Translation, Cross-Cultural Adaptation, and Psychometric Evaluation of the Thai Family Management Measure Among Caregivers of Children with Chronic Conditions

Highlights

Public health relevance—How does this work relate to a public health issue?

• Childhood chronic conditions require sustained family management, making culturally appropriate measurement relevant to family-centered pediatric care.
• The Thai FaMM was adapted to examine family management among primary caregivers, including parents and other family members.

Public health significance—Why is this work of significance to public health?

• The findings provide preliminary, scale-specific measurement evidence, with limitations in several items and dimensions.
• Additional ordinal-data analyses did not uniformly corroborate the original dimensional structures, supporting further evaluation.

Public health implications—What are the key implications or messages for practitioners, policy makers and/or researchers in public health?

• Responses may inform discussion with caregivers, but clinical screening, intervention selection, and separate dimensional scoring require further validation.
• Condition Management Effort requires refinement, and measurement equivalence across caregiving and cultural contexts remains unestablished.

Abstract

The Family Management Measure (FaMM) assesses family management of childhood chronic conditions, but evidence on its measurement properties in Thailand is limited. This methodological study translated, culturally adapted, and evaluated the Thai FaMM among 350 primary family caregivers. Translation followed International Society for Pharmacoeconomics and Outcomes Research Task Force guidelines; psychometric evaluation followed the Consensus-based Standards for the Selection of Health Measurement Instruments framework. Independent exploratory and confirmatory subsamples comprised 150 and 200 caregivers. Child’s Daily Life was unidimensional after post hoc residual-covariance respecification. Multidimensional first-order models of Condition Management Ability, Family Life Difficulty, View of Condition Impact, and Parental (Caregiver) Mutuality fitted better than one-factor alternatives; second-order models did not improve fit. Internal consistency varied across scales and dimensions; nine items across these five scales had corrected item–total correlations below 0.30. Reported scale-level test–retest intraclass correlations ranged from 0.92 to 1.00. Condition Management Effort had low internal consistency (Cronbach’s α = 0.28), failed original exploratory extraction, and an inadmissible confirmatory solution. These preliminary, scale-specific findings require further validation before separating dimensional scoring or clinical use.

1. Introduction

An estimated 10–30% of children and adolescents live with long-term conditions that require ongoing management to maintain health, prevent disease progression, and optimize quality of life [1]. Similar epidemiological trends in Thailand have intensified demands on pediatric healthcare and family support systems [2]. Children with chronic conditions commonly experience physical limitations alongside emotional and behavioral challenges, including anxiety, depression, social isolation, irritability, and difficulties with attention, which are often associated with symptom burden, treatment demands, and disruptions in normal developmental experiences [3,4,5,6,7,8]. Because these challenges frequently exceed children’s developmental capacity for independent illness management, families assume primary responsibility for integrating treatment into everyday life. Management therefore depends on clinical care and how families perceive, organize, and sustain condition management in daily routines [9,10].

Family-centered pediatric nursing recognizes the family as the primary caregiving context [11,12]. Effective family management is associated with child health, psychosocial well-being, treatment adherence, and healthcare utilization [13]. Valid measurement can inform support needs, individualized interventions, and evaluation of change. Instruments must account for cultural differences in family roles, beliefs, and caregiving practices. This study therefore adapted and evaluated the Thai Family Management Measure (FaMM), examining how the Family Management Style Framework’s constructs were represented in this context.

Background and Theoretical Basis

The Family Management Style Framework explains how families define, organize, and respond to childhood chronic illness [14]. The framework conceptualizes family management through three interrelated components: family definition of the situation, encompassing perceptions of the child’s condition, the child’s identity, and family views of illness management; management behaviors, referring to the strategies and routines families employ in everyday care; and perceived consequences, representing families’ evaluations of the effects of the condition on the child and family life [15,16]. These components conceptualize family management as a dynamic, context-dependent process that extends beyond caregiver burden or family functioning to capture how families integrate chronic illness into daily life [17].

The FaMM operationalizes the Family Management Style Framework through six empirically derived scales: Condition Management Ability, Condition Management Effort, Family Life Difficulty, View of Condition Impact, Child’s Daily Life, and Parental Mutuality [18]. Together, they represent the framework’s three components [17]. The ease-oriented scales—Child’s Daily Life, Condition Management Ability, and Parental Mutuality—address adaptive processes and strengths. The difficulty-oriented scales—Condition Management Effort, Family Life Difficulty, and View of Condition Impact—address management challenges. These complementary orientations support assessment of family capacities and strain [13].

Although the FaMM has demonstrated satisfactory psychometric performance across Western and selected non-Western populations [13], no validated Thai-language version is currently available. This gap matters because family-centered care has become increasingly integrated into Thai pediatric nursing education, practice, and research [19]. Thai families are characterized by collectivist values, multigenerational caregiving, hierarchical family relationships, and strong cultural expectations regarding family obligation and harmony [20]. These characteristics may influence how families perceive childhood illness, organize caregiving responsibilities, and evaluate the consequences of chronic illness. Consequently, linguistic translation alone cannot ensure conceptual and measurement equivalence, making rigorous cross-cultural adaptation essential to preserve the theoretical meaning of the Family Management Style Framework constructs and ensure valid assessment within the Thai context [13,15,18,21]. Evidence remains limited on whether its underlying latent constructs are represented consistently across different sociocultural settings. Cross-cultural evaluation can examine linguistic and conceptual correspondence and how family management constructs are operationalized across settings.

Structural validity informs interpretation of reliability because dimensionality determines how scores represent the intended constructs. The original FaMM development study identified six scales through EFA [18], but comprehensive evaluations across all six scales in other cultural settings remain limited [13]. Whether these structures are retained after translation therefore requires evaluation. Previous studies have predominantly used Cronbach’s α [13], although its interpretation can be affected by multidimensionality and violations of its assumptions [22]. Complementary coefficients, including McDonald’s ω, should be interpreted alongside the measurement structure [22,23,24]. Reporting additional coefficients does not itself compensate for weak factors or establish that overall and dimensional scores are equally defensible.

This study translated, culturally adapted, and evaluated selected measurement properties of the Thai FaMM among primary family caregivers. Structural validity preceded interpretation of internal consistency; test–retest reliability was assessed separately. Agreement between administrations by different research assistants was a supplementary procedural assessment. The study examined construct representation in this sample, without establishing equivalence across cultural or caregiving groups.

2. Materials and Methods

2.1. Study Design

This methodological study comprised translation and cultural adaptation followed by psychometric evaluation, using a cross-sectional design with repeated measures for reliability assessment. The consensus-based standards for the selection of health measurement instruments (COSMIN) framework guided evaluation of structural validity, internal consistency, and test–retest reliability [25]. Reliability was interpreted alongside structural validity and item and dimensional limitations. Agreement between administrations by different research assistants was a supplementary procedural assessment. Construct validity hypothesis testing, measurement invariance, measurement error, responsiveness, and scale-level floor and ceiling effects were not evaluated; item response distributions were examined.

• Phase 1: Translation and cross-cultural adaptation of the Family Management Measure

2.2. Instrument

Family Management Measure

The FaMM is a 53-item self-report instrument assessing integration of childhood condition management into family life [18], with six separately scored scales based on the Family Management Style Framework [17]. Originally, all parents complete 45 items, and those living with an adult partner complete eight additional Parental Mutuality items. This study included parents and other relatives (e.g., grandparents or older siblings) responsible for daily care. “Partner” references encompassed identified family co-caregivers. Primary caregivers completed all 53 items, reporting their own perceptions of collaboration with the identified co-caregiver.

The scales contain 5 Child’s Daily Life, 12 Condition Management Ability, 4 Condition Management Effort, 14 Family Life Difficulty, 10 View of Condition Impact, and 8 Parental Mutuality items. Responses range from 1 (strongly disagree) to 5 (strongly agree). Original item-to-scale assignments and designated reverse scoring were retained. Each scale is scored independently, with higher scores reflecting its intended direction; no global FaMM score was calculated.

Throughout this report, the ease-oriented scales (Child’s Daily Life, Condition Management Ability, and Parental Mutuality) indicate adaptive management and strengths at higher scores. Higher scores on the difficulty-oriented scales (Condition Management Effort, Family Life Difficulty, and View of Condition Impact) indicate greater management challenges.

Previous studies reported Cronbach’s α of 0.51–0.91, test–retest reliability of 0.71–0.94, and expected relationships with family functioning, child behavior, and functional status supporting construct validity [13,18].

2.3. Translation and Cross-Cultural Adaptation Process

Translation and cross-cultural adaptation followed the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) Task Force guidelines [26]. The process was coordinated by the co-principal investigator (J.S.), and involved by a multidisciplinary panel comprising five clinical and methodological experts and two bilingual linguistic experts. The panel’s expertise included pediatric nursing, family-centered care, chronic illness, and psychometric evaluation. The process included preparation, forward translation, reconciliation, back translation, back-translation review, harmonization, cognitive debriefing, finalization, and documentation. Throughout the process, the aim was to preserve the conceptual meaning of the Family Management Style Framework while supporting semantic, experiential, and conceptual correspondence and cultural appropriateness.

Fourteen items across five FaMM scales were culturally adapted to improve conceptual equivalence while preserving the underlying constructs. Changes addressed semantic clarity, caregiving terminology, and culturally appropriate expressions. Item 7’s “roller coaster” metaphor became “ups and downs like a boat riding the wave” to convey fluctuations in the child’s condition. “Partner” references were broadened to reflect extended-family caregiving in Thailand. Table S3 documents changes in emphasis and implications for conceptual correspondence, plus wording checks for items 13 and 32 that were not counted as substantive adaptations.

Translation-stage cognitive debriefing involved face-to-face interviews with 10 caregivers: three caring for children with cardiovascular and seven with non-cardiovascular conditions. Caregivers completed the pre-final version using think-aloud methods; probes addressed meaning, relevance, the metaphor, instructions, and response options. Feedback informed minor wording changes and provided qualitative evidence of comprehensibility. This activity preceded and was distinct from pilot testing. Tables S1–S3 detail the adaptation procedures.

2.4. Content Validity

Content validity of the translated Thai FaMM was evaluated by an independent panel comprising two family nursing researchers and three pediatric nurse specialists. Each translated item was rated using four-point ordinal scales for relevance (1 = not relevant; 4 = very relevant), clarity (1 = not clear; 4 = very clear), simplicity (1 = not simple; 4 = very simple), and absence of ambiguity (1 = doubtful meaning; 4 = clear meaning) [27]. Higher ratings indicated more favorable evaluations for all criteria; the ambiguity criterion therefore did not require reverse coding. Item-level content validity indices (I-CVIs) were calculated as the proportion of experts providing favorable ratings for each criterion. Criterion-specific scale-level indices (S-CVI/Ave) were calculated by averaging the corresponding I-CVIs within each FaMM scale [28]. Item-level content validity indices ranged from 0.80 to 1.00, and criterion-specific S-CVI/Ave values ranged from 0.95 to 1.00. Descriptive averages across the four criteria ranged from 0.96 to 1.00. These averages were not modified kappa coefficients. Detailed results are presented in Table S4.

• Phase 2: Psychometric evaluation of the Family Management Measure

2.5. Participants and Setting

Participants were primary family caregivers of children aged 3 months to 12 years with physician-diagnosed chronic conditions. Eligible caregivers (1) assumed primary responsibility for the child’s daily care and condition management, (2) were actively involved in caregiving, (3) resided with the child, and (4) were aged 18 years or older. Children’s diagnoses included congenital heart disease, diabetes, obesity, thalassemia, asthma, epilepsy, chronic kidney disease, cancer, and other chronic conditions. Each primary caregiver identified one co-caregiver who also participated in the child’s care. Paid caregivers were excluded.

Participants were recruited using convenience sampling from the pediatric inpatient and outpatient units of one tertiary teaching hospital and one general hospital in southern Thailand. Eligible caregivers were approached during inpatient admissions or routine outpatient visits. Only one primary caregiver was enrolled for each child. Primary caregivers completed all 53 items, including the eight Parental (Caregiver) Mutuality items. For these eight items, respondents reported their own perceptions of their relationship with the identified co-caregiver.

The research protocol specified an a priori recruitment target of 530 caregivers, corresponding to 10 participants per item for the 53-item FaMM. This planning approach was informed by published sample-size guidance for factor analysis [29]. Recruitment yielded 373 caregivers, below the planned target. Following the exclusion of 23 observations identified through multivariate outlier screening, the final analytic sample comprised 350 caregivers, allocated to independent exploratory (n = 150) and confirmatory (n = 200) subsamples. Sample-size adequacy was considered separately for each analysis in relation to model complexity, communalities, and the number of indicators per factor, rather than inferred solely from the combined sample size or participant-to-item ratio [30]. The smaller-than-planned sample, particularly the exploratory subsample and low communalities of several items, warranted caution when interpreting structural stability.

2.6. Data Collection Procedures

2.6.1. Research Assistant Training

Two registered nurses experienced in pediatric and family-centered care completed training in study procedures, FaMM theory, ethics, and standardized administration, including competency-based role-play. Agreement between their repeated questionnaire administrations was subsequently examined as a supplementary procedural assessment.

2.6.2. Pilot Testing

The translated FaMM was pilot tested with 20 caregivers (5 cardiovascular and 15 non-cardiovascular cases) for clarity, cultural appropriateness, and feasibility. Questionnaire completion was followed by structured cognitive debriefing, distinct from translation-stage interviews. No substantive concerns required modification. The questionnaire remained unchanged, and pilot participants were retained in the final sample.

2.6.3. Main Data Collection

Between November 2024 and June 2025, consenting caregivers completed paper questionnaires during hospitalization or outpatient visits. Standardized forms recorded demographic and clinical information. Each primary caregiver completed all 53 FaMM items, including Mutuality items referring to the identified co-caregiver.

A randomly selected subsample of 60 caregivers repeated the questionnaire after two weeks for a test–retest assessment. In a separate random subsample of 40, each caregiver completed it under the supervision of two research assistants no more than 30 min apart. This was a supplementary procedural assessment, not evidence of interobserver reliability of the self-report instrument.

2.7. Statistical Analysis

SPSS Statistics 26.0 supported descriptive analyses and original EFA; AMOS 24.0 supported CFA. Additional ordinal-data EFA during revision used Python (version 3.12.13), as specified below. Descriptive statistics summarized participant characteristics and item distributions.

Squared Mahalanobis distances across 53 items screened the original 373 observations. Values exceeding 90.573 (df = 53, p < 0.001) identified potential multivariate outliers [31]. Excluding 23 observations (6.2%) left 350 cases, with no missing analysis data. Exclusion aimed to limit unusual response patterns’ influence on conventional maximum likelihood analyses but did not establish erroneous responses. Mahalanobis distance itself is outlier-sensitive because it uses the sample mean and covariance matrix [32]. A computer-generated sequence assigned participants to nonoverlapping exploratory (n = 150) and confirmatory (n = 200) subsamples, allowing evaluation of structures in participants not involved in extraction. Reported item skewness and kurtosis were below 3.0 and 10.0, respectively [33]. These univariate checks did not establish multivariate normality.

Original scale-specific EFA used maximum likelihood extraction based on Pearson correlations. Maximum likelihood was selected because initial univariate screening suggested approximate normality; this method permits model-fit evaluation and statistical inference when its distributional assumptions are adequately met [34]. Promax oblique rotation was applied to multifactor solutions because the underlying dimensions were expected to correlate [35]. Sampling adequacy and factorability were assessed using the Kaiser–Meyer–Olkin statistic, with ≥0.60 as the adopted criterion, and Bartlett’s test of sphericity [35,36]. Factor retention was guided by examination of eigenvalues and visual inspection of the scree plot. KMO and Bartlett’s test assessed factorability and did not determine the number of factors retained. In the original analyses, extracted communalities (h²) represented the proportion of item variance accounted for by the retained factors, with values ≥ 0.40 considered acceptable for item representation. Values below 0.40 indicated a need for further evaluation rather than automatic exclusion [35].

Revision-stage exploratory analyses used the original exploratory subsample (n = 150), with each FaMM scale analyzed separately. Existing reverse scoring was retained. Two-step polychoric correlations were estimated from ordinal responses using empirical marginal thresholds [37]. Unobserved response categories were omitted from threshold estimation without merging observed categories. Principal axis factoring used squared multiple correlations as initial communalities and Promax oblique rotation for multifactor solutions. Factor retention was evaluated using permutation parallel analysis. A total of 1000 datasets were generated by independently permuting responses within each item, preserving observed response-category frequencies. Observed eigenvalues of the polychoric matrices, with squared multiple correlations on the diagonal, were compared with the corresponding 95th-percentile null eigenvalues. Successive leading eigenvalues exceeding this criterion determined the suggested factor count. Component-based parallel analysis using unit diagonals was also examined to assess sensitivity to the retention procedure.

Parallel-analysis recommendations considered convergence, communalities, loading patterns, factor correlations, and interpretability. No matrix smoothing or continuity correction was applied. Principal axis iterations used a maximum of 10,000 iterations and a maximum absolute communality-change tolerance of 10⁻⁷; Promax rotation used a power of 4. The additional analyses used a custom Python implementation (version 3.12.13) with NumPy 2.3.5 and SciPy 1.17.0, with random seed 20260908. Numerical specifications, validation checks, and analysis code are provided in the Supplementary Analysis Materials.

Original ML EFA structures were evaluated using maximum likelihood CFA in the confirmatory subsample (n = 200). Alternative structures suggested by the additional ordinal-data EFA were not evaluated in the confirmatory sample. The original EFA and CFA treated the five-category item responses as approximately continuous. The ordinal response format and observed item skewness were therefore considered limitations when interpreting conventional maximum likelihood estimates and fit statistics [38]. Model fit was evaluated using complementary indices [39,40,41]. Comparative fit index (CFI) and Tucker–Lewis index (TLI) values ≥ 0.90 indicated acceptable fit, and values ≥ 0.95 indicated good fit. Root mean square error of approximation (RMSEA) values ≤ 0.08 indicated acceptable fit, and values ≤ 0.05 indicated good fit; 90% confidence intervals were reported. Standardized root mean square residual (SRMR) values < 0.08 indicated good fit. Thresholds were interpretive guidelines, not absolute rules. The chi-square statistic was reported and interpreted alongside approximate fit indices because of its sensitivity to sample size. Absolute standardized loadings ≥ 0.30 were the minimum acceptable magnitude; direction was evaluated against content and scoring. Lower loadings indicated poor performance even for retained items. Loadings and communalities were interpreted alongside corrected item–total correlations, internal consistency, and content. Retention, acceptable global fit, or overall reliability did not establish item or dimensional scoring adequacy.

Each respecification began with the hypothesized model. All selected residual covariances were then added simultaneously, and the respecified model was estimated once [42]. No further modification rounds followed. These modifications were introduced post hoc, and the hypothesized and respecified models were evaluated in the same confirmatory sample. Improvements in fit were therefore interpreted as within-sample findings requiring independent replication. Second-order CFA examined hierarchical representations of relationships among first-order dimensions [43]. First-order and second-order models were compared descriptively using their fit statistics and degrees of freedom. Identical global fit was not interpreted as independent confirmation of a higher-order structure, and interpretation considered the restrictions imposed by each model specification [44].

Average variance extracted (AVE) was estimated from the final reported first- or second-order CFA models using standardized item loadings for the unidimensional Child’s Daily Life scale and the constituent first-order factors of the multidimensional scales. AVE was not estimated for Condition Management Effort because its CFA solution was inadmissible. AVE was calculated by dividing the sum of squared standardized factor loadings by that sum plus the sum of standardized residual variances. Values ≥ 0.50 were used as an interpretive benchmark for factor-specific AVE, indicating that the corresponding factor accounted for at least half of its indicators’ variance on average [45]. For multidimensional scales, pooled summaries were the item-count-weighted mean of the constituent first-order AVEs. This summary described the average indicator variance accounted for by the first-order factors collectively; it was not interpreted as evidence of unidimensionality, variance attributable exclusively to the higher-order factor, or the adequacy of an overall scale score. Estimates were interpreted alongside structural validity, item performance, and reliability and did not independently establish external construct validity or the adequacy of separate dimensional scores.

Internal consistency was assessed using McDonald’s omega total (ω_total) and Cronbach’s α for scales and applicable dimensions. Omega total estimates the proportion of score variance attributable to all common factors represented in the specified measurement model and accommodates unequal factor loadings [22,23,24]. It was not interpreted as evidence that a higher-order factor alone adequately represented the scale score or that multidimensionality was automatically resolved. Cronbach’s α was interpreted cautiously because its equality to score reliability depends on assumptions including essential tau-equivalence and uncorrelated measurement errors. Essential tau-equivalence requires equal unstandardized loadings on a common factor while allowing item intercepts to differ; this assumption was not established in the present study [46]. Interpretation considered structural validity and item performance. Corrected item–total correlations were examined using ≥0.30 as the study’s criterion for acceptable item discrimination, with lower values indicating items requiring further evaluation [47].

Test–retest reliability was estimated using intraclass correlation coefficients (ICCs) derived from a two-way mixed-effects model with absolute agreement [48]. Agreement between repeated administrations conducted by different research assistants was estimated using a two-way random-effects ICC with absolute agreement and treated as a supplementary procedural assessment. ICCs were reported with 95% confidence intervals. Point estimates were interpreted as poor (<0.50), moderate (0.50 to <0.75), good (0.75 to <0.90), or excellent (≥0.90), with confidence intervals considered when assessing precision [48].

Revision-stage CFA sensitivity analyses evaluated the original unidimensional structures of Condition Management Ability, Parental (Caregiver) Mutuality, Family Life Difficulty, and View of Condition Impact in the confirmatory sample. Each model retained the original scale items, specified a single latent factor, and included no correlated item residuals. All models were estimated using maximum likelihood. Their fit was compared descriptively with that of the corresponding original EFA-derived first-order models before the addition of correlated item residuals. Comparisons focused on CFI, TLI, and RMSEA to distinguish differences associated with dimensionality from those associated with subsequent model respecification. These analyses tested dimensionality, not robustness to ordinal or robust CFA estimation. Sensitivity to the inclusion of the 23 excluded observations was not evaluated.

3. Results

3.1. Participant Characteristics

The 350 primary caregivers were predominantly women (89.1%) and mothers (78.3%); 75.1% were married or partnered. Mean age was 38.8 years and caregiving duration was 4.85 years. Most had at least secondary education, 57.7% reported sufficient income, and 68.0% identified a parent as secondary caregiver (Table 1).

Table 1. Characteristics of participants (n = 350).

Characteristics	M ± SD	n (%)
Family Caregivers
Gender
Women		312 (89.1)
Men		38 (10.9)
Age in years	38.76 ± 8.36
21–40		213 (60.9)
41–60		132 (37.7)
>60		5 (1.4)
Relationship with the children
Mother		274 (78.3)
Father		35 (10.0)
Grandparent		26 (7.4)
Aunt/uncle		13 (3.7)
Older sibling		2 (0.6)
Marital status
Single		31 (8.9)
Married or partnered		263 (75.1)
Divorced, separated, or widowed		56 (16.0)
Education
Below primary		2 (0.6)
Primary education		51 (14.6)
Secondary education		150 (42.8)
Diploma		45 (12.9)
Bachelor’s degree or higher		102 (29.1)
Occupation
Manual labor		111 (31.7)
Agricultural worker		66 (18.8)
Homemaker		73 (20.9)
Retailer		71 (20.3)
Government		27 (7.7)
Private business		2 (0.6)
Household income
Insufficient		50 (14.3)
Sufficient		202 (57.7)
Comfortable		98 (28.0)
Duration of caring, years	4.85 ± 3.10
Secondary caregiver
Parent		238 (68.0)
Grandparent		83 (23.8)
Aunt or uncle		10 (2.8)
Other (e.g., older sibling or other relatives)		19 (5.4)
Children
Gender
Girl		170 (48.6)
Boy		180 (51.4)
Age in years	7.76 ± 2.48
Infant (<1 year)		1 (0.3)
Toddler (1–3 years)		6 (1.7)
Pre-school age (3–6 years)		84 (24.0)
School age (7–12 years)		259 (74.0)
Chronic condition classification
Congenital heart disease		16 (4.6)
Diabetes		12 (3.4)
Obesity		8 (2.3)
Thalassemia		78 (22.3)
Asthma		55 (15.7)
Epilepsy		44 (12.6)
Thyroid		33 (9.4)
Cancer		25 (7.1)
Chronic renal disease		25 (7.1)
Anemia		17 (4.9)
Attention-deficit/hyperactive disorder		4 (1.1)
Multiple chronic conditions		10 (2.9)
Other		23 (6.6)
Number of chronic diseases
Single		340 (97.1)
Multiple		10 (2.9)
Age at first-time diagnosis, years	3.00 ± 2.90

Note. Single-condition diagnostic categories and the separate multiple-condition category account for all 350 children. Percentages use the full sample denominator and may not total 100% because of rounding.

Children’s mean age was 7.8 years; 340 (97.1%) had one chronic condition, and 10 (2.9%) had multiple conditions, reported separately (Table 1). Thalassemia and asthma predominated; 16 children (4.6%) had congenital heart disease within single-condition categories. Each exploratory and confirmatory subsample included five children with multiple conditions (Table S5).

3.2. Item Descriptive Analysis

Item statistics (Tables S6 and S7) and response percentages (Table S8) varied across scales. All ease-oriented items had means above the midpoint (M > 3.00); 22 of 28 difficulty-oriented items were below it (M < 3.00). Thirty-six items (67.9%) had percentages ≥60% in Table S8’s summary column, ranging from 2.6% for Item 8 to 89.1% for Item 25. Interpretation depends on wording and scoring direction; these patterns do not indicate uniformly favorable family management.

3.3. Ease-Oriented Scales’ Structural Validity, Internal Consistency Reliability, and Item Analysis

Figure 1, Figure 2 and Figure 3 show final ease-oriented CFA models derived from original ML EFA. Table 2 reports original/ordinal EFA and final CFA loadings; Table 3 reports CFA fit; Table S6 reports item statistics, corrected item–total correlations, reliability, and AVE. Figure S1 provides additional CFA diagrams (panels S1.1–S1.4).

Figure 1. Final CFA model of the Child’s Daily Life scale.

Figure 2. Final CFA model of the Condition Management Ability scale.

Figure 3. Final CFA model of the Parental (Caregiver) Mutuality scale.

Table 2. Original maximum likelihood EFA loadings, additional polychoric-based principal axis factor loadings, communalities, and CFA loadings for the ease-oriented Thai Family Management Measure scales.

	Original EFA		Ordinal-Data EFA			ML CFA Loading
Scale/Dimension/Item	Original ML EFA Loading	Original Reported h2	Ordinal PAF F1	Ordinal PAF F2	Ordinal Extracted h2
Child’s Daily Life
Item 1	0.57	0.20	0.49	—	0.24	0.53
Item 5	0.58	0.39	0.68	—	0.47	0.59
Item 10	0.42	0.37	0.60	—	0.36	0.35
Item 16	0.66	0.36	0.67	—	0.45	0.67
Item 20	0.79	0.69	0.89	—	0.80	0.81
Condition Management Ability
Perceived caregiving efficacy dimension
Item 13	0.64	0.56	0.80	0.05	0.65	0.73
Item 18	0.46	0.24	0.58	−0.19	0.34	0.53
Item 19	0.78	0.64	0.78	−0.008	0.61	0.69
Item 25	0.58	0.40	0.72	0.03	0.52	0.52
Item 41	0.56	0.34	0.56	0.07	0.33	0.63
Perceived future planning and resources dimension
Item 4	0.60	0.42	0.57	−0.03	0.32	0.64
Item 14	0.98	0.99	0.70	0.004	0.49	0.78
Item 15	0.45	0.28	0.44	0.17	0.24	0.36
Perceived management challenges dimension
Item 17	0.41	0.21	−0.28	0.49	0.29	0.39
Item 27	0.68	0.47	−0.02	0.76	0.58	0.79
Item 28	0.70	0.53	0.17	0.72	0.58	0.76
Item 34	0.48	0.24	0.14	0.51	0.30	0.43
Parental (Caregiver) Mutuality
Perceived caregiving teamwork dimension
Item 46	0.46	0.26	0.609	−0.01	0.36	0.50
Item 48	0.73	0.64	0.805	0.09	0.77	0.78
Item 49	0.59	0.55	0.692	0.16	0.66	0.64
Item 50	0.93	0.91	1.09	−0.15	0.97	0.76
Item 53	0.77	0.67	0.80	0.12	0.81	0.74
Perceived caregiving concordance dimension
Item 47	0.36	0.23	0.15	0.49	0.37	0.56
Item 51	0.95	0.93	−0.09	0.97	0.82	0.81
Item 52	0.53	0.39	0.01	0.79	0.65	0.78

Note. Original ML EFA and additional polychoric PAF used n = 150; CFA used n = 200. Additional EFA presents admissible one-factor Child’s Daily Life and two-factor Ability and Mutuality solutions. Ordinal loadings are pattern coefficients, with Promax rotation for multifactor solutions. Original dimension headings organize items without implying equivalence to ordinal factors; F1/F2 are solution-specific. Standardized CFA loadings are from respecified Child’s Daily Life and second-order Ability and Mutuality models, which did not test ordinal-EFA alternatives. Table S10 reports other candidate solutions and factor correlations. A dash indicates no second factor. Abbreviations. CFA, confirmatory factor analysis; EFA, exploratory factor analysis; FaMM, Family Management Measure; h², communality; ML, maximum likelihood; PAF, principal axis factoring.

Table 3. Fit index values for the Family Management Measure.

FaMM	Model	χ2 Goodness-of-Fit Test				CFI	TLI	RMSEA	90% CI	p Close	SRMR
		χ2	df	p	CMIN/df
Ease-Oriented Scales
Child’s Daily Life	A1. Hypothesized, first-order	26.80	5	<0.001	5.36	0.90	0.80	0.14	0.09–0.20	0.002	0.06
	A2. Respecified, first-order	2.73	4	0.604	0.68	1.00	1.01	0.00	0.00–0.09	0.785	0.01
Condition Management	B1. Hypothesized, first-order	101.72	51	<0.001	1.99	0.91	0.88	0.07	0.05–0.09	0.047	0.08
Ability	B2. Respecified, first-order	85.23	48	<0.001	1.78	0.93	0.91	0.06	0.04–0.08	0.165	0.07
	B3. Respecified, second-order	91.59	50	<0.001	1.83	0.92	0.90	0.06	0.04–0.08	0.121	0.07
Parental (Caregiver)	C1. Hypothesized, first-order	42.35	19	0.002	2.23	0.96	0.94	0.07	0.04–0.11	0.068	0.04
Mutuality	C2. Hypothesized, second-order	42.35	19	0.002	2.23	0.96	0.94	0.07	0.04–0.11	0.068	0.04
Difficulty-Oriented Scales
Family Life Difficulty	D1. Hypothesized, first-order	146.27	74	<0.001	1.98	0.91	0.89	0.07	0.05–0.09	0.027	0.06
	D2. Respecified, first-order	129.99	72	<0.001	1.81	0.93	0.91	0.06	0.04–0.08	0.100	0.05
	D3. Respecified, second-order	135.75	74	<0.001	1.83	0.92	0.90	0.06	0.04–0.08	0.080	0.06
Condition Management Effort ‡	E1. Hypothesized, one-factor; inadmissible solution	36.35	2	<0.001	18.18	0.34	−0.99	0.29	0.22–0.38	<0.001	0.13
View of Condition Impact	F1. Hypothesized, first-order	44.06	34	0.116	1.30	0.96	0.94	0.04	0.00–0.06	0.706	0.06
	F2. Hypothesized, second-order	44.06	34	0.116	1.30	0.96	0.94	0.04	0.00–0.06	0.706	0.06
Sensitivity Analyses of the Original Unidimensional Scales
Condition Management Ability	S3.1. Hypothesized, one-factor	228.38	54	<0.001	4.23	0.68	0.61	0.13	0.11–0.14	<0.001	0.11
Parental (Caregiver) Mutuality	S3.2. Hypothesized, one-factor	95.29	20	<0.001	4.76	0.86	0.81	0.14	0.11–0.16	<0.001	0.06
Family Life Difficulty	S3.3. Hypothesized, one-factor	202.48	77	<0.001	2.63	0.84	0.81	0.09	0.07–0.11	<0.001	0.07
View of Condition Impact	S3.4. Hypothesized, one-factor	71.02	35	<0.001	2.03	0.84	0.80	0.07	0.05–0.09	0.067	0.07

Note. CFA used subsample B (n = 200). “Respecified” models added post hoc residual covariances in the same confirmatory sample; improvements require independent replication. TLI can exceed 1, so 1.01 does not independently indicate overfitting. Mutuality and Impact models added no residual covariances; identical first- and second-order fit does not independently favor a hierarchy. Second-order Ability and Difficulty each had two additional degrees of freedom relative to respecified first-order models. The RMSEA p-value tests close fit. Effort was unsupported. ^‡ Effort yielded an inadmissible solution. Software-reported fit statistics are shown for transparency, not as evidence of a valid model. Abbreviations. CFI = comparative fit index, CI = confidence interval, CMIN/df = minimum discrepancy per degree of freedom, FaMM = Family Management Measure, TLI = Tucker and Lewis index, RMSEA = root mean square error of approximation, SRMR = standardized root mean square residual.

3.3.1. Child’s Daily Life Scale

Original ML EFA supported a unidimensional structure for Child’s Daily Life (KMO = 0.73; Bartlett’s χ² = 168.64, p < 0.001), explaining 38.0% of the total variance, with factor loadings ranging from 0.42 to 0.79. Originally reported communalities were below 0.40 for Items 1, 5, 10, and 16 (Table 2). The hypothesized CFA model showed mixed fit, with TLI and RMSEA outside the prespecified acceptable thresholds. Global fit improved following the post hoc addition of a residual covariance between Items 1 and 10 (Table 3; Figure 1). Standardized loadings in the respecified model ranged from 0.35 to 0.81. Same-sample identification and evaluation leave structural support provisional pending independent replication.

Internal consistency was satisfactory (ω_total = 0.75; α = 0.74), with corrected item–total correlations of 0.38–0.64. AVE was approximately 0.37, below 0.50 (Table S6). Item-deletion checks did not improve internal consistency or establish reduced-item structural validity or content coverage; all five items remained.

3.3.2. Condition Management Ability Scale

Original ML EFA identified three Ability factors (KMO = 0.74; Bartlett’s χ² = 481.46, p < 0.001), explaining 44.7% of variance; loadings were 0.41–0.98. Originally reported communalities were below 0.40 for Items 15, 17, 18, 34, and 41 (Table 2). The hypothesized first-order TLI fell below the acceptable threshold. Simultaneously adding three post hoc residual covariances yielded a broadly acceptable fit, loadings of 0.32–0.81, and inter-factor correlations of 0.17–0.67 (Figures S1.2 and S1.3).

Second-order fit remained broadly acceptable but slightly poorer than the respecified first-order fit: χ² increased by 6.36, with degrees of freedom increasing from 48 to 50 (Table 3). Standardized item loadings ranged from 0.36 to 0.79; loadings for the future planning and resources Items 4, 14, and 15 were 0.64, 0.78, and 0.36, respectively (Table 2; Figure 2). Second-order loadings were 0.86, 0.76, and 0.49. Management challenges had the lowest second-order loading (λ = 0.49; R² = 0.24), indicating that the higher-order factor accounted for approximately 24% of its variance.

Overall reliability was ω_total = 0.74 and α = 0.76. Caregiving efficacy was satisfactory (ω_total = 0.77; α = 0.76); future planning/resources (ω_total = 0.60; α = 0.56) and management challenges (ω_total = 0.68; α = 0.65) were lower. Corresponding AVEs were approximately 0.39, 0.38, and 0.38, all below 0.50; pooled AVE was 0.387 (Table S6). Item 17’s corrected item–total correlation was 0.22. Low AVEs, uneven dimensional reliability, and weak item performance limit separate scoring despite acceptable global fit.

3.3.3. Parental (Caregiver) Mutuality Scale

Original ML EFA identified a two-factor structure for Parental (Caregiver) Mutuality (KMO = 0.84; Bartlett’s χ² = 570.57, p < 0.001), explaining 57.2% of the variance, with loadings ranging from 0.36 to 0.95. Originally reported communalities were below 0.40 for three items: 46, 47, and 52 (Table 2). The first-order CFA model demonstrated broadly acceptable fit without adding correlated item residuals (Figure S1.4). Standardized loadings ranged from 0.50 to 0.81, and the inter-factor correlation was 0.70. The first-order and second-order representations produced identical fit statistics and degrees of freedom; global fit therefore did not distinguish between them (Table 3). Second-order loadings were 0.96 and 0.73 (Figure 3). The loading for caregiving teamwork approached unity (λ = 0.96; reported R² = 0.93), indicating that approximately 93% of its variance was attributed to the higher-order factor under this model specification.

Reliability was satisfactory overall (ω_total = 0.84; α = 0.84), for teamwork (ω_total = 0.81; α = 0.81), and for concordance (ω_total = 0.75; α = 0.74). Corrected item–total correlations were 0.39–0.69. AVEs were approximately 0.47 and 0.52, respectively; only concordance met 0.50. Pooled AVE was 0.496 (Table S6). Variance explained differed despite satisfactory reliability, and identical fit did not favor a hierarchy.

3.4. Difficulty-Oriented Scales’ Structural Validity, Internal Consistency Reliability, and Item Analysis

Figure 4 and Figure 5 show final Difficulty and Impact CFA models derived from original ML EFA. Table 3 reports CFA fit; Table 4 reports original/ordinal EFA and interpretable CFA loadings; Table S7 reports item statistics, corrected item–total correlations, reliability, and AVE. Effort had no interpretable CFA solution. Figure S2 provides additional diagrams (panels S2.1–S2.3).

Figure 4. Final CFA model of the Family Life Difficulty scale.

Figure 5. Final CFA model of the View of Condition Impact scale.

Table 4. Original maximum likelihood EFA loadings, additional polychoric-based principal axis factor loadings, communalities, and CFA loadings for the difficulty-oriented Thai Family Management Measure scales.

	Original EFA		Ordinal-Data EFA			ML CFA Loading
Scale/Dimension/Item	Original ML EFA Loading	Original Reported h2	Ordinal PAF F1	Ordinal PAF F2	Ordinal Extracted h2
Family Life Difficulty
Perceived family life disruption dimension
Item 22	0.81	0.68	0.82	−0.08	0.70	0.72
Item 31	0.53	0.49	0.77	−0.004	0.60	0.73
Item 33	0.47	0.32	0.65	0.10	0.42	0.62
Item 36	0.31	0.13	0.26	−0.28	0.16	0.23
Item 39	0.55	0.37	0.51	−0.46	0.52	0.50
Item 42	0.50	0.41	0.76	0.11	0.57	0.69
Item 43	0.63	0.52	0.80	0.08	0.64	0.74
Item 45	0.79	0.68	0.80	−0.20	0.72	0.66
Perceived emotional life difficulty dimension
Item 2	0.68	0.47	0.50	0.02	0.25	0.53
Item 6	0.49	0.43	0.66	−0.03	0.45	0.72
Item 9	0.43	0.46	0.77	0.19	0.60	0.71
Perceived normalcy of family life dimension
Item 23	0.57	0.33	0.30	0.64	0.47	0.71
Item 38	0.78	0.67	0.55	0.80	0.85	0.56
Item 44	0.21	0.07	−0.08	0.24	0.07	−0.03
Condition Management Effort ‡
Item 3	NI	0.10	0.67	—	0.46	NI
Item 7	NI	0.09	0.45	—	0.21	NI
Item 12	NI	0.03	−0.18	—	0.03	NI
Item 35	NI	0.05	0.29	—	0.08	NI
View of Condition Impact
Perceived long-term life difficulty dimension
Item 11	0.60	0.37	0.64	−0.05	0.42	0.70
Item 21	0.19	0.06	−0.25	0.13	0.09	0.16
Item 24	0.54	0.29	0.57	−0.02	0.33	0.42
Item 29	0.65	0.46	0.70	0.30	0.55	0.57
Item 30	0.72	0.52	0.75	0.09	0.57	0.67
Item 37	0.59	0.35	0.66	0.03	0.44	0.56
Perceived manageability of condition dimension
Item 8	0.39	0.19	0.27	−0.54	0.39	0.25
Item 26	0.58	0.34	0.13	0.67	0.45	0.75
Item 32	0.13	0.02	0.040	0.14	0.02	0.18
Item 40	0.73	0.52	−0.08	0.79	0.65	0.46

Note. Original ML EFA and additional polychoric PAF used n = 150; CFA used n = 200. Additional EFA presents two-factor Difficulty and Impact and one-factor Effort solutions. Original dimension headings organize items without implying ordinal-factor equivalence; F1/F2 are solution-specific. Bold ordinal extracted communalities are below 0.40. Difficulty and Impact CFA loadings are from second-order models, which did not test ordinal-EFA alternatives. ^‡ Effort’s original communalities are initial estimates: two-factor ML extraction stopped because the Hessian was not positive definite. Its one-factor CFA was inadmissible, including a negative Item 35 residual variance. ML loadings are therefore marked NI. Ordinal one-factor extraction was performed separately. Abbreviations. CFA, confirmatory factor analysis; EFA, exploratory factor analysis; FaMM, Family Management Measure; h², communality; ML, maximum likelihood; NI, no interpretable estimate; PAF, principal axis factoring.

3.4.1. Family Life Difficulty Scale

Original ML EFA identified a three-factor structure for Family Life Difficulty (KMO = 0.86; Bartlett’s χ² = 666.72, p < 0.001), explaining 43.4% of the variance. Loadings ranged from 0.21 to 0.81; Item 36 loaded at 0.31, whereas Item 44 loaded at 0.21, below the prespecified 0.30 criterion. Originally reported communalities were below 0.40 for five items: 23, 33, 36, 39, and 44 (Table 4). The hypothesized first-order CFA model had a TLI slightly below the acceptable threshold. Following the simultaneous post hoc addition of two residual covariances, the respecified model demonstrated broadly acceptable fit, with inter-factor correlations ranging from 0.45 to 0.73.

Second-order fit remained broadly acceptable, although χ² increased by 5.76 and degrees of freedom increased from 72 to 74 relative to the respecified first-order model (Table 3). Second-order loadings were 0.79, 0.87, and 0.67. Most standardized item loadings ranged from 0.50 to 0.74; however, Item 36 loaded at 0.23, and Item 44 had an essentially null loading of −0.03 (Table 4; Figure 4).

Overall internal consistency was satisfactory (McDonald’s ω_total = 0.82; Cronbach’s α = 0.81). Dimensional estimates were satisfactory for family life disruption (ω_total = 0.83; α = 0.83), lower for emotional life difficulty (ω_total = 0.68; α = 0.67), and poor for normalcy of family life (ω_total = 0.38; α = 0.25). Corresponding AVE estimates were approximately 0.40, 0.43, and 0.27, all below the 0.50 benchmark. The descriptive pooled scale-level value was 0.38 (Table S7). Corrected item–total correlations were below 0.30 for Items 23, 36, 38, and 44, including a near-zero negative correlation for Item 44 (−0.01). Overall fit and reliability did not extend uniformly to items and dimensions. The low AVE, poor reliability, and weak item performance of the normalcy dimension did not support its separate scoring.

3.4.2. Condition Management Effort Scale

Effort showed substantial limitations. KMO was below criterion (0.55), although Bartlett’s test was significant (χ² = 22.30, p = 0.001). Two-factor ML extraction stopped because the Hessian was not positive definite. Original communalities of 0.03–0.10 were initial, not extracted, estimates (Table 4). One-factor CFA was also inadmissible, including a negative Item 35 residual variance. Neither ML EFA nor CFA loadings were interpreted, and AVE was withheld. Software-reported CFA fit appears in Table 3 for transparency.

Internal consistency was low (α = 0.28). Only Item 3 met corrected item–total correlation ≥0.30 (0.33); Items 7, 12, and 35 had 0.29, 0.06, and −0.12 (Table S7). Additional ordinal EFA converged but retained weak items (Section 3.5). These findings did not support the Effort score interpretation.

3.4.3. View of Condition Impact Scale

Original ML EFA identified long-term life difficulty and condition manageability (KMO = 0.73; Bartlett’s χ² = 234.86, p < 0.001), explaining 31.6% of variance. Loadings were 0.13–0.73, with Items 21 and 32 below 0.30. Only Items 29, 30, and 40 had originally reported communalities ≥0.40 (Table 4).

First-order CFA fit was broadly acceptable without residual covariances; the factor correlation was 0.46. First- and second-order fit and df were identical (Table 3, models F1 and F2). Second-order item loadings were mostly 0.42–0.75, but Items 8, 21, and 32 loaded at 0.25, 0.16, and 0.18. Higher-order loadings were 0.63 and 0.72 (Table 4; Figure 5).

Overall reliability was modest (ω_total = 0.66; α = 0.60), with ω_total = 0.69 and α = 0.67 for long-term difficulty and ω_total = 0.43 and α = 0.14 for manageability. Corresponding AVEs were approximately 0.29 and 0.21, below 0.50; the pooled AVE was 0.26 (Table S7). Items 8, 21, 32, and 40 had corrected item–total correlations below 0.30. Deletion marginally improved internal consistency, but manageability remained unsatisfactory. These checks did not establish reduced-item structural validity or content coverage. Low AVE, reliability, and item performance did not support separate manageability scoring.

3.5. Additional Ordinal-Data Exploratory Analyses

All six polychoric matrices were positive definite without smoothing. Common-factor parallel analysis suggested two Child’s Daily Life, three Ability, one Effort, and two each Difficulty, Mutuality, and Impact factors (Table 5). Two-factor Child’s Daily Life did not converge and produced an improper communality; three-factor Ability was improper (Item 14 h² = 1.09). One-factor Child’s Daily Life and two-factor Ability alternatives were admissible.

Table 5. Factor-retention recommendations and solution diagnostics from additional ordinal-data exploratory analyses of the Thai Family Management Measure.

Scale	Original ML EFA Structure	Common-Factor PA: Suggested Factors	Component PA: Suggested Components	Polychoric-Based PAF Findings and Implications
Ease-Oriented Scales
Child’s Daily Life	1 factor	2	1	The two-factor solution did not converge and produced an improper communality. The one-factor solution was admissible, with loadings of 0.49–0.90. Factor retention remained uncertain.
Condition Management Ability	3 factors	3	2	The three-factor solution was improper: Item 14 had a communality of 1.09. A two-factor alternative was admissible, with a factor correlation of 0.13, but requires further confirmation.
Parental (Caregiver) Mutuality	2 factors	2	1	The two-factor solution was admissible, with a factor correlation of 0.70. Item 50’s communality approached 1 (0.97). Retention recommendations differed between procedures.
Difficulty-Oriented Scales
Family Life Difficulty	3 factors	2	2	Both retention procedures suggested two dimensions. In the two-factor solution, the largest absolute pattern loadings for Items 36 and 44 were 0.28 and 0.24, respectively.
Condition Management Effort	Extraction failed	1	1	The one-factor solution converged, but Item 12 loaded negatively (−0.18) and Item 35 remained below 0.30 (0.29). Successful extraction did not resolve the scale’s measurement limitations.
View of Condition Impact	2 factors	2	2	Both procedures suggested two dimensions. Items 21 and 32 remained weak, with the largest absolute pattern loadings of 0.25 and 0.14, respectively. Item 8 loaded oppositely to Items 26 and 40 on the second factor.

Note. Additional analyses used polychoric matrices and Promax rotation for multifactor solutions. PA used 1000 independent within-item permutations preserving category frequencies, with a 95th-percentile criterion. Common-factor PA used squared multiple correlations on the diagonal; component PA used unit diagonals. Retention was based on convergence, communalities, loadings, and interpretability. Communality >1 indicates an improper solution; admissibility alone does not establish validity. Reverse scoring was retained. Abbreviations. ML = maximum likelihood; PA = parallel analysis; PAF = principal axis factoring.

Component parallel analysis suggested one Child’s Daily Life and Mutuality component and two Ability components, indicating retention sensitivity. Both procedures suggested two Difficulty dimensions instead of the original three. Table S9 reports parallel-analysis eigenvalues; Table S10 reports pattern loadings, communalities, factor correlations, and alternative solutions. Ordinal communalities below 0.40 occurred for Child’s Daily Life Items 1 and 10 (0.24, 0.36); Ability Items 4 (0.32), 15 (0.24), 17 (0.29), 18 (0.34), 34 (0.30), and 41 (0.33); and Mutuality Items 46 and 47 (0.36, 0.37).

Effort’s one-factor loadings were 0.67, 0.45, −0.18, and 0.29 for Items 3, 7, 12, and 35; communalities for Items 7, 12, and 35 were below 0.40 (0.21, 0.03, 0.08). Difficulty Items 2, 36, and 44 had communalities of 0.25, 0.16, and 0.07; Items 36 and 44 also had absolute pattern loadings below 0.30 on both factors. Impact Items 8, 21, 24, and 32 had communalities of 0.39, 0.09, 0.33, and 0.02; Items 21 and 32 had absolute loadings below 0.30 on both factors.

Ordinal analyses did not uniformly corroborate original structures. Retaining low-communality items did not establish adequacy. Section 3.3 and Section 3.4 report CFA of original-analysis models; ordinal-EFA alternatives were not evaluated in the confirmatory sample.

3.6. Test–Retest Reliability

Test–retest ICCs (n = 60) were 0.92 (95% CI = 0.85–0.95) for Mutuality, 0.98 (95% CI = 0.97–0.99) for Child’s Daily Life, and 1.00 (95% CI = 1.00–1.00) for Ability, Difficulty, Effort, and Impact (Table S4).

3.7. Agreement Between Repeated Administrations

In the separate subsample (n = 40), reported ICCs were 1.00 across items and scales administered by different assistants. These short-interval estimates describe procedural agreement, not independent evidence of FaMM interobserver reliability.

3.8. Sensitivity Analyses

Original one-factor Ability, Mutuality, Difficulty, and Impact models did not meet acceptable global-fit criteria (Figure S3, panels S3.1–S3.4). All had lower CFI/TLI and higher RMSEA than EFA-derived first-order alternatives without residual covariances. Comparisons favored multidimensionality in this sample but did not distinguish first- and second-order representations or resolve item and reliability limitations.

3.9. Overall Psychometric Performance

Performance varied across scales (Table 6). Respecified unidimensional Child’s Daily Life had satisfactory internal consistency and item–total correlations. Multidimensional Ability, Mutuality, Difficulty, and Impact models showed broadly acceptable fit and outperformed one-factor alternatives. Reliability and item performance remained limited, particularly for future planning/resources, management challenges, normalcy, and manageability.

Table 6. Measurement models, psychometric findings, and implications of the Thai Family Management Measure.

FaMM Scale	Original Model	Model Evaluated in Thai Sample	Reliability (ω/α)	Psychometric Conclusion	Measurement Contribution	Clinical Implications
Ease-Oriented Scales
Child’s Daily Life	Unidimensional	One-factor CFA with one post hoc residual-covariance pair	0.75/0.74	Improved respecified fit and satisfactory internal consistency; AVE below 0.50; independent replication required	Original ML EFA supported one factor; additional ordinal analyses indicated uncertainty in factor retention	Responses may inform discussion of everyday life; clinical utility remains unestablished
Condition Management Ability	Unidimensional	Three correlated dimensions: caregiving efficacy, future planning and resources, and management challenges; second-order representation also evaluated	0.74/0.76	Broadly acceptable CFA fit; lower reliability in two dimensions, weak ITC for Item 17, and all dimensional AVEs below 0.50	Original three-factor CFA fitted better than the one-factor alternative; ordinal EFA yielded an improper three-factor solution and an admissible two-factor alternative	Separate dimensional scores require further validation before guiding targeted care
Parental (Caregiver) Mutuality	Unidimensional	Two correlated dimensions: caregiving teamwork and caregiving concordance; second-order representation also evaluated	0.84/0.84	Broadly acceptable CFA fit and satisfactory internal consistency; only concordance met the dimensional AVE benchmark	Two-factor CFA fitted better than the one-factor alternative; first- and second-order models had identical fit, and ordinal retention recommendations differed	Responses may inform discussion of caregiving collaboration; interpret them in relation to the identified co-caregiver
Difficulty-Oriented Scales
Family Life Difficulty	Unidimensional	Three correlated dimensions: family life disruption, emotional life difficulty, and normalcy of family life; second-order representation also evaluated	0.82/0.81	Broadly acceptable CFA fit and satisfactory overall internal consistency; poor normalcy reliability, low dimensional AVEs, and an essentially null loading for Item 44	Original three-factor CFA fitted better than the one-factor alternative; both ordinal parallel-analysis procedures suggested two dimensions	Normalcy scores and poorly performing items are not supported as independent clinical indicators
Condition Management Effort	Unidimensional	Original ML EFA extraction failed; one-factor CFA was inadmissible; additional ordinal one-factor EFA converged	—/0.28	Low internal consistency and persistent item limitations; no interpretable CFA-based AVE reported	Successful ordinal extraction did not resolve the scale’s measurement limitations; wording, scoring, content coverage, and dimensionality require reassessment	Current findings do not support clinical interpretation of the scale score
View of Condition Impact	Unidimensional	Two correlated dimensions: long-term life difficulty and condition manageability; second-order representation also evaluated	0.66/0.60	Broadly acceptable CFA fit, but modest overall internal consistency, poor manageability reliability, low dimensional AVEs, and weak items	Two-factor CFA fitted better than the one-factor alternative; higher-order fit was identical, and item limitations persisted in ordinal EFA	Manageability scores and poorly performing items are not supported as independent clinical indicators

Note. Table 2 and Table 4 report EFA/CFA loadings, Table 3 CFA fit, Tables S6 and S7 reliability/AVE, and Tables S9 and S10 ordinal analyses. Reliability refers to overall scale scores. One-factor comparisons preceded residual-covariance respecification. Ordinal-EFA alternatives were not confirmed. Relative fit, identical first- and second-order fit, and item retention do not establish dimensional scoring adequacy. Clinical thresholds, responsiveness, and intervention-selection utility were unestablished. A dash indicates unreported omega total. Abbreviations: AVE, average variance extracted; CFA, confirmatory factor analysis; EFA, exploratory factor analysis; FaMM, Family Management Measure; ITC, corrected item–total correlation; ML, maximum likelihood; ω_total, McDonald’s omega total; α, Cronbach’s alpha.

Nine items across these five scales had corrected item–total correlations below 0.30: one Ability, four Difficulty, and four Impact items. All Child’s Daily Life and Mutuality items met the criterion (Tables S6 and S7).

Second-order models did not improve fit: Mutuality and Impact had identical fit, while Ability and Difficulty had slightly poorer fit with two additional df. Hierarchical superiority was therefore unestablished. Effort had inadequate structural validity and internal consistency.

4. Discussion

The culturally adapted Thai FaMM showed preliminary, scale-specific measurement support. Original analyses identified unidimensional Child’s Daily Life and four multidimensional scales, whereas Effort was unsupported. Acceptable CFA fit coexisted with low communalities, weak loadings, and uneven dimensional reliability. Additional ordinal EFA did not uniformly corroborate these structures. Although multidimensional first-order models outperformed one-factor alternatives, second-order models offered no superior fit. The findings warrant further examination of Family Management Style Framework measurement [14,15] without establishing culturally specific differences or adequate separate dimensional scores. The dimensions may inform Family Management Style Framework operationalization [14,15], distinguishing present competence from future preparedness, emotional difficulty from disrupted routines, and teamwork from interpersonal concordance. These patterns are compatible with dynamic, context-dependent family management [16,17], but do not establish cultural specificity or superior hierarchical measurement.

Evaluation of higher-order structures requires attention to model specification, identification, and the restrictions imposed on relationships among first-order factors [44]. For Parental (Caregiver) Mutuality and View of Condition Impact, the first-order and second-order representations produced identical global fit statistics without adding correlated item residuals. Global fit therefore did not distinguish these representations or independently confirm a higher-order structure. For Condition Management Ability and Family Life Difficulty, the second-order models had two additional degrees of freedom and slightly lower CFI and TLI values, although approximate fit remained broadly acceptable. These hierarchical representations offer a possible theoretical account of relationships among dimensions, but their interpretation depends on the imposed constraints and the measurement quality of the constituent dimensions. Moreover, multidimensionality alone does not establish the suitability of overall scale scores or separate dimensional scores; additional evidence concerning the reliable variance represented by each score is required [49]. For Parental (Caregiver) Mutuality, the near-unit loading for caregiving teamwork suggested substantial overlap with the proposed higher-order factor under the fitted specification. For Condition Management Ability, the higher-order factor explained approximately 24% of the variance in management challenges, indicating uneven representation of its dimensions. The weak first-order correlation of 0.17 within this scale also warrants caution when interpreting a common higher-order construct. These estimates do not independently establish factor distinctiveness or justify separate dimensional scores.

The exploratory solutions explained modest variance (31.6% to 44.7%) across View of Condition Impact, Child’s Daily Life, Family Life Difficulty, and Condition Management Ability scales, compared with 57.2% in Parental (Caregiver) Mutuality. These proportions should be interpreted alongside the observed loadings and communalities. Low communalities and weakly defined factors can reduce the stability of factor recovery, particularly in smaller samples [30]. In this study, these concerns, together with uneven dimensional reliability, limit the strength of support for the proposed measurement structures. Moreover, identifying multiple factors does not itself justify separate dimensional scores; their interpretation requires evidence concerning the reliable variance represented by those scores [49]. Independent replication is therefore needed to evaluate structural stability and scoring implications.

Sociocultural characteristics offer one possible explanation for the multidimensional structures identified in this sample. Family management is embedded within cultural expectations regarding caregiving roles, family relationships, and collective responsibility [13]. Multigenerational caregiving and shared family responsibilities may be relevant to how caregivers perceive management competence, future preparedness, disruption of family routines, and caregiving collaboration. However, these cultural influences were not directly tested in the present study. Similar observations in Turkish and Brazilian adaptations [50,51] suggest that although the Family Management Style Framework is broadly transferable across cultures, the measurement structure through which its constructs are operationalized should not be assumed to be universally invariant [14]. An alternative explanation concerns item wording and keying direction. The management challenges dimension comprised reverse-keyed Condition Management Ability items, the family life normalcy dimension comprised reverse-keyed Family Life Difficulty items, and condition manageability contained predominantly reverse-keyed items. Although these items were scored according to the original FaMM instructions, correct reverse scoring does not exclude wording-related response effects. The observed groupings may therefore reflect substantive content, wording direction, or both. Because these contributions were not separately evaluated, the findings should not be attributed solely to Thai sociocultural characteristics. Further evaluation of wording effects and measurement invariance is needed to clarify the interpretation and cross-cultural comparability of the proposed dimensions.

ML sensitivity comparisons preceded residual-covariance additions, so poorer one-factor fit was not attributable to subsequent respecification. Better relative fit nevertheless establishes neither structural stability nor scoring adequacy. Respecification requires substantive explanations alongside diagnostics: meaning and presentation order can produce residual dependence [52], but conceptual relationships or modification indices alone do not justify it. Child’s Daily Life items 1–10 may connect peer participation with judgements of everyday-life similarity. Difficulty Items 33–36 connect obtaining help with balancing care and family life; Items 36–39 connect balance with school functioning. These links are indirect, particularly 36–39, and Item 36 remained weak. Ability Items 19–34 concern responding to condition changes and establishing routines; items 41–15 concern goals and financial resources. Clear wording duplication is absent, and the basis for 15–27 remains uncertain. These retrospective interpretations establish neither residual-dependence causes nor original selection criteria. Same-sample post hoc searches may capitalize on chance and require independent validation [53].

Ordinal EFA provided mixed evidence of structural stability. Although common-factor parallel analysis suggested the original factor counts for several scales, the three-factor Condition Management Ability solution was improper, and both parallel-analysis procedures suggested two dimensions for Family Life Difficulty rather than the original three. Other procedural disagreements indicated retention sensitivity. Retention recommendations should therefore be considered alongside solution admissibility, loading patterns, and interpretability [34]. These findings qualify the original structural conclusions and support confirmatory evaluation of competing models. They do not establish the proposed dimensions as separate clinical subscales, which require additional evidence supporting their score interpretation [49]. Ordinal analyses identified representation limitations across all scales. Communalities below 0.40 warrant examination of an item’s relationship to the retained factors and its intended content rather than automatic deletion [35]. Magnitude matters: Item 8 in View of Condition Impact was close to the guideline (h² = 0.39), whereas items 12 and 35 in Condition Management Effort, item 44 in Family Life Difficulty, and items 21 and 32 in View of Condition Impact had communalities below 0.10. These particularly low values indicate that the evaluated factors accounted for little of the corresponding underlying item-response variance.

Six items remained below 0.40 in the two-factor Condition Management Ability solution, indicating that the admissible alternative still provided limited representation of several items. Child’s Daily Life and Parental (Caregiver) Mutuality also contained items below the guideline, showing that representation problems extended beyond the weakest scales. Low communalities do not establish whether the problem arises from item wording, distinct but relevant content, or an inadequate factor structure. Moreover, communalities and the number of strong indicators per factor influence the sample size needed for stable factor recovery, warranting caution with the exploratory sample [30]. Decisions about revision or removal should therefore combine content review and caregivers’ interpretations with further evaluation of competing measurement structures. Neither retention nor deletion was established to improve measurement. Item 44 (“We know when our child needs to be a child like the other children”) warrants particular attention because its final CFA loading was essentially null (−0.03), accompanied by a near-zero corrected item–total correlation (−0.01) and low communality in the additional ordinal EFA (0.07). Its adapted wording concerns recognizing when the child wants to be like other children, which may differ from the family-level normalcy represented by the other items in this dimension. This is a possible content mismatch, rather than an established explanation for its performance. Retention in the reported model does not support its contribution to scoring. The effects of removing item 44 on model fit, reliability, and content coverage were not evaluated and remain unresolved.

This study evaluated structural validity before interpreting internal consistency, consistent with recommendations for evaluating health measurement instruments [25]. Interpretation considered item performance and dimensional reliability alongside global fit. Internal consistency was assessed using McDonald’s ω_total and Cronbach’s α, with interpretation informed by the measurement structure and assumptions underlying their calculation [22,54]. Variable internal consistency precluded endorsing all five scales with broadly acceptable fit. AVE further qualified the measurement findings. Only the caregiving concordance dimension of Mutuality met the 0.50 benchmark; the remaining factor-specific estimates indicated that the corresponding factors accounted for less than half of their indicators’ variance on average [45]. Child’s Daily Life and caregiving teamwork demonstrated satisfactory internal consistency despite AVEs below this benchmark, illustrating that internal consistency and average indicator variance explained provide distinct information. The particularly low AVEs for family life normalcy and condition manageability, together with their poor internal consistency and weak item performance, further limited support for separate dimensional scores. Pooled scale-level summaries were interpreted descriptively because they could obscure differences between dimensions and did not establish variance attributable exclusively to the higher-order factor or justify overall scale scoring. These findings support further evaluation of item content and competing measurement structures; low AVE alone does not determine whether an item should be removed.

Alpha is calculable for Likert responses, but reliability interpretation assumes essential tau-equivalence and uncorrelated errors, neither established here. Omega total allows unequal loadings but remains model-dependent; neither coefficient establishes unidimensionality or separate scoring adequacy [46]. Five-category responses and item skewness remain concerns despite reporting omega. Cross-study comparisons must consider item composition, scoring, distributions, and sample characteristics. Brazilian [51] and Thai Effort coefficients therefore describe low consistency within each sample, rather than test cross-cultural differences.

Overall and dimensional performance differed particularly for Family Life Difficulty and View of Condition Impact. Their family life normalcy and condition manageability dimensions had low internal consistency, comparable to that of the unsupported Condition Management Effort scale (Table S7). These dimensions therefore should not be considered adequately supported for separate scoring merely because the broader models showed acceptable global fit. Evidence supporting dimensional scores requires evaluation of the reliable variance represented by those scores [49]. Similarly, items with CFA loadings below the study’s criterion of 0.30 remained poorly performing indicators despite their inclusion in the reported models. Their retention and contribution to the proposed hierarchical structures require further evaluation, considering both measurement performance and content coverage.

Item endpoint concentrations warrant further evaluation. These patterns do not establish scale-level floor or ceiling effects. Such assessment requires examining the proportions of respondents obtaining the minimum or maximum possible score on each scale [55]. Endpoint concentrations may indicate limited differentiation among respondents and should be examined alongside item content and score distributions. Moreover, the inclusion of non-parent caregivers broadens the sample’s caregiving perspectives but does not establish measurement equivalence across caregiver relationships. Equivalence requires explicit evaluation of measurement invariance across the relevant groups [56].

Reported test–retest reliability suggests temporal consistency [18,57], but does not resolve structural or internal-consistency limitations, particularly Effort’s unsupported structure. Agreement between administrations by trained assistants suggests consistent responses under standardized procedures, although the ≤30 min interval allowed possible recall. This provides limited procedural evidence, not established reproducibility across clinicians in routine practice.

Effort had failed original exploratory extraction, inadmissible CFA, and α = 0.28. Brazilian Effort reliability was also low (α = 0.36) [51], without establishing a shared cause. Wording, content coverage, and sample contributions remain unresolved. Item 7’s initial communality was 0.09 and corrected item–total correlation 0.29. Its inadmissible CFA loading was uninterpretable, so these findings cannot establish improved comprehension or conceptual correspondence from the metaphor. Diagnostic heterogeneity was not tested. Cognitive interviews should inform evaluation of wording, scoring, content, and dimensionality before score use.

In sum, this study evaluated the translation, cultural adaptation, structural validity, and reliability of the Thai FaMM among family caregivers of children with chronic conditions. These findings provide preliminary Thai FaMM evidence with uneven item and dimensional performance. Condition Management Effort was not supported, and several proposed dimensions demonstrated limited reliability or low average indicator variance explained. Differences between the original and additional exploratory analyses further qualify the proposed structures.

4.1. Strengths

This study combined systematic translation and cultural adaptation following ISPOR guidance with evaluation of selected measurement properties informed by the COSMIN framework [25,26]. Forward and backward translation, expert review, and cognitive debriefing provided complementary evidence concerning the wording and comprehensibility of the Thai version. Including both parental and non-parental caregivers broadened the caregiving experiences represented in the sample.

Independent exploratory and confirmatory subsamples allowed the original EFA-derived structures to be evaluated in participants not involved in their extraction. Additional ordinal-data EFA and parallel analysis examined sensitivity to estimation and factor-retention procedures, while comparisons with the original one-factor CFA models provided further information about dimensionality. Reporting item performance, dimensional reliability, and AVE alongside global fit made the limitations of the proposed models explicit. These strengths do not remove the need for independent replication or establish equivalence across caregiving and cultural contexts.

4.2. Limitations

Several limitations should be acknowledged. First, inadequate Effort structure and internal consistency preclude clinical score interpretation; wording, scoring, content, and dimensionality require review. Poor normality and reliability likewise limit separate scores despite acceptable broader model fit.

Second, the exploratory sample comprised 150 participants, and several items had low communalities, potentially limiting the stability of the extracted structures. Sample-size adequacy cannot be established solely from the total sample size or a participant-to-item ratio [30]. Additional polychoric-based principal axis analyses and permutation parallel analysis examined sensitivity to the ordinal response format and factor-retention procedure. Nevertheless, some solutions were nonconvergent, improper, or sensitive to the retention method, and several items had sparse response categories. Polychoric estimation assumes underlying bivariate normal response variables and therefore does not eliminate distributional assumptions [37]. The additional analyses did not evaluate ordinal or robust CFA estimation or test the suggested alternative structures in the independent confirmatory sample.

Third, same-sample post hoc residual additions may yield optimistic fit. Although this sample was independent of the exploratory sample, it did not provide independent validation of the subsequent modifications. The respecified models therefore require replication. Sensitivity to the exclusion of 23 observations identified using Mahalanobis distance was not evaluated. These observations may include valid but uncommon caregiving experiences; consequently, the influence of their exclusion on factor structures, reliability estimates, and representation of the target population remains uncertain. Item-removal sensitivity analyses were not conducted using a consistent framework across scales. The available checks therefore do not establish whether removing problematic items, including Item 44, would improve structural validity and reliability while preserving content coverage. The additional ordinal EFA and dimensionality comparisons did not evaluate this question, and the findings do not support recommendations for reduced-item versions.

Fourth, the confirmatory analyses used maximum likelihood estimation, treating the five-category responses as approximately continuous. The requested CFA using ordered indicators, WLSMV estimation, and theta parameterization was not performed. Consequently, the sensitivity of the reported factor loadings, standard errors, and fit statistics to ordinal estimation remains unknown. The additional polychoric-based EFA does not resolve this limitation. Further evaluation should apply an appropriate ordinal estimator to the same substantive CFA models before separately examining alternative structures suggested by the additional exploratory analyses.

Fifth, measurement invariance was not evaluated across caregiving contexts, including parental versus non-parental caregivers, caregivers of boys versus girls, and different childhood chronic conditions. Consequently, the findings do not establish equivalent measurement across these groups or justify interpreting between-group score differences solely as differences in family management. Cross-cultural measurement equivalence also remains unestablished.

Sixth, wording-related method effects were not explicitly modeled. Although designated items were reverse-scored according to the original FaMM instructions, several identified dimensions comprised predominantly reverse-keyed items. The analyses therefore could not distinguish the contributions of substantive content and wording direction to the observed factor structures. Sociocultural explanations for these structures consequently remain tentative.

Finally, translation permission was obtained, but developer endorsement was not. Independent developer review could scrutinize adapted wording and conceptual correspondence.

4.3. Implications for Practice

The cross-culturally adapted Thai FaMM may offer a structured approach to discussing how families manage childhood chronic conditions in everyday life. Its potential clinical contribution lies in considering both family strengths and management difficulties. However, interpretation should reflect the evidence and limitations for each scale; the present findings do not establish effectiveness for clinical screening, selecting interventions, or improving care outcomes.

The ease-oriented scales—Child’s Daily Life, Condition Management Ability, and Parental (Caregiver) Mutuality—address perceptions of the child’s everyday life, management capability, and collaboration in caregiving. The difficulty-oriented scales address perceived management effort, disruption of family life, and concerns about the condition and its future implications. These topics may inform dialog with caregivers alongside clinical assessment. Interpretation of Mutuality responses should consider the relationship of the identified co-caregiver to the respondent.

The present findings do not support clinical interpretation of Condition Management Effort scores or separate scores for dimensions with inadequate reliability, particularly family life normalcy and condition manageability. Other proposed dimensional scores also require further confirmation before being used to guide targeted interventions. Individual item responses may provide topics for discussion, but no item-level clinical thresholds or screening accuracy were established; responses should therefore not be treated as validated clinical flags or independently determine care decisions. Further research should evaluate clinical interpretability, responsiveness, and usefulness in routine family-centered care.

4.4. Implications for Research

Preliminary findings warrant further Thai FaMM evaluation. Original and ordinal exploratory differences require further structural evaluation. A priority is to evaluate the original CFA models using ordered indicators and an appropriate ordinal estimator, such as WLSMV [58]. Retaining the same substantive model specifications would help distinguish the influence of estimation method from changes in model structure. Alternative structures suggested by the additional EFA should then undergo confirmatory evaluation, followed by independent replication. Before recommending separate dimensional scores for research use, further analyses should establish whether these scores provide reliable information beyond that represented by the broader scale [49].

Cognitive interviews or focus groups should examine Effort item interpretation and content coverage, informing wording clarification, scoring review, or theory-informed redevelopment before quantitative evaluation. Normalcy and manageability also need refinement given low reliability and AVE. Further analyses should examine wording effects, systematic item removal, and sensitivity to retaining unusual but potentially valid responses.

Once defensible models are established, evaluate invariance across caregiving contexts, including parental and non-parental caregivers, caregivers of boys and girls, and different childhood chronic conditions. Group-specific sample sizes and response-category frequencies should inform feasibility, and ordinal multi-group analyses should use appropriate identification procedures to evaluate configural, metric, scalar, and strict invariance [59]. Any partial invariance and its implications for the intended comparisons should be reported transparently. Separate evaluations across healthcare settings and cultural groups are also needed. Longitudinal studies should assess responsiveness and invariance over time before interpreting score changes as changes in family management. Further evaluation of predictive validity and clinical interpretability would clarify the usefulness of scale and dimensional scores in observational and intervention research and establish when comparisons across populations and over time are justified.

5. Conclusions

The culturally adapted Thai FaMM provides preliminary, scale-specific evidence, with uneven item and dimensional performance. Condition Management Effort was unsupported, and additional ordinal analyses qualified the original structures. Refinement, ordinal CFA, and independent validation remain necessary before recommending separate dimensional scoring, between-group comparisons, or clinical decision-making. These limitations should guide interpretation of the current scores.