A Motivational Telephone Intervention to Reduce Early Dropout in Cardiac Rehabilitation: A Feasibility Pilot Study

Abstract

Purpose:

Cardiac rehabilitation (CR) improves outcomes, yet early dropout is common. The purpose of the study was to determine if a motivational telephone intervention among patients at risk for nonadherence would reduce early dropout.

Methods:

We performed a randomized double-blind pilot study with the intervention group receiving the telephone intervention 1-3 d after outpatient CR orientation. The control group received the standard of care, which did not routinely monitor attendance until 2 wk after orientation. The primary outcome was the percentage of patients who attended their second exercise session as scheduled. Secondary outcomes included attendance at the second CR session at any point and total number of sessions attended. Because not everyone randomized to the intervention was able to be contacted, we also conducted a per-protocol analysis.

Results:

One-hundred patients were randomized to 2 groups (age 62 ± 15 yr, 46% male, 40% with myocardial infarction) with 49 in the intervention group. Patients who received the intervention were more likely to attend their second session as scheduled compared with the standard of care (80% vs 49%; RR = 1.62; 95% CI, 1.18 - 2.22). Although there was no difference in total number of sessions between groups, there was a statistically significant improvement in overall return rate among the per-protocol group (87% vs 66%; RR = 1.31; 95% CI, 1.05 - 1.63).

Conclusions:

A nursing-based telephone intervention targeted to patients at risk for early dropout shortly after their CR orientation improved both on-time and eventual return rates. This straightforward strategy represents an attractive adjunct to improve adherence to outpatient CR.

CONDENSED ABSTRACT

This study evaluated the effect of a telephone intervention on patients who were considered at risk for nonadherence to cardiac rehabilitation. We found that patients who received the intervention were more likely to attend their second session as scheduled compared with the program’s standard of care.

Cardiac rehabilitation (CR) provides benefits such as reducing mortality, while improving exercise tolerance and psychosocial wellbeing.¹ Although CR is a class I recommendation, it remains underutilized and early dropout is common. Only 32% of eligible Medicare patients participate in CR and 75% dropout prior to 36 sessions.² Evidence demonstrates that attending the maximum amount of sessions lowers risk and improves survival in the older population;^3,4therefore, an intervention focusing on improving adherence is essential to optimize risk reduction and secondary prevention. At our institution we found that the second CR session was poorly attended after the initial orientation. Nearly 30% of patients would not return to CR after orientation; thus, we wanted to focus our research efforts around this period of time.

Literature has shown that telephone interventions in combination with other techniques like face-to-face encounters have been successful in overcoming obstacles such as diminishing barriers and improving perceived benefits from CR participation, increasing self-control and enhancing the perception of controllability.^5–19 However, little is known about the utility of a single phone call after orientation in CR to improve adherence to the second session. We decided to utilize a telephone intervention versus other methods for a variety of reasons. First, this department would require additional resources to support innovative technology such as text messaging or electronic correspondence. Secondly, much of our population did not have the capacity to use such technology. Given our resources, demands, patient population and what was has been reported in the literature, a telephone call was deemed appropriate.

Our primary objective was to determine if a telephone intervention delivered shortly after CR orientation would improve initial patient return rates among those at-risk for nonadherence. A secondary aim was to examine the relationship between a developed risk stratification tool and total sessions attended. We hypothesized that a telephone intervention delivered to patients at risk for nonadherence would improve return rates and patients with a higher risk score would likely not return.

METHODS

Study Design

This was a single-center, single-blinded randomized-controlled trial conducted at a large CR program at Baystate Medical Center in Springfield, MA. Participants were randomized to either the intervention or usual control group. The Institutional Review Board (IRB) reviewed and approved this study. Given that the intervention was a change in process of care rather than in actual care received, risks were judged to be low and were not any higher than usual care, thus the IRB waived requirements for consent after full consideration of risk and benefits (IRB #804761–6). This study design also ensured participants were blinded to the study hypothesis, eliminating any potential Hawthorne effect.

Study Participants and Setting

Between January 2016 and December 2016, we included sequential participants age ≥21 yr referred to CR. Initial orientation is defined as the first visit to the program for patients, which includes intake and baseline exercise measurements. The second exercise session is defined as the patient’s next exercise session after their orientation. Qualifying diagnoses for CR included: myocardial infarction (MI) with or without percutaneous coronary intervention (PCI); heart failure; stable angina; valve repair/replacement; and coronary artery bypass graft (CABG) surgery. We grouped diagnosis based on surgical, PCI or other. We excluded patients scheduled to return the next business day, because this would not allow time to administer the intervention. Per standard program routines, we recorded demographics, insurance, risk factors, diagnosis and procedures, American Association of Cardiovascular and Pulmonary Rehabilitation exercise-risk category, time to enrollment and exercise capacity.

To increase our potential impact and utilize resources efficiently, we focused efforts on participants identified at-risk for nonadherence rather than all participants. Not finding a suitable risk stratification tool in the literature, we created a tool by including barriers identified in the literature regarding nonadherence. Eight barriers were selected: lack of transportation; people who live alone; being elderly; financial hardship; levels of stress; depression; current smokers; and patients with a diagnosis of a MI, heart failure or angina (Table 1).^20-26 Lack of transportation was defined as the patient expressing concerns related to travel, distance and/or being reliant on someone. We defined elderly as ≥75 yr of age. Financial hardship was accounted for if the patient expressed concern regarding copayments or insurance coverage. Schedule conflicts were accounted for if the patient expressed worry regarding their schedule, particularly those who needed to return to work. Stress level was defined by the patient if they reported caregiver strain, work-related stress or generalized life stressors during their orientation. We defined depression by the presence of documented diagnosis of depression or Patient Heath Questionnaire (PHQ9) score ≥10. Nicotine dependence was defined as a current smoker or having quit within the last 3 mo. Diagnosis of MI, heart failure or stable angina was determined from the patient’s medical record. Not finding any suitable weighting factors in the literature, we gave each factor an equal weight of 1 point. At-risk for non-adherence was defined as a score of ≥3 but it was also recognized that this tool needed to be validated. CR staff completed the risk score at the time of the orientation for all patients.

Table 1.

Risk Criteria for Nonadherence to Cardiac Rehabilitation

Criteria	Scorea
Diagnosis of MI, PCI, heart failure or stable angina pectoris
Current smoker or quit <3 mo ago
Diagnosis of depression and/or score >10 on PHQ9
Reported levels of work stress, financial stress or general stress
Travel issues identified
Recognition of schedule conflicts
Lives alone and/or no support group
>75 yr of age
Total Score: 0-2 Low Risk 3-5 Medium Risk 6-8 High Risk

Abbreviations: MI, myocardial infarction; PCI, percutaneous coronary intervention; PHQ9, Patient Health Questionnaire.

^aEach criteria category is weighted equally as 1 point.

Randomization and Blinding

Participants were randomized using random sequence generation and allocation concealment. Our statistical department created a randomized numbers log and placed group assignment into opaque sequential envelopes. When a patient met inclusion criteria, the envelope was opened to determine assignment. All CR providers were blinded to the intervention and were unaware who was randomized into the intervention arm. The purpose of blinding the staff was to assure that, if the patient did not return, they still received the standard of care, which consisted of a routine short phone call, typically ≥2 wk after their initial orientation that was not semistructured or purposeful in the delivery.

Study Intervention

Participants in the intervention group were telephoned as soon as possible, usually within 1–3 days after their orientation and always prior to their second session. There was 1 primary clinician who performed all the calls to ensure consistency. The intervention call was attempted on the next business day, with up to 3 attempts until their next attendance at CR. If after the third attempt they were unable to be reached, patients were no longer contacted. If the patient was not home, messages were not left and if another individual answered and the participant was unavailable, a generic message was left regarding confirmation of the next scheduled visit. This was done because the primary clinician was not always available to take return phone calls and we did not want to unblind patients or CR staff to group assignment. If the participant did not return on their anticipated second session, they followed the standard of care protocol.

The intervention was guided by the Transtheoretical Model (TTM), as it provided recognition of the participant’s motivation and gave a way to create meaningful communication.²⁷ The goal was not to attempt to advance patients to the next stage of change but, rather, to attempt to identify the level of motivation. During the phone call, we used a semistructured script to reinforce benefits of exercise, individualized risk factors and personal goals that were established at orientation (SDC Appendix A). During the intervention, we explored the patient’s enthusiasm and created awareness of how their lifestyle behaviors and risk factors should be modified. This was done by reviewing risk factors and explaining relevance to their disease process. By identifying ambivalence to change, the caller could take appropriate steps to initiate support (SDC Appendix B).

Outcome Measures

Because we recognized that a single phone call would most likely have transient effects, our primary outcome was return on the scheduled next visit (second session), as determined after the initial orientation versus long-term attendance or completion of the program. We defined scheduled next visit based on when the participant stated they would return to the program and if they did return on that date. CR staff members blinded to group allocation documented anticipated next visit based on when the participant said they would return, thus avoiding any bias in outcomes assessment.

Our primary aim was to reduce early dropout, however we wanted to get a complete picture of the intervention, and so we recorded several secondary outcomes. These included overall return, length of intervention, total sessions, completion of CR and change in estimated functional metabolic equivalents (METs). Overall return was defined as the patient returning to the program despite anticipated date of return. Total sessions of CR (out of a potential of 36) were measured by all sessions completed, including orientation. METs were estimated using an online calculator with calibrated exercise equipment comparing baseline MET to last day of completion.²⁸ Overall, completion was defined as having a formal exit assessment at the end of the program, even if the patient did not attend all 36 sessions. Length of the intervention call was measured in minutes.

Statistical Analysis

Because this was initially planned as a pilot study, our primary interest was feasibility and effect. Therefore, we did not do a statistical power analysis. We initially recruited 60 participants; however, after 4 mo we found enrollment was proceeding quickly and minimal resources were needed for implementation. Thus, we increased our sample size to 100 given the projected timeframe. This expanded our ability to do hypothesis testing and these changes were approved by the IRB prior to expanding enrollment. No interim analysis was done to inform this decision.

Analysis was completed using intent-to-treat (ITT) and per-protocol analysis. We included a per-protocol group since some patients did not receive the intervention despite 3 attempts. The main reason this occurred was because the time between orientation and the second scheduled session of CR was short, often just a few days, and not all patients could be reached during this timeframe. For the per-protocol analysis, patients in the intervention group who could not be reached were analyzed with the usual care group.

Data were entered into a secured database (Redcap) and analyzed using JMP 12.0.1 (SAS Institute). Normally distributed variables were expressed as mean ± standard deviation and as proportions. Skewed variables were reported with a median and interquartile range. Differences between the 2 groups were analyzed using a 2-sample t-test, relative risk ratio, and the Wilcoxon rank sum test, as appropriate. Linear regression was used to examine the relationship of risk score to total number of CR sessions. A 2-sided P value of <.05 was considered statistically significant.

RESULTS

Participant Flow and Followup

From January to December 2016, among 491 patients screened in our program, we identified and randomized 100 eligible patients with 49 in the intervention group and 51 in the usual care group (Figure 1). Of the 49 who were randomized to the intervention, 11 (22%) were unable to be reached. These 11 participants did not receive the intervention as planned and were considered part of the usual care group for the per-protocol analysis. On average, the telephone intervention lasted 4.8 ± 2.9 min. Baseline characteristics by group are reported in Table 2. Across multiple characteristics, the 2 groups were not statistically different other than body mass index (BMI). The average participant was 62 yr of age, female, had a PCI, and was insured by Medicare.

Figure 1.

CONSORT patient flow diagram. Abbreviations: CONSORT, Consolidated Standards for Reporting of Trials; ITT, intention-to-treat.

Table 2.

Baseline Patient Characteristics^a

	Intervention (ITT) (n = 49)	Usual Care (n = 51)	P Value
Age, yr	62 ± 14	61 ± 17	.61
Male sex	19 (39)	27 (53)	.16
Diagnosis and/or procedure			.33
Surgical	13 (27)	12 (24)
Percutaneous coronary intervention	30 (52)	27 (46)
Other	16 (12)	12 (24)
Insurance			.89
Medicare	21 (42)	25 (49)
Medicaid	7 (14)	8 (16)
Commercial	11 (22)	9 (18)
Other	10 (20)	9 (17)
Risk factors
Body mass index, kg/m2 (n = 83)	28 ± 6	31 ± 7	.046
Hypertension	39 (80)	40 (78)	.89
Hyperlipidemia	40 (82)	42 (82)	.92
Diabetes mellitus (n = 88)	16 (38)	16 (35)	.75
Smoking (n = 99)			.56
Current	7 (15)	9 (18)
Former	18 (38)	14 (27)
Never	23 (48)	28 (55)
AACVPR risk category			.054
Low	3 (6)	3 (6)
Medium	14 (29)	5 (10)
High	32 (65)	43 (84)
Daily physical activity, min (n = 79)	14 ± 16	16 ± 15	.62
Baseline exercise capacity, METs (n = 82)	2.4 ±0.8	2.4 ± 0.9	.81
Time to enroll, d	24 (13 - 34.5)	26 (14 - 57)	.40
Risk score for nonadherence	3 (3 - 4)	3 (3 - 4)	.27

Abbreviations: AACVPR, American Association of Cardiovascular and Pulmonary Rehabilitation and Prevention; ITT, intention-to-treat; METs, metabolic equivalents from baseline exercise assessment.

^aData reported as number (%), mean ± standard deviation or median (interquartile range).

Primary Outcome

In the ITT analysis, patients who received the intervention were more likely to have their second session of CR as scheduled compared with standard of care (80% vs 49, P = .002). Overall return rate was somewhat, though not statistically significant higher in the intervention group compared with the usual care group (81% vs 67). However, in the per-protocol analysis, overall return rate demonstrated a moderate effect size (87% versus 66%, P = .02) (Table 3).

Table 3.

Analysis of Primary and Secondary Outcomes^a

	ITT (n = 49)	Usual Care (n = 51)	RR (95% CI)	P Value	Per-Protocol (n = 38)	Usual Care (n = 62)	RR (95% CI)	P Value
Primary Outcome
Return as expected	39 (80)	25 (49)	1.62 (1.18-2.22)	.002	32 (84)	32 (52)	1.63 (1.23-2.15)	.0001
Secondary Outcomes
Eventual return	40 (81)	34 (67)	1.22 (0.97-1.54)	.09	33 (87)	42 (66)	1.31 (1.05-1.63)	.02
Completed CR	23 (47)	23 (45)	1.04 (0.68-1.59)	.85	18 (46)	28 (45)	1.04 (0.68-1.62)	.83
Total CR sessions	8 (2.5-31)	7 (2-36)		.92	8.5 (3-34.5)	6.5 (2-36)		.59
Δ Estimated METs	0.75 ± 0.85	0.68 ± 0.95		.71	0.78 ± 0.97	0.68 ± 0.86		.65
Days to 2nd session	3.36 ± 1.38	3.40 ± 0.89		.86	3.02 ± 2.21	3.40 ± 0.89		.32
Length of intervention, min	4.75 ± 2.86

Abbreviations: CI, confidence interval; CR, cardiac rehabilitation; ITT, intention-to-treat; METs, metabolic equivalents; RR, relative risk; Δ, change.

Data reported as number (%), mean ± standard deviation or median (interquartile range), unless otherwise noted.

Secondary Outcomes

As expected, there was no difference between groups when comparing total sessions of CR completed or other secondary outcomes. Post-hoc analysis showed that our risk stratification tool was predictive of total number of CR sessions. Patients with higher risk scores had fewer sessions and, overall, the total number of session was much lower than the usual mean value of 14 in our program. For each additional risk factor for nonadherence, there was an associated reduction in estimated sessions of CR attended of 5, such that a patient with a score of 4 (vs 2) would be expected to attend approximately 10 fewer sessions of CR (R² = 0.07; P = .008) (Figure 2).

Figure 2.

Correlation of risk score for nonadherence at cardiac rehabilitation (CR). Session of CR = 33.2 - 5.1 * risk score. R² = 0.07, P = .0075.

DISCUSSION

Consistent with our hypothesis, this randomized controlled blinded pilot study demonstrated that a telephone intervention delivered shortly after orientation significantly reduced early dropout from CR compared to the usual care. The impact was larger among patients who actually received the intervention and at earlier time points (next scheduled appointment vs overall return rate). The protocol was easy to deliver and did not require additional resources. Having a nurse deliver the intervention had value. Although the intervention was aimed at encouraging adherence, it was also considered to be a medium for patients to ask clinical questions such as: questions regarding their discharge and often medication reconciliation questions. This intervention can also be used as an educational tool, as we also reviewed individualized risk factors for CVD. Thus, a nurse or clinical provider should be delivering the intervention compared to administrative support. It appears that a single, nurse-led telephone intervention given shortly after a patient’s initial exercise session at CR improves return rates and attenuates early dropout.

Our results have significant financial and clinical implications as our current environment in CR continues to shift towards coordination of care across different health care settings to improve outcomes and reduce cost in capitated or shared risk insurance environments.²⁹ Prior literature has demonstrated a strong dose-response between number of CR sessions attended and long-term outcomes. The more sessions attended, the lower risk of death and adverse events compared to those who attend fewer sessions.^{4, 27} Thus, an intervention to improve return and adherence is essential in this era of health care reform²⁹ and to improve patient outcomes.

To our knowledge, this is the first time that a single telephone intervention has been evaluated in CR to improve return rates and to prevent early dropout. This is consistent with prior nursing telephone interventions that were successful in helping patients overcome barriers after discharge and in promoting CR adherence. However, unlike our study, these protocols were used in combination with other methods such as face-to-face meetings and informatics, like text messaging that required additional resources to implement.^5–19

We estimate this intervention is likely to be cost-effective and easy to implement. Our protocol appeared highly feasible, although there were some issues in reaching patients in the intervention group prior to their scheduled return. Moreover, the telephone intervention averaged less than 5 min. Thus, if a similar CR program were interested in using this intervention, we expect no additional resources would be required.

Although not previously validated, our risk score did identify patients at risk for nonadherence. Notably, median attendance among our group was only 7 sessions, which is much lower than our program median of 14–16 sessions per patient.³¹ This is possibly due to our inclusion criteria of identifying patients who were at risk for nonadherence. The overall average at-risk score was higher in those who did not return to the program in the intervention group. The use of this tool allowed us to focus our efforts on those at risk.

A major strength of this study includes the randomization and blinding. The intervention was delivered without any patient expectation and they behaved as patients normally would without knowing they were being studied. Moreover, this study design essentially eliminated the Hawthorne effect or the chance that patients changed their behavior because they were being observed. Staff was also blinded to those patients receiving the intervention to ensure all patients received the standard of care. As a result, the likelihood of residual bias or unmeasured confounding on our results was likely to be low. However, the design of the study did not account for phone calls made by the patient into the program that may have altered outcomes. For example, if a patient called into the program to say he/she had to postpone their anticipated date of return, this may not have been documented in the patient’s chart; therefore, this change in date was not accounted for in the study.

There were limitations that should be discussed. The main limitation of this study was the lack of a power analysis. This was initially designed as a pilot, examining effect and feasibility. However, because recruitment was proceeding quickly, this allowed an expansion of the trial and the addition of formal statistical testing. Still, sample size was small and there is a possibility that our positive results reflect this study design or other possible random sampling error. Another limitation is that there were 11 patients in the intervention group who were unable to be reached. If this pilot were to be replicated, ensuring accuracy of contact information and best available contact time would be useful to assure delivery of the intervention.

Another limitation was that a single individual delivered the intervention and it may not work as well when implemented by other staff members or at other centers. A semistructured script was used to ensure consistency with delivery of the intervention. Although the TTM was the underlying framework for the script, we cannot draw conclusions that the intervention used based on the subject’s level of change was the cause of improving adherence to the program. The TTM model has limitations within its framework; for example, ignoring the subject’s social setting, such as socioeconomic status and income. In addition, the framework does not have an established criterion of how to determine a person’s stage of change and assumes that individuals make rational plans in their decision making process which is not always accurate. Another limitation is this trial was implemented in a single-center and performed in western Massachusetts, which may have populations different than other geographic areas. Lastly, the stratification tool used to identify patients at-risk for nonadherence was not validated.

Given the clinical significance of our findings, other CR programs may want to reproduce our pilot on a larger scale or consider developing a clinical policy of integrating a telephone intervention after a patient’s exercise orientation. Programs may even want to consider expanding this to a series of phone calls or other various communication strategies to improve early dropout rates and overall completion. Further research is needed in developing a valid and reliable risk stratification tool in the CR community. Currently, there is no tool available for risk stratification for nonadherence.

Conclusion

We demonstrated that a nursing telephone intervention delivered after CR orientation was both highly feasible and effectively reduced early dropout. Although there were issues in reaching patients in the intervention group prior to their scheduled return, resources needed were minimal to make a significant impact. This intervention has the potential to be more effective than a recorded reminder, text message or e-mail since it allows an opportunity to create a partnership with the participant and the clinician with open, 2-way communication. This straightforward strategy represents an attractive adjunct to the current management of outpatient CR patients.