HBSS was used as the foundation for this adaptation. Originally developed by Hughes in 1984 based on social support theory, the HBSS encompasses three dimensions: emotional, instrumental, and informational support, comprising 30 items rated on a 4-point Likert scale (1 = no help provided to 4 = sufficient help provided). The HBSS was selected based on its theoretical alignment with Sherriff et al.’s five domains of paternal breastfeeding support and its established psychometric properties. The original scale demonstrated robust reliability, with corrected split-half coefficients of 0.85, 0.85, and 0.89 for the three dimensions, respectively, and Cronbach’s alpha coefficients ranging from 0.84 to 0.88.

Translation procedures followed the Brislin translation model, adhering to internationally accepted standards for the cross-cultural adaptation of health measurement instruments (4). This process involved five systematic steps to ensure semantic, conceptual, and functional equivalences. Two independent public health graduate students specializing in maternal and child health performed forward translations from English to Chinese. Both translators were native Chinese speakers with advanced English proficiency and a clinical background in maternal health, ensuring cultural competence during the translation process. A synthesis meeting involving both translators and a third bilingual expert produced a preliminary Chinese version through consensus discussions and the resolution of discrepancies.

Back translation was subsequently performed by two independent doctoral-level translators (one medical doctor and one English literature specialist) who were blinded to the original version in English. The back-translation team had no prior exposure to the original scale content to minimize translation bias and ensure its objectivity. A bilingual committee comprising the translation and back-translation teams and methodological experts systematically compared all versions to identify discrepancies and achieve conceptual equivalence between the source and target languages (19).

A quantitative assessment of translation equivalence was conducted using the Translation Validity Index (TVI). Two independent bilingual experts rated the semantic equivalence of each item on a 4-point scale (1 = not equivalent to 4 = highly equivalent). Inter-rater agreement was calculated using weighted kappa (κw = 0.84, indicating excellent agreement) (20, 21). Items with TVI scores <3.0 underwent additional revision cycles until consensus was reached. The mean TVI across all items was 3.72 (SD = 0.31), indicating a high translation equivalence.

Content validity was systematically evaluated using a structured expert panel process and established content validity assessment procedures. The expert panel comprised six specialists: a chief obstetrician with over 15 years of breastfeeding counseling experience, an associate professor of public health specializing in maternal health, a public health professor with expertise in scale development methodology, and three certified international lactation consultants. Each expert independently rated the relevance, clarity, cultural appropriateness, and comprehensiveness of the items using a standardized 4-point evaluation form, providing both quantitative ratings and qualitative feedback for each item.

The panel members evaluated each item across multiple dimensions, including relevance to the intended construct, clarity of expression, simplicity and comprehensibility, and cultural appropriateness for the Chinese context. Experts also assessed whether the items adequately captured the theoretical framework of paternal breastfeeding support and provided recommendations for cultural modifications while maintaining the construct fidelity.

The expert panel evaluation yielded the following content validity indices:

Based on Lynn’s criteria for six experts, items with I-CVI < 0.78 (items 3, 7, 8, and 13) were modified to enhance cultural relevance while maintaining construct fidelity. These modifications included adapting item 3 from “partner helps with household chores” to “partner manages household tasks to give me more time for breastfeeding,” reflecting the indirect support valued in the Chinese culture.

A structured pilot testing phase was conducted to evaluate item comprehension and response process validity and to identify potential sources of measurement errors. A convenience sample of 30 postpartum women at 42 days post-delivery participated in the pilot test. Pilot participants were selected to represent diverse educational backgrounds (ranging from primary education to postgraduate) and socioeconomic statuses to ensure broad applicability across the target population. Prior to completing the scale, the participants received standardized instructions about the study purpose and provided written informed consent for both questionnaire completion and cognitive interview participation.

Cognitive interviews were conducted immediately following scale completion using a semi-structured think-aloud protocol based on response-process validity principles. The interview guide was developed through a review of breastfeeding-support literature and COSMIN-aligned content domains, and then refined through expert consultation to ensure cultural and clinical relevance. Participants were asked to verbalize their understanding of the instructions, item wording, and response options, and to identify ambiguous or culturally incongruent phrasing. The interviews were audio-recorded with permission. Transcription was completed by one trained research assistant and independently verified against the recordings by a second investigator to ensure accuracy and completeness of the data.

The cognitive interview transcripts were analyzed using thematic content analysis. Two members of the research team independently reviewed the transcripts, generated preliminary codes, and grouped the codes into themes. Any discrepant interpretations or apparently contradictory feedback were resolved through discussion and consensus with reference to the original audio recordings when necessary. The themes were finalized after iterative refinement, and saturation was achieved after 22 interviews, with no new themes emerging in the final eight interviews.

Based on the integrated findings from pilot testing, expert feedback, and cognitive interview analysis, several items were refined to improve their clarity and cultural alignment. For example, wording was adjusted to better reflect common postpartum practices and family roles in China, and examples were added where needed to reduce ambiguity in the interpretation of items. These refinements informed the development of a preliminary 30-item PBSS for subsequent quantitative psychometric evaluation.

The validation study was conducted at the Women’s Hospital, School of Medicine, Zhejiang University, a tertiary care facility serving both urban and rural populations in Hangzhou, China. This setting was selected to ensure diverse sociodemographic representation and the generalizability of the findings to the broader Chinese postpartum population. The study timeline extended from October 2018 to October 2019, allowing for adequate participant recruitment and consideration of seasonal variations.

The sample size calculation was based on established psychometric guidance for exploratory factor analysis, including the widely used recommendation of approximately 10 participants per item for initial scale development (22, 23). Accordingly, the 30-item preliminary PBSS required a minimum of 300 participants to be validated. We intentionally retained this conservative target to accommodate anticipated exclusions (e.g., incomplete questionnaires or response bias) and preserve adequate analytical power following iterative item reduction. Sample adequacy for factor analysis was subsequently evaluated empirically using the Kaiser-Meyer-Olkin (KMO) statistic (24) (threshold ≥0.60) and Bartlett’s test of sphericity.

A convenience sampling approach was employed to recruit 300 postpartum women who attended routine 42-day postnatal checkups. This timing was specifically chosen because it represents a critical period for establishing breastfeeding patterns, and maternal perceptions of paternal support are highly relevant for continued breastfeeding success. While some methodological guidance suggests inflating the target sample by approximately 10% to reduce the impact of missing data from non-response, our initial recruitment target of 300 was selected to maximize feasible enrollment during the study period and provide a practical buffer for anticipated exclusions. After data quality screening and exclusion of incomplete or invalid questionnaires, the final analytic sample remained adequate for the final PBSS-22 structure. We acknowledge that our urban tertiary hospital sample may not fully represent rural populations or those with lower socioeconomic status, which may limit generalizability.

Inclusion criteria:

Exclusion criteria:

A maternal age range of 22–35 years was selected to focus on the most common reproductive age group attending our postpartum services and to minimize heterogeneity from age-related psychosocial and obstetric factors that could confound support perception during the psychometric evaluation. Women younger than 22 years may face distinct vulnerabilities in partner support dynamics and breastfeeding decision-making, whereas pregnancies at age ≥35 years are commonly categorized as advanced maternal age and may involve higher obstetric complexity and different postpartum support requirements. By restricting the age range, we sought to enhance sample comparability while maintaining clinical relevance for routine postpartum care settings (26).

The mode of delivery was recorded in detail, including elective and emergency cesarean sections. Given evidence that emergency operative delivery may be experienced as psychologically distressing and could influence postpartum perceptions and support needs (27, 28), we retained these participants to reflect real-world clinical diversity and explicitly discussed this consideration in the interpretation and limitations.

A standardized data collection protocol was implemented to ensure consistency and minimize measurement errors in all assessments. Trained research assistants who were blinded to the study hypotheses and theoretical expectations administered the surveys using a secure electronic scanning system. Data collectors underwent comprehensive 4-h training sessions covering the study procedures, participant interaction protocols, ethical considerations, and techniques to minimize response bias.

Eligible participants received comprehensive information about the study objectives, procedures, and significance before the questionnaires were distributed. Data were collected in private consultation rooms to ensure confidentiality, minimize distractions, and allow for clarification of questions without influencing responses. All participants completed the questionnaires on-site with immediate electronic submission to prevent missing data and ensure data integrity of the data.

Multiple quality control measures were implemented, including the following:

The primary outcome was the psychometric performance of the PBSS, assessed using multiple validity and reliability indices following established psychometric standards. Demographic and clinical characteristics, including maternal and paternal ages, parity, delivery mode, breastfeeding practices, educational attainment, occupation, and household income, were systematically collected. Breastfeeding exclusivity was assessed using standardized WHO definitions, with exclusive breastfeeding defined as the infant receiving breast milk only, without liquids or solids other than vitamins, minerals, or medications.

Additional contextual variables were collected to assess potential confounding factors, including family support structure, previous breastfeeding experience, attendance at breastfeeding education classes, and perceived breastfeeding self-efficacy, using validated measures.

All analyses followed a pre-specified statistical analysis plan developed prior to data collection to ensure methodological rigor and minimize analysis bias. Data organization and analysis were performed using Excel 2016 and SPSS version 20.0. Statistical significance was set at p < 0.05, with effect sizes reported alongside significance testing to enhance clinical interpretation. Visual figures, including boxplots and other graphical representations, were designed using R Studio (version 4.3.2) to provide a clear and precise visualization of the data distributions and outcomes (29).

Comprehensive item analysis employed eight analytical approaches systematically applied to ensure rigorous item selection: response distribution analysis to assess option validity and identify floor or ceiling effects; independent-samples t-tests to compare high and low total score groups to evaluate item discrimination; Spearman correlation analysis for item sensitivity assessment; Pearson correlation analysis between item scores and expert ratings to evaluate representativeness; correlation analysis between items to assess independence and identify redundancy; exploratory factor analysis for factor loading determination and structural assessment; and Cronbach’s alpha calculation with item-total correlations for internal consistency evaluation.

Item retention criteria were established a priori based on the established psychometric literature and empirical testing.

These thresholds were validated through sensitivity analysis, confirming optimal balance between scale brevity and measurement precision (6, 30, 31).

The validity assessment followed established psychometric standards with multiple sources of evidence, as recommended by the contemporary measurement theory. Structural validity was explored through exploratory factor analysis using principal component extraction with promax rotation to identify the correlated factors. Factor retention was determined using multiple criteria, including eigenvalues >1.0, scree plot examination, parallel analysis, and theoretical interpretability of the factor solutions. Although confirmatory factor analysis (CFA) is recommended as a subsequent step to confirm the factor structure—ideally in an independent validation sample (32)—the present study focused on initial scale development and item reduction within a single dataset. Therefore, CFA was not performed and identified as a priority for future external validation.

Content validity was assessed by calculating the Content Validity Index (CVI) at both the item (I-CVI) and scale levels, including the scale-level CVI for universal agreement (S-CVI/UA) and average agreement (S-CVI/Ave). Content validity thresholds were set at I-CVI ≥ 0.78 and S-CVI ≥ 0.90, based on the established guidelines for panels of six experts.

Reliability evaluation included an internal consistency assessment through Cronbach’s alpha coefficient calculation for the total scale and individual dimensions, with coefficients ≥0.70 considered acceptable and ≥0.80 indicating good reliability (33). Split-half reliability was assessed using the Spearman-Brown prophecy formula to evaluate temporal stability, and item-total correlations were examined to identify items that could compromise internal consistency.

This study was approved by the Institutional Review Board of the Women’s Hospital, School of Medicine, Zhejiang University (Protocol Number: WHSMZU-193370, approved September 15, 2018). All procedures were conducted in accordance with the Declaration of Helsinki and Chinese regulations for human subjects research. Written informed consent was obtained from all participants after providing comprehensive information about the study’s purpose, procedures, voluntary participation, and withdrawal rights. Participants were assured that non-participation or withdrawal would not affect their clinical care or treatment. No compensation was provided to ensure voluntary participation of the participants. The full study protocol, participant information sheet, and all data collection instruments were reviewed and approved prior to the recruitment of participants.

Data confidentiality was maintained through secure electronic storage, password protection, and restricted access to authorized research personnel. All participants were assigned unique study identification codes, and personal identifiers were stored separately from the study data to ensure anonymity.

Given the cross-sectional nature of the study and non-invasive questionnaire-based data collection, the risk to the participants was minimal. However, provisions were made for referral to appropriate support services if the participants experienced distress related to breastfeeding challenges during the study period.

A preliminary assessment of the data’s suitability for factor analysis yielded highly favorable results. The Kaiser-Meyer-Olkin (KMO) (24, 35) measure of sampling adequacy achieved 0.927, substantially exceeding the recommended threshold of 0.6 and indicating excellent data appropriateness for factor extraction procedures. Bartlett’s test of sphericity demonstrated statistical significance (χ² = 3138.993, df = 231, p < 0.001), confirming that the correlation matrix possessed sufficient structure for meaningful factor extraction and rejecting the null hypothesis of identity matrix relationships.

Exploratory factor analysis employing principal component extraction with promax rotation was conducted without predetermined factor number constraints to allow for the emergence of an empirical factor structure. The analysis identified four distinct factors with eigenvalues exceeding 1.0, accounting for 66.551% of the total variance in the dataset. Individual factor eigenvalues were 9.847, 2.156, 1.423, and 1.240 for factors 1 through 4, respectively; Factor 1 explained 44.8% of the variance, Factor 2 contributed 9.8%, Factor 3 added 6.5%, and Factor 4 accounted for 5.6% of the total variance explained. The scree plot and variance explained by each factor are shown in Figure 3.

A detailed examination of the factor loadings revealed several items requiring systematic removal due to inadequate psychometric properties that compromised the factor interpretability and structural clarity. Items 9 (“partner researches breastfeeding information”) and 26 (“partner burps baby after feeding”) demonstrated problematic multiple loadings across factors, with loading differences less than 0.2, violating the established criteria for clear factor assignment. Specifically, item 9 loaded 0.52 on factor 1 (emotional support) and 0.48 on factor 3 (informational support), while item 26 showed loadings of 0.45 on factor 2 (practical assistance) and 0.43 on factor 4 (childcare support), creating an interpretational ambiguity.

Item 13 (“partner attends breastfeeding education classes”) and 1 (“partner expresses positive attitudes about breastfeeding”) exhibited factor loadings below the acceptable threshold of 0.4 across all four extracted factors, indicating an insufficient relationship with any underlying construct and a poor structural fit. Item 13 showed a maximum loading of 0.32 on factor 2, while Item 1 achieved a maximum loading of 0.35 on factor 1, both falling substantially below acceptable standards.

Items 28 (“partner helps with night wakings”) and 5 (“partner shares in feeding decisions”) demonstrated weak primary factor loadings and exhibited concerns in previous sensitivity and representativeness analyses. Item 28 loaded 0.38 on factor 4, while Item 5 showed 0.36 loading on factor 2, both marginally below acceptable thresholds. Following a comprehensive team discussion incorporating expert consultation and psychometric evidence, these items were systematically removed to enhance their factor clarity and interpretability.

Factor-level reliability analysis was conducted by calculating Cronbach’s alpha coefficients and an inter-item correlation assessment within each extracted factor. Item 14 (“partner assists with household cleaning”) demonstrated weak correlations with other items within its designated factor, achieving a multiple correlation squared value below 0.3, indicating a poor fit within the factor structure. Specifically, item 14 showed correlations ranging from 0.18 to 0.27 with other items in the practical assistance factor, which was substantially below the recommended minimum of 0.30 for factor cohesion. Combined with previous evidence of low representativeness, poor sensitivity, and ceiling effects, Item 14 was removed after comprehensive team discussion and expert consultation.

Ultimately, the systematic item removal process retained 22 items that demonstrated robust psychometric properties, clear factor assignments, and theoretical coherence within the four-dimensional framework of paternal breastfeeding support (Table 3).

The final 22-item PBSS demonstrated a coherent four-factor structure through exploratory factor analyses. Factor loadings for all retained items consistently exceeded 0.4, with 18 items (81.8%) demonstrating strong loadings above 0.6, indicating robust relationships between the items and their respective factors. The extracted factors aligned conceptually with theoretical expectations and demonstrated clear interpretability within the context of paternal breastfeeding support behaviors (11, 19, 31). As noted above, CFA is recommended as the next stage of validation and will be undertaken in a separate, independent sample to confirm this structure (32).

Factor 1: Emotional Support encompassed eight items with factor loadings ranging from 0.938 to 0.687, capturing paternal behaviors related to encouragement, reassurance, and emotional availability during breastfeeding challenges. This factor explained 44.8% of the total variance and included items such as “listens to my concerns about breastfeeding” (loading = 0.938), “provides encouragement when breastfeeding becomes difficult” (loading = 0.891), and “shows understanding when I feel frustrated” (loading = 0.847). These items reflect the emotional dimensions of support that directly impact maternal confidence and their persistence.

Factor 2: Informational Support comprised four items with loadings ranging from 0.862 to 0.610, representing paternal provision of knowledge, advice, and information-seeking behaviors related to breastfeeding practices. This factor contributed 9.8% of the variance and included items addressing “seeks information about breastfeeding techniques” (loading = 0.862), “shares helpful knowledge about breastfeeding” (loading = 0.798), “researches solutions to breastfeeding problems” (loading = 0.734), and “provides advice based on learning” (loading = 0.610), which captured the cognitive and educational aspects of paternal support.

Factor 3: Practical Assistance included six items with loadings ranging from 0.857 to 0.547, reflecting tangible support through household management, meal preparation, and daily living assistance that creates conducive environments for breastfeeding success. Contributing to 6.5% of the variance, this factor captured concrete supportive behaviors such as “helps with meal preparation” (loading = 0.857), “manages household responsibilities” (loading = 0.823), “assists with laundry and cleaning” (loading = 0.789), and “runs errands and handles logistics” (loading = 0.547), representing instrumental support that reduces maternal burden and facilitates breastfeeding.

Factor 4: Childcare Support contained 4 items with loadings from 0.964 to 0.468, addressing paternal involvement in direct infant care activities that complement and support maternal breastfeeding efforts. This factor explained 5.6% of the variance and included items related to “soothes baby when crying” (loading = 0.964), “changes diapers promptly” (loading = 0.823), “provides nighttime infant care” (loading = 0.756), and “assists with baby positioning during feeding” (loading = 0.468), reflecting hands-on parenting support that enhances the success of breastfeeding. The complete factor structure, including all item loadings and statistical parameters, is presented in Table 3. The factor loading matrix with a visual representation of the item-factor relationships is presented in Figure 4.