Introduction

Social isolation, defined as the absence of or limited access to social resources and relationships, has been extensively linked to negative health outcomes across psychological and physical domains [1–3]. Social isolation elevates the risk of mental health issues, such as depression, anxiety, and suicidal ideation [4], as well as physiological conditions, including cardiovascular disease and immune dysfunction [5]. Moreover, the growing reliance on digital interaction in contemporary society may have altered the nature of physical and social networks, potentially leading to emerging patterns ofsocial isolation [6,7]. Accordingly, there is an urgent need for tools capable of comprehensively evaluating social isolation and its associated networks to guide targeted interventions.

Social networks encapsulate individuals’ connections and the nature of these interactions, encompassing both quantitative and qualitative dimensions [8,9]. Robust social networks are consistently associated with improved mental and physical health outcomes [10–12]. Despite the well-established importance of social networks conceptual clarity on the distinctions and interrelations between social isolation and social networks remains insufficient in many studies [13,14]. While social isolation generally refers to an objective lack of social ties, social networks involve both structural aspects such as network size and density and functional aspects, including emotional support and perceived quality of relationships [13,15,16]. Addressing the multidimensionality of social connectedness, including emotional, structural, and functional layers, is thus essential [17]. To align with theoretical frameworks such as Berkman & Glass (2000), Cacioppo & Hawkley (2009), and the UCLA Loneliness Model, the SISN subscales are conceptually mapped as follows. The emotional layer captures perceived closeness and affective support, the structural layer measures network size, frequency of contact, and embeddedness, and the functional layer assesses the perceived availability of support and quality of interactions. The SISN incorporates two primary domains: Objective Isolation (7 items) and Social Network (12 items), designed to comprehensively represent these layers. This mapping ensures conceptual consistency and clarifies the measurement logic of each SISN subdomain [2,3].

In addition to these conceptual challenges existing tools such as the Lubben Social Network Scale (LSNS) and other brief instruments have been widely used [18]. However, they often emphasize quantitative metrics like the number of contacts or frequency of interactions while neglecting qualitative and culturally sensitive dimensions [19,20]. Furthermore, these scales may not sufficiently capture context-specific nuances, limiting their utility in diverse cultural settings such as Korea [21]. Therefore, there is a critical need to develop new tools that provide comprehensive, culturally adapted assessments that address these gaps [22].

The newly developed Social Isolation and Social Network (SISN) Scale aims to bridge these gaps by integrating both quantitative and qualitative aspects of social connectivity, emphasizing cultural relevance and conceptual depth. Specifically, the SISN considers emotional closeness, perceived support, and structural embeddedness to comprehensively profile social connectedness [23].

Tools assessing social isolation often rely on self-reported data or frequency-based interaction measures [24]. Although these methods are valuable, they lack sensitivity to the broader dynamic aspects of social connectivity, leading to potential limitations in reliability and construct validity [25]. Existing tools frequently fail to address critical diversity factors such as cultural or age-specific variations [24]. Developing a new tool that bridges these gaps and provides a multidimensional, culturally sensitive measurement of social isolation and network structure is fundamental for overcoming these challenges.

Reliability and validity are the two pillars of psychometric evaluation and form the foundation of standardized measurement tools [26]. Reliability, evaluated through internal consistency and test-retest methods, ensures measurement stability and internal coherenceof the tool over time [27]. Specifically, internal consistency measures whether individual items within the tool align cohesively to assess the intended constructs [28]. Conversely, validity gauges the extent to which a tool successfully measures the target theoretical framework. This study emphasized both reliability and validity by employing established psychometric methods to support clinical applicability and cross-cultural generalizability [29].

Receiver operating characteristic (ROC) curve analysis is a widely recognized statistical method for evaluating the diagnostic accuracy of measurement tools [30]. This technique assesses the balance between sensitivity and specificity, making it particularly useful for validating psychological and social assessments [31]. By implementing ROC curve analysis, the present study explored whether the newly designed tool could provide complementary value to traditional instruments such as the LSNS in detecting and distinguishing between varying levels of social isolation and network patterns [32]. This approach enhances understanding of the tool’s strengths and limitations in various cultural contexts, thus informing future adaptation and validation efforts.

The study of social isolation and connectedness requires a multidisciplinary perspective encompassing psychology, sociology, public health, and data science [33,34]. Effective tool development not only requires rigorous psychometric evaluation but also ensures relevance to real-world applications that can inform public health initiatives [35]. The present study aligns with the ongoing demand for holistic methodologies that address complex social phenomena using a coordinated cross-disciplinary framework [36].

This study sought to establish the reliability and validity of a newly developed social isolation and network profiling tool. Specifically, it evaluated psychometric reliability through internal consistency and test-retest analysis while validating diagnostic efficiency through ROC curve analysis. The ultimate goal was to provide a standardized, nuanced instrument for assessing social isolation and connectivity that has both clinical relevance and public health implications. This study addresses a critical gap in the existing toolkit for social evaluation and lays the groundwork for future interdisciplinary approaches to alleviate the adverse effects of social disconnection.

Methods

Instruments

Social Isolation and Social Network Scale (SISN).

The SISN scale was developed through a rigorous two-phase process involving systematic literature review and expert consensus validation using a modified Delphi technique [23]

Phase 1: Literature-based Item Development: The initial item pool was generated through a comprehensive literature review based on established theoretical frameworks. The scale development was grounded in Berkman and Syme’s (1979) [37] social network theory, Cornwell and Waite’s (2009) [5] conceptual distinction between objective and subjective isolation, and Cacioppo and Hawkley’s (2009) [38] framework for perceived social isolation. This systematic process yielded 35 items organized into three conceptual domains: objective social isolation (7 items), subjective social isolation (10 items), and social network (15 items).

Phase 2: Expert Consensus Validation: A modified Delphi survey was conducted with 23 multidisciplinary experts (70% occupational therapists, 17% physical therapists, 9% nurses, 4% social workers) with over 5 years of clinical experience in Korea. Through two rounds of evaluation using Content Validity Ratio (CVR) analysis with a threshold of ≥0.37, three items were removed due to insufficient consensus and one new item regarding social network service (SNS) usage was added based on expert recommendations. All final items achieved CVR ≥ 0.37, with the overall CVR reaching 0.87, demonstrating high consensus (0.31) and convergence (0.87) [23].

Final Scale Structure: The final SISN consists of 19 items across two primary dimensions: Objective Isolation (7 items) and Social Network (12 items). The Objective Isolation domain includes items on frequency of leaving home, daily contact, SNS usage, and perceived disconnection. The Social Network domain measures family and friend network size, emotional closeness, availability of support, and quality of interactions. All items are rated on a five-point Likert scale ranging from 0 (“not at all”) to 4 (“very much”), with higher scores reflecting stronger social connectedness and more robust network integration. No reverse coding was required, and domain scores were calculated as item means. Thus, the observed subscale means (e.g., 0.791–1.746 for the Social Network domain) fall within the expected 0–4 scale range. (S1 File).

Lubben Social Network Scale – Korean version (LSNS-K).

The LSNS-K was used to examine the concurrent validity of the SISN. The LSNS was originally developed as a general measure of social ties to capture key features of older adults’ social networks. The Korean version was validated through a systematic translation and validation process. The LSNS-K consists of 10 self-administered items that measure three content areas: family and friend networks, interdependent social support, and living arrangements. The scale has demonstrated good psychometric properties in the Korean older adult population, with a Cronbach’s alpha of 0.75 and a four-week test-retest reliability coefficient of 0.78 (p < 0.0001) [39]. The LSNS-K was selected for the concurrent validity assessment because of its established validity and reliability in measuring social network characteristics among older Korean adults, making it an appropriate comparison measure for the newly developed SISN scale.

Table 1 demonstrates a comparative analysis of the SISN against commonly used social isolation and network assessment tools, highlighting the unique features and domains covered by each instrument. This comparison demonstrates how the SISN addresses gaps in existing tools by incorporating both objective and subjective dimensions of social connectedness, with particular attention to cultural relevance and emotional quality of relationships.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Comparative features of social isolation and network assessment tools.

https://doi.org/10.1371/journal.pone.0338522.t001

Results

Participants

Table 2 presents the demographic characteristics and descriptive statistics of the 350 participants. The mean age of participants was 68.95 years (SD = 3.98). Additionally, 227 (64.9%) participants had college degrees. The mean SISN was 3.32 (SD = 0.72) and the mean LSNS-K was 2.06 (SD = 0.73).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Demographic characteristics and descriptive statistics of community-dwelling korean older adults (N = 350).

https://doi.org/10.1371/journal.pone.0338522.t002

Reliability

The Cronbach’s alpha for the total SISN scale was 0.94, indicating excellent internal consistency (Table 3). While this high coefficient raises considerations about potential item redundancy, the two-dimensional factor structure confirmed through CFA and the distinct factor loadings (ranging from 0.521–0.917 for social isolation and 0.519–0.916 for social network) suggest that items contribute meaningfully to their respective constructs without excessive overlap. Additionally, the strong model fit indices and the theoretical distinctiveness of the two subscales support the appropriateness of the current item structure.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Internal Consistency and test-retest reliability of the social isolation and social network scale (N = 350).

https://doi.org/10.1371/journal.pone.0338522.t003

Specifically, Cronbach’s alpha was 0.92 for the social isolation subscale (7 items) and 0.94 for the social network subscale (12 items). Composite reliability was also calculated to provide a more robust estimate of internal consistency, yielding values of 0.93 for the social isolation subscale and 0.95 for the social network subscale, which were consistent with the Cronbach’s alpha values and further confirmed excellent internal consistency.

Test-retest reliability was assessed in 150 participants who completed the SISN twice over a two-week interval (Table 3). The intraclass correlation coefficient (ICC) for the total scale was 0.929 (95% confidence interval [CI]: 0.902–0.948, p < 0.001), indicating excellent temporal stability. The subscales also demonstrated strong test-retest reliability, with ICC values of 0.873 (95% CI: 0.825–0.908, p < 0.001) for the social isolation subscale (seven items) and 0.918 (95% CI: 0.887–0.941, p < 0.001) for the social network subscale (12 items).

Construct validity

The construct validity of the SISN was evaluated via CFA using the Mplus software. Initial analysis compared the fit of a single-factor model against the hypothesized two-factor model, with the two-factor model demonstrating superior fit. A second-order factor model was also tested but did not provide significantly better fit than the first-order two-factor model, supporting the retention of the more parsimonious two-factor structure.

Analyses were conducted separately for the social isolation (7 items) and social network (12 items) subscales. For the social isolation subscale, the CFA results demonstrated good model fit: χ² (14) = 38.151, p < .001; RMSEA = 0.070 (90% CI: 0.044–0.097); CFI = 0.983; TLI = 0.974; SRMR = 0.028. All items showed significant factor loadings with z-values ranging from 9.17 to 10.99 (all p < .001) ranging from 0.790–0.917 (p < .001), with standardized regression weights (β) ranging from 0.271–0.838.

For the social network subscale, the CFA results also indicated good model fit: χ² (44) = 106.295, p < .001; RMSEA = 0.064 (90% CI: 0.048–0.079); CFI = 0.976; TLI = 0.964; SRMR = 0.033. All items demonstrated significant factor loadings with z-values ranging from 9.21 to 13.92 (all p < .001), confirming the hypothesized factor structure of the social network dimension. Examination of modification indices revealed no theoretically justified modifications that would substantially improve model fit, supporting the appropriateness of the final model specification.

Item analysis revealed that all items meaningfully contributed to their respective subscales. The mean scores for social isolation items ranged from 3.117–3.997, whereas those for social network items ranged from 0.791–1.746. All items were measured on a 5-point Likert scale. Item-total correlations were satisfactory, indicating good discriminant properties of the individual items. The results of the CFA support the two-dimensional structure of the SISN, confirming that the instrument effectively measured both social isolation and social network constructs as distinct but related dimensions of social connectedness (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Confirmatory factor analysis fit indices for the social isolation and social network scale (N = 350).

https://doi.org/10.1371/journal.pone.0338522.t004

ROC analysis

ROC analysis was performed to determine the cutoff score of the SISN that can identify older adults with social isolation risk using the LSNS-K as the reference standard. ROC analysis revealed that the SISN demonstrated good discriminative ability, with an area under the curve (AUC) of 0.900 (95% CI: 0.866–0.933), which was statistically significant (p < .001). This indicates good diagnostic accuracy for identifying social isolation risk as defined by the LSNS-K reference standard. This indicates that the SISN had an good discriminative validityfor identifying individuals at risk of social isolation.

Using the Youden index to maximize both sensitivity and specificity, the optimal cutoff score was determined to be 3.24. At this threshold, the SISN demonstrated 81.8% sensitivity and 88.2% specificity (Table 5, Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Diagnostic accuracy of the social isolation and social network scale for identifying social isolation risk.

https://doi.org/10.1371/journal.pone.0338522.t005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. ROC curve for the SISN to investigate the optimal cut-off score.

https://doi.org/10.1371/journal.pone.0338522.g001

Correlation analysis between the SISN and LSNS-10 scores showed a strong positive correlation (r = 0.785, p < .001), further confirming that the SISN measures social isolation and social network constructs in a manner highly consistent with the established LSNS-10 measure. Based on the ROC analysis results, SISN mean scores < 3.24 indicated a high risk for social isolation, whereas scores ≥ 3.24 suggested a low risk for social isolation (good social connectedness).

Discussion

This study evaluated the psychometric properties of the SISN Scale among community-dwelling older adults in South Korea. The results demonstrated preliminary psychometric evidence, supporting the utility of the scale as a robust assessment tool for measuring SISN dimensions. These findings are consistent with those of prior research and highlight the importance of capturing the multidimensional aspects of social connectedness in aging populations [42].

Notably, the SISN’s distinct yet interconnected subscales offer essential insights for clinical applications and public health interventions aimed at reducing social isolation and enhancing networks among older adults.

The reliability indices of the SISN were particularly strong, with a Cronbach’s alpha of 0.94 for the total scale and similarly high values for the subscales of social isolation and social networks (0.92 and 0.94, respectively). While these results suggest high internal consistency, the elevated Cronbach’s alpha may also indicate potential item redundancy.

Further analyses, such as inter-item correlations, item-total correlations, and item response theory (IRT) modeling, could help identify and refine overlapping items [40]. In particular, Future research should explore whether a reduced item set could maintain psychometric robustness while enhancing practical utility.

These findings are consistent with similar results from previous studies on social isolation scales that found high reliability in similar psychometric assessments [42]. However, contrasting evidence from a study by Shankar et al. [43] reported moderately lower reliability of tools assessing social isolation, suggesting that cultural or contextual factors may influence reliability outcomes. To address this, the revised discussion now clarifies that these cultural factors should not be assumed but rather examined through measurement invariance testing across diverse demographic and cultural subgroups [44].

This discrepancy may arise from the differing operational definitions of social isolation and diverging measurement practices across studies. Shankar et al. [43] highlighted the potential for overreliance on Western-centric measurement frameworks, which may not fully account for the collectivist cultural norms predominant in countries such as South Korea.

By contrast, the present study, conducted in a collectivist society, demonstrates how culturally sensitive adaptations of measurement tools can yield more robust and reliable outcomes. These comparisons underscore the need for cross-cultural measurement invariance testing and further scale refinement to ensure generalizability across different cultures [42].

Construct validity was well-supported by a CFA, with fit indices (RMSEA, CFI, and TLI) meeting or exceeding benchmarks. These outcomes reinforce the SISN’s hypothesized two-dimensional structure comprising the social isolation and social network subscales. This finding supports previous research suggesting that social connectedness encompasses distinct, yet interrelated constructs [5,45].

However, Western studies have reported greater overlap between these domains [46,47], raising questions about whether collectivist cultural contexts, such as South Korea, facilitate clearer separation. This nuance has been emphasized to avoid speculative cultural interpretations and to propose future cross-cultural invariance testing as a research priority.For example, Cornwell and Waite [45] argued that the boundaries between these constructs are blurred when assessing populations with high levels of interdependence, such as older adults who live in communal settings. This discrepancy may be attributed to contextual differences in social organizations. Older adults in South Korea often rely on multigenerational households to establish community networks, potentially contributing to a clearer demarcation between social isolation and networks. Further comparative analyses across diverse living arrangements are required to clarify these differences.

Although CFA supports structural validity, complementary psychometric methods such as Rasch analysis or differential item functioning (DIF) analysis can further examine item fairness and functioning across demographic subgroups, including sex, socioeconomic status, and educational levels [48]. These advanced approaches will help enhance the scale’s applicability and fairness.

The SISN demonstrated high diagnostic accuracy in identifying older adults at risk of social isolation (AUC = 0.900). While this strong discriminatory ability aligns with prior studies [21], other research reported lower accuracy in cross-cultural contexts due to diverse social norms [49]. The high AUC in this study may reflect contextual relevance to South Korean older adults, where social roles emphasize familial and close social bonds. However, reliance on LSNS-K as a reference standard in ROC analysis could introduce circularity, as both scales target similar constructs [18,23]. This limitation should be addressed in future comparative validation studies using more independent criterion measures.

Despite its contributions, this study has some limitations. First, the study sample was limited to older adults in South Korea, which may restrict the generalizability of the findings. Given that the cultural norms related to social connectedness vary across countries and regions, the interpretation of “support” or “network” may differ substantially in rural or migrant populations, where resources and social structures are less urban-centric. This context-specific variability highlights the need for further validation studies in diverse geographic and socio-cultural settings are required to broaden the applicability of SISN. Coyle and Dugan (2012) highlighted the potential inability of culturally neutral tools to capture subtle differences in social relationships and emphasized the necessity of developing tools that account for cultural contexts [50].

Second, Second, convenience sampling via an online research panel limits sample representativeness and generalizability. Voluntary panel participation likely overrepresented older adults with greater social engagement or online experience, while underrepresenting those with limited digital access or higher social isolation. Furthermore, a very high proportion of respondents (95%) had at least a high school education, suggesting that the scale was tested largely in groups with relatively high social literacy and survey familiarity. These sampling biases constitute a critical limitation and may affect the reliability and construct validity analyses of the SISN, particularly for populations with lower educational attainment.

Third, this study employed a cross-sectional design, which prevented the assessment of changes in social connectedness over time. Although confirming the temporal stability of the SISN is a significant achievement, future research should incorporate longitudinal designs to evaluate how sensitively the scale responds to changes in the social environment. For instance, assessing whether a SISN adequately detects network changes following intervention programs is a critical task for future research [42].

Finally, it is important to further refine the assessment of item-level performance through advanced psychometric approaches such as Rasch modeling or Differential Item Functioning analysis. These techniques would allow for the evaluation of the fairness of the scale across demographic subgroups, including sex, socioeconomic status, and education level, thereby strengthening the applicability of the SISN to diverse populations.

Conclusion

In conclusion, the SISN is a preliminarily validated and culturally informedtool for assessing social isolation and social networks among older South Korean adults. The findings of this study contribute to the growing literature on social isolation assessment while highlighting the need for culturally sensitive adaptations and longitudinal research. Despite certain limitations, the evidence of reliability, construct validity, and classification accuracyfor the SISN underscores its potential utility in both research and practice. Future research should include cross-cultural measurement invariance testing, alternative validation benchmarks, and longitudinal studies to strengthen generalizability and ecological validity. These steps will enhance the scale’s applicability to diverse aging populations and support broader implementation in clinical and community settings.

Supporting information