Social support as a moderator of healthcare adherence and distress in long-term hematopoietic cell transplantation survivors

Abstract

Background

Treatment with hematopoietic cell transplantation (HCT) has potentially severe effects on physical and psychosocial functioning. Poor social support has been linked with physical morbidity and mortality as well as psychological distress in HCT survivors. This study tested a theory-driven hypothesis that social support buffers adverse effects of health stressors of comorbidities and graft-versus-host disease (cGVHD) on distress and adherence to recommended healthcare among long-term HCT survivors.

Methods

This cross-sectional study analyzed baseline data from a randomized controlled trial in adult survivors 3–18 years post-HCT. Data included medical records and patient-reported outcomes including cancer and treatment distress (CTXD), healthcare adherence (HCA), comorbidity index, cGVHD, ENRICHD Social Support Instrument (ESSI), Social Activity Log, and Health Self-Efficacy. We tested hypothesized models for HCA and CTXD using blocked hierarchical linear regressions.

Results

Among the 781 HCT survivors completing baseline assessment, 38% had ≥ 3 comorbidities, 8% had moderate–severe cGVHD, 30% reported low social support, 30% reported elevated distress, and 49% reported low healthcare adherence. Social support and self-efficacy were directly related to both adherence and distress. Regression models supported the hypothesized moderated relationships for distress but not for healthcare adherence.

Conclusions

The two tested models confirm that the health stressors of comorbidities and cGVHD are moderated by better social support and self-efficacy in their associations with lower distress but without moderating effects for healthcare adherence.

Implications for Cancer Survivors

Social support and self-efficacy confer protective benefits on healthcare adherence and psychological distress. Interventions are needed that focus on maintaining social networks or finding new networks if necessary.

Clinical trial registration number

NCT00799461.

Introduction

Advances in hematopoietic cell transplantation (HCT) for hematologic malignancies have improved survival rates with an increasing number of long-term survivors [1, 2]. However, HCT remains a treatment with high rates of late toxicities, and consequently, survivors require ongoing healthcare monitoring [3–6]. Long-term complications after HCT include chronic graft-versus-host disease (cGVHD) for recipients of allogeneic HCT. GVHD is an immunologic reaction where the donor immune cells identify the body as foreign and attack it, causing inflammation and other effects that can be life-threatening. It can manifest anywhere in the body and often involves the skin, liver, eyes, mouth, sinuses, and gut [5, 7]. Other complications include higher rates of cardiometabolic syndromes, subsequent malignancies, other comorbidities than are seen in other cancer survivors, and high rates of long-term psychological distress [5]. These healthcare needs require ongoing surveillance and treatment.

Studies have investigated adherence to survivorship recommendations in HCT survivors with common findings that half of survivors are non-adherent and results identifying specific characteristics in survivors with low adherence or high-risk behaviors [8–10]. However, none of these studies included social support as a possible explanatory variable associated with healthcare adherence. Meanwhile, research on the effects of social support on other health outcomes has revealed convincing links between poor social support and psychological distress [11–15]. In patients with cancer, a high level of social support is associated with fewer psychological symptoms and greater well-being, self-efficacy, and quality of life [16–21]. Studies in HCT recipients have shown that long-term survivors who have greater psychological distress are more likely to report problems with social functioning or lack of support from within their social networks; however, research has not addressed whether social support moderates self-efficacy and distress in HCT survivors [22–25].

To address these gaps in understanding the role of social support in healthcare adherence and distress in HCT long-term survivors, we developed a conceptual model based on the “buffering hypothesis” (Fig. 1) [26]. The model derives from the influential theory on social support and health outcomes which posits that social support protects people from the potential adverse effects of stressful events by having beneficial influences on psychological well-being and consequently on health outcomes [26]. Our model explains that social support functions as a buffering moderator that enhances coping by increasing the likelihood that survivors will appraise threatening situations as manageable because they believe they have support to assist if needed. This support in turn reinforces the coping process, thereby increasing self-efficacy by enhancing confidence that survivors can manage health demands. Therefore, social support and, in turn, self-efficacy buffer the relationship between health stressors and health outcomes [24, 27]. This model guides the approach in the present study.

Fig. 1.

Hypothesized conceptual model of the role of social support in healthcare adherence and distress for cancer survivors. As tested in this study, this model is based on the “buffering hypothesis” (Cohen & Willis 1985) and posits that social support protects people from the potential adverse effects of stressful events, thereby having beneficial influences on health outcomes. Note that although the diagram is displayed sequentially in the order of blocks entered into the model, the measures for this study, other than covariates, were collected concurrently

The purpose of this study is to examine whether our conceptual model based on the buffering hypothesis is supported in long-term HCT survivors. Specifically, we hypothesized that greater social support and self-efficacy would moderate the adverse effects of stressful health demands, as represented by cGVHD and comorbidities in our analyses, on healthcare adherence and psychological distress [26].

Methods

Design

This is a secondary analysis of baseline data, prior to randomization, from a randomized controlled trial [28]. The trial provided an online secure patient portal and web site (INSPIRE) with or without problem/solving treatment telehealth calls for survivors 3–18 years post-HCT. Within the model tested, medical record data used to test the model were collected prospectively, whereas patient-reported outcomes (PRO), including current comorbidities and active cGVHD, were collected at baseline assessment. Thus, the analytic design is largely cross-sectional.

Participants and procedures

Eligible participants were identified through the transplant center’s research database at the Fred Hutchinson Cancer Research Center in Seattle. Eligibility included a hematologic malignancy diagnosis, 3–18 years since their most recent transplant, currently age 18 or older, living in the USA or Canada, and with internet access and adequate English to complete the assessment. Participants were excluded if they had recurrence or subsequent malignancy within 2 years prior to enrollment. All survivors who were eligible based on the research database were approached by letter and could opt in or out, or directly go to the URL in the letter to register, consent and complete the baseline survey. Those who did not opt out were contacted by phone to determine their interest and assist them with logging into the URL. The baseline survey included screening to confirm eligibility based on no recurrence or subsequent malignancy and internet access. A detailed description of the study design and sampling frame has been published previously [28]. All procedures were reviewed and approved by the Fred Hutchinson Cancer Research Center’s institutional review board. Participants completed assessments online using a study-built, secure PRO program.

Measures

Medical records provided information on diagnosis, transplant details, years post-transplant, and history of cGVHD. The PRO included sociodemographic characteristics and the measures described below.

Outcomes

Healthcare adherence (HCA) was assessed as the proportion of recommended surveillance tests and exams completed according to the frequency recommended for HCT long-term survivors [6]. The 15 healthcare items in the scoring each have a recommended frequency, adapted for age, gender, and treatment exposures [6]. The HCA score is the proportion of recommended tests adhered to within a year of the recommended frequency, from 0 to 1.00. The measure was developed for the study though is derived from that described by Khera et al. [8]. Validity is supported by those who report not having had a physical exam in the previous 2 years (when many tests could be done) having lower HCA scores than those who report having had a physical exam (p < 0.0001) [6].

Psychological distress was assessed with cancer and treatment distress (CTXD) which is a 22-item measure of distress or worry related to cancer events [29]. Responses range from no distress = 0 to severe distress = 3 for items such as “not knowing what the future will bring” and “being a burden to other people.” The measure has a total mean score as used in these analyses and six subscales: uncertainty, health burden, family strain, identity, medical demands, and finances. The internal consistency is strong (α = 0.95), and it correlated in this cohort with the mental component summary of the SF-36, r = − 0.70 [29].

Covariates

Covariates included gender (male/female), age (continuous), education (< 4 years in college/≥ 4 years in college), annual income (< $40,000/≥ $40,000), marital status (married or partnered/single, divorced, or similar), years post-transplant (< 10 years/≥ 10 years), and donor source (autologous/allogeneic).

Moderators

Stressful health demands (health stressors) were assessed by comorbidities and severity of active cGVHD. The study used the PRO version of the Charlson Comorbidity Index, which has 22 items with acceptable reliability relative to Charlson medical record scores [30]. We used a cut point of ≥ 3 comorbidities indicating a health stressor. The cGVHD measure is based on PRO indicating current severity of cGVHD from none = 0 to severe = 3. The score was dichotomized, with moderate or severe indicating a health stressor.

Social support was assessed using two measures, the Social Activity Log (SAL) and the ENRICHD Social Support Instrument (ESSI). SAL is a 13-item self-report measure designed to capture the frequency and diversity of social activities outside of daily responsibilities in long-term HCT survivors; it has high reliability and validity [31]. The SAL was developed specifically for the HCT population since existing measures of social activities were designed for elderly or dementia adults and include both solitary and social behaviors. Means of three subscales that were defined by principal components analysis are combined into a total score with equal weighting of the subscale scores for frequent contact items (e.g., “got emails, texts, letters, or notes from people you know, but do not live with”), moderate contact items (e.g., “had friends or family come to visit”), and infrequent contact items (e.g., “went to a movie, concert, theater, or other entertainment event”). Instructions indicate to report “a number for how many times you did these activities in the past month.” Response options range from 0 = not at all to 6 = 6 or more times. The ESSI has six items indicating availability of social support. Items are summed for a total score, with higher scores indicating greater social support and a total score < 18 indicating low social support [32]. Item examples include “Is there someone you can count on to listen to you when you need to talk?” and “Is there someone available to help you with daily chores?” Responses range from 0 = none of the time to 4 = all of the time. Although it is brief and does not have subscales for the different domains of support, the ESSI does include items on emotional support, advice, and assistance. It has been used in a variety of medical populations, including with cancer survivors, and demonstrates high reliability and validity (α = 0.88 with this sample) [33, 34]. Supporting the distinct domains measured, the SAL and ESSI were modestly correlated, r = 0.29. These measures were chosen to capture both availability of perceived social support and engagement in distinct social activities.

Coping process was assessed using the Health Self-Efficacy (HSE) measure which was modified from the General Self-Efficacy (GSE) by the addition to each item of “health” or “medical.” The HSE (like the GSE) is a 10-item measure with higher scores indicating greater self-efficacy [35]. Items include “I can usually handle whatever health or medical difficulties come my way” and “it is easy for me to stick to my aims and accomplish my health goals.” The HSE has demonstrated high validity and reliability with α = 0.91 for this sample and with all items loading on a single factor with 59% of the variance explained [35].

Statistical analysis

Data analysis was carried out using IBM SPSS Statistics for Windows, version 25. The demographic and clinical characteristics of participants were summarized using percentages for categorical variables and mean (SD) for continuous variables. Dependent/outcome variables were the HCA and CTXD as continuous scores. Potential covariates and moderators were identified based on the conceptual model. We performed univariate analyses of the hypothesized sociodemographic and clinical covariates and moderators with the outcomes of adherence and distress using an independent t test for binary hypothesized moderator variables, ANOVA for multiple categorical variables with more than two levels, and Pearson correlations between continuous variables. Multicollinearity was checked using variance inflation factor (VIF) analyses, and we identified no VIF > 3 between covariates or moderator variables and dependent variables. In order to maximize power, covariates including demographic and clinical variables were retained in the final regression models only if they were associated with adherence or distress at p < 0.10 in univariate analyses. Hierarchical regression analysis for each outcome (adherence, distress), in which sets of variables were added sequentially, was conducted in 4 blocks. Within each hierarchical regression, block 1 included covariates of demographic and clinical variables, block 2 added health stressors of comorbidities and cGVHD, block 3 added social support (SAL and ESSI) as moderators of health stressors, and block 4 added self-efficacy as a moderator of health stressors. p values < 0.05 were used to determine statistical significance.

Results

Of 1775 HCT survivors approached, 1322 met eligibility criteria and were approached [36]. Of these, 781 (59%) consented to participate, completed the baseline assessment, and are included in analyses.

The study population mean age was 52 years (range of 18–79), most were male (56%), 20% were rural residents, and 75% received an allogeneic donor HCT (Table 1). Mean years post-transplant were 9 (SD: 4.5), with 22% less than 5 years and 42% 5 to 9 years post-transplant. Comorbidity scores ≥ 3 were reported by 38% of the participants, while 10% of the allogeneic transplant recipients reported current moderate to severe cGVHD (8% of the entire sample), and 30% had elevated distress above the cut point (> 0.90). Approximately half (51%) reported adhering to ≥ 80% of healthcare recommendations for HCT survivors. Furthermore, 30% scored below the ESSI cut point for low perceived social support (< 18). Health stressors of comorbidities and cGVHD were unrelated to social support on the ESSI or SAL (r < 0.03 for all).

Table 1.

Demographic and clinical characteristics of study participants.

Characteristic	Total sample N = 781 Value
Gender, female, n (%)	344 (44)
Age, mean (SD)	52 (13)
Range	18–79
Categories, n (%)
18–40	125 (16)
40–65	542 (69)
65 or older	114 (15)
Rural, n (%)	154 (20)
Race, n (%)
White/Caucasian	700 (89)
Native American/Alaska Native	3 (4)
African American	5 (6)
Asian	13 (2)
Other non-White or mixed	9 (1)
Unknown	51 (7)
Ethnicity: Hispanic or Latino, n (%)	13 (2)
Unknown	38 (5)
Education, n (%)
High school or less	203 (26)
4 years in college or more	514 (66)
Unknown	64 (8)
Income, n (%)
Below $ 40.000 per year	112 (14)
$ 40.000-$79.000 per year	212 (27)
$80.000 and above per year	369 (47)
Unknown	88 (11)
Marital status, n (%)
Married or living with a partner	554 (71)
Single, separated, divorced, or widowed	167 (21)
Unknown	60 (8)
Diagnosis, n (%)
Acute leukemia	235 (30)
Chronic myeloid leukemia	230 (29)
Lymphoma	157 (20)
MDS	79 (10)
Multiple myeloma	59 (8)
Donor source, n (%)
Autologous	196 (25)
Allogeneic	585 (75)
Years post-transplant, n (%)
< 10 years	499 (64)
≥ 10+ years	282 (36)
Current chronic GVHD, n (%)
None/mild	708 (91)
Moderate/severe	60 (8)
Unknown	13 (1)
Comorbidity, n (%)
< 3 comorbid disease	461 (59)
≥ 3 comorbid diseases	298 (38)
Unknown	22 (3)
Cancer and treatment distress (CTXD) > 0.90, *n (%)	234 (30)
CTXD continuous score, mean (SD)	0.65 (0.56)
Healthcare adherence (HCA) < 0.80, *n (%)	382 (49)
HCA continuous score, mean (SD)	0.74 (0.17)
ENRICHED Social Support Instrument (ESSI) < 18, *n (%)	235 (30)
ESSI continuous score, mean (SD)	18.9 (5)
Social Activity Log
Mean (SD)	2.8 (1.0)
Health Self-Efficacy
Mean (SD)	3.2 (0.5)

GVHD graft-versus-host disease, N/n number, MDS myelodysplastic syndrome

^*Established cut point for impaired scores

Univariate analyses

Table 2 illustrates the univariate relationships between hypothesized covariates and moderators associated with distress and healthcare adherence. The ESSI, SAL, and self-efficacy were directly correlated with both adherence and distress (p ≤ 0.001) and met the criteria of p < 0.10 for inclusion in the regression for adherence and distress as moderators. All demographic and clinical covariates also met the criteria of p < 0.10 for inclusion in the regressions for both adherence and distress, with the exceptions for distress of education, marital status, and urban vs rural residence, which had p ≥ 0.10 in univariate analyses with distress. In addition, both social support and social activity were associated with self-efficacy (ESSI: r = 0.24, p < 0.01; SAL: r = 0.20, p < 0.001).

Table 2.

Univariate relationship between hypothesized factors associated with distress and adherence

Variable	Health care adherence		Cancer and treatment distress
	Test value	p value	Test value	p value
Covariates
Gender	8.144a	< 0.001	3.320a	0.001
Age (< 40, 40–64, > 65)	14.67b	< 0.001	3.048b	0.048
Years post-transplant (< 10, ≥ 10)	2.084a	0.038	3.063a	0.002
Income $ (< 40.000, ≥40.000)	4.266a	< 0.001	2.676a	0.008
Education (< 4 years in college, ≥ 4 years in college)	2.670a	0.008	1.369a	0.171
Marital status (married/partner, single/divorced, widowed)	3.302a	< 0.001	1.091a	0.276
Donor source (autologous, allogeneic)	3.304a	0.001	1.658a	0.098
Urban\rural	1.907a	0.057	0.021a	0.983
Health stressors
Comorbidity (< 3, ≥ 3)	3.695a	< 0.001	4.735a	< 0.001
cGVHD (moderate/severe, mild/none)	2.683a	0.009	5.196a	< 0.001
Support
ESSI (continuous)	0.134c	0.001	0.340c	< 0.001
SAL (continuous)	0.139c	< 0.001	0.310c	< 0.001
Coping process
HSE (continuous)	0.166c	< 0.001	0.404c	< 0.001

An association is measured using an

^aindependent t test,

^bone-way ANOVA,

^cPearson correlation

cGVHD chronic graft-versus-host disease, ESSI ENRICHED Social Support Instrument, SAL Social Activity Log, HSE Health Self-Efficacy

Hierarchical regression analysis

Hierarchical regression analyses were conducted for each of the continuous outcome variables healthcare adherence (Table 3) and cancer and treatment distress (Table 4).

Table 3.

Multiple linear regression analysis predicting healthcare adherence

	Block 1		Block 2		Block 3		Block 4
	β	p value	β	p value	β	p value	β	p value
Covariates
Gender (ref = female)	−0.32	< 0.001	−0.32	< 0.001	−0.32	< 0.001	−0.31	< 0.001
Age (continuous)	0.27	< 0.001	0.25	< 0.001	0.254	< 0.001	0.23	< 0.001
Years post-transplant (ref = ≥ 10 years)	0.12	0.001	0.11	0.002	0.11	0.002	0.11	0.003
Income (ref = ≥ 40.000 $)	−0.14	< 0.001	−0.16	< 0.001	−0.16	< 0.001	−0.15	< 0.001
Education (ref = ≥4 years in college)	−0.07	0.041	−0.07	0.032	−0.08	0.065	−0.08	0.03
Marital status (ref = married or partnered)	−0.003	0.942	−0.009	0.828	−0.014	0.43	−0.009	0.82
Donor source (ref = allogeneic)	−0.16	< 0.001	−0.15	< 0.001	−0.14	0.001	−0.14	< 0.001
Urban\rural (ref = urban)	−0.05	0.148	−0.04	0.187	−0.044	0.225	−0.04	0.211
Health stressors
Comorbidity score ≥ 3 (ref = < 3)			−0.09	0.016	−0.161	0.201	0.14	0.48
cGVHD moderate-severe (ref = none/mild)			−0.05	0.16	−0.081	0.609	0.34	0.111
Social support moderators
ESSI (continuous)*comorbidity score ≥ 3 (ref = < 3)					0.136	0.24	0.134	0.253
ESSI (continuous)*cGVHD moderate-severe (ref = none/mild)					0.161	0.26	0.103	0.48
SAL (continuous)*comorbidity score > 3 (ref = < 3)					0.131	0.152	0.129	0.158
SAL (continuous)*cGVHD moderate-severe (ref = none/mild)					−0.022	0.793	−0.041	0.629
Cognitive process moderators
HSE (continuous)*comorbidity score > 3 (ref = < 3)							−0.02	0.91
HSE (continuous)*cGVHD moderate-severe (ref = none/mild)							0.34	0.066
F for block	22.61	< 0.001	19.14	0.017	14.18	0.001	12.66	0.001
R2 Δ for block	0.21		0.01		0.02		0.02
F for full model							12.66	0.001
R2 for full model							0.24

β standardized regression coefficient, cGVHD chronic graft-versus-host disease, ESSI ENRICHED Social Support Instrument, SAL Social Activity Log, HSE Health Self-Efficacy

Table 4.

Multiple linear regression analysis predicting cancer and treatment distress

	Block 1		Block 2		Block 3		Block 4
	β	p value	β	p value	β	p value	β	p value
Covariates
Gender (ref = male)	0.12	0.002	0.12	0.001	0.13	< 0.001	0.13	< 0.001
Age	− 0.004	0.924	−0.05	0.202	−0.05	0.12	−0.009	0.794
Years post-transplant (≥ 10 years)	0.09	0.023	0.06	0.116	−0.06	0.108	0.07	0.047
Income (ref = ≥ 40.000$)	0.12	0.003	0.08	0.033	0.03	0.432	−0.035	0.324
Donor source (ref = allogeneic)	0.05	0.21	0.1	0.01	0.06	0.091	0.09	0.014
Health stressors
Comorbidity score > 3 (ref = < 3)			0.15	< 0.001	0.16	< 0.001	1.791	< 0.001
CGVHD moderate-severe (ref = none/mild)			0.22	< 0.001	0.18	< 0.001	0.455	0.034
Social support moderators
ESSI (continuous)*comorbidity score > 3 (ref = < 3)					−0.567		−0.459	< 0.001
ESSI (continuous)*CCGVHD moderate-severe (ref = none/mild)					0.047		0.173	0.236
SAL (continuous)*comorbidity score > 3 (ref = < 3)					−0.408	< 0.001	−0.37	< 0.001
SAL (continuous)*CGVHD moderate-severe (ref = none/mild)					0.033	0.7	0.07	0.41
Coping process moderators
HSE (continuous)*comorbidity score > 3 (ref = < 3)							−0.883	< 0.001
HSE (continuous)*CCGVHD moderate-severe (ref = none/mild)							−0.511	0.006
F for block	6.068	< 0.001	12.599	< 0.001	13.81	< 0.001	15.792	< 0.001
R2 for block	0.04		0.07		0.06		0.05
F for full model							36.18	< 0.001
R2 Δ for full model							0.22

β standardized regression coefficient, cGVHD chronic graft-versus-host disease, ESSI ENRICHED Social Support Instrument, SAL Social Activity Log, HSE Health Self-Efficacy

Social support as a moderator of health stressors on healthcare adherence

In the final model for HCA, after controlling for sociodemographic and clinical covariates, the health stressor of comorbidities, but not cGVHD, added to the model for adherence (p = 0.016). However, with the moderators of social support and health self-efficacy in the model, neither the health stressors of comorbidities or cGVHD nor the social support or self-efficacy moderators of health stressors showed significant effects on healthcare adherence (p > .05). In the final model, covariates of female gender, older age, fewer years post-transplant, higher income, and allogeneic donor explained the greatest proportion of variance in adherence (R² Δ = 0.24, p < 0.001).

Social support as a moderator of health stressors on cancer and treatment distress

In the final model for CTXD, in the first block tested, the sociodemographic and clinical covariates of female gender and < 10 years post-transplant were associated with greater distress (R² Δ = 0.04, p < 0.001), while greater health stressors (R² Δ = 0.07, p = 0.001) added to explained variance. The social support variables had a moderating effect on comorbidities but not for current GVHD, while self-efficacy moderated both comorbidities and current GVHD symptoms for distress (p < .05).

Discussion

Our results were consistent with our hypothesized model of social support and self-efficacy as buffer of the impact of health stress on distress. However, that was not the case for healthcare adherence where the moderators did not interact with health stressors to moderate adherence. We found that social support and self-efficacy were associated with each other and that social support and self-efficacy were directly associated with both adherence and distress. Nonetheless, in the final model for adherence, neither the health stressors themselves nor the hypothesized moderators of support or self-efficacy on those health stressors were significant. In contrast, social support and self-efficacy independently buffered the effects of the health stressors of cGVHD and comorbidities on distress even after controlling for major sociodemographic and clinical covariates. This is to our knowledge the first study to have tested a model that demonstrates social support as a buffer for psychological distress in HCT survivors. These findings also clarify factors associated with adherence, though not through the moderation theory we proposed.

An HCT late effects initiative organized by the National Institute of Health (NIH) in 2016 recognized that major and unique health and psychosocial challenges continue for long-term survivors after evidence of their underlying disease has been eradicated by HCT [5, 37, 38]. Our results confirmed that over 40% of the long-term survivors of HCT reported comorbidities and/or cGVHD that would be expected to be major health stressors [39]. Because we did not find evidence for the moderating effects of social support and self-efficacy on health stressors as a pathway for improved healthcare adherence, other variables that were not assessed in the current study may moderate the effect on healthcare adherence such as type of healthcare insurance. Alternatively, social support and self-efficacy may moderate important factors other than comorbidities and cGVHD, such as disparities of lower income or education in their impact on adherence.

Extensive literature in cancer survivors suggests that social support reduces the impact of the disease and promotes long-term health [11, 13, 16, 17, 21, 24]. Although it remains to be tested prospectively, our study identifies the association between perceived social support, social inactivity, and self-efficacy as a possible mechanism through which this support improves health. Evidence for the buffering model, as proposed by Cohen and elaborated in our model, has also been demonstrated in a review of prospective studies which defines a direction of effect for support influencing health outcomes [26]. Interestingly, perceived support appears to be a stronger predictor of health-related outcomes than other measures of social function [40]. We found both low levels of perceived social support and social inactivity associated with greater distress, although the relationship was weaker for social inactivity and non-adherence. These findings are consistent with previous studies where the effectiveness of support was found to be an important contributor to buffering distress among partnered survivors within the first 3 years after HCT [22]. Another study in newly transplant recipients found that receiving effective support from a partner predicted lower later distress [23].

Our results emphasize the importance of continuing to evaluate perceived social support and social inactivity, and to provide support interventions in long-term survivors, and possibly also for the family caregivers who bare the continuing role of providing primary support for these HCT survivors. Survivors may benefit from interventions that facilitate maintaining or expanding their social support networks because they need these networks not only to help with healthcare tasks, as is the focus during treatment and recovery, but also to provide engagement in life beyond healthcare. The return to “normal” social activities may help to reduce worries and focus on health and thereby reduce cancer and treatment-related distress. Our model supports an explanatory theory that social support enhances coping by increasing the likelihood that HCT survivors will appraise threatening situations as manageable because they believe they have support and therefore they can handle major health events if necessary, and this belief then translates to improved adherence behaviors and improved distress [27]. However, future research is needed to directly examine whether in fact these appraisals differ in survivors with and without support.

Strengths of this study include the large sample size and participants with a long range of time since transplantation, with prospective collection of medical data and comprehensive patient-reported outcomes. Limitations include the cross-sectional design which increases the risk of reverse causation or confounding variables accounting for the findings and cannot assure causal relationships. Since our results cannot determine whether social support improves adherence or distress, the question remains whether people who are more socially engaged lead healthier lives than those who are isolated or conversely people who are healthy lead more socially engaged lives than unhealthy people [40]. In addition, greater distress may reduce one’s perception of available support and ability to engage in social activities. There exists a potential bias in the study sample as participants were enrolled for a randomized clinical trial for which internet access was required, although 90% of the US population has internet access and only 12% of those approached indicated they did not have internet and email access and were not comfortable using computers [36]. This is a single-center study, and the results might vary for other centers in the USA as well as internationally. Limitations also include a somewhat homogenous patient population, consistent with most transplantation centers in the USA, with a high proportion of White, more-educated, and higher-income patients. This limits the external validity of the findings until replicated in other settings, particularly in other cultures and countries.

This study confirms the proposed model that social support and self-efficacy confer unique protective benefits on psychological distress. Given that perceived social support, social inactivity, and self-efficacy can be modified, these findings have valuable implications for clinical practice and research. More importantly, these implications may be applicable to other cancer survivors, though it remains unclear whether the findings will apply equally to other cancer populations. Many HCT survivors report changes in their social networks during their disease trajectory. Therefore, in survivorship clinical practice, social support and confidence in being able to manage healthcare needs are important to evaluate along with promoting long-term survivors’ knowledge of recommended healthcare. By evaluating and promoting the use of available social connections, clinicians may help to reduce survivors’ distress and non-adherence to healthcare. In research, clinical trials are needed that focus on improving social support networks in vulnerable patients with major health stressors and low support. Possible interventions could focus on maintaining or re-establishing social networks as well as finding new networks if their past ones have fragmented or no longer seem relevant. Specifically, survivors may benefit from strategies to maintain their connections to their extended family and friends, co-workers, and employers, as well as identifying which survivors need couples or family therapy. Survivors may also be helped by connecting with other survivors in similar situations, especially for socially isolated individuals. For this latter group in particular, social media modalities could be helpful. Clinical trials are needed that demonstrate improvements in maintaining or building these social supports and self-efficacy in long-term survivors of HCT, with related improvements in health outcomes.