Predictors of completion and response to a psychological intervention to promote health behavior adherence in heart failure

Abstract

Objective:

Most individuals with heart failure (HF) struggle to adhere to one or more health behaviors, and interventions to promote adherence are time-intensive and costly. In this analysis, we examined the predictors of engagement and response related to a telephone-delivered health behavior intervention for individuals with HF.

Method:

Using data from two pilot trials (N=25) of a behavioral intervention for individuals with New York Heart Association (NYHA) class I-II HF, we examined predictors of intervention engagement and response using linear and mixed effects regression analyses. Predictors included medical (NYHA class, physical health-related quality of life [HRQoL], and HF symptoms) and intervention (ease and usefulness/utility ratings of the first intervention exercise) characteristics. Outcomes included percentage of sessions completed, accelerometer-measured physical activity, and sodium intake.

Results:

Lower physical HRQoL and more frequent HF symptoms were associated with completion of more sessions. In contrast, more frequent HF symptoms and higher NYHA class were associated with less physical activity improvement. Finally, participants’ ratings of the first session’s utility were associated with greater improvements in physical activity at follow-up.

Conclusions:

These findings suggest that while individuals with greater functional impairment are more engaged in a behavioral intervention, they may be less able to increase physical activity in response to the program. Furthermore, the perceived utility of an initial session may predict longer-term behavior change. Larger studies are needed to clarify the presence of additional predictors and determine how they can be used to better tailor health behavior interventions.

Clinical Trials Registration:

NCT02938052, NCT03220204

Introduction

Adherence to cardiovascular health behaviors, including physical activity and a low sodium diet, is associated with better cardiovascular health outcomes among individuals with heart failure (HF) [1–3]. Despite these benefits, over 60% of patients with HF struggle to engage in these health behaviors [3–6]. Both sociodemographic and medical factors have been linked to nonadherence in HF. Among individuals hospitalized for HF, lower education levels and less severe HF symptoms are associated with reduced engagement in overall self-care [7], and lower body mass index, older age, reduced self-efficacy, and lower education are associated with less physical activity one month later [8]. Furthermore, male gender, lower income, lower education, type 2 diabetes, and higher body mass index/obesity have been linked to impaired adherence to a low-sodium diet in patients with HF [4, 9, 10].

Interventions to improve adherence to health behaviors in HF have substantial limitations. Low-intensity interventions that focus primarily on education about self-care have had mixed success related to adherence [11, 12] and limited impact on cardiovascular outcomes [13, 14]. More intensive interventions, such as formal exercise training programs or cardiac rehabilitation, are more effective and associated with improved cardiovascular health outcomes [3, 15–17]. However, these interventions often require multiple in-person visits, are attended by a minority of individuals) [17, 18], and have limited availability [19].

Identifying patient and intervention characteristics that are associated with engagement in and response to health behavior interventions may help to better target these interventions to those individuals most likely to benefit from them. To date, there is limited information about factors associated with engagement in HF-specific health behavior interventions. In one study of individuals with HF engaging in a home walking program, lower medical comorbidity, higher body mass index, and shorter duration of HF were associated with more frequent exercise completion during the program [20]. In a second study of patients with HF in a formal exercise training program, new HF diagnosis was associated with more frequent attendance at exercise training sessions, and new HF diagnosis and being physically active at baseline were associated with a higher likelihood of meeting recommended levels of physical activity at the end of the program [21]. To our knowledge, no studies have examined the predictors of engagement and response to interventions that target multiple health behaviors in patients with HF.

Over the past several years, our team has developed a positive psychology- (PP-) based intervention to promote health behavior adherence in HF. PP involves the systematic completion of activities (e.g., recalling positive events) [22] to cultivate positive psychological constructs (e.g., optimism, positive affect) that are independently associated with both health behavior adherence and health outcomes [23–26]. In this intervention, PP was paired with HF education and motivational interviewing to set goals related to HF-specific health behaviors. In single-arm (N=10) and randomized, controlled pilot trials (N=45), this PP-based intervention was feasible, well-accepted, and associated with promising improvements in well-being and health behavior adherence [27, 28]. Using data from these two pilot trials, we aimed to identify factors that predicted intervention completion, physical activity, and sodium intake.

Method

In this project, we examined data from two pilot trials of a positive psychology- (PP-) based intervention to promote health behavior adherence. The first was a single-arm trial (N=10) of a 9-week version of the intervention, while the second was a three-arm, randomized trial (N=45; we examined the 15 participants receiving the PP-based program) of a 12-week version of the program. Both trials were approved by the Partners Healthcare Institutional Review Board and prospectively registered on ClinicalTrials.gov (NCT02938052, NCT03220204). All participants provided full informed consent prior to participation.

Participants

Individuals with New York Heart Association (NYHA) [29] class I-III HF (confirmed via chart review and outpatient provider as needed) and suboptimal adherence to physical activity, diet, or medications (i.e., score of ≤15 on the Medical Outcomes Study Specific Adherence Scale [30]) were eligible to participate. Exclusion criteria included cognitive impairment (measured by a 6-item cognitive screening tool [31]), medical illness likely to lead to death within 6 months, inability to engage in physical activity, lack of telephone access, and inability to read or speak in English. Potential participants were identified through searches of the electronic medical record using an IRB-approved process, then were contacted via opt-out letters and subsequent phone calls. Interested and eligible patients were invited to an in-person study visit to provide informed consent and complete baseline measures. Full details of recruitment and enrollment have been published previously [27, 28].

Intervention

After providing baseline information, participants were assigned to a treatment condition. In the proof-of-concept trial, all participants received the 9-week, PP-based intervention. In the larger pilot trial, participants were randomized to receive the 12-week PP-based intervention, a motivational interviewing-based educational control condition, or treatment as usual. For this analysis, only those individuals in the PP-based intervention group (n=15) were included.

Participants completed weekly, 30- to 45-minute phone sessions with a study trainer, then completed a PP exercise and worked towards one or more health behavior goals between sessions. PP exercises aimed at promoting gratitude (e.g., writing a letter of gratitude), using strengths (e.g., using a strength in a new way), or cultivating meaning (e.g., writing about one’s good life in the future). The goal-setting portion of the program utilized motivational interviewing principles and focused on physical activity, a low sodium diet, and medication adherence. Participants learned to track these health behaviors, set specific and realistic goals to improve adherence, problem-solve barriers to health behavior completion, and identify resources to reach their goals.

All participants received a treatment manual that contained information related to the PP and health behavior topics discussed each week, as well as a step counter to assist with the setting of physical activity goals. In both studies, the intervention was delivered by a physician (CC) who had received training in PP and health behavior goal setting in trials of similar interventions in patients with acute coronary syndrome [32, 33]. Fidelity scales were used to ensure that all relevant intervention content was covered during the appropriate phone sessions. Full intervention details are available in prior publications [27, 28].

Outcomes

Engagement:

Intervention engagement was measured by the percentage of intervention sessions completed, with session completion defined as completion of the intervention phone call, PP exercise, and setting of a new health behavior goal.

Adherence:

Adherence measures included overall physical activity (steps/day), moderate to vigorous physical activity (MVPA; minutes/day), and sodium intake. Activity outcomes were assessed using Actigraph GT3X+ accelerometers (Actigraph LLC, Pensacola, FL), which were worn by participants for one week at baseline and immediately following intervention completion. The cutoff for MVPA was set at 1952 counts/minute [34], and 8+ hours of wear time for 4+ days was considered adequate [35, 36]. Sodium intake was measured using the Scored Sodium Questionnaire (SSQ) [37], a validated measure of sodium intake that correlates significantly with urinary sodium and has been used in patients with medical illness [27, 37] (Cronbach’s alpha=0.73 in validation sample and 0.60 in our sample).

Predictors

Participant characteristics:

Sociodemographic and medical characteristics, including age, gender, race/ethnicity, NYHA class, and age-adjusted Charlson Comorbidity Index score [38], were obtained through participant interview and review of medical records. Physical health-related quality of life (HRQoL) was measured using the Medical Outcomes Study Short Form-12 (SF-12) [39]. The SF-12 is a valid (relative validity to discriminate between major and minor physical illness = 0.93) and reliable (test-retest reliability = 0.89) measure [39] that has been used in cardiac populations [40]. Finally HF symptom severity, symptom burden, symptom stability, and physical limitations were assessed by the corresponding scores on the Kansas City Cardiomyopathy Questionnaire (KCCQ) [41]. The KCCQ is valid and reliable [41, 42], with the KCCQ clinical summary score correlating significantly with NYHA class [41] and Cronbach’s alphas of subscales ranging from 0.62 to 0.95 [41]. The Medical Outcomes Study Short Form-12 and KCCQ had high internal consistency in our sample (Cronbach’s alpha=0.86 and 0.92, respectively).

Intervention characteristics:

To examine early predictors of improvements in health behavior adherence, we included participant ratings of the ease and utility of the first PP exercise/session. These were measured on a 0-10 Likert scale (i.e., 0 = not easy/helpful, 10=very easy/helpful).

Statistical Analysis

Baseline participant characteristics were summarized using descriptive statistics. To examine the associations between participant characteristics and intervention engagement, we performed linear regression analyses, with percentage of sessions completed as the outcome variable. To identify predictors of overall physical activity, MVPA, and sodium intake, we performed mixed effects regression analyses, with a categorical effect of time and an unstructured covariance matrix. We included time, the predictor, and a time*predictor interaction term, with the time*predictor interaction term representing the impact of the predictor on change in the outcome over time. Additionally, we calculated standardized coefficients by dividing each predictor and outcome variable by its standard deviation; these standardized coefficients reflect the number of standard deviations the outcome variable changes in response to a one-standard deviation change in the predictor variable. Mixed effects regression was chosen for these analyses, as it allowed for the inclusion of individuals with some missing data. Statistical analyses were performed in Stata Version 15.1 (College Station, TX), and significance was set at alpha=.05. Due to the exploratory nature of the analyses, we did not correct for multiple comparisons.

Results

Twenty-five participants received the PP-based intervention and were included in our analyses to examine predictors of intervention engagement and sodium intake, while 24 were included in our analyses to identify predictors of MVPA and overall physical activity, as one participant did not provide adequate physical activity data. Participants were 66.4 (SD 9.3) years old, 76% were male, and 88% were White (see Table 1). All participants had NYHA class I-II symptoms at the time of enrollment, with NYHA class II being more frequent (68%).

Table 1.

Characteristics of included participants

Characteristic	Mean (SD)
Age	66.4 (9.3)
Male gender (N [%])	19 (76)
Race/ethnicity (N [%])
White	22 (88)
Black/African American	3 (12)
New York Heart Association class (N [%])
Class I	8 (32)
Class II	17 (68)
Medical comorbidity (age-adjusted Charlson Comorbidity Index)	5.9 (2.7)
Physical HRQoL (SF-12 PCS; range 0-100)*	44.9 (9.8)
Physical limitations (KCCQ physical limitation score; range 0-100)**	87.0 (14.1)
Symptom stability (KCCQ symptom stability score; range 0-100)**	48.0 (6.9)
Symptom frequency (KCCQ symptom frequency score; range 0-100)**	79.0 (17.3)
Symptom burden (KCCQ symptom burden score; range 0-100)**	78.3 (20.1)

Key: HRQoL = health-related quality of life; KCCQ = Kansas City Cardiomyopathy Questionnaire; SF-12 PCS = Medical Outcomes Study Short Form-12 Physical Component Score

^*Higher scores indicate better physical HRQoL.

^**Higher scores on the KCCQ indicate fewer symptoms, less physical limitation, or improving symptoms.

Results of the linear regression analyses to identify predictors of intervention engagement are provided in Table 2. Physical HRQoL and HF symptom frequency were negatively associated with completion of intervention exercises (physical HRQoL: B=−1.69, 95% CI −3.15, −0.22 p=.026; symptom frequency: B=−0.95, 95% CI −1.78, −0.12, p=.027), suggesting that individuals with lower physical HRQoL and more frequent HF symptoms were more engaged in the intervention and completed more sessions. The remaining participant and intervention characteristics were not associated with intervention engagement.

Table 2.

Predictors of intervention engagement (percentage of sessions completed^***)

Predictor	B*	95% CI	P	β **
Age	0.01	−1.72, 1.74	.99	0.00
Male gender	−9.02	−45.54, 27.51	.62	−0.11
Non-Hispanic white	−24.62	−71.72, 22.47	.29	−0.22
NYHA class II (compared to class I)	6.94	−26.55, 40.44	.67	0.09
Age-adjusted Charlson score	1.84	−4.07, 7.76	.53	0.13
Physical HRQoL (SF-12 PCS)	−1.69	−3.15, −0.22	.026	−0.44
Physical limitations (KCCQ)	−0.76	−1.85, 0.32	.16	−0.29
Symptom stability (KCCQ)	−0.94	−3.22, 1.34	.40	−0.18
Symptom frequency (KCCQ)	−0.95	−1.78, −0.12	.027	−0.44
Symptom burden (KCCQ)	−0.57	−1.32, 0.19	.14	−0.31
Ease of first session	1.43	−6.25, 9.11	.70	0.08
Utility of first session	−1.24	−8.54, 6.06	.73	−0.07

Key: HRQoL = health-related quality of life; KCCQ = Kansas City Cardiomyopathy Questionnaire; NYHA = New York heart Association; SF-12 PCS = Medical Outcomes Study Short Form-12 Physical Component Score

^*Unstandardized coefficient of the predictor in linear regression model

^**Standardized coefficient of the predictor in linear regression model

^***Session completion was defined as completion of the intervention phone call, PP exercise, and setting of a new health behavior goal

In the mixed effects regression models (see Table 3), several relationships were found. Less frequent HF symptoms and higher participant ease and utility ratings of the first session were significantly associated with greater improvements in MVPA (symptom frequency: B=0.27, 95% CI 0.02, 0.52, p=.037; ease: B=2.53, 95% CI 0.16, 4.91, p=.037; utility: B=2.48, 95% CI 0.45, 4.50, p=.017). Similarly, less frequent HF symptoms, lower HF symptom burden, and greater participant utility scores were associated with greater increases in overall physical activity (symptom frequency: B=44.1, 95% CI 16.5, 71.7, p=.002; symptom burden: B=32.6, 95% CI 8.8, 56.3, p=.007; utility: B=295.4, 95% CI 44.5, 546.4, p=.021), and higher NYHA class (indicative of more severe HF symptoms) was associated with smaller increases in physical activity (B=−1272.6, 95% CI −2403.6, −141.6, p=.027). No significant relationships were found between participant/intervention characteristics and sodium intake, though participant utility ratings of the first session trended towards being associated with greater reductions in sodium intake (B=−3.16, 95% CI −6.66, 0.33, p=.076).

Table 3.

Predictors of health behavior adherence

Predictor	B*	95% CI	P	β **
Moderate to vigorous physical activity (minutes/day)
Age	0.02	−0.54, 0.59	.93	0.01
Male gender	3.26	−7.54, 14.07	.55	0.05
Non-Hispanic white	3.39	−9.67, 16.46	.61	0.04
NYHA class II (compared to class I)	−3.23	−13.40, 6.94	.53	−0.05
Age-adjusted Charlson score	−0.37	−2.19, 1.44	.69	−0.03
Physical HRQoL (SF-12 PCS)	0.06	−0.41, 0.54	.79	0.02
Physical limitations (KCCQ)	0.09	−0.23, 0.42	.59	0.04
Symptom stability (KCCQ)	0.24	−0.38, 0.85	.45	0.05
Symptom frequency (KCCQ)	0.27	0.02, 0.52	.037	0.15
Symptom burden (KCCQ)	0.15	−0.07, 0.37	.17	0.10
Ease of first session	2.53	0.16, 4.91	.037	0.16
Utility of first session	2.48	0.45, 4.50	.017	0.17
Physical activity (steps/day)
Age	18.4	−50.9, 87.7	.60	0.04
Male gender	−865.0	−2153.0, 423.0	.19	−0.08
Non-Hispanic white	544.6	−1063.2, 2152.4	.51	0.04
NYHA class II (compared to class I)	−1272.6	−2403.6, −141.6	.027	−0.13
Age-adjusted Charlson score	−119.0	−338.7, 100.7	.29	−0.07
Physical HRQoL (SF-12 PCS)	24.6	−33.3, 82.4	.41	0.05
Physical limitations (KCCQ)	20.1	−19.4, 59.6	.32	0.06
Symptom stability (KCCQ)	48.1	−25.9, 122.0	.20	0.07
Symptom frequency (KCCQ)	44.1	16.5, 71.7	.002	0.16
Symptom burden (KCCQ)	32.6	8.8, 56.3	.007	0.14
Ease of first session	152.1	−167.5, 471.7	.35	0.06
Utility of first session	295.4	44.5, 546.4	.021	0.13
Sodium intake (Scored Sodium Questionnaire)
Age	0.34	−0.52, 1.20	.44	0.12
Male gender	10.81	−6.75, 28.37	.23	0.18
Non-Hispanic white	−3.54	−25.72, 18.65	.76	−0.05
NYHA class II (compared to class I)	−7.98	−23.81, 7.85	.32	−0.15
Age-adjusted Charlson score	1.50	−1.25, 4.24	.29	0.15
Physical HRQoL (SF-12 PCS)	0.36	−0.40, 1.13	.35	0.14
Physical limitations (KCCQ)	−0.04	−0.58, 0.49	.87	−0.02
Symptom stability (KCCQ)	−0.65	−1.67, 0.38	.21	−0.17
Symptom frequency (KCCQ)	−0.22	−0.64, 0.20	.31	−0.15
Symptom burden (KCCQ)	0.03	−0.34, 0.40	.88	0.02
Ease of first session	−2.86	−7.00, 1.27	.17	−0.22
Utility of first session	−3.16	−6.66, 0.33	.076	−0.25

Key: HRQoL = health-related quality of life; KCCQ = Kansas City Cardiomyopathy Questionnaire; NYHA = New York heart Association; SF-12 PCS = Medical Outcomes Study Short Form-12 Physical Component Score

^*Unstandardized coefficient of the time*predictor interaction term in mixed effects regression model

^**Standardized coefficient of the time*predictor interaction term in mixed effects regression model

Discussion

In this secondary analysis, several important predictors of intervention engagement and response were identified. First, individuals with more frequent HF symptoms and lower physical HRQoL engaged more fully in the PP-based health behavior intervention. Second, HF symptom frequency and burden were associated with smaller increases in physical activity but were not associated with changes in sodium intake. Finally, participant ratings of the utility of the initial PP exercise were associated with improvements in physical activity outcomes, with a similar trend found for sodium intake.

The findings related to intervention engagement extend those in the existing literature. To our knowledge, only one study has examined factors that predict engagement in a HF-focused exercise training program [21]. In that analysis, new HF diagnosis—but not disability or functional status—was associated with more frequent attendance at exercise training sessions. The current study suggests that individuals with more frequent HF symptoms and lower physical HRQoL also may be particularly interested in a health behavior intervention. The discrepancy in findings related to functional status may be attributed, at least in part, to the current intervention’s additional focus on well-being. PP interventions have been shown to improve well-being and reduce depression [43], and the PP component of the program may have promoted engagement by helping participants feel better mentally [27, 28], especially those with greater physical limitation and symptom frequency. Given that past studies have noted associations between disease severity and higher odds of dropout from cardiac rehabilitation programs [44, 45], the results of the current study suggest there may be benefit to developing interventions that can accommodate patients with greater physical limitations and symptom frequency.

While participants with more frequent HF symptoms and lower physical HRQoL were more likely to engage in the program, patients with less frequent symptoms demonstrated greater increases in MVPA and overall activity. Further, those with lower symptom burden and lower NHYA class (in this instance, class I vs. class II) additionally showed greater increases in overall activity, suggesting that patients who were healthier at baseline were more capable of improving physical activity outcomes. These findings are similar to those examining the efficacy of exercise training programs in HF, which have found that individuals with greater medical comorbidity, higher NYHA class, longer HF duration, and lower functional capacity are less able to complete assigned exercises during exercise training and less likely to meet recommended levels of activity post-intervention [20, 21]. The ability to engage in exercise training appears to lead to downstream benefits as well, as better baseline fitness and health status are associated with greater increases in exercise capacity and cardiorespiratory fitness in response to cardiac rehabilitation programs [46, 47]. Patients who are healthier at baseline may, therefore, be better equipped to engage in and benefit from physical activity-based training programs.

In contrast to the findings related to physical activity, participant characteristics were not associated with changes in sodium intake. Our limited sample size may have contributed to these findings, and larger studies are needed to replicate them. However, if HF symptom severity, frequency, and burden do not affect the ability to reduce sodium intake, this analysis would support the inclusion of diet as a topic for a multicomponent intervention for HF. Indeed, past studies have demonstrated the benefits of nutrition interventions for HF. In a recent systematic review, nutrition-related interventions were found to be effective at improving both dietary adherence and cardiac outcomes, such as readmission rates [48].

Finally, participants who rated the utility of the initial PP exercise more highly tended to have greater increases in physical activity and reductions in sodium intake over the course of the program. While one previous study found that greater increases in optimism following an initial PP exercise were associated with a greater likelihood of completion of a health behavior program [49], no studies have examined whether responses to an initial session are associated with improvements in adherence. This information could be used to identify individuals who are less likely to respond and who might benefit from a different intervention approach.

These analyses have several limitations. The small sample size limits the statistical power of the analyses and prevented the inclusion of multiple variables in the analysis. The sample itself was largely White and mostly male, which limits its generalizability. Furthermore, since all individuals had NYHA class I or II symptoms, the results may not generalize to individuals with more severe HF symptoms. Analyses of larger trials are needed to more definitively identify predictors of intervention engagement and response.

Despite these limitations, this analysis identified several patient- and intervention-level characteristics that were associated with both intervention engagement and improvement in health behavior adherence. More frequent HF symptoms and lower physical HRQoL were associated with greater intervention engagement, suggesting that more symptomatic individuals may be eager to engage in health behavior interventions. Furthermore, more frequent and burdensome HF symptoms were associated with smaller increases in physical activity, suggesting that it may be more difficult for individuals with more severe symptoms to increase activity in response to these programs. Finally, the utility of the initial PP exercise was associated with improvements in physical activity and could be used as an early marker of future physical activity benefit. This information could be incorporated into future health behavior interventions, to target interventions towards individuals who are more likely to respond, and to identify individuals for whom changes in intervention content may be helpful. This type of tailoring has the potential to increase engagement, promote health behavior adherence, and improve health outcomes in patients with HF.