Trajectories and determinants of family resilience following radical surgery for gynecological malignancies: A prospective longitudinal study

Abstract

Objective

This study aimed to examine the longitudinal trajectories of family resilience within six months after radical surgery for gynecological malignancies and to identify predictors associated with these trajectories.

Methods

A total of 243 patients with gynecological malignancies undergoing radical hysterectomy were recruited through convenience sampling at a tertiary hospital in Guangzhou, China, between May 2024 and March 2025. Family resilience was assessed at four time points (preoperatively, and at 1, 3, and 6 months post-surgery) using the Chinese version of the Family Resilience Assessment Scale. Additional measures included the Perceived Social Support Scale, the Perceived Stress Scale-14, and the Family APGAR Scale. Growth mixture modeling (GMM) was applied to identify distinct family resilience trajectories, and binary logistic regression was used to determine predictors of trajectory membership.

Results

Mean family resilience scores were 71.40 ​± ​9.84 (baseline), 68.42 ​± ​10.66 (1 month), 68.99 ​± ​12.43 (3 months), and 68.65 ​± ​13.65 (6 months). Two distinct trajectory classes emerged: a “high-level slow-increase” group and a “low-level continuous-decline” group. Higher baseline monthly family income and greater perceived social support predicted membership in the high-level slow-increase group, whereas higher baseline perceived stress predicted membership in the low-level continuous-decline group..

Conclusions

Family resilience trajectories following radical surgery for gynecological malignancies demonstrate significant heterogeneity. Early identification of individuals with low family income, low perceived social support, and high perceived stress is essential to guide timely, targeted interventions that promote adaptive family functioning.

Introduction

Gynecological malignancies, such as cervical, endometrial and ovarian cancers, pose a serious global threat to women's health, with clear trends for a rising incidence and younger onset.¹ While radical hysterectomy, the standard treatment for early-stage gynecological malignancies, has improved five-year relative survival rates to 50% to 82%,^2,3 the extensive tissue resection and nerve damage associated with this technique often lead to organ dysfunction and various potential complications, including lymphedema, bladder dysfunction and sexual dysfunction.4, 5, 6 These complications can trigger negative emotions in patients, including anxiety, depression, stigma and social isolation.⁷ This dual burden of physical and psychological distress results in a persistently low quality-of-life for patients in the years following surgery.⁸ Consequently, there is a critical need for research to identify ways to help patients to cope with the challenges posed by surgery and to successfully reintegrate into the family and society with a high quality-of-life.

Within this context, the family emerges as a paramount source of support for women navigating the complex aftermath following gynecological cancer surgery. The efficacy of this support largely depends on the level of family resilience.⁹ Family resilience refers to the characteristics and abilities of a family to flexibly utilize both internal and external resources to achieve positive adjustment and adaptation when facing crises or stress.¹⁰ When a family encounters a crisis event, such as a member being diagnosed with cancer, family resilience acts as a protective resource, encouraging the patient and family members to respond positively, communicate constructively, maintain a stable family structure, and ensure normal functionality.^9,11 This helps them to better buffer the trauma and pain caused by adversity and promotes adaptation and recovery.¹¹ Previous research has shown that the level of family resilience perceived by cancer patients is significantly and positively correlated with their own sense of meaning in life.¹² Patients with higher levels of family resilience are more likely to achieve meaning in their disease experience, that is, to see the positive significance and growth generated by adverse events.¹² Crucially, studies have indicated that due to the loss of reproductive organs and societal stigmatization surrounding women's reproductive health issues, the levels of family resilience among patients undergoing surgery for gynecological cancer are relatively low, resulting in diminished coping capabilities within this population.¹³

As a psychological trait, family resilience is influenced by multiple factors. Walsh's family resilience model identifies three key factors: an individual's perception and understanding of stress, the ability to maintain family functional stability in adversity, and the availability of social and economic resources.¹⁴ A systematic review of cancer patients and their family members further confirmed that family resilience is influenced by demographic and clinical factors (such as age, residence, family monthly income and marital status), personal strengths and resources, family stressors, family coping strategies (such as family functioning and social support), and the outcomes of family adaptive resilience.¹⁵ Generally, higher levels of family resilience are associated with reduced stress, stronger social support, and healthier family functioning. However, despite synthesizing 21 high-quality studies, the conclusions of this previous systematic review have limited generalizability due to restrictive inclusion criteria and population specificity.¹⁵ Further research is still needed to investigate the diversity of family resilience and its influencing factors across different patient populations.

It is important to consider that family resilience is not a fixed ability but rather a dynamic quality that changes with the family's interactions with the outside world.¹⁶ This view was confirmed in post-operative breast cancer patients¹⁷ and families with premature infants.¹⁸ Moreover, family systems theory emphasizes that each family is unique, which leads to significant heterogeneity in the patterns of family resilience.¹⁴ However, previous research on family resilience in the gynecological malignancy population were predominantly limited to cross-sectional surveys or qualitative studies.^13,15 While these studies have established the importance of family resilience and identified relevant influencing factors, their methodologies offer limited insight into how it evolves over time. Specifically, the longitudinal adaptive process and the potential divergence of resilience trajectories remain unexplored. This gap is particularly noteworthy given that the postoperative period involves ongoing adjustment and recovery. Furthermore, considerable variation exists in disease progression, treatment response, and psychological outcomes following radical surgery.^3,6,19 Thus, investigating how family resilience changes during the early postoperative phase and identifying distinct trajectory patterns is essential.

The growth mixture model (GMM), as an established statistical model, integrates the characteristics of growth curve modeling and latent class analysis, and can effectively identify latent heterogeneity and latent classes within data, thereby revealing the growth trajectories and corresponding characteristics of different individuals.²⁰ Therefore, in the present study, we conducted a 6-month follow-up study of patients undergoing radical surgery for gynecological malignancies, and utilized GMM to identify and describe the heterogeneous trajectory categories of their family resilience. Based on the core dimensions of the Walsh family resilience model,¹⁴ we investigated the predictive role of sociodemographic characteristics, disease-related information, stress perception, perceived social support, and family functioning on the trajectory of family resilience. Our overall aim was to provide theoretical and empirical support for a smooth transition during the postoperative recovery period of patients with gynecological malignancies.

Methods

Study design

This study was a prospective and longitudinal study and the research process complied strictly with the STROBE checklist (Appendix Table 1).

Participants and setting

Convenience sampling was used to recruit patients who underwent radical surgery for gynecological malignancies at a tertiary care hospital in Guangzhou, China, from May 2024 to March 2025.

The inclusion criteria were as follows: (1) age ≥ 18 years; (2) first diagnosed with cervical, ovarian or endometrial cancer and requiring radical hysterectomy (these three types of cancer were selected because they are the top three most common gynecological malignancies, representing a significant proportion of cases that require surgical intervention for treatment¹); (3) the provision of informed consent and voluntary participation in the study. The exclusion criteria were: (1) severe mental or cognitive impairments hindering comprehension or expression; (2) other serious comorbidities (e.g., severe cardiac, liver, or renal failure).

According to the sample size estimation method advocated by Kendall, the sample size should be at least 5–10 times the number of study variables.²¹ Given that the present study involved 28 variables and considering a 20 percent attrition rate, the estimated sample size was at least 175 participants. However, when applying the Bayesian information criterion (BIC) as the primary indicator for model selection,²² the minimum sample size required for GMM with four or more follow-up time points was 200 cases. Therefore, the required sample size for this study was at least 200 participants.

Data collection

Based on the standard follow-up time windows recommended by the national comprehensive cancer network (NCCN) guidelines for postoperative gynecological malignancies,² this study established the time points for investigation as preoperative (T0), 1 month post-surgery (T1), 3 months post-surgery (T2), and 6 months post-surgery (T3). Another important reason for this choice was that, through a literature review and clinical investigations,^4,7,8 it was found that patients within six months post-surgery faced multidimensional health challenges, including physiological reconstruction of function, initiation of adjuvant therapy, frequent emotional fluctuations, and difficulties in role adaptation.

Before data collection began, all researchers involved in data collection (Zhang X and Li X) underwent standardized training, which included standardized procedures for questionnaire distribution during baseline surveys and the standardization of various follow-up protocols, ensuring consistency in their methods.

The procedures for data collection were as follows. First, patients were approached in the gynecology ward once clinicians had determined that radical hysterectomies were required. Second, all patients who met the inclusion criteria and voluntarily agreed to participate in the study signed an informed consent form and provided their contact information (e.g., telephone number). At this time, the patients completed paper-based questionnaires covering demographic characteristics, disease-related information, family resilience, perceived social support, perceived stress, and family functioning (T0). The specific time for the questionnaire survey was determined by the patient, and it took approximately 10–20 minutes to complete. To ensure the completeness of the questionnaire, the researchers were always present and checked for missed items immediately after completion. Third, the researchers assessed family resilience levels at 1-month post-surgery (T1), 3-months post-surgery (T2), and 6-months post-surgery (T3) using the Chinese Version of the Family Resilience Assessment Scale through outpatient visits or telephone follow-ups.

Measurement tools

General information questionnaire

Collected the sociodemographic characteristics of participants, including age, place of residence, educational level, marital status, employment status, family monthly income, primary caregiver, number of caregivers, and payment method for medical expenses. Through the hospital's medical record system, collected the disease-related information of participants, including cancer type, stage, family history, chronic diseases, and whether chemotherapy is needed.

The Chinese version of the Family Resilience Assessment Scale

The Chinese version of the Family Resilience Assessment Scale (C-FRAS) was used to evaluate family resilience. The scale was developed by Bu based on the Walsh's family resilience framework, incorporating key characteristics of traditional Chinese culture.²³ The scale consisted of four dimensions: tenacity, harmony, openness, and supportiveness, with 20 items in total. It used a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). The total score ranged from 20 to 100, with higher scores indicating greater levels of family resilience. The scale had been validated for use in female cancer populations and demonstrated good reliability and validity.¹² In this study, the Cronbach's α for the C-FRAS was 0.910.

The Perceived Social Support Scale

The Perceived Social Support Scale (PSSS), developed by Zimet, was used to assess perceived social support.²⁴ The scale included three dimensions: family support, friend support, and other support, with a total of 12 items. It used a seven-point scoring system, ranging from 1 to 7, from low to high. The total score ranged from 12 to 36 (indicating a low support status), 37 to 60 (indicating a moderate support status) and 61 to 84 (indicating a high support status). A higher total score indicated a higher level of perceived social support. The PSSS had demonstrated good reliability and validity internationally,²⁴ and in this study, the Cronbach's α of this scale was 0.965.

The Perceived Stress Scale-14

The Perceived Stress Scale-14 (PSS-14), developed by Cohen,²⁵ was used to assess perceived stress. It included two dimensions: feelings of loss of control and tension, with 14 items in total. Each item was scored on a scale from 0 (never) to 4 (very often), and 7 items (4, 5, 6, 7, 9, 10, 13) were reverse-scored. The total score ranged from 0 to 56, with higher scores indicating higher levels of perceived stress. The scale had shown good reliability and validity.²⁵ In this study, the Cronbach's α for the PSS-14 was 0.937.

The Family APGAR Scale

The Family APGAR Scale, developed by Smilkstein,²⁶ assessed five dimensions of family functioning: adaptation, participation, growth, emotional, and problem-solving. Each item represented a single dimension, using a three-point rating scale. The total score ranged from 0 to 10, with higher scores indicating better family functioning. Scores of 8–10 indicated good family functioning, scores of 4–7 indicated moderate dysfunction, and scores of 0–3 indicated severe dysfunction. This scale had been widely used internationally and had good psychometric properties.²⁶ In this study, the scale's Cronbach's α was 0.829.

Data analysis

Data analysis in this study was conducted using SPSS version 25.0 and Mplus version 8.0. The normality of continuous variables was verified via the Shapiro–Wilk test and Kolmogorov–Smirnov test (Appendix Table 2). Continuous variables with a normal distribution were expressed as mean and standard deviation, while those that were not normally distributed were represented by median and interquartile range (IQR). Categorical variables were presented as frequencies and percentages. Harman's single-factor test was conducted to assess common method bias. One-way analysis of variance (ANOVA), Chi-squared tests, Fisher's exact test and Mann–Whitney U test were used to explore whether there were differences in baseline data between the follow-up group and the lost to follow-up group. The significance level for all the above methods was set at 0.05.

The general estimation equation (GEE) was used to analyze the overall trend of family resilience, while GMM was employed to explore the trajectory characteristics of family resilience over time. The analysis included data from 210 patients who completed all follow-up visits, with parameters estimated using the maximum likelihood robust estimator (MLR). When values were missing from the dataset, we used Mplus software and the full information maximum likelihood (FIML) estimator to fit the model. The model evaluation indices were as follows: (1) Information criteria: Akaike information criterion (AIC), Bayesian information criterion (BIC), and adjusted bayesian information criterion (aBIC). Smaller values of AIC, BIC, and aBIC were considered to indicate a better model fit. (2) Likelihood ratio test indicators: These tests were used to compare the fit of latent class models, including the Lo-Mendell-Rubin likelihood ratio test (LMR) and the bootstrapped likelihood ratio test (BLRT). A P-value of < 0.05 indicated that dividing the model into k categories fits significantly better than k-1 categories. (3) Classification accuracy indicators: An entropy value closer to 1 was considered to imply more accurate classification. Based on all these indices, we ultimately, selected the best-fitting model. Subsequently, stratified analysis was utilized to explore the impact of tumor heterogeneity.

Finally, univariate analysis (χ² test, Fisher's exact test, one-way ANOVA, or Mann–Whitney U test) was used for the preliminary screening of predictors of changes in family resilience trajectories based on baseline data. To avoid missing potential predictive variables, the significance level for the univariate analysis was set at 0.1. To identify predictors of trajectories, significant factors in above analysis were entered into binary logistic regression as independent variables and trajectories classifications were dependent variables (P ​< ​0.05).

Ethical considerations

This study was approved by the Ethics Committee of Southern Hospital, Southern Medical University (Approval No. NFEC-2024-479). Before the study commenced, all patients were informed about the purpose of the research, the data collection process, the benefits and risks of participation, confidentiality, and other related details, and they provided written informed consent. Throughout the study, patients had the right to request termination or withdrawal at any time. If a patient experienced negative emotions during the study, we terminated the study and, if necessary, contacted a professional counseling team for the patient.

Results

Participant characteristics

At the beginning of the study, we recruited 264 patients who met the inclusion and exclusion criteria. Of these, 21 patients refused to participate, resulting in a response rate of 92 percent. A total of 243 patients agreed to participate and signed the informed consent form, and 210 patients completed all follow-up visits (Fig. 1). Only patients who completed the follow-up were included in our final analysis. A comparison of the baseline information between the follow-up group and the lost-to-follow-up group revealed no statistically significant differences, with specific data presented in Table 1.

Fig. 1.

Flow chart of data collection of participants.

Table 1.

Sociodemographic and disease-related characteristics of the participants (N ​= ​243).

Variables	Follow-up group (n ​= ​210)	Lost to follow-up group (n ​= ​33)	χ2/F/U	P
	Mean ​± ​SD/n (%)/ Median (IQR)	Mean ​± ​SD/n (%)/ Median (IQR)
Age(years)a	48.80 ​± ​8.26	49.85 ​± ​8.28	0.459b	0.499
Place of residence			0.045c	0.831
Urban area	74 (35.2%)	11 (33.3%)
Rural area	136 (64.8%)	22 (66.7%)
Marital status			1.133d	0.287
Married	185 (88.1%)	31 (93.9%)
Unmarried/divorced/widowed	25 (11.9%)	2 (6.1%)
Educational level			2.250c	0.522
Primary school and less	85 (40.5%)	12 (36.4%)
Junior middle school	42 (20.0%)	10 (30.3%)
Senior middle school	47 (22.4%)	5 (15.2%)
Bachelor's degree or higher	36 (17.1%)	6 (18.2%)
Working condition			1.117d	0.553
Employed	82 (39.0%)	10 (30.3%)
Retired	20 (9.5%)	4 (12.1%)
Unemployed	108 (51.4%)	19 (57.6%)
Family monthly income (RMB)			3.554c	0.169
> ​10,000	45 (21.4%)	3 (9.1%)
5000–10,000	74 (35.2%)	16 (48.5%)
< ​5000	91 (43.3%)	14 (42.4%)
Primary caregiver			1.636d	0.919
Spouse	106 (50.5%)	17 (51.5%)
Children	45 (21.4%)	8 (24.2%)
Parents	7 (3.3%)	0 (0.0%)
Relatives	36 (17.1%)	5 (15.2%)
Siblings	9 (4.3%)	1 (3.0%)
Others	7 (3.3%)	2 (6.1%)
Number of caregivers			0.236c	0.627
One	105 (50.0%)	18 (54.5%)
Multiple	105 (50.0%)	15 (45.5%)
Method of payment for medical expenses			0.455d	0.935
Health insurance	78 (37.1%)	11 (33.3%)
Public health care	9 (4.3%)	1 (3.0%)
Out-of-pocket	23 (11.0%)	3 (9.1%)
Social security	100 (47.6%)	18 (54.5%)
Type of cancer			4.741c	0.093
Cervical cancer	84 (40.0%)	12 (36.4%)
Ovarian cancer	53 (25.2%)	14 (42.4%)
Endometrial cancer	73 (34.8%)	7 (21.2%)
Cancer stage			0.283d	0.982
Ⅰ	109 (51.9%)	17 (51.5%)
Ⅱ	63 (30.0%)	11 (33.3%)
Ⅲ	34 (16.2%)	5 (15.2%)
Ⅳ	4 (1.9%)	0 (0.0%)
Family history			0.017d	0.895
Yes	14 (6.7%)	2 (6.1%)
No	196 (93.3%)	31 (93.9%)
Chronic diseases			1.331c	0.249
Yes	56 (26.7%)	12 (36.4%)
No	154 (73.3%)	21 (63.6%)
Chemotherapy			1.800c	0.180
Yes	76 (36.2%)	8 (24.2%)
No	134 (63.8%)	25 (75.8%)
Perceived social support	53 (18)	49.39 ​± ​11.22	1.445e	0.149
Perceived stress	31 (22)	30.82 ​± ​11.39	0.970e	0.332
Family functioning	6 (3)	6 (2)	1.264e	0.206

SD, standard deviation; IQR, interquartile range.

^aThe result of the homogeneity of variance test showed P ​= ​0.725.

^bOne-way ANOVA.

^cChi-square Pearson test.

^dFisher's exact test.

^eMann-Whitney U test.

Common method bias

Analysis revealed that at T0, six common factors with eigenvalues > 1 were identified, with the first factor accounting for 28.255 percent of the total variance. At T1, four common factors with eigenvalues > 1 were extracted, with the first factor explaining 32.875 percent of the variance. By T2, three factors with eigenvalues > 1 were recognized, with the first factor contributing 35.036 percent to the variance. Finally, at T3, two factors with eigenvalues > 1 were detected, with the first factor explaining 37.358 percent of the variance. Since more than one factor was extracted at each time point, and the variance explained by the first factor was below the critical threshold of 40 percent²⁷ these findings suggest that common method bias was not a serious concern in this study.

Different trajectories of family resilience

The family resilience scores at the four time points in this study were 71.40 ​± ​9.84 (T0), 68.42 ​± ​10.66 (T1), 68.99 ​± ​12.43 (T2), and 68.65 ​± ​13.65 (T3), respectively. The GEE results indicated statistically significant differences in family resilience scores across the four time points (Wald χ² ​= ​277.782; P ​< ​0.001).

Based on this, we fitted six GMM models and compared indices of model fit, as detailed in Table 2. As the number of categories increased from 1 to 6, the values of Log-Likelihood (LL), AIC, BIC, and aBIC gradually decreased. The BLRT test was statistically significant for all six models, while the LMR test showed statistical significance only for the two-class and five-class models. However, the five-class model featured two classes with proportions less than 5 percent, which resulted in a higher complexity compared to the two-class model. Based on this information, the two-class model was selected as the best fit model.

Table 2.

Model fit information of GMM (N ​= ​210).

GMM	LL	AIC	BIC	aBIC	Entropy	LMR	BLRT	Probability of class
1	−2334.352	4686.703	4716.827	4688.310	–	–	–	1
2	−2303.766	4631.531	4671.697	4633.674	0.943	0.013	< 0.001	0.624/0.376
3	−2291.195	4612.391	4662.597	4615.069	0.936	0.206	< 0.001	0.581/0.371/0.048
4	−2277.077	4590.153	4650.401	4593.367	0.948	0.589	< 0.001	0.362/0.028/0.562/0.048
5	−2263.885	4569.770	4640.060	4573.520	0.945	0.037	< 0.001	0.562/0.052/0.043/0.324/0.019
6	−2255.048	4558.097	4638.427	4562.381	0.936	0.414	< 0.001	0.190/0.048/0.338/0.386/0.019/0.019

GMM, growth mixture modeling; LL, Log-Likelihood Ratio; AIC, Akaike information criterion; BIC, Bayesian information criterion; aBIC, adjusted Bayesian information criterion; LMR, Lo-Mendell-Rubin likelihood ratio test; BLRT, bootstrapped likelihood ratio test.

The two trajectories of family resilience in the above models were shown in Fig. 2. The Class 1 trajectory exhibited an upward trend (s ​= ​0.700, P ​< ​0.001), while the Class 2 trajectory exhibited a downward trend (s ​= ​−2.613, P ​< ​0.001). Based on the characteristics of each trajectory, Class 1 was named the high-level-slow-increasing class, and Class 2 was named the low-level-continuous-decreasing class. Due to limitations imposed by sample size, we performed stratified analysis only for the cervical cancer group (n ​= ​84) and the cancer stage Ⅰ group (n ​= ​109). The fitting results were not significantly different from those of the overall sample (n ​= ​210), which to some extent rules out the impact of tumor heterogeneity on the findings of this study (Appendix Table 3).

Fig. 2.

Two trajectories of family resilience in patients after gynecological malignant tumor radical surgery.

Class differences in baseline variables

Next, we conducted univariate analysis with demographic characteristics, disease-related data, perceived social support, perceived stress, and family functioning as independent variables, and family resilience classification as the dependent variable. The results were shown in Table 3. Statistically significant differences (P ​< ​0.1) were found between the two groups of patients in terms of age, place of residence, marital status, education level, family monthly income, number of caregivers, cancer stage, chronic diseases, chemotherapy, perceived social support, perceived stress, and family functioning.

Table 3.

Class differences in demographic characteristics, disease-related data, perceived social support, perceived stress, and family functioning (N ​= ​210).

Variables	Class 1 (n ​= ​131)	Class 2 (n ​= ​79)	χ2/F/U	P
	Mean ​± ​SD/n (%)/ Median (IQR)	Mean ​± ​SD/n (%)/ Median (IQR)
Age (years)a	49.91 ​± ​8.33	46.96 ​± ​7.87	6.431b	0.012f
Place of residence			3.031c	0.082
Urban area	52 (39.7%)	22 (27.8%)
Rural area	79 (60.3%)	57 (72.2%)
Marital status			5.652c	0.017f
Married	110 (84.0%)	75 (94.9%)
Unmarried/divorced/widowed	21 (16.0%)	4 (5.1%)
Educational level			9.877c	0.020f
Primary school and less	60 (45.8%)	25 (31.6%)
Junior middle school	30 (22.9%)	12 (15.2%)
Senior middle school	23 (17.6%)	24 (30.4%)
Bachelor's degree or higher	18 (13.7%)	18 (22.8%)
Working condition			3.276c	0.194
Employed	45 (34.4%)	37 (46.8%)
Retired	13 (9.9%)	7 (8.9%)
Unemployed	73 (55.7%)	35 (44.3%)
Family monthly income (RMB)			22.369c	< 0.001f
> 10,000	39 (29.8%)	6 (7.6%)
5000–10,000	50 (38.2%)	24 (30.4%)
< ​5000	42 (32.1%)	49 (62.0%)
Primary caregiver			7.920d	0.152
Spouse	58 (44.3%)	48 (60.8%)
Children	28 (21.4%)	17 (21.5%)
Parents	6 (4.6%)	1 (1.3%)
Relatives	27 (20.6%)	9 (11.4%)
Siblings	6 (4.6%)	3 (3.8%)
Others	6 (4.6%)	1 (1.3%)
Number of caregivers			4.566c	0.033f
One	73 (55.7%)	32 (40.5%)
Multiple	58 (44.3%)	47 (59.5%)
Method of payment for medical expense			1.024d	0.806
Health insurance	47 (35.9%)	31 (39.2%)
Public health care	7 (5.3%)	2 (2.5%)
Out-of-pocket	15 (11.5%)	8 (10.1%)
Social security	62 (47.3%)	38 (48.1%)
Type of cancer			0.224c	0.894
Cervical cancer	51 (38.9%)	33 (41.8%)
Ovarian cancer	33 (25.2%)	20 (25.3%)
Endometrial cancer	47 (35.9%)	26 (32.9%)
Cancer stage			8.991d	0.023f
Ⅰ	67 (51.1%)	42 (53.2%)
Ⅱ	33 (25.2%)	30 (38.0%)
Ⅲ	28 (21.4%)	6 (7.6%)
Ⅳ	3 (2.3%)	1 (1.3%)
Family history			1.676c	0.196
Yes	11 (8.4%)	3 (3.8%)
No	120 (91.6%)	76 (96.2%)
Chronic diseases			3.819c	0.051
Yes	41 (31.3%)	15 (19.0%)
No	90 (68.7%)	64 (81.0%)
Chemotherapy			13.929c	< 0.001f
Yes	60 (45.8%)	16 (20.3%)
No	71 (54.2%)	63 (79.7%)
Perceived social support	58.06 ​± ​8.39	43 (9)	9.667e	< 0.001f
Perceived stress	19 (16)	39 (6)	9.855e	< 0.001f
Family functioning	7 (2)	5 (2)	9.564e	< 0.001f

SD, standard deviation; IQR, interquartile range.

^aThe result of the homogeneity of variance test showed P = 0.560.

^bOne-way ANOVA.

^cChi-square Pearson test.

^dFisher's exact test.

^eMann-Whitney U test.

^fFactors with P ​< ​0.1 were included in the regression, and we identified them with P ​< ​0.1 in bold.

Predictors of family resilience trajectories

Finally, binary logistic regression analysis was performed with significant variables identified by the univariate analysis as independent variables and family resilience classification as the dependent variable. Analysis showed that patients with family monthly income > 10,000 (odds ratio [OR] ​= ​0.077, 95% confidence interval [CI]: 0.010–0.606) and high social support (OR ​= ​0.890, 95% CI: 0.799–0.992) were more likely to experience the high-level-slow-increase trajectory, while patients with high perceived stress (OR ​= ​1.200, 95% CI: 1.094–1.316) were more likely to experience the low-level-constant-decline trajectory (Table 4).

Table 4.

Predictors of family resilience trajectories by binary logistic regression.

Predictors	B	SE	Wald	P	OR	OR (95% CI)
						Lower	Upper
Age	0.039	0.039	0.991	0.320	1.040	0.963	1.124
Place of residence
Urban area	−0.294	0.648	0.206	0.650	0.745	0.209	2.653
Rural area	Reference
Marital status
Married	0.962	0.941	1.045	0.307	2.618	0.414	16.565
Unmarried/divorced/widowed	Reference
Educational level			1.891	0.595
Primary school and less	−1.271	1.007	1.592	0.207	0.281	0.039	2.021
Junior middle school	−0.469	0.981	0.229	0.633	0.626	0.092	4.278
Senior middle school	−0.722	0.875	0.681	0.409	0.486	0.088	2.699
Bachelor's degree or higher	Reference
Family monthly income (RMB)			6.380	0.041a
> 10,000	−2.567	1.054	5.928	0.015a	0.077	0.010	0.606
5000-10,000	−1.338	0.702	3.632	0.057	0.262	0.066	1.039
< 5000	Reference
Number of caregivers
One	0.176	0.571	0.096	0.757	1.193	0.390	3.650
Multiple	Reference
Cancer stage			2.897	0.408
Ⅰ	−3.156	3.568	0.782	0.276	0.043	0.000	46.371
Ⅱ	−2.194	3.529	0.386	0.534	0.112	0.000	112.510
Ⅲ	−2.933	3.524	0.693	0.405	0.053	0.000	53.206
Ⅳ	Reference
Chronic diseases
Yes	0.066	0.703	0.009	0.925	1.068	0.270	4.235
No	Reference
Chemotherapy
Yes	−1.607	0.867	3.437	0.064	0.200	0.037	1.096
No	Reference
Perceived social support	−0.117	0.055	4.440	0.035a	0.890	0.799	0.992
Perceived stress	0.182	0.047	14.883	< 0.001a	1.200	1.094	1.316
Family functioning	−0.296	0.312	0.898	0.343	0.744	0.404	1.371

OR, odds ratio; CI, confidence interval. Reference group: Class 1.

^aBolded the P-values with statistical significance in the logistic regression.

Discussion

Main findings

Based on GMM, we explored the trajectories of family resilience and identify specific predictors for these trajectories during the six months following radical surgery in patients with gynecological malignancies. Our findings provide preliminary empirical data to inform potential interventions to facilitate a smooth transition during the postoperative recovery period for these patients.

Our analysis revealed significant differences in family resilience across four time points (71.40 ​± ​9.84, 68.42 ​± ​10.66, 68.99 ​± ​12.43, and 68.65 ​± ​13.65), with lower resilience compared to a previous study of breast cancer patients in an urban population.¹⁶ This difference underscores the socioeconomic disadvantages of our cohort, which predominantly consisted of rural residents with lower education levels and incomes. Rural families face systematic barriers to post-discharge care, long travel distances for medical appointments, and financial strain from treatment costs coupled with lost wages. Lower education levels may impede comprehension of medical information and self-care protocols, potentially increasing anxiety and reducing family resilience. Financial stress, including transportation costs and income loss, may also force families to balance treatment priorities with essential needs, further affecting family resilience. Throughout the survey period, the lowest family resilience level was observed one month post-surgery (T1). Two key factors likely contribute to this: first, the one-month period post-surgery marks a transition from hospital care to home, where patients face the dual challenge of recovering from surgery while transitioning from the role of patient to that of family member. Second, this period coincides with the initiation of chemotherapy for most patients, whose severe side effects lead to physical discomfort and negatively impact their treatment adherence. The absence of adequate professional support and the lack of guidance at this time likely exacerbates negative emotional states. Future interventions can focus on this critical one-month period by establishing a tiered psychological intervention system. This system may include: symptom management (e.g., regular check-ins with mental health professionals, standardized assessments like depression/anxiety scales, and tailored therapeutic exercises); family role reconstruction (e.g., workshops on communication skills, shared decision-making, and balancing caregiving, delivered online or in person); and emotional support network development (e.g., online peer forums, local group information, and peer check-in calls).

A major contribution of this study was that GMM identified two distinct trajectories of family resilience following radical surgery for gynecological malignancies. The optimal model revealed that 62.38 percent of patients in this study belonged to the high-level-slowly-increasing group, while 37.62 percent belonged to the low-level-continuously-decreasing group. Binary logistic regression analysis indicated that family monthly income and perceived social support were significant predictors of membership in the high-level-slowly-increasing group, while perceived stress was a predictor for the low-level-continuously-decreasing group. Specifically, patients with a higher family monthly income were more likely to belong to the high-level-slowly-increasing group, in line with previous research findings.²⁸ A higher family income may facilitate access to superior health care resources—including personalized treatment plans, rehabilitation programs, and comprehensive nursing services²⁹—while also alleviating financial stress related to medical expenses. This economic stability allows patients to focus more fully on recovery after radical hysterectomy. Furthermore, reduced financial pressure enables family members to provide more consistent emotional support,²⁹ fostering a stable environment that enhances the patient's psychological well-being and adaptation. Based on these findings, health care professionals are encouraged to prioritize support for low-income families through implementing clinical decision-making protocols (e.g., cost-efficacy balanced treatment algorithms; standardized cost-benefit assessment tools; tiered therapy options based on payment ability), and advocating for institutional support mechanisms (e.g., sliding-scale fee structures, medication assistance programs, transportation subsidies, and financial counseling services).

Radical hysterectomy represents a significant traumatic event for patients with gynecological malignancies, and family resources alone are insufficient to meet the psychological adjustment needs of these patients, making external support crucial.³⁰ Perceived social support, which includes emotional and practical help from family, friends, and social networks, plays an essential role in mitigating feelings of loneliness and helplessness, improving coping mechanisms, and fostering positive behavioral changes.³¹ In this study, patients with higher levels of perceived social support were more likely to exhibit a more stable trajectory of family resilience. This suggests that high levels of social support may contribute not only to a patient's well-being but also to more positive family dynamics, which, in turn, might guide family resilience trajectories toward the high-level-slowly-increasing group. These findings were consistent with previous studies on the psychological coping of cancer patients,³² thus highlighting the potential value of supportive interventions in improving family resilience. Therefore, it may be beneficial for health care institutions, communities, and relevant social organizations to consider developing comprehensive social support networks and continuity care systems to help patients with gynecological cancer reintegrate into their families and society.

In addition, our analysis showed that perceived stress was a significant predictor of membership in the low-level-continuously-decreasing group. For each one-point increase in the perceived stress score, the odds of being categorized into the low-level-continuous-decline group increased by 20 percent (OR ​= ​1.200, p ​< ​0.001). Perceived stress refers to an individual's evaluation of stress in various life situations.³³ Women are generally more prone to chronic stress and are more sensitive to the impact of life events. Research by Salgado et al.³⁴ shows that high levels of perceived stress consistently exerts negative impact on a woman's occupational well-being and quality-of-life. High levels of perceived stress over the long-term can lead to psychological health issues, including anxiety disorders, reduced self-esteem, and a loss of sense of control.³⁵ These negative emotions could adversely affect emotional health and potentially increase the burden and anxiety of family members,¹³ leading to tension in family relationships. This dynamic might establish a cycle that could further challenge family resilience and be associated with its decline in some cases. Given this association, investigating the provision of psychological support and stress management training for patients and their families might be warranted to help manage perceived stress. Furthermore, fostering stronger communication and cooperation among family members could potentially help build a more supportive family environment, which may mitigate risks for decline in family resilience.

Implications for nursing practice and research

In this study, we identified two distinct trajectories of family resilience in patients following radical surgery for gynecological malignancies, providing significant insights for clinical health care professionals. Health care providers should pay more attention to patient groups with low family income, low perceived social support, and high perceived stress levels, particularly during the critical one-month post-discharge period. Nurses can help patients address various issues that arise after surgery by providing personalized treatment plans, medication assistance, multi-level social networks, stress management training, psychological support, and other intervention strategies, thus enabling patients to return to their families and society with a higher quality-of-life. Nurses are encouraged to collaborate with oncologists, psychologists, and social workers to establish interdisciplinary teams, explore and develop more targeted support strategies for patients after radical gynecological surgery, and help them to identify healthy ways to maintain high levels of family resilience.

Limitations

This study had several limitations that need to be considered when interpreting our findings. First, the use of a convenience sample from a single tertiary care hospital could have introduced selection bias, particularly referral filter bias. In addition, due to the lack of external validation and regional differences, the findings of our study may not be generalizable to other cities in China. Future studies should employ stratified and multi-center sampling across diverse health care settings (e.g., urban/rural hospitals, different insurance coverage systems) to capture the full spectrum of patients undergoing radical surgery for gynecological cancer. Second, although this study preliminarily explored the impact of tumor heterogeneity on the trajectories of family resilience through stratified analysis, the small sample size in some sub-groups may lead to overfitting in these models. It is recommended that future studies expand the sample size to investigate the trajectories of family resilience among patients with different types and stages of cancer. Third, we only collected follow-up data from preoperative to six months post-surgery. This is a relatively short time span. Future research could extend the follow-up period to obtain more comprehensive and dynamic data. Fourth, this study only focused on the predictive effects of the variables and did not track the dynamic changes of variables such as perceived social support and perceived stress. Future research could explore this direction in greater depth.

Conclusions

In this study, we used GMM to identify two trajectories of family resilience in patients after radical surgery for gynecological malignant tumors. Our findings indicate that family resilience is lowest one month after surgery, during the transition from hospital to home care and the initiation of chemotherapy. High family monthly income and social support predict higher and more stable family resilience, while high levels of perceived stress predicts lower and more fluctuating resilience. Interventions should focus on this critical period, providing psychological support and stress management. Addressing financial and social needs, especially for low-income families, is crucial. Health care providers should implement targeted support strategies to improve recovery outcomes. Future research should refine interventions for vulnerable groups.

CRediT authorship contribution statement

Xinru Zhang: Conceptualization, Data curation, Formal analysis, Methodology, Writing – original draft, Writing – review & editing, Visualization. Xi Li: Data curation, Formal analysis. Yiteng Chen: Data curation, Writing – review & editing, Conceptualization, Methodology. Xindi Wang: Data curation, Formal analysis, Writing – review & editing. Yu Guo: Data curation. Weiqing Ruan: Writing – review & editing. All authors have contributed to the article and approved the submitted version.

Ethics statement

The study was approved by the Ethics Committee of Southern Hospital, Southern Medical University (Approval No. NFEC-2024-479) and was conducted in accordance with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. All participants provided written informed consent.

Data availability statement

The data that support the findings of this study are available from the corresponding author, WR, upon reasonable request.

Declaration of generative AI and AI-assisted technologies in the writing process

No AI tools/services were used during the preparation of this work.

Funding

This work was supported by the Guangdong Science and Technology plan project (Grant No. 2022A1414020007). The funders had no role in considering the study design or in the collection, analysis, interpretation of data, writing of the report, or decision to submit the article for publication.

Declaration of competing interest

The authors have no conflict of interest to declare.

Appendix A. Supplementary data

The following are the Supplementary data to this article.