whereas pain was less influential. This discrepancy may be attributed to differences in data sources. While previous studies relied on structured, self-reported questionnaires, the present study utilized unstructured clinical data extracted from electronic medical records, focusing on patients’ chief complaints at the time of their initial clinical consultation. Accordingly, the symptoms identified in this analysis are more likely to reflect those perceived as particularly distressing and disruptive, rather than merely present. This distinction emphasizes the clinical relevance of our findings, as it highlights symptoms that directly motivate individuals to seek medical attention.

By prioritizing spontaneously reported complaints rather than predefined symptom checklists, the present study provides a complementary perspective to the existing literature and offers practical insights into patient-centered care for climacteric syndrome. Identifying symptoms that patients themselves consider most burdensome may assist clinicians in developing more targeted and responsive therapeutic strategies.

From a clinical perspective, the identification of pain and hot flush as central nodes within the symptom network has important implications for the assessment and management of climacteric syndrome. Given their high connectivity, these symptoms may serve as clinically meaningful indicators of overall symptom burden. Prioritizing the assessment of pain and hot flush during initial intake evaluations and follow-up visits may facilitate early identification of patients at risk of broader symptom aggravation. Furthermore, incorporating these central symptoms into structured symptom diaries or digital monitoring tools may enable more efficient tracking of symptom co-occurrence and longitudinal changes in routine clinical practice.

The modular structure identified in the postmenopausal symptom network further provides a useful framework for holistic clinical intervention. In particular, the clustering of sleep disorder, dryness, and fatigue suggests that these symptoms frequently co-occur and may share underlying mechanisms or reinforcing pathways. Rather than addressing each symptom in isolation, clinicians may consider bundled, nonpharmacological management strategies targeting multiple symptoms simultaneously. Such approaches may include integrated education on sleep hygiene, lifestyle modification, fatigue management, and symptom-focused self-care guidance, tailored to the postmenopausal stage.

Overall, these findings support the clinical utility of symptom network analysis as a decision-support tool for stage-specific and patient-centered care. By identifying central symptoms and clinically relevant symptom clusters, this approach may assist clinicians in prioritizing assessments, structuring symptom monitoring, and designing more comprehensive and efficient management strategies for women experiencing climacteric symptoms. Another significant aspect of our study is that data were extracted based on the primary symptoms reported by patients, facilitating the collection of detailed information. For instance, hot flush is a sensation of warmth that is often accompanied by facial flushing and sweating [34] and can occur with or without facial flushing. Therefore, hot flush, sweating, and facial flushing were all considered distinct symptoms in our data analysis. Additionally, since our study focused on the patients who visited the hospital specifically for menopausal symptom treatment, our findings have higher applicability in clinical settings compared to earlier research.

This study has several limitations. First, the relatively small sample size and limited data made it challenging to analyze the correlations among age, duration since menopause, and hormone therapy use. Second, because the study was based on medical records from a single Korean Medicine hospital, the sample consisted predominantly of Korean women receiving care within a specific clinical context. Consequently, the identified symptom network structure may not be readily generalizable to broader populations, including patients treated in general hospitals, primary care or gynecology clinics, community-based settings, or populations from different cultural or ethnic backgrounds, given well-documented cross-ethnic differences in menopausal symptom profiles. In addition, the sociodemographic characteristics and healthcare-seeking patterns specific to this setting may have contributed to the network structure identified in the current study. Potential omissions may have occurred while rephrasing and categorizing the symptoms extracted from medical records. These factors may have contributed to the lack of statistical significance in the difference in symptom frequency between premenopausal and postmenopausal groups. These limitations underscore the need for cautious interpretation of the findings and highlight the need for future studies using larger, multicenter datasets across diverse clinical settings and countries to enhance external validity. Nevertheless, this study offers valuable insights into the symptom patterns and interrelationships of climacteric syndrome using network analysis, revealing central symptoms and modular patterns across premenopausal and postmenopausal stages. These findings have the potential to inform targeted, stage-specific treatment strategies and provide a foundation for further research, ultimately contributing to improved clinical care and management of menopausal women.

Conclusion

By analyzing medical records of patients with climacteric syndrome, this study comprehensively clarifies the symptom patterns in menopausal women, provides valuable data for developing targeted and stage-specific treatments, and offers guidance for future research on managing climacteric syndrome.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ICD

International Statistical Classification of Diseases and Related Health Problems

SSC

Specific symptom counting

GSC

General symptom counting

Author contributions

S-J Choi: methodology, investigation, writing—review and editing; B Kang: data curation, formal analysis, investigation, writing—original draft; C-E Kim: investigation, formal analysis, supervision; D-I Kim: conceptualization, investigation, supervision.

Funding

This research was supported by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number: RS-2025-02307524).

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare that there are no competing interests.

Ethical approval and consent to participate

This study followed the guidelines of the Declaration of Helsinki and Tokyo for humans. This study was approved by the Institutional Review Board (IRB) of Dongguk University Ilsan Korean Medicine Hospital (approval number: DUIOH 2022–03-003). The IRB waived the requirement for obtaining informed consent from the patients owing to the retrospective nature of this study. Throughout the entire study process, the personal information of the subjects was treated as strictly confidential except for the researchers.

Consent for publication

Not applicable.