Primary caregiver-reported family resilience in children with cancer in central China: a latent profile analysis

Abstract

Background

Childhood cancer can disrupt family functioning, increase caregiver psychological distress, and impair caregiver quality of life. While family resilience is crucial for adaptation, most research has focused on individual-level factors, neglecting heterogeneity and multilevel influences on family resilience.

Methods

Guided by the Social Ecological Model (SEM), this cross-sectional observational study used latent profile analysis (LPA) to identify distinct profiles of family resilience among caregivers of children with cancer and to explore factors associated with these profiles. Between July 2022 and March 2024, 292 caregivers were recruited. Family resilience was measured using the Family Resilience Assessment Scale. LPA was employed to identify resilience profiles, and binary logistic regression was used to explore influencing factors.

Results

Two latent profiles were identified: the Low Resources–Low Positivity profile (86%) and the High Internal Resilience profile (14%). The Low Resource–Low Positivity profile demonstrated generally lower scores, especially in utilizing social and economic resources and maintaining a positive outlook. The High Internal Resilience profile showed higher scores across all family resilience dimensions, particularly in communication/problem solving, positive outlook, and meaning-making, while the use of external social and economic resources remained relatively lower. Univariate analysis showed significant differences between profiles in residence, number of siblings, caregiver education, individual resilience, social support, caregivers’ physical and psychological well-being and child communication (caregiver-reported). Binary logistic regression identified having more than one child (OR = 3.184, 95% CI: 1.437 ~ 7.057, P = 0.004) and higher individual resilience (OR = 1.095, 95% CI: 1.028 ~ 1.165, P = 0.005) as significant predictors of High Internal Resilience profile.

Conclusion

This study identified two distinct family resilience profiles among caregivers of children with cancer. Limited use of social and economic resources was common, while caregiver resilience and having multiple children predicted higher family resilience. Interventions should enhance caregiver coping capacity, support one-child families through peer and family programs, and improve access to social support, flexible employment, and affordable care to strengthen family resilience.

Clinical trial number

Not applicable.

Introduction

Family resilience refers to a family’s ability to adapt, maintain, or restore stability when facing significant stress or adversity [1]. As a systemic construct, it emphasizes the dynamic and interactive processes through which families cope collectively with challenges, particularly in high-stress health contexts such as pediatric oncology. Family resilience includes multiple dimensions, such as family communication, family cohesion, attitudes toward adversity, spirituality, a shared sense of responsibility, and resources utilization [2]. Prior research has shown that higher family resilience helps caregivers manage the physical, emotional, and financial burdens of cancer treatment and sustain family functioning [3]. In pediatric oncology, where children depend heavily on their families for daily care and emotional support, family resilience is especially critical. The diagnosis and treatment of childhood cancer not only impose immense psychological pressure on families [4], but also impairing caregivers’ physical health, work capacity and financial stability [5]. These stressors can, in turn, affect the child’s treatment and recovery [6]. Emerging evidence suggests that maintaining adequate family resilience remains a challenge for many families of children with chronic illnesses, which may adversely affect caregiving effectiveness and family functioning [7, 8]. As a particularly burdensome chronic condition, pediatric cancer therefore deserves focused investigation.

To better understand the multilevel influences on family resilience, the Social Ecological Model (SEM) offers a valuable conceptual framework. The SEM posits that health and well-being are influenced by multiple interacting systems at different levels [9]. At the microsystem level, individual factors such as the patient’s physical symptoms, the caregiver’s psychological distress and personal resilience have been associated with family resilience [10, 11]. However, the roles of child-related clinical and demographic characteristics (e.g., diagnosis type, recurrence, child’s age, caregiver’s education) remain underexplored. At the mesosystem level, family relationships and contextual factors, such as marital status, sibling dynamics, and perceived social support, have been linked to family adaptability and cohesion [11–13]. However, additional contextual variables, including residential setting, caregiver employment, and children’s school attendance, remain underexamined, particularly in non-Western settings. At the macrosystem level, broader social systems and cultural values shape how families respond to adversity. In Chinese society, values such as collectivism, filial piety, and family harmony may encourage mutual support and joint coping within families [14]. However, many families still face systemic challenges, such as disparities in healthcare coverage, limited access to psychosocial services in rural areas, and insufficient financial assistance from the government and general public [15, 16]. These barriers may reduce families’ ability to access resources and weaken their overall family resilience. Although more difficult to quantify, such macro-level influences provide critical context for understanding variations in family resilience within the SEM framework.

Despite the complexity of these multilevel influences, most previous studies have used variable-centered approaches, which assume that all families respond similarly to adversity. This approach may overlook important heterogeneity in how families adapt [17]. In contrast, person-centered methods such as latent profile analysis (LPA) can identify distinct family subgroups based on shared patterns across multiple dimensions of family resilience [18–20]. Since families often show both strengths and vulnerabilities across different resilience dimensions, identifying latent profiles is particularly valuable in pediatric oncology.

By applying LPA within the SEM framework, this study aims to (a) identify distinct family resilience profiles among families of children with cancer and (b) examine how multilevel contextual factors are associated with profile membership. Guided by the SEM and existing literature, we propose the following research questions: (a) Are there distinct latent profiles of family resilience among families of children with cancer, and what are their key characteristics across family resilience dimensions? (b) How are microsystem, mesosystem, and macrosystem factors associated with membership in different family resilience profiles? We hypothesize that: (a) Distinct latent profiles of family resilience exist, each characterized by unique patterns across family resilience dimensions; (b) Multilevel contextual factors are significantly associated with profile membership. These findings will inform the development of targeted interventions and health policies to enhance family resilience and improve the overall quality of life for families affected by childhood cancer.

Methods

Study design

This study utilized a cross-sectional observational design.

Participants

Primary family caregivers were defined as the family member who assumed the main responsibility for the child’s daily care and medical decision-making, such as the child’s mother, father, or grandparent [21]. The inclusion criteria were as follows: being the primary family caregiver of a child (under 18 years old) with cancer, being actively involved in the child’s daily care, being able to read and understand the questionnaire, and providing informed consent to participate in the study. The exclusion criteria included: caregivers were excluded if they reported a history of severe mental illness or if research staff observed signs of cognitive or behavioral difficulties during the screening that might affect their ability to complete the questionnaire.

Sample size

Sample size was estimated using G*Power based on multiple linear regression with 22 predictors, assuming a medium effect size (f² = 0.15), an alpha level of 0.05, and a statistical power of 0.95. This yields a minimum required sample size of 230 participants. Although the primary outcome is categorical, this method provided a conservative estimate for model complexity and ensured sufficient power. After identifying two latent profiles, binary logistic regression was used. The final sample of 292 ensured adequate power.

Data collection

Participants were recruited using convenience sampling from July 2022 to March 2024 at four tertiary hospitals: three in Wuhan, Hubei Province, and one in Taiyuan, Shanxi Province. All participating hospitals are regional pediatric oncology centers that provide specialized diagnosis, treatment, and follow-up care for children with cancer. During the child’s inpatient stay, trained research staff reviewed medical records to identify children under 18 with a confirmed cancer diagnosis. For each eligible child, their primary caregiver (e.g., mother, father, or grandparent) was identified and approached in person at the pediatric oncology ward. The research staff explained the study’s purpose, procedures, and confidentiality protections, answered any questions, and invited the caregiver to participate. Those who expressed interest were assessed for eligibility based on inclusion and exclusion criteria, and written informed consent was obtained prior to enrollment.

Measurements

Demographic and disease-related questionnaire

A self-designed questionnaire was used to collect the following information: children’s age, gender, place of residence, number of siblings, current school enrollment status, education level, diagnosis, time since diagnosis, current treatment stage, and recurrence status. Additionally, it included information about the primary caregiver’s age, education level, marital status, employment status, relationship to the child, average daily caregiving hours, monthly income, receipt of financial assistance/support, and number of permanent household residents.

The family resilience assessment scale (FRAS)

Family resilience was measured using the Family Resilience Assessment Scale (FRAS), originally developed by Sixbey and translated into Chinese in 2018 [22, 23]. The Chinese FRAS includes 44 items across four dimensions: family communication and problem solving (27 items), utilizing social and economic resources (8 items), maintaining a positive outlook (6 items), and ability to make meaning of adversity (3 items). Each item is rated on a 4-point Likert scale (1 = strongly disagree to 4 = strongly agree), with higher scores indicating greater perceived family resilience. Total scores range from 44 to 176, and dimension scores were calculated as the mean of items within each subscale. The Chinese version has demonstrated good reliability (Cronbach’s α = 0.96) in caregivers of children with chronic conditions [23]. In this study, mean scores of the four dimensions were used in the LPA. These dimensions capture key theoretical domains of family resilience and are suitable for identifying heterogeneity in family responses.

The quality of life scale-family (QOL-Scale)

The Quality of Life (Family Version) is a 37-item instrument designed to assess the quality of life of a family member caring for a patient with cancer [24]. In this study, we used the Chinese version of this instrument, which contains 35 items and four dimensions: physical well-being, psychological well-being, social concerns, and spiritual well-being [25]. It has been widely used in cancer population research in China, and the Cronbach’ s α is 0.69 [26]. Mean scores of the four dimensions were used for both univariate and multivariate analyses of family resilience profiles.

The perceived social support scale (PSSS)

The Perceived Social Support Scale, a revised version of the Multidimensional Scale of Perceived Social Support [27], was translated and adapted in Chinese in 2005 [28]. The 12-item scale measures perceived social support from families, friends and significant others using a 7-point Likert scale, with higher scores indicating greater perceived support. To reduce conceptual overlap with the FRAS and minimize participant burden, we excluded the four-item family support subscale from the PSSS. For instance, the PSSS item like “My family is willing to help me make decisions” closely overlaps with FRAS items such as “We are able to make decisions about how to solve problems” and “We are able to solve family problems together.” Including both may introduce multicollinearity and inflate associations. We therefore retained only the subscales for support from friends and significant others, which have demonstrated good reliability (Cronbach’s α = 0.88) and have been widely validated in Chinese cancer populations [29]. Their mean scores were used in univariate and multivariate analyses.

The connor-davidson resilience scale (CD-RISC-10)

The 10-item CD-RISC-10 measures individual resilience using a 5-point Likert scale ranging from 0 (“never”) to 4 (“always”) [30]. Total scores are calculated by summing all item scores, with higher scores reflecting greater resilience. The Chinese version has demonstrated good internal consistency (Cronbach’s α = 0.85) [31] and has been widely applied among caregivers of children with cancer in China [32].

Pediatric quality of life inventory 3.0 (PedsQL 3.0)

The PedsQL 3.0 cancer module is designed to evaluate the health-related quality of life of children and adolescents with cancer, and it includes 27 items across eight dimensions: pain, nausea, procedural anxiety, treatment anxiety, worry, cognitive problems, perceived physical appearance, and communication [33]. Items are rated on a 5-point Likert scale ranging from 0 (“never”) to 4 (“almost always”), with higher scores indicating better quality of life. The Chinese version has been validated, with Cronbach’s α ranging from 0.70 to 0.90 [34], and has been widely used in China [35]. In this study, caregiver-proxy reports were used due to concerns about the reliability of self-report in younger children. Mean scores across the eight dimensions were used for both univariate and multivariate analyses of family resilience profiles.

Ethical approval

This study was approved by the Ethics Committee of Wuhan University School of Medicine (approval number IRB2022018). All participants, including parents, grandparents, and other primary caregivers, provided written informed consent before participation. The consent process ensured that participants fully understood the study purpose, procedures, voluntary nature, and their right to withdraw at any time. Confidentiality was maintained by removing personally identifiable information and assigning unique identification codes to each questionnaire. All data were securely stored on password-protected computers accessible only to the research team. Participants received a small token of appreciation valued at approximately 10 RMB (about 1.5 USD), intended only as a gesture of gratitude without any coercive influence. To address potential distress, participants were informed that they could pause or withdraw from the study at any time. If participants exhibited signs of distress or requested help, they were provided with information about available psychological counseling services for further support.

Statistical analysis

Data were independently entered by two researchers and cross-checked using EpiData 3.1. Then, the data were exported to Excel documents. The data analysis was conducted using Mplus 7.0 and SPSS 23.0. The LPA was conducted using the four dimensions of the Family Resilience Assessment Scale as continuous indicators. Starting from a one-profile solution, more profiles were gradually added. Model selection was based on multiple fit indices, including the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and sample-size adjusted BIC (aBIC), with lower values indicating better model fit [36]. Classification quality was assessed using entropy, with values above 0.80 suggesting good classification accuracy [37]. Additionally, the Lo-Mendell-Rubin (LMR) and bootstrap likelihood ratio tests (BLRT) were used to compare models, with a P-value < 0.05 suggesting the model with more profiles was better [38].

Continuous variables were presented as mean ± standard deviation, while categorical variables were expressed as frequencies and percentages. To explore the relative strengths and weaknesses of each profile across resilience domains, paired-sample t-tests were conducted within each latent profile to compare mean scores across the four FRAS dimensions. Potential confounders were identified based on literature review and theoretical relevance. Variables significantly associated with latent profiles in univariate analyses (P < 0.05) were included in the multivariate logistic regression model to adjust for their potential confounding effects. This approach allowed us to isolate the independent association of each variable with profile membership while accounting for shared variance across predictors. Profile comparisons were conducted using the chi-square test for categorical variables and one-way ANOVA analysis for continuous variables. Variables showing statistical significance were entered into a binary logistic regression model to identify the factors influencing latent family resilience profiles among caregivers of children with cancer. Model-based R² statistics and Hosmer-Lemeshow goodness-of-fit test were reported to evaluate model fit. A two-tailed P-value of < 0.05 was considered statistically significant.

Missing data were observed in three variables: daily caregiving hours (continuous, n = 3), receipt of financial support (categorical, n = 7) and pediatric quality of life (caregiver proxy-report, continuous, n = 24), out of a total of 292 participants. Missing data were minimal (< 10%) and handled using mean imputation for continuous variables and mode imputation for categorical variables.

Results

Demographic and disease-related characteristics of the sample

Of the 332 caregivers approached, 292 agreed to participate and 40 declined due to lack of time or interest, and the response rate was 87.95% (292/332). The average age of the children was 7.28 ± 4.45 while caregivers had an average age of 36.98 ± 6.72. Regarding the types of cancer, 130 children (44.5%) were diagnosed with acute leukemia, 82 children (28.1%) with other hematologic malignancies, and 80 children (27.4%) with solid tumors. Among the children, 80.1% had an education level of primary school or below, with the majority (76.7%) not currently attending school. More than half of children with cancer have siblings (59.2%). Additionally, 66.4% of caregivers were unemployed (Table 1).

Table 1.

Characteristics of the participants (n = 292)

Variables	Frequency (%)
Sociodemographic and clinical characteristics
Patient’s age (years, mean ± SD)	7.28 ± 4.45
Patient’s gender
Male	191 (65.4)
Female	101 (34.6)
Place of residence
Urban	141 (48.3)
Rural	151 (51.7)
Number of siblings
0	119 (40.8)
≥ 1	173 (59.2)
Current school enrollment status
Yes	68 (23.3)
No	224 (76.7)
Patient’s education level
Primary school and below	234 (80.1)
Junior high school	44 (15.1)
High school and above	14 (4.8)
Diagnosis
Acute leukemia	130 (44.5)
Other hematologic malignancies	82 (28.1)
Solid tumors	80 (27.4)
Time since diagnosis
≤ 6 months	191 (65.4)
7 months to 1 year	42 (14.4)
1 to 3 years	48 (16.4)
≥ 3 years	11 (3.8)
Current treatment stage
Pre-chemotherapy	51 (17.5)
During chemotherapy	175 (60.0)
Preparing for the next phase of chemotherapy	43 (14.7)
Other	23 (7.9)
Recurrence status
No	274 (93.8)
Yes	18 (6.2)
Caregiver’s age (years, mean ± SD)	36.98 ± 6.72
Caregiver’s education level
Junior high school and below	124 (42.5)
College degree	111 (38.0)
Bachelor’s degree and above	57 (19.5)
Caregiver’s marital status
Married	273 (93.5)
Single	19 (6.5)
Caregiver’s employment status
No	194 (66.4)
Yes	98 (33.6)
Caregiver’s relationship to the child
Parent	211 (72.3)
Grandparent	62 (21.2)
Other	19 (6.5)
Monthly income
0 ~ 3000 RMB	100 (34.2)
3001 ~ 4999 RMB	87 (29.8)
≥ 5000 RMB	105 (36.0)
Receipt of financial support (government and public support)
Yes	119 (40.8)
No	166 (56.8)
Number of permanent household residents
≤ 3	67 (22.9)
4	103 (35.3)
≥ 5	122 (41.8)
Daily caregiving hours (mean ± SD)	20.18 ± 6.18
Scores of psychosocial scale and quality of life (mean ± SD)
Individual resilience	24.63 ± 7.44
Social support
Friends	18.78 ± 5.18
Others	20.29 ± 4.72
Caregivers’ quality of life
Physical well-being	5.18 ± 2.20
Psychological well-being	4.03 ± 1.66
Social concerns	3.24 ± 1.83
Spiritual well being	6.18 ± 1.74
Pediatric quality of life (caregiver proxy-report)
Pain	64.04 ± 25.10
Nausea	51.83 ± 23.47
Procedural anxiety	41.13 ± 28.94
Treatment anxiety	50.88 ± 28.43
Worry	36.59 ± 31.57
Cognitive problems	54.03 ± 22.12
Perceived physical appearance	61.84 ± 25.02
Communication	57.62 ± 25.66

Note: 1 RMB ≈ 0.14 US dollar

Latent profiles of family resilience

LPA was conducted using the scores from the four dimensions of the Family Resilience Assessment Scale. Four models were tested, and their fit indices are shown in Table 2. As the number of categories increased, the AIC, BIC, and aBIC values decreased gradually. However, the LMR test did not show a statistically significant result for the three-profiles model (P > 0.05), suggesting that adding a third profile did not significantly improve model fit over the two-profile model. In a third profile model, one profile accounted for only 5.5% of the sample (approximately 15 participants out of 292), which is too small to ensure reliable results and may introduce bias. Considering both statistical indicators and theoretical interpretability, the two-profile model was selected as the optimal solution. The analysis showed high classification accuracy, with average posterior probabilities of 0.992 for Profile 1 and 0.952 for Profile 2. These values reflect the likelihood that individuals were correctly assigned to their respective latent profiles based on their observed response patterns, indicating good model fit and clear distinction between classes (Table 3). The deviance value of 0.883 reflects the extent to which the LPA model deviates from a saturated model. Lower values indicate better fit, and this value suggests that the identified profiles sufficiently represent the observed data patterns.

Table 2.

Fit statistics for LPA models

Profile	AIC	BIC	aBIC	Entory	P		Profile probability
					LMR	BLRT
1	1415.473	1444.887	1419.518	——	——	——	——
2	1051.637	1099.435	1058.209	0.958	<0.001	<0.001	0.85959	0.14041
3	936.445	1002.626	945.544	0.951	0.0538	<0.001	0.80479	0.05479	0.14041
4	883.830	968.395	895.457	0.951	0.1028	<0.001	0.05479	0.78082	0.10959

Note: AIC = Akaike Information Criterion, BIC = Bayesian Information Criterion, aBIC = adjusted BIC, LMR = the Lo-Mendell-Rubin adjusted likelihood ratio test, BLRT = bootstrap likelihood ratio test

Table 3.

The probability assigned to latent profiles

Latent profiles	The probability assigned to latent profiles
	P1	P2
P1	0.992	0.008
P2	0.048	0.952

Note: P1 = Low Resources–Low Positivity profile; P2 = High Internal Resilience profile

Figure 1 presented the mean FRAS dimension scores (range 0 ~ 4) for each latent profile, with higher scores indicating greater family resilience. Profile 1 (n = 251, 86%) showed generally lower scores, with significantly lower ratings on “utilizing social and economic resources” and “maintaining a positive outlook” compared to the other dimensions (Table 4). This profile was labeled the Low Resources–Low Positivity profile to reflect its limited external support and diminished optimism. Profile 2 (n = 41, 14%) demonstrated higher overall scores, especially on the three internally focused dimensions: “family communication and problem solving,” “maintaining a positive outlook,” and “making meaning of adversity.” These three dimensions did not differ significantly within the profile (P > 0.05), whereas “utilizing social and economic resources” remained significantly lower than each of these internal dimensions (all P<0.001). This pattern supports the label High Internal Resilience profile, indicating strong internal coping capacity despite limited external resource use.

Fig. 1.

Two latent profiles of family resilience identified by LPA

Table 4.

Paired-sample t-test results for FRAS dimension scores within each latent profile

Dimension Pair (A – B)	Profile 1: Low Resources–Low Positivity (n = 251)		Profile 2: High Internal Resilience (n = 41)
		P	t	P
Communication vs. Social & Economic Resource	13.255	<0.001	5.358	<0.001
Communication vs. Positive Outlook	12.668	<0.001	2.004	0.052
Communication vs. Making Meaning	-0.199	0.842	-0.037	0.971
Social & Economic Resources vs. Positive Outlook	-2.773	0.006	-5.131	<0.001
Social & Economic Resources vs. Making Meaning	-10.058	<0.001	-5.248	<0.001
Positive Outlook vs. Making Meaning	-9.371	<0.001	-1.247	0.219

Comparison of participant characteristics in two latent profiles

Univariate analysis revealed significant differences between the two latent profiles in place of residence, number of siblings, caregiver’s education level, individual resilience, and social support (including support from friends and others). Significant differences were also observed in caregivers’ physical and psychological well-being, as well as in the communication dimension of the child’s quality of life (Table 5).

Table 5.

Differences in participant characteristics in two latent profiles of family resilience

Variables	P1 (n = 251) N (%)	P2 (n = 41) N (%)	χ2/t	Effect size (Cohen’s d / Cramer’s V)	P
Patient’s age (mean ± SD)	7.33 ± 4.57	6.95 ± 3.72	0.255 a	0.085b	0.614
Patient’s gender
Male	163 (85.3)	28 (14.7)	0.175	0.024	0.410
Female	88 (87.1)	13 (12.9)
Place of residence
Urban	113 (80.1)	28 (19.9)	7.644	0.162	0.005
Rural	138 (91.4)	13 (8.6)
Number of siblings
0	93 (78.2)	26 (21.8)	10.145	0.186	0.001
≥ 1	158 (91.3)	15 (8.7)
Current school enrollment status
Yes	58 (85.3)	10 (14.7)	0.032	0.011	0.497
No	193 (86.2)	31 (13.8)
Patient’s education level
Primary school and below	198 (84.6)	36 (15.4)	2.421	0.100	0.307
Junior high school	39 (88.6)	5 (11.4)
High school and above	14 (100.0)	0 (0.0)
Diagnosis
Acute leukemia	111 (85.4)	19 (14.6)	0.217	0.027	0.909
Other hematologic malignancies	70 (85.4)	12 (14.6)
Solid tumors	70 (87.5)	10 (12.5)
Time since diagnosis
≤ 6 months	163 (85.3)	28 (14.7)	0.244	0.050	0.865
6 months to 1 year	36 (85.7)	6 (14.3)
1 to 3 years	43 (89.6)	5 (10.4)
≥ 3 years	9 (81.8)	2 (18.2)
Current treatment stage
Pre-chemo	44 (86.3)	7 (13.7)	0.368	0.062	0.776
During chemo	149 (85.1)	26 (14.9)
Preparation for next chemo	39 (90.7)	4 (9.3)
Other	19 (82.6)	4 (17.4)
Recurrence status
No	237 (86.5)	37 (13.5)	1.064	0.060	0.235
Yes	14 (77.8)	4 (22.2)
Caregiver’s age (mean ± SD)	36.84 ± 6.32	37.88 ± 8.85	0.845 a	0.155 b	0.359
Caregiver’s education level
Junior high school and below	114 (91.9)	10 (8.1)	7.844	0.163	0.020
College degree	93 (83.8)	18 (16.2)
Bachelor’ s degree and above	44 (77.2)	13 (22.8)
Caregiver’s marital status
Married	233 (85.3)	40 (14.7)	1.297	0.067	0.222
Single	18 (94.7)	1 (5.3)
Caregiver’s employment status
No	171 (88.1)	23 (11.9)	2.287	0.089	0.093
Yes	80 (81.6)	18 (18.4)
Caregiver’s relationship to the child
Parent	182 (86.3)	29 (13.7)	0.425	0.054	0.654
Grandparent	54 (87.1)	8 (12.9)
Other	15 (78.9)	4 (21.1)
Monthly income
0 ~ 3000 RMB	88 (88.0)	12 (12.0)	3.311	0.110	0.184
3001 ~ 4999 RMB	78 (89.7)	9 (10.3)
≥ 5000 RMB	85 (81.0)	20 (19.0)
Receipt of financial support (government and public support)
Yes	104 (87.4)	15 (12.6)	0.526	0.043	0.292
No	140 (84.3)	26 (15.7)
Number of permanent household residents
≤ 3	56 (83.6)	11 (16.4)	3.244	0.103	0.209
4	85 (82.5)	18 (17.5)
≥ 5	110 (90.2)	12 (9.8)
Daily caregiving hours (mean ± SD)	20.03 ± 6.12	21.07 ± 5.98	1.005 a	0.168 b	0.317
Individual resilience (mean ± SD)	23.72 ± 6.85	30.20 ± 8.49	29.298 a	0.913 b	<0.001
Social support (mean ± SD)
Friends	18.18 ± 5.04	22.46 ± 4.47	26.180 a	0.862 b	<0.001
Others	19.77 ± 4.48	23.51 ± 4.90	23.905 a	0.823 b	<0.001
Caregivers’ QOL (mean ± SD)
Physical well-being	5.07 ± 2.19	5.84 ± 2.19	4.395 a	0.353 b	0.037
Psychological well-being	3.92 ± 1.61	4.76 ± 1.76	9.331 a	0.515 b	0.002
Social concerns	3.17 ± 1.81	3.66 ± 1.91	2.526 a	0.270 b	0.113
Spiritual well being	6.13 ± 1.69	6.48 ± 2.03	1.404 a	0.195 b	0.237
Pediatric QOL (mean ± SD)
Pain	63.99 ± 25.27	64.33 ± 24.46	0.006 a	0.013 b	0.936
Nausea	51.07 ± 23.09	56.06 ± 24.85	1.573 a	0.213 b	0.211
Procedural anxiety	41.07 ± 28.25	41.42 ± 32.86	0.005 a	0.012 b	0.943
Treatment anxiety	52.17 ± 27.20	43.50 ± 34.13	3.091 a	-0.306 b	0.075
Worry	36.51 ± 31.76	36.99 ± 30.86	0.008 a	0.015 b	0.929
Cognitive problems	53.10 ± 21.91	59.22 ± 22.81	2.677 a	0.278 b	0.103
Perceived physical appearance	61.19 ± 24.62	65.45 ± 27.17	1.006 a	0.170 b	0.317
Communication	56.28 ± 25.57	65.04 ± 25.15	4.091 a	0.344 b	0.044

Note: 1 RMB ≈ 0.14 US dollar; Chemo = chemotherapy; QOL = quality of life; ^a indicates results of t-test; ^b Cohen’s d for continuous variables

Predictors of family resilience in two latent profiles

To adjust for potential confounders, we included in the multivariate model all variables that were significantly associated with latent profiles in univariate analyses (P < 0.05), based on both statistical and theoretical considerations. Collinearity diagnostics showed that all variables have VIF values less than 3, indicating no multicollinearity issue. Table 6 showed the results of binary logistic regression analysis, using the Low Resources–Low Positivity profile as the reference group. Families with only one child were more likely to belong to the Low Resources–Low Positivity profile (OR = 3.184, 95% CI: 1.437 ~ 7.057, P = 0.004,). Higher levels of individual resilience were associated with increased odds of being assigned to the High Internal Resilience profile (OR = 1.095, 95% CI: 1.028 ~ 1.165, P = 0.009). The model explained a substantial proportion of variance in profile membership (Nagelkerke R² = 0.325). The Hosmer–Lemeshow goodness-of-fit test yielded χ² (8) = 15.885, P = 0.044, suggesting marginal misfit. However, given the known sensitivity of this test to large sample sizes, and the model’s overall performance, the fit was considered acceptable for interpretation.

Table 6.

Binary logistic regression analysis of latent profiles of family resilience among caregivers of children with cancer

Variables	β	SE	Wald χ²	OR	95% CI	P
Residence (Ref: Rural)
Urban	0.443	0.448	0.981	1.558	0.648 ~ 3.745	0.322
Number of siblings (Ref: 0)
≥ 1	1.158	0.406	8.135	3.184	1.437 ~ 7.057	0.004
Caregiver’s education level (Ref: Junior high school and below)
College degree	-0.005	0.497	<0.001	0.995	0.376 ~ 2.636	0.992
Bachelor’s degree and above	0.285	0.604	0.223	1.330	0.407 ~ 4.347	0.637
Individual resilience	0.090	0.032	7.997	1.095	1.028 ~ 1.165	0.005
Physical well-being	-0.094	0.113	0.682	0.911	0.729 ~ 1.137	0.409
Psychological well-being	0.124	0.161	0.595	1.132	0.826 ~ 1.552	0.440
Social support (friends)	0.052	0.070	0.555	1.053	0.919 ~ 1.207	0.456
Social support (others)	0.109	0.076	2.094	1.116	0.962 ~ 1.294	0.148
Child’s QOL: Communications	0.008	0.008	0.934	1.008	0.992 ~ 1.023	0.334

Note: Nagelkerke R² = 0.325; Hosmer-Lemeshow goodness-of-fit test: χ² (8) = 15.885, P = 0.044

Discussion

Guided by the Social Ecological Model, this study identified two distinct family resilience profiles among caregivers of children with cancer: the Low Resources–Low Positivity profile and the High Internal Resilience profile. This reflected different configurations of internal strengths and external resources. The more prevalent Low Resources–Low Positivity profile showed both diminished external support and a weaker positive outlook, whereas the High Internal Resilience profile demonstrated strong intrafamilial coping capacities despite limited access to social and economic support. These findings highlight the multidimensional and context-dependent nature of family resilience and align with prior research emphasizing the interplay between internal family processes and broader systemic factors. Notably, caregivers who rely primarily on internal strengths may be vulnerable to emotional exhaustion over time if contextual support remains insufficient. These results highlight the need for tailored interventions that enhance both internal resilience and external support systems. They also support multi-level interventions that strengthen individual resilience, consider family dynamics, and address systemic gaps in social and economic support.

The consistently low scores in the “utilizing social and economic resources” dimension across both profiles reflect a common challenge faced by families of children with cancer—insufficient access to comprehensive social support. Families often face significant logistical and financial challenges [39], especially when traveling long distances for treatment. To reduce the strain of travel, many families rent accommodations near hospitals, which can isolate them from their local communities [40]. Additionally, stigma or privacy concerns surrounding a cancer diagnosis may lead some parents to withhold information, further limiting informal social support [41]. Notably, the “High Internal Resilience” profile, despite demonstrating strong internal coping and family support, also showed lower utilization of external resources. This may be influenced by cultural factors that emphasize family privacy, self-reliance, and stigma associated with seeking outside assistance, leading these families to depend more on internal strengths rather than formal support systems [42]. These findings underscore the need to enhance support systems that bolster family resilience. At the mesosystem level, strengthening peer support groups, caregiver networks, and community-based services can help families build emotional and practical resources. At the macrosystem level, policy efforts should focus on expanding insurance reimbursement, streamlining financial aid procedures, and developing more inclusive social welfare programs [43].

In the High Internal Resilience profile, the score for “maintaining a positive outlook” was comparable to those of “family communication and problem-solving” and “empowering the meaning of adversity.” In contrast, the Low Resources–Low Positivity profile showed significantly lower scores for “maintaining a positive outlook,” although these scores remained slightly but significantly higher than those for “utilizing social and economic resources”. This suggests that families with lower resilience may struggle to sustain a positive outlook, likely due to heightened uncertainty about disease outcomes and insufficient resources and emotional support. These challenges hinder their ability to translate optimism into effective coping behaviors [44]. Conversely, in High Internal Resilience families, a positive outlook not only provides emotional support but also motivates proactive behaviors, such as seeking external resources and adopting adaptive coping strategies. In the Chinese culture context, proverbs such as “When one door closes, another opens” and “Success depends on human effort” may reflect a belief in perseverance and self-efficacy [45]. Such cultural values may influence how families respond to adversity, including serious challenges like childhood cancer. While our study did not directly assess cultural beliefs, prior literature suggests that such value systems can influence coping orientations [46].

However, a positive outlook alone is insufficient to address the underlying causes of low resilience. Effective interventions should combine psychological and informational support to help caregivers address their fears, treatment uncertainties, and increase their confidence [47]. At the mesosystem level, multidisciplinary care teams—integrating physicians, psychologists, and social workers—are essential for delivering comprehensive, family-centered support.

In the High Internal Resilience profile, high scores on both “family communication and problem-solving” and “finding meaning in adversity” suggest that effective intrafamily coordination and the capacity to derive meaning from hardship are key processes in fostering resilience. Good communication enable families to resolve conflicts, strengthen emotional bonds, and better manage stress [48]. In contrast, families with lower resilience often experience communication difficulties, leading to misunderstandings and reduced adaptability. Additionally, some families may suppress their feelings to maintain peace and avoid causing trouble for others, which can make open communication more difficult [49]. Therefore, efforts to improve family communication should encourage honesty and open expression of feelings while respecting individual comfort levels. Balancing emotional expression with family harmony can enhance collaboration and resilience.

Finding meaning in adversity means perceiving challenges as opportunities for growth, learning, and strengthening family ties. This positive appraisal transforms suffering into hope and strength, supporting families in facing difficulties together [50]. Conversely, inability to find meaning or tendencies toward self-blame may increase stress and impede recovery [51]. Supportive measures should facilitate emotional openness, acceptance of circumstances, and help families identify shared values and goals to build meaning in adversity.

The findings also revealed that individual caregiver resilience—a microsystem-level factor— was a significant predictor of family resilience. This finding aligns with existing literature, highlighting the crucial role of caregivers’ resilience in managing caregiving demands and fostering the family’s capacity to adapt to adversity [52]. These results underscore the need for healthcare interventions that focus on enhancing caregivers’ psychological resilience through targeted support such as counseling services, coping skills training, and stress management programs.

Interestingly, while univariate analysis suggested that families with only children were more prevalent in the High Internal Resilience profile, multivariate analysis showed that in mesosystem level, families with multiple children were more likely to demonstrate higher resilience. This may due to the emotional and practical support that siblings provide, which can reduce the caregiving burden [53]. In contrast, families with only children may face greater emotional pressure, as the sole child often bears the family’s hopes and expectations. However, the impact of family structure, specifically the number of children, on family resilience remains underexplored in literature. Further studies are needed to clarify the relationship between family composition and family resilience, especially in the context of childhood cancer. The findings can inform the development of tailored interventions for different family structures.

While place of residence, caregivers’ education, psychological and physical well-being, social support, and child’s communication functioning (as reported by caregivers) were significant in univariate analysis, they lost significance in the multivariate model. This may be partially explained by overlapping variance among predictors, where shared explanatory power reduces individual effects when variables are analyzed together. It may also reflect suppression effects, where the inclusion of certain variables (e.g., individual resilience) clarifies the relationship between predictors and outcome by accounting for irrelevant variance. Additionally, some factors may exert indirect influences through more proximal variables. For example, higher education may enhance caregivers’ access to information and coping resources [54], which in turn strengthen individual resilience and ultimately promote family resilience. Similarly, residing in rural areas may limit service access and increase financial stress [55], which can impair coping capacity. Caregivers with better mental and physical health may be more likely to manage caregiving demands effectively, indirectly strengthening family resilience [56]. Moreover, the effects of social support may depend on family structure; having more than one child can provide additional emotional support and distribute caregiving responsibilities, fostering a more adaptive environment [57]. The communication dimension, as reported by caregivers, reflected the child’s functional and interpersonal abilities rather than a modifiable caregiver or family-level factor. Its significance in univariate analysis may reflect an overall better family dynamic or child functioning in more resilient families. However, once individual resilience and family structure were considered, this association becomes less prominent. These findings suggest that the observed univariate associations may be mediated or confounded by microsystem-level factors such as caregiver resilience and mesosystem-level factors like family structure. Future research using path analysis or structural equation modeling could further elucidate these complex interrelationships [58].

To better interpret the findings, it is essential to acknowledge the limitations of the current study. First, the cross-sectional design limits causal inference; associations between SEM-level factors and family resilience profiles are correlational. Longitudinal studies are needed to examine changes over time and clarify the directionality of these relationships. Second, convenience sampling from three hospitals may limit generalizability, and potential clustering effects by hospital were not addressed; future studies should consider multilevel modeling to improve accuracy. Third, reliance on self-reported data from only primary caregivers may introduce recall and social desirability biases and provide an incomplete picture of family dynamics by excluding other family members. Fourth, although guided by the SEM, this study focused mainly on individual and family-level factors, with limited assessment of community and societal influences. Finally, despite good model fit in latent profile analysis, possible classification errors remain, suggesting that advanced analytical methods could enhance future research.

Clinical Implications.

The predominance of the “Low Resources–Low Positivity” profile among families of children with cancer highlights urgent needs across multiple levels of the caregiving ecosystem. Health systems and policymakers should invest in preventive psychosocial services that are both widely accessible and affordable, especially in resource-limited settings. These efforts should include systematic screening for caregiver distress, individualized psychological support, and expanded access to community-based services. Policy reforms are also needed to address structural barriers such as financial strain, employment constraints, and gaps in social protection—through initiatives like flexible work policies, housing assistance for displaced families, and improved coverage for psychosocial care [59].

Guided by the Social Ecological Model, our findings further suggest tailored intervention targets across different levels. At the microsystem level, caregiver psychological resilience was a key determinant of family functioning. Healthcare providers should incorporate regular mental health assessments and offer targeted programs—such as cognitive-behavioral therapy, mindfulness training, and expressive writing—to support coping [60]. At the mesosystem level, family structure (e.g., having an only child) may heighten caregiver burden; thus, services such as peer-support groups, sibling-inclusive care planning, and family counseling can strengthen intra-family resilience [61]. At the macrosystem level, the universally low utilization of external resources underscores the need for broader social and policy-level supports to close existing service gaps.

Conclusion

Using latent profile analysis within the SEM framework, this study identified two distinct profiles of family resilience among caregivers of children with cancer: one marked by low resources and outlook, and another characterized by strong internal coping strategies. Across both profiles, poor utilization of social and economic resources emerged as a key area for improvement. Individual caregiver resilience and having more than one child were significant predictors of higher family resilience. These findings call for multilevel interventions: enhancing stress coping, promoting sibling involvement and peer support, and expanding financial and community support. Future research is suggested to use longitudinal, multi-informant designs and develop SEM-guided interventions to better support affected families.

Abbreviations

SEM

Social Ecological Model

LPA

Latent profile analysis

RMB

Renminbi (1 RMB = 0.14 USD)

AIC

Akaike information criterion

BIC

Bayesian information criterion

aBIC

ajusted BIC

LMR

Lo-mendell-rubin likelihood ratio

BLRT

Bootstrap likelihood ratio test

Chemo

Chemotherapy

QOL

Quality of life

Author contributions

B.L.: Conceptualization, Software, Writing - Original Draft, Formal Analysis, Writing - Review & Editing; D.S.: Conceptualization, Methodology, Formal Analysis, Writing - Review & Editing; S.P.: Conceptualization, Methodology, Interpretation of Data, Writing - Review & Editing; L.W.: Conceptualization, Writing - Review & Editing; L.X.W.: Resources, Data Curation, Writing - Review & Editing; X.Y.: Validation, Interpretation of Data, Writing - Review & Editing; J.T.X.: Data Curation, Validation, Interpretation of Data, Writing - Review & Editing; H.Z.: Visualization, Interpretation of Data, Writing - Review & Editing; B.X.Y.: Supervision, Resources, Writing - Review & Editing, Funding Acquisition; Q.L.: Supervision, Resources, Writing - Review & Editing, Funding Acquisition.

Funding

This study was funded by the National Natural Science Foundation of China (No. 82103025).

Data availability

The data underlying this article may be shared upon reasonable request from the corresponding author.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of Wuhan University School of Medicine (IRB2022018). All procedures were conducted in accordance with relevant guidelines and regulations and adhered to the ethical principles outlined in the Declaration of Helsinki. Written informed consent was obtained from all participants prior to data collection. Participants were informed that their involvement was voluntary and that they could withdraw from the study at any time without any consequences.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.