Distinct dyadic quality of life profiles among patient-caregiver dyads with advanced lung cancer: a latent profile analysis

This study aimed to identify the heterogeneity of dyadic quality of life (QoL) profiles, determine whether these profiles differ in terms of demographic and medical factors, neuroticism, resilience, and family functioning, and explore the combined effect of patient and caregiver neuroticism, resilience, and family functioning on dyadic QoL profiles. A cross-sectional study was conducted with 304 advanced lung cancer patient-caregiver dyads. Self-report questionnaires were administered to patient-caregiver dyads to assess demographic and medical characteristics, neuroticism, resilience, family functioning, and QoL. The latent profile analysis identified four subgroups of dyadic QoL: patient-low-caregiver-high profile (38.82%), patient-high-caregiver-high profile (22.37%), patient-high-caregiver-low profile (19.74%), and patient-low-caregiver-low profile (19.08%). Additionally, when both patients and their caregivers had a high level of neuroticism or low level of resilience and low family functioning, compared with only member having them, there was a higher risk of poorer dyadic QoL. Our study identified the four heterogeneities of dyadic QoL profiles among advanced lung cancer patient-caregiver dyads. Future dyadic interventions should consider the heterogeneity of dyadic QoL in this population and prioritize patient-caregiver dyads at risk of poor dyadic QoL. Furthermore, when high neuroticism, low resilience, or family functioning coexist between patients and their caregivers, both parties exhibit much lower dyadic QoL.


Introduction
In China, lung cancer is the most common cancer and the leading cause of cancer-related deaths.Approximately 810,000 people were newly diagnosed with lung cancer, and 710,000 people died of lung cancer in 2020 [1].Significantly, the majority of new lung cancer cases are diagnosed at an advanced stage, with 57% developing distant metastases [2,3].
Quality of life (QoL) is an individual's perception of physical, psychological, and social well-being [4].The challenges or adversities caused by cancer can lead to QoL in both patients and their caregivers.As lung cancer and its treatment have adverse effects on both the patients and their caregivers, the main focus of cancer care and its research has shifted from the individual to the dyadic level of patient-caregiver dyads [5].Furthermore, the interdependence theory [6] and empirical research have confirmed the interdependence of QoL in patient-caregiver dyads, calling for treating the patient-caregiver dyad as a unit both conceptually and methodologically [7,8].
The traditional method of studying dyadic QoL is a variable-centered approach, which is useful for clarifying associations among variables and unraveling contributors to specific QoL outcomes but is limited in capturing the heterogeneity in QoL within patient-caregiver dyads [9].Distinct dyadic QoL profiles among patient-caregiver dyads may be unidentified.Specifically, prior studies indicated that patients with cancer and their caregivers experienced positive (e.g., better family closeness and healthier lifestyles) and negative aspects (e.g., disease progress and increasing caregiving demands) in the process of adapting to cancer [10][11][12], which had different effects on their QoL.However, not each dyad has the same experiences, which leads to variations in specific QoL dimensions across dyads.For instance, Lee and Lyons (2019) found a congruent pattern between patients and caregivers of non-small cell lung cancer, which was characterized by almost identical moderate mental health in patient-caregiver dyads and a disparate pattern characterized by the better mental health of the patients compared with the caregivers [13].Increasing evidence [13,14] acknowledges heterogeneity across dyads; however, the specific profiles of dyadic QoL are still unclear.Identifying the heterogeneity of these dyadic QoL profiles can ascertain vulnerable subgroups and identify their risk and protective factors.
According to the transactional stress theory [15], personal and environmental characteristics affect an individual's adaptive outcomes (i.e., QoL) when exposed to adversity.Neuroticism is a crucial personal characteristic characterized by more stress, negative emotions, and emotional instability [16].Accumulative evidence suggests that patients with cancer and caregivers with higher neuroticism scores have an increased risk of poor QoL [17,18].Resilience, another important personal characteristic, is an individual's ability to recover from challenges and major traumatic events [19].Cancer patients' and their caregivers' resilience has been positively linked to their own increased QoL [20,21].Family functioning is a crucial environmental protective characteristic [22].Individuals living in families with positive family functioning are prone to feel loved, accepted, gain more positive resources, and this is an important predictor of better QoL among patients with cancer and caregivers [23,24].Given that the above-mentioned personal (i.e., neuroticism, resilience) and environmental characteristics (i.e., family functioning) play important roles in predicting QoL, it is helpful to examine the independent effects of these factors on the distinct dyadic QoL profiles.Additionally, when high neuroticism coexists between patients and their caregivers, both parties are more reactive to stress and emotions, which degrades their QoL.In contrast, when high resilience or family functioning coexists between patients and their caregivers, both parties show positive adaption in the face of adversity or acquire more intra-and extra-family resources for better adaption, which underpins their QoL.The question arises whether there is a higher risk of poorer dyadic QoL profiles when both the patients and caregivers have high levels of neuroticism or low levels of resilience or family functioning, simultaneously, compared with only one party having them.However, a relative dearth of research explored the combined effect of these factors (e.g., the transactional effect of one patient's high resilience and the caregiver's low resilience) on dyadic QoL profiles.
This study aimed to (1) identify heterogeneity of dyadic QoL profiles among advanced lung cancer patient-caregiver dyads, (2) determine whether these profiles differ in terms of demographic and medical factors, personal characteristics (i.e., neuroticism and resilience), and environmental characteristics (i.e., family functioning); and (3) explore the combined effect of patients' and caregivers' neuroticism, resilience, and family functioning on dyadic QoL profiles.

Study design and participants
A cross-sectional survey was conducted in the departments of radiotherapy and chemotherapy at three tertiary hospitals in Jinan, Shandong Province, China, from March 2019 to February 2020.Patient and caregiver inclusion criteria included: age 18 years or older, fluent in written or spoken Chinese, diagnosed with lung cancer stages III or IV, and caregiver identified by the patient as the primary caregiver.The exclusion criteria included unwillingness to participate in this study, psychiatric disorders (e.g., depressive disorder, anxiety disorder, bipolar disorder, schizophrenia), or cognitive disorders.Of the 380 patient-caregiver dyads eligible for the study, 29 refused to participate, and 47 failed to complete the questionnaires.Thus, 304 patient-caregiver dyads were included in this study.This study was approved by the university ethics committee and was conducted following the tenets of the Declaration of Helsinki.Informed consent was obtained from all participants included in this study.

Neuroticism
The 8-item Neuroticism subscale of the 44-item Big Five Inventory (BFI) [25] was used to assess neuroticism.Responses were rated on a five-point Likert scale ranging from "strongly disagree" = 1 to "strongly agree" = 5.Higher scores were related to higher levels of neuroticism.Cronbach's α for the subscales were 0.79 in patients and 0.76 in caregivers, respectively.

Resilience
The 10-item Connor-Davidson Resilience Scale [26] was used to assess resilience.Responses were rated on a fivepoint Likert scale ranging from "Not true at all" = 0 to "True almost always" = 4. Higher scores were associated with greater resilience.Cronbach's α of the questionnaire was 0.93 in patients and 0.92 in caregivers, respectively.

Family functioning
The General Functioning subscale of the Family Assessment Device [27] was used to assess participants' perceived overall functioning of a family system.The 12-item scale was rated on a four-point Likert scale ranging from "strongly disagree" = 1 to "strongly agree" = 4. Higher scores reflected greater family functioning.Cronbach's α of the questionnaire was 0.85 in patients and 0.81 in caregivers.

QoL
The Medical Outcomes Study Short Form-8 Health Survey (SF-8) [28] was used to assess QoL.SF-8 comprises eight dimensions: general health perceptions, physical functioning, physical role, bodily pain, vitality, social functioning, mental health, and emotional role.Responses were rated on a five-point Likert scale.The raw scores were transformed into standard scores of 0-100, with higher scores associated with better QoL.Cronbach's α of the SF-8 was 0.90 in patients and 0.89 in caregivers, respectively.

Covariates
The covariates analyzed in the study consisted of demographic and medical characteristics such as patients with advanced lung cancer and their caregivers' age, sex, marriage, education, employment status, lung cancer type, Eastern Cooperative Oncology Group (ECOG) performance, caregivers' relationship with patients (see Table 1).

Data analysis
Latent profile analysis (LPA) is a person-centered approach that determines heterogeneity profiles of dyadic QoL, categorizing them as diverging subgroups that are internally homogeneous [29] based on the patterns of means on observed variables (i.e., the eight dimensions of QoL in this study).The optimal number of profiles was evaluated using model-fit indices, such as the Akaike information criterion (AIC), Bayesian information criterion (BIC), sample-sizeadjusted BIC (SSABIC), entropy, the Vuong-Lo-Mendell-Rubin likelihood ratio test (LMR), and the bootstrapped likelihood ratio test (BLRT).The lower values of AIC, BIC, SSABIC indicate a better fit of the model.A higher entropy value represents better classification accuracy, with an entropy value above 0.80, indicating adequate classification accuracy of more than 90% of the participants [30].A significant p value for LMR and BLRT indicates a better model fit with the k-profile model compared to the k-1 profile model.
After the optimal number of profiles was obtained, the one-way analysis of variance (ANOVA) and chi-square tests were conducted to identify differences in demographic and medical characteristics, neuroticism, resilience, and family functioning among different dyadic QoL profiles.Multinomial logistic regression analysis was conducted to evaluate the unique association of each predictor variable that was significant in the one-way ANOVA or chi-square tests with profile membership.To examine the combined effect, we initially divided patients' and caregivers' scores of neuroticism, resilience, and family functioning into high (above mean) and low (below mean) groups according to the mean split method.Then, we created a new composite variable (e.g., neuroticism) including 4 combinations of patient neuroticism (high vs. low) and caregiver neuroticism (high vs. low) to explore further the combined effect of patients' and caregivers' neuroticism, resilience, and family functioning on dyadic QoL profiles were employed using multinomial logistic regression analysis.Statistical analyses were performed using SPSS 26.0 and Mplus 7.4 with a significance level of 5% (2-tailed).

Results
The mean age of the advanced lung cancer patient-caregiver dyads was 58.50 (SD = 10.11) years for patients and 45.47 (SD = 11.85) for caregivers.Table 1 depicts the detailed demographic and medical characteristics.

Latent profile analyses
LPA models with one to six profiles were estimated (see Table 2) to identify the optimal profile solution.The fourprofile solution fitted the data significantly better than the three-profile solution; more specifically, the four-profile solution demonstrated lower values of AIC, BIC, and SSABIC, higher entropy than the three-profile solution, and LMR and BLRT became significant when reaching the four-profile solution.Although the five-profile solution demonstrated lower values of AIC, BIC, and SSABIC than the four-profile solution, the nonsignificant P value for LMR further confirmed that the five-profile solution did not significantly improve over the four-profile solution.Additionally, the entropy values in the four-profile solution were the highest, indicating that the four-profile solution provided fair classification accuracy.
Four profiles of dyadic QoL are presented graphically in Fig. 1.In group 1 (n = 118, 38.82%), labeled as patientlow-caregiver-high profile, patients reported low levels of QoL, but caregivers reported a higher level of QoL than  patients.Group 2 (n = 68, 22.37%), labeled as patient-highcaregiver-high profile, was characterized by both patients and their caregivers reporting higher levels of QoL relative to the other profiles.In group 3 (n = 60, 19.74%), labeled as patient-high-caregiver-low profile, patients reported a comparatively higher level of QoL than their caregivers, who reported low levels of QoL.Group 4 (n = 58, 19.08%), labeled as patient-low-caregiver-low profile, was characterized by patients and their caregivers reporting lower levels of QoL compared with the other profiles.

Comparisons of demographic, medical characteristics, neuroticism, resilience, and family functioning among the four profiles
The results showed that patients' economic situation, ECOG performance status, caregivers' age, marriage, economic situation, relationship to the patient, and the neuroticism, resilience, and family functioning of patients and their caregivers significantly differed across the four profiles (see Table 1).
Multinomial logistic regression was further conducted to explore whether the significant variables in the univariate analysis predicted profile membership (Table 3).Considering that no caregivers whose marital status was single/ divorced/widowed were classified into a patient-low-caregiver-low profile, marriage was omitted from further analyses.The patient-high-caregiver-high profile was used as the reference group.Patients whose income was less than the expenditure (OR = 5.94; P = 0.006) were more likely to belong to the patient-low-caregiver-high profile compared to those with income more than the expenditure.Patients with higher scores on ECOG performance status and neuroticism were associated with increased odds of membership to the patient-low-caregiver-high profile and patient-low-caregiverlow profile.Elderly caregivers (OR = 1.07;P = 0.015) were more likely to be categorized as having a patient-high-caregiver-low profile.Moreover, caregivers with low resilience tended to have increased odds of belonging to the patienthigh-caregiver-low and patient-low-caregiver-low profiles.Patients with high family functioning were more likely to be categorized as a patient-high-caregiver-low profile.

Combined effect of patient neuroticism, resilience, family functioning and caregiver neuroticism, resilience, and family functioning associated with distinct dyadic QoL profiles
As Table 4 shows, the patient-high-caregiver-high profile was used as the reference group.The subgroup of patienthigh-caregiver-low neuroticism was 3.21 times more likely to belong to the patient-low-caregiver-high profile than the dyads of patient-low-caregiver-low neuroticism.The subgroup of patients and their caregivers who both had high neuroticism was 3.92 times more likely to be classified as the patient-high-caregiver-low profile.The subgroups of patientlow-caregiver-high neuroticism, patient-high-caregiver-low neuroticism, and patient-high-caregiver-high neuroticism had 4.25, 9.17, and 10.87 times the likelihood of belonging to the patient-low-caregiver-low profile, respectively.That is, both the patients and their caregivers with a high level of neuroticism simultaneously had a higher likelihood of belonging to the patient-low-caregiver-low profile, compared with one member with high neuroticism.Furthermore, compared to patient-high-caregiver-high resilience, the subgroup of patient-low-caregiver-high resilience and patient-low-caregiver-low resilience was 2.73, 7.27 times, respectively, more likely to belong to the patient-low-caregiver-high profile.In addition, the subgroup of patient-high-caregiver-low resilience and the subgroup of patients and their caregivers with low resilience were 4.09, 8.62 times more likely to belong to the patient-highcaregiver-low profile, respectively.Finally, the subgroup of patient-high-caregiver-low resilience was 4.33 times more likely to belong to the patient-low-caregiver-low profile, and the subgroup of patient-low-caregiver-low resilience was 11.34 times more likely to belong to the patient-lowcaregiver-low profile.Therefore, the dyads of both patients and their caregivers with low resilience increased their likelihood of belonging to the patient-low-caregiver-low profile than only one party with low resilience (OR = 11.34 vs. OR = 4.33 vs. OR = 2.22).One interesting finding is that only caregivers with lower resilience were found to be at a higher likelihood of being classified in the patient-lowcaregiver-low profile than only patients with lower resilience (OR = 4.33, P = 0.013 vs. OR = 2.22, P = 0.173).
As for the combined effect of family functioning, the odds of both patients and their caregivers with low family functioning being classified in the patient-high-caregiver-low and patient-low-caregiver-low profiles were 5.45 and 2.94 times, respectively.

Discussion
This study utilized a person-centered approach (i.e., LPA) to capture the heterogeneity of dyadic QoL profiles and explored their associations with different characteristics Table 4 The combined effect of patient-caregiver dyads' neuroticism, resilience, and family functioning on latent profile membership The patient-high-caregiver-high profile was used as the reference group Model adjusted for significant demographic variables (patient's economic situation, ECOG performance status; caregiver's age, economic situation, relationship to patient) among patients with advanced lung cancer and their caregivers.

CI confidence interval
Our findings revealed four dyadic QoL profiles.First, most advanced lung cancer patient-caregiver dyads (n = 118, 38.82%) were divided into patient low-caregiver-high-profile, in which patients scored low, but caregivers scored high on the QoL indicators.This result is expected, given that most prior researchers have found worse QoL among patients with cancer than their caregivers [31][32][33].Second, a patient-high-caregiver-high profile was observed.Third, an important and interesting finding is the existence of such a patient-high-caregiver-low profile characterized by patients scoring high on the QoL indicators but caregivers scoring low on the QoL indicators.This finding is difficult to trace using the traditional variable-centered approach and suggests that clinical practitioners should pay more attention to this vulnerable profile, especially the caregivers, because of the lower QoL.Fourth, a patient-low-caregiver-low-profile was also identified.The heterogeneity of dyadic QoL profiles highlights the importance of adopting tailored dyadic interventions rather than a "one-size-fits-all" approach.For instance, the dyads in patient-low-caregiver-high profile may benefit from adopting a patient-focused, caregiver-assisted dyadic intervention.The dyads in the patient-low-caregiverlow profile may benefit from adopting both patient-and caregiver-focused dyadic interventions.Future research could explore the effectiveness of personalized dyadic interventions in heterogeneous profiles.
The demographic and medical characteristics of the dyadic QoL profiles differed.Specifically, patients with higher financial burdens had an increased likelihood of belonging to the patient-low-caregiver-high profile.A study indicated that 72.7% of patients with lung cancer reported catastrophic health-related spending, and 37% reported healthcare costs exceeding the annual household income [34].Our results revealed that 70.1% of the patients had incomes less than their expenditures.The association between high financial burden and poor QoL in this study is consistent with prior research [35].Patients with higher scores of ECOG performance status were more likely to belong to the patient-low-caregiver-high profile and patientlow-caregiver-low profile.Patients with cancer with higher scores on ECOG performance status have more difficulty in completing activities of daily living and, therefore, experience QoL deterioration [36].Older caregivers were more likely to belong to the patient-high-caregiver-low profile.This finding may be owing to elderly caregivers' decline in physical health (e.g., degenerative changes), which is also associated with lower QoL [37,38].
Patients in the patient-low-caregiver-high profile and patient-low-caregiver-low profile were more neurotic.Unsurprisingly, a person with high neuroticism may regard the diagnosis of cancer as more threatening or severe, have more negative health perception, and thus report lower QoL [16,17,39].Caregivers with low resilience were more likely to be classified in the patient-high-caregiver-low and patientlow-caregiver-low profiles.Resilience is a critical psychological asset that embodies the personal qualities needed for individuals to thrive in adversity [40,41].Caregivers with low resilience tended not to utilize the resources to adapt to the adversity they encountered, thereby decreasing their QoL.Moreover, patients who perceived better family functioning were more likely to be included in the patient-highcaregiver-low profile.Patients with better family functioning perceived a strong emotional bond with their family.They could acquire more support from family members, leading to better QoL consistent with prior work [23].
This study found that the dyads of both patients and caregivers with high neuroticism were associated with the highest odds of belonging to the patient-low-caregiver-low profile, followed by the dyads of only patients with high neuroticism, and the dyads of only caregivers with high neuroticism.This finding indicated that both the patients and their caregivers with a high level of neuroticism simultaneously had lower dyadic QoL, compared with only one of the dyads being neurotic.Additionally, the dyads of only patients with high neuroticism were found to be at a higher likelihood of being classified in the patient-low-caregiver-low profile than the dyads of only caregivers with high neuroticism.Patients with advanced lung cancer and their caregivers are inclined to react as emotional units.Patients' neuroticism has a negative effect on their own QoL, as does spillover on their caregivers' QoL [42].In this study, compared to caregivers' neuroticism, patients' neuroticism generated greater spillover effects on the QoL of patient-caregiver dyads.
Interestingly, only caregivers with lower resilience showed a higher likelihood of belonging to the patient-low-caregiver-low profile than only patients with low resilience.Caregivers' heavy caregiving tasks (e.g., form treatment regimens and manage medications) place high demands on their physical and emotional reserves, and resilience not only helps them facilitate the caregiving task but transmits from them to the patients through information exchange and communication.[40,43].Moreover, both patients and caregivers with low resilience showed the highest risk of reducing dyadic QoL compared with only one of the dyads with low resilience.This finding suggests that identifying the dyads of both patients and caregivers with low resilience may be conducive to implementing preventive interventions to mitigate their higher risk of poor QoL.
Furthermore, patients and caregivers with low family functioning were more likely to belong to the patient high-caregiver low profile or patient low-caregiver low profile.A previous study suggested that patients' perceived family functioning may differ from that of their caregivers, even though they belong to the same family [43,44].Thus, assessing family functioning from the perspectives of both patients and their caregivers could truly reflect the functioning of the family [45].Both patients and caregivers in this study reported poorer family functioning, which was related to ineffective communication patterns and poor health-related behaviors (e.g., physical activity, screening physical examination), thus reducing their QoL.
This study has certain limitations.First, causal directionality needs to be cautiously interpreted with this cross-sectional design.Second, this study focused on modifiable psychological variables that could be addressed through an intervention.However, several factors (e.g., symptom distress, caregiver burden, and coping strategy) that predict dyadic QoL were omitted.Future research should incorporate these factors to develop a robust predictive model of dyadic QoL.Third, this study assessed participants' perceived family functioning rather than objective family functioning.However, this study assessed family functioning from multiple perspectives and greatly reduced the error.Future studies should specify extended or nuclear families when assessing family functioning.

Conclusion
This study captured the four heterogeneities of dyadic QoL profiles among advanced lung cancer patient-caregiver dyads and explored their influencing factors.Future dyadic interventions should consider the heterogeneity of dyadic QoL in this population and prioritize patient-caregiver dyads at risk of poor dyadic QoL.Furthermore, when high neuroticism, low resilience, or family functioning coexist between patients and their caregivers, both parties exhibit much lower dyadic QoL.

Table 1
Demographic, medical characteristics, neuroticism, resilience, and family functioning of four latent profiles

Table 3
Potential predictors of latent profile membershipThe patient-high-caregiver-high profile was used as the reference group