Perceived Stress and Depressive Symptoms as Predictors of Decisional Conflict in Dyads Considering a Left Ventricular Assist Device

Abstract

Background:

The decision to pursue a left ventricular assist device (LVAD) commits loved ones to major caregiving responsibilities and, often, medical decision making. How emotional domains overlap within patients and their caregivers and contribute to conflict around the decision to pursue LVAD remains largely unexplored.

Methods and Results:

The associations within and between persons in patient-caregiver dyads considering LVAD were estimated in a specific type of structural equation model known as the Actor-Partner Interdependence Model. This model tested whether each person’s depression and stress predicted their own decisional conflict (actor effects) as well as their partner’s decisional conflict (partner effects). At the time of study enrollment when a formal LVAD evaluation was initiated, 162 patient-caregivers dyads completed assessments of decisional conflict using the Decisional Conflict Scale, depressive symptoms using the Patient Health QuestionNaire-2, and stress using the Perceived Stress Scale. Across both models, decisional conflict was significantly correlated within patient-caregiver dyads (β=0.47 and β=0.44, for depression and perceived stress respectively p’s<0.001). Greater perceived stress in both the patient (β=0.18; p<0.05) and caregiver (β=0.28; p<0.001) was significantly related to greater decisional conflict (both actor effects). Greater patient depressive symptoms were related to greater patient decisional conflict (β=0.16; p<0.05); whereas caregiver depressive symptoms was not related to their own decisional conflict (β=0.07; p=0.37). There were no partner effects identified between decisional conflict and perceived stress or depressive symptoms.

Conclusion:

Patient and caregiver conflict over the decision to pursue an LVAD was highly correlated in this sample, with greater perceived stress significantly predicting greater decisional conflict in both patients and caregivers. Depressive symptoms in patients also predicted greater patient decisional conflict. No partner effects were identified in predicting decisional conflict. These results contribute to a larger body of work acknowledging the importance of patient-caregiver well-being in serious illness.

Clinical Trial Registration:

Clinicaltrials.gov; Unique Identifier: NCT02344576

Patients and their caregivers are faced with many challenges and complex decisions during the course of chronic progressive disease. Heart failure is a chronic disease that often requires engagement among the patient-caregiver pair or dyad. Although sometimes rewarding,¹ caregiving in heart failure can come with psychological stress and depression, which are also common among patients.^2–4 In this setting, major medical decisions are further complicated, often leading to higher levels of decisional conflict. Decisional conflict is defined as the personal perception of uncertainty about which course of action to take when the choice among competing options involves risk, regret, or challenge to a person’s values.⁵

For patients with end-stage heart failure who are not candidates for cardiac transplantation, the decision to pursue a destination therapy left ventricular assist device (DT LVAD) may be presented. In order for a patient to be considered for this therapy, many programs require the formal identification of a caregiver—with some caregivers needing to provide care for an indefinite length of time. This makes the decision to pursue a DT LVAD a mutual decision between patient and caregiver. From previous research, we know that the decision to pursue a DT LVAD can be highly emotional, with patients and caregivers reporting feelings of being confused and overwhelmed.^{6, 7} While LVAD technology improves and may prolong survival and improve quality of life, complications such as stroke, bleeding, and infection remain common.⁸ Additionally, the burdens of post-implant care and adjustment to life with an LVAD can be difficult for both the patient and the caregiver. Further, the reasons for transplant ineligibility most often include advanced age, frailty, comorbidity, and psychosocial barriers—all of which can be compounded by LVAD therapy.

Patient-caregiver dyadic research has tended to focus on quality of life, health behaviors, and healthcare utilization.^9–11 While some researchers have advocated for examination of provider-patient dyads in medical decision making,¹² much less attention has been paid to patient-caregiver dyads in the context of decision making, despite the importance of family members’ role in reaching decisions in the face of serious illness.¹³ Notably, little is known about how much decisional conflict is shared within dyads in their decisions to pursue DT LVAD, and if emotions such as depression and stress have an effect on decisional conflict surrounding this decision.

Using dyadic data from the DECIDE-LVAD trial^{14, 15} and applying a specific kind of structural equation model known as the Actor-Partner Interdependence Model (APIM), we evaluated: 1) how patients’ and caregivers’ depression and stress were related to their own decisional conflict (actor effects) and 2) how patients’ depression and stress were related to caregivers’ decisional conflict and vice versa (partner effects).

METHODS

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

Design and Sample

The DECIDE-LVAD trial was a randomized controlled trial aimed at assessing the effectiveness of a shared decision support intervention for patients and caregivers considering DT LVAD placement at 6 LVAD programs across the United States. Full methods and details are published elsewhere.^14–16 For the DECIDE-LVAD trial, patient eligibility criteria included age 18 years or older, end-stage heart failure, and active consideration for a DT LVAD. Eligible patients were identified by the study team at each site during the time DT LVAD evaluation was initiated. Caregivers were then identified by the patient or clinical team and were defined as the person who would provide the majority of caregiving following the potential DT LVAD implantation. Only one caregiver per patient was enrolled. Patients and caregivers were enrolled prior to the clinical team’s formal LVAD education; baseline surveys were also administered before formal LVAD education, typically within 48 hours of enrollment. All data for this analysis was extracted from the baseline assessments. The DECIDE-LVAD study was overseen by the institutional review board at the University of Colorado and approved by institutional review boards at all sites. All participants provided written informed consent.

Measures

Decisional Conflict: Outcome Variable

The baseline 16-item Decisional Conflict Scale (DCS) was used to measure decisional conflict for patients and caregivers.¹⁷ The DCS has been reported as the most widely used measure in decision making research¹⁸ and is psychometrically sound in its widespread use. Individuals were asked to reflect on the decision to pursue DT LVAD implantation and then respond to statements using a 5-point Likert scale (0=strongly agree, 4=strongly disagree). The questions were slightly modified for the caregiver participants to address their “loved one’s” treatment decision rather than using the words “me” or “my”. Scale scores were obtained by multiplying the mean score (across the 16 items) by 25 so that the scale ranges from 0 to 100. Higher scores represent higher levels of decisional conflict with no defined score cut-points. The psychometrics for the DCS have been supported in other populations.¹⁷

Depressive Symptoms

The Patient Health Questionnaire (PHQ-2) was used to screen for depressive symptoms.¹⁹ The measure is comprised of the first two items of the Patient Health Questionnaire-9 and includes questions about the degree to which an individual has experienced a depressed mood and anhedonia over the past two weeks. Individuals were asked two questions on a 4-point scale (0=not at all, 3=nearly every day). The scale score, computed as a sum of the two items, ranged from 0–6. A positive screen for depression is 3, however the current analyses involved the continuous scoring of depressive symptoms. The psychometrics for the PHQ-2 have been supported in other populations.²⁰

Perceived Stress

The 10-item Perceived Stress Scale (PSS-10)²¹ was used to measure stress in patients and caregivers considering DT LVAD. The scale is designed to understand how unpredictable, uncontrollable, and overloaded an individual finds their life. Individuals were asked to reflect on the last month and respond to statements on a 5-point Likert scale (0=never, 4=very often). There are 6 negatively stated questions and 4 positively stated questions; positive items were reverse-scored. Higher scores represent greater perceived stress with no defined score cut-points. The psychometrics for the PSS-10 have been supported in other populations.²²

Analytical Approach

Descriptive statistics and correlations were computed in Statistical Programming for the Social Sciences V.25²³ and structural equation models were estimated in Mplus 8.²⁴ All variables were examined for characteristics of normality to verify the appropriate use of parametric statistics. Chi-square tests were used to test differences between patients and caregivers on categorical demographic characteristics of interest. Model fit was determined through the use several standard fit indices including: chi-square, Root Mean Square Error of Approximation (RMSEA), Comparative Fit Index (CFI) and Standardized Root Mean Residual (SRMR). Statistical significance of parameter estimates was based on an alpha level of 0.05.

Baseline data for patients and their caregivers—including PHQ-2, PSS-10, and DCS scores—were compared with bivariate correlations within patient/caregiver dyads and means were compared between patient and caregivers with paired samples t-tests. With the understanding that patients and caregivers are interdependent and likely to be similar on the variables of interest, our analytic approach accounted for the dyadic design by employing a statistical model in which the shared variance between partners is adequately handled. An approach that can simultaneously account for both within person relationships (also known as actor effects) and relationships between patients and caregivers (also known as partner effects) is the Actor Partner Interdependence Model (APIM).²² This approach treats the dyad as the unit of analysis which allows for the simultaneous estimation of the contribution of patients’ perceived stress and depressive symptoms on their own decisional conflict (actor effects), while also estimating the impact of patients’ depressive symptoms/perceived stress on caregivers’ decisional conflict (partner effects). In the final models tested, covariates of interest were added as predictors of decisional conflict to estimate the effects of stress and depressive symptoms on decisional conflict while controlling for covariates. A priori covariates included variables previously shown to relate to shared well-being within patient-caregiver dyads,²⁵ including patient and caregiver age,¹¹ sex²⁶ and relationship between patient and caregiver (spousal versus nonspousal).¹⁰ We also investigated patients’ severity of illness (Interagency Registry for Mechanically Assisted Circulatory Support [INTERMACS] profile and time since diagnosis) upon enrollment as covariates. In addition, we examined patient and caregivers’ level of education and income as potential covariates. Our interest was in the contribution of each of the emotional domains to decisional conflict, thus two separate APIM models, each predicting decisional conflict, were tested: one for depressive symptoms and one for perceived stress.

RESULTS

Patient and Caregiver Characteristics

In the DECIDE-LVAD trial, a total of 248 patients and 182 caregivers were enrolled, which included 166 patient-caregiver dyads. Of those, 162 dyads provided full baseline data, which were used for these analyses. Average age for patients was 64.2 (SD=9.8) and caregivers was 61.0 (SD=11.6). Most patients had heart failure diagnosed for 4 or more years (75.7%), many were recruited out of the ICU (25.9%), and their average INTERMACS score was 3.02 (SD=1.24). Participants were predominantly Caucasian (79.9% and 82.6% for patients and caregivers, respectively). The majority of patient-caregiver dyads were spouses (75.3%) with the majority of caregivers being the wives of patients (64.2%) within this spousal group. In addition, most caregivers lived with the patient (82.7%). Additional patient and caregiver characteristics are displayed in Table 1.

Table 1.

Demographics of patients and caregivers

	No. (%)
	Patients (n=162)	Caregivers (n=162)
Age, mean (SD)	64.2 (9.8)	61.0 (11.6)
Sex, female	27 (16.7%)	139 (85.8%)
Race/Ethnicity
White, non-Hispanic	127 (79.9%)	133 (82.6%)
Black	23 (14.5%)	19 (11.8%)
Other	9 (5.7%)	9 (5.6%)
Some college or more	96 (59.6%)	107 (66.9%)
Employment
Employed	17 (10.8%)	67 (42.4%)
Retired	86 (54.8%)	62 (39.2%)
On Disability	40 (25.5%)	12 (7.6%)
Other	14 (8.9%)	17 (10.8%)
Annual household income <$40,000	64 (42.7%)	54 (36.5%)
Married	126 (78.3%)	136 (84.5%)
First diagnosed with heart failure
Within past 2 years	17 (11.2%)	-
2-4 years	20 (13.2%)	-
4 or more years	115 (75.7%)	-
INTERMACS score*
1	10 (7.2%)	-
2-3	90 (64.7%)	-
4-7	39 (28.1%)	-
Comorbidities†
Peripheral vascular disease	7 (4.3%)	-
Major stroke	2 (1.2%)	-
Severe diabetes	18 (11.1%)	-
Chronic renal disease	42 (25.9%)	-
Pulmonary disease	13 (8.0%)	-
Liver dysfunction	7 (4.3%)	-
History of solid organ or blood cancer	13 (8.0%)	-
Enrollment location
Outpatient	40 (24.7%)	-
Inpatient (non-ICU)	80 (49.4%)	-
ICU	42 (25.9%)	-
Caregiver relationship to patient
Wife	-	104 (64.2%)
Husband	-	18 (11.1%)
Daughter	-	12 (7.4%)
Sister	-	9 (5.6%)
Mother	-	5 (3.1%)
Partner	-	5 (3.1%)
Son	-	2 (1.2%)
Other‡	-	7 (4.3%)
Caregiver lives with patient	-	134 (82.7%)

ICU, intensive care unit; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; SD, standard deviation.

^*INTERMACS score profile descriptions: 1=critical cardiogenic shock, 2=progressive decline, 3=stable but inotrope dependent, 4=resting symptoms, 5=exertion intolerant, 6=exertion limited, 7=Advanced New York Heart Association function class III.

^†Used from INTERMACS pre-implant data collection form, section “Concerns and Contraindications”.

^‡Answer options that had less than 1% were combined to create this “other” category.

Patient and Caregiver Variables

Decisional Conflict

The average decisional conflict for patients was mean=21.07 (SD=16.99) and mean=20.25 (SD=16.92) for caregivers; which were not significantly different (t(161)=.60; p=0.55). Internal consistency reliability in this sample was excellent for both patients and caregivers (α=0.95 for both). Decisional conflict was highly correlated between patients and caregivers (r=0.48, p<0.001). All bivariate correlations are reported in Table 2.

Table 2.

Bivariate correlations of patient and caregiver depressive symptoms, perceived stress, and decisional conflict

	1.	2.	3.	4.	5.	6.
1. Patient Depressive Symptoms	-	-	-	-	-	-
2. Caregiver Depressive Symptoms	.12	-	-	-	-	-
3. Patient Perceived Stress	.57*	.069	-	-	-	-
4. Caregiver Perceived Stress	.082	.50*	.16†	-	-	-
5. Patient Decisional Conflict	.14	.055	.18†	.16†	-	-
6. Caregiver Decisional Conflict	.032	.055	.096	.27†	.48*	-

^*p<.001.

^†p<.05.

Depressive Symptoms

Average PHQ-2 scores were mean=1.78 (SD=1.69) for patients and mean=0.9 (SD=1.42) for caregivers, demonstrating that patients had significantly higher depressive symptom scores than their caregivers (t(156)=5.43; p=0.48). Internal consistency reliability was acceptable for both patients and caregivers (α=0.71 and 0.86, respectively). Bivariate correlations revealed that patient and caregiver PHQ-2 scores were not significantly related (r=0.12; p=0.13).

Perceived Stress

Average perceived stress score for patients was mean=15.63 (SD=6.32) and mean=15.16 (SD=6.32) for caregivers, which were not significantly different (t(159)=.70; p=0.55). Internal consistency reliability was good for both patients and caregivers (α=0.80 and 0.86, respectively). Perceived stress was significantly related within dyads (r=0.16, p<0.05).

Actor Partner Interdependence Models

The two hypothesized models were tested and covariates of interest were examined. Among covariates tested, greater patient age was related to greater patient decisional conflict (r=0.24; p <0.05). Spousal caregivers had lower decisional conflict (mean=18.74; SD=16.35) than caregivers who were not spouses of patients (mean=24.86; SD=17.98; t(160)=2.00; p=.047). No other covariates related to patient or caregivers’ decisional conflict and inclusion of additional covariates (i.e., age, sex, income, education level, spousal caregiver versus not and patient INTERMACS score) weakened overall model fit. Therefore, analyses included only these two covariates: patient age and caregiver relationship type.

Depressive Symptoms Predicting Decisional Conflict

Greater patient depressive symptoms were related to greater patient decisional conflict (β=0.16; p<0.05), whereas caregiver depression was not related to their own decisional conflict (β=0.07; p=0.37). Patient depression was not related to caregiver depression (β=0.12; p=0.12), while decisional conflict was significantly related among patient and caregivers (β=0.47; p<0.001). Patient depression was not related to caregiver decisional conflict (β=0.042; p=0.59) nor was caregiver depression related to patient decisional conflict (β=0.07; p=0.39). All standardized parameter estimates, controlling for patient age and patient-caregiver relationship, are presented in Figure 1.

Figure 1.

Depressive symptoms and decisional conflict within dyads facing LVAD.

All parameter estimates presented are standardized coefficients. Straight arrows can be interpreted as linear regression coefficients such that the value represents the strength and direction of a predicted effect, controlling for the other predictor of decisional conflict. Curved arrows represent correlations among exogenous variables (depression) and represent correlations among residualized variability among endogenous variables (conflict). Each estimate can be interpreted as greater values equates to a larger effect and controls for other relationships in the model. Final models covaried for patient relationship to caregiver (spouse versus not) and patient age. All results remained when covarying for caregiver age, patient/caregiver sex, and patient INTERMACS score at baseline, but these covariates significantly worsened model fit thus were not included in the final model. Fit statistics: X² ₍₉₎ = 64.1, p < .001; CFI = 1.00, RMSEA = .00, SRMR= .00.

*p<.001; ^†p<.05

Perceived Stress Predicting Decisional Conflict

Greater patient perceived stress was related to greater patient decisional conflict (β=0.18; p<0.05) and greater caregiver stress was related to greater caregiver decisional conflict (β=0.28; p<0.001). Patient perceived stress was significantly related to caregiver perceived stress (β=0.16; p<0.05), while decisional conflict was significantly related among patient and caregivers (β=0.44; p<0.001). Patient perceived stress was not related to caregiver decisional conflict (β=0.069; p=0.36), while caregiver perceived stress did not reach significance in its relationship to patient decisional conflict (β=0.13; p=0.074). All standardized parameter estimates, controlling for patient age and patient-caregiver relationship, are presented in Figure 2.

Figure 2.

Perceived stress and decisional conflict within dyads facing LVAD

All parameter estimates presented are standardized coefficients. Straight arrows can be interpreted as linear regression coefficients such that the value represents the strength and direction of a predicted effect, controlling for the other predictor of decisional conflict. Curved arrows represent correlations among exogenous variables (stress) and represent correlations among residualized variability among endogenous variables (conflict). Each estimate can be interpreted as greater values equates to a larger effect and controls for other relationships in the model. Final models covaried for patient relationship to caregiver (spouse versus not) and patient age. All results remained when covarying for caregiver age, patient/caregiver sex, and patient INTERMACS score at baseline, but these covariates significantly worsened model fit thus were not included in the final model. Fit statistics: X² ₍₉₎ = 78.8, p< .001; CFI = 1.00, RMSEA = .00, SRMR= .00.

*p<.001; ^†p<.05

DISCUSSION

Our overarching goal was to examine decisional conflict within patient-caregiver dyads facing the decision to pursue LVAD therapy. We found that decisional conflict was highly correlated between patients and caregivers. Several actor effects emerged: higher perceived stress was significantly related to one’s own decisional conflict for patients and caregivers and greater depressive symptoms were significantly related to greater decisional conflict in patients. In caregivers, greater depressive symptoms were not related to caregiver decisional conflict. The lack of actor effect among caregivers could be related to the fact that caregivers experienced significantly fewer depressive symptoms than patients. There were no statistically significant partner effects (patients to caregivers and vice versa) identified. Although caregiver perceived stress did not reach significance in its relationship to patient decisional conflict, a comparison of perceived stress between patient and caregivers was significant although small (β=0.16) suggesting that caregiver stress may add to the prediction of patient decisional conflict.

A number of interesting associations emerged from examination of potential covariates and decisional conflict. Greater patient age was related to greater patient decisional conflict. This could be explained by the fact that older patients likely have more co-morbidities than younger patients²⁷ and may have higher risk profiles for surgical mortality or LVAD complications. Older patients may also be more willing to accept their mortality and choose a palliative or hospice course rather than major surgery. We also found that spousal caregivers had lower decisional conflict than caregivers who were not spouses of patients. Several possible explanations for this may exist, including that the additional burden placed on non-spousal caregivers who may not live with the patient or have families of their own may lead to higher conflict. Another possibility is that spouses experience more emotional overlap, which may lead to higher levels of distress, but also lead to lower decisional conflict as they share the emotional experience.²⁸

Dyads and Shared Decision Making

To examine decision making within patient-caregiver dyads, a common approach is to look at concordance within dyads on treatment preferences.²⁹ Qualitative data has also attempted to classify different typologies regarding the level of patient and caregiver dyadic overlap (e.g., are interdependent).^{30, 31} Areas such as self-care, advance care planning, and communication have been shown to be affected by incongruence or lack of communication within dyads.^{32, 33} Incongruence within dyads has also affected psychosocial well-being and distress.³⁴ The current analysis does not directly assess congruence; however, it does demonstrate that both patients and caregivers are conflicted during the decision making process and, importantly, that this conflict is shared within dyads. Recent recommendations for shared decision making in heart failure highlight that we do not necessarily know which emotional domains are relevant to decision making.³⁵ The current results suggest that perceived stress and depressive symptoms may be particularly relevant to decision making within patients and caregivers.³⁶

For patients with end-stage heart failure and their caregivers, the decision to pursue an LVAD is highly emotional and often presented within the context of “life or death”. One area of increased focus is the use of shared decision making in these emotionally-charged and preference-sensitive decisions. The high emotional valence in facing serious decisions—such as pursuing an LVAD—also underscores that emotional factors may be more related to decisions than cognitive and intellectual factors.³⁶ Assessing and integrating emotional readiness of the patient and family is vital to effective communication within the healthcare setting.³⁵ Owing to the uniqueness of each individual patient and caregiver, it is important to recognize factors that may influence a patient or caregivers’ willingness to participate in shared decision making.

Heart Failure as a Shared Emotional Contagion

We found that in addition to patient and caregiver decisional conflict being significantly related during the time of considering LVAD, these individuals also had overlapping levels of perceived stress. This suggests that the emotional aspects of the decision-making process are experienced by both patients and caregivers. While no partner effects were observed, the shared perceived stress symptoms can perhaps be explained by theories of emotional contagion in which emotions and related behaviors directly trigger similar emotions and behaviors in others.³⁷ Not surprisingly, the decision to pursue an LVAD is a time of uncertainty; therefore, emotions and behaviors may be at extremes. As clinicians, if we want to engage in shared decision making, it is important that we acknowledge the contribution of emotions—which may be shared between patients and caregivers. The tension remains in presenting options honestly while also attending to emotional distress of patients and caregivers.

Limitations

There are a number of limitations to the current study that deserve further comment. We only investigated perceived stress and depressive symptoms as potential emotional contributors to decisional conflict. As a result, we were not able to tease out various components of depression (anhedonia, cognitive symptoms, etc.) and their respective contributions to decisional conflict. A more comprehensive assessment of depression – or emotional well-being overall – may uncover different contributors to decisional conflict within dyads facing LVAD. There also may be medical variables, such as longevity of prior therapy or symptom burden, not captured in our current analyses that may be related to decision making. We were also unable to compare across patient-caregiver dyads in which caregivers were not enrolled and could only compare characteristics between patients whose caregiver provided full data, despite the caregiver requirement for consideration of LVAD placement. Thus, there may be differences across patients in the larger sample of this study.

CONCLUSIONS

Patient and caregiver conflict over the decision to pursue an LVAD was highly correlated. In both patients and caregivers, greater perceived stress significantly predicted greater decisional conflict. Depressive symptoms in patients also predicted greater patient decisional conflict. There were no partner effects identified. Examining contributors from the larger social milieu which often impact patients’ decisions is an important consideration in serious illness. Future research should not only examine dyadic relationships over time, but also consider larger social contextual factors that may also impact decisional conflict.

What is Known.

• Despite the importance of family members’ role in reaching decisions in the face of serious illness, there is limited research focused on patient-caregiver dyads in the context of decision making.

• Little is known about how decisional conflict is shared within dyads in their decisions to pursue a destination therapy left ventricular assist device, and if emotions such as depression and stress have an effect on decisional conflict surrounding this decision.

What the Study Adds.

• Patient and caregiver conflict over the decision to pursue a destination therapy left ventricular assist device was highly correlated.

• In both patients and caregivers, greater perceived stress significantly predicted greater decisional conflict and greater depressive symptoms in patients also predicted greater patient decisional conflict.

• No partner effects were identified in predicting decisional conflict.