Author + information
- Received October 18, 2018
- Revision received November 25, 2018
- Accepted December 9, 2018
- Published online June 24, 2019.
- Diann E. Gaalema, PhDa,b,∗ (, )@vtcenterbh,
- Rebecca J. Elliott, BAa,
- Patrick D. Savage, MSc,
- Jason L. Rengo, MSc,
- Alex Y. Cutler, BSa,
- Irene Pericot-Valverde, PhDa,
- Jeffrey S. Priest, PhDd,
- Donald S. Shepard, PhDe,
- Stephen T. Higgins, PhDa,b and
- Philip A. Ades, MDc
- aDepartment of Psychiatry, University of Vermont, Burlington, Vermont
- bDepartment of Psychology, University of Vermont, Burlington, Vermont
- cDivision of Cardiology, University of Vermont Medical Center, Burlington, Vermont
- dDepartment of Medical Biostatistics, University of Vermont, Burlington, Vermont
- eHeller School for Social Policy and Management, Brandeis University, Waltham, Massachusetts
- ↵∗Address for correspondence:
Dr. Diann E. Gaalema, Department of Psychiatry, University of Vermont, 1 South Prospect Street, OH3 UHC MS 482, Burlington, Vermont 05405.
Objectives This study sought to examine the efficacy of financial incentives to increase Medicaid patient participation in and completion of cardiac rehabilitation (CR).
Background Participation in CR reduces morbidity, mortality, and hospitalizations while improving quality of life. Lower-socioeconomic status (SES) patients are much less likely to attend and complete CR, despite being at increased risk for recurrent cardiovascular events.
Methods A total of 130 individuals enrolled in Medicaid with a CR-qualifying cardiac event were randomized 1:1 to receive financial incentives on an escalating schedule ($4 to $50) for completing CR sessions or to receive usual care. Primary outcomes were CR participation (number of sessions completed) and completion (≥30 sessions completed). Secondary outcomes included changes in sociocognitive measurements (depressive/anxious symptoms, executive function), body composition (waist circumference, body mass index), fitness (peak VO2) over 4 months, and combined number of hospitalizations and emergency department (ED) contacts over 1 year.
Results Patients randomized to the incentive condition completed more sessions (22.4 vs. 14.7, respectively; p = 0.013) and were almost twice as likely to complete CR (55.4% vs. 29.2%, respectively; p = 0.002) as controls. Incentivized patients were also more likely to experience improvements in executive function (p < 0.001), although there were no significant effects on other secondary outcomes. Patients who completed ≥30 sessions had 47% fewer combined hospitalizations and ED visits (p = 0.014), as reflected by a nonsignificant trend by study condition with 39% fewer hospital contacts in the incentive condition group (p = 0.079).
Conclusions Financial incentives improve CR participation among lower-SES patients following a cardiac event. Increasing participation among lower-SES patients in CR is critical for positive longer-term health outcomes. (Increasing Cardiac Rehabilitation Participation Among Medicaid Enrollees; NCT02172820)
Cardiac rehabilitation (CR) is a structured program of supervised exercise and risk factor control that is standard of care following a major cardiac event such as myocardial infarction or coronary revascularization (1). Attendance at CR following a major cardiac event is associated with a 26% reduction in cardiovascular mortality, a 31% reduction in 1-year hospital readmissions (2,3), and a 21% to 34% reduction in 5-year all-cause mortality (4). CR is also beneficial for individuals who have experienced chronic systolic heart failure (5,6), although participation rates for these patients is extremely low (7).
Lower-socioeconomic status (SES) patients bear a higher proportion of morbidity and mortality resulting from cardiovascular disease (8). Lower-SES patients have high-risk cardiac profiles including low fitness and high rates of smoking, obesity, and diabetes mellitus (9) that result in increased rates of cardiac events (9,10). Lower-SES patients also have higher rates of chronic systolic heart failure (11) and worse post-hospitalization outcomes, with a 1-year mortality rate double that of more affluent patients (9,10,12). These SES disparities are largely accounted for by modifiable behaviors (e.g., smoking, physical inactivity) (9,10,12,13). Thus, increased risk following a cardiac event in lower-SES individuals is likely modifiable by attending secondary prevention interventions such as CR.
Despite the proven benefits of CR, overall attendance rates for eligible patients have been low, ranging from 14% to 34% (14–16). Attendance among lower-SES patients is even poorer. Within the population of Medicare patients (≥65 years of age), those who are also enrolled in Medicaid (i.e., lower SES) attend CR at the rate of 20% to 45% of higher-SES patients (4,15,16). This pattern is similar in studies using education as an SES marker, where those with limited educational attainment were one-third less likely to attend CR (14,17). Part of this lower attendance could be attributable to disparities in initial CR referral (18).
Patients must also be retained in CR. Overall, completion rates for patients who have started CR are estimated at approximately 75% (19,20), but disparities exist. Lower-SES patients complete fewer sessions than their higher-SES counterparts (21). Given that completing more CR sessions is associated with greater health benefits (4,22), the crucial outcome should not be to increase enrollment alone but to maximize the number of sessions completed. In the latest systematic review of interventions to improve adherence, only 1 study aimed to increase the number of CR sessions completed, and the intervention was unsuccessful (23). Additionally, in the present authors' knowledge, not a single study, other than the present one reporting pilot data from the current trial (24), has focused on increasing CR participation in lower-SES patients.
Incentive-based interventions can be a highly efficacious approach to altering health behaviors among disadvantaged populations. One approach, contingency management, involves providing financial incentives contingent on objective evidence of change in behavior. The use of financial incentives first garnered attention as a method that increased treatment participation and abstinence from drug use among cocaine-dependent outpatients (25). The incentives-based model was subsequently shown to be effective at increasing treatment participation and abstinence from a wide variety of substances (26). Additionally, the use of financial incentives has been successful in promoting adherence to health care visits in predominantly lower SES populations (27,28). In CR, where health effects appear to be dose-dependent (4,22), the ability to sustain participation would be of considerable clinical benefit.
Lower-SES patients have higher-risk cardiac profiles and worse outcomes following an acute cardiac event. Efficacious interventions for getting these patients into proven treatments are needed. Financial incentives are efficacious for modifying other health behaviors in lower-SES populations, especially for promoting attendance at health care visits. Thus, the present randomized clinical trial was designed to test the efficacy of financial incentives for increasing CR participation among recently hospitalized cardiac patients with Medicaid insurance.
The present study was a 2-arm randomized clinical trial (NCT02172820). All participants were expected to complete assessments at baseline and at 4 months following baseline assessment and were offered compensation and travel reimbursement for assessments. Participants randomized to the incentive intervention received financial incentives for completion of CR sessions. Patients randomized to the control condition did not earn incentives. The CR program has been described in detail elsewhere (1,29) but, briefly, consists of up to 36 outpatient exercise sessions held, generally, in 2 to 3 weekly sessions. The program is progressive and adaptive, with an eventual goal of exercise duration of 45 min at 70% to 85% of baseline peak exercise heart rate. The program also includes educational sessions dealing with stress management, healthy nutrition, medication use, symptom recognition, heart failure education, benefits of exercise, and risk factor control. For this CR program, Medicaid was billed for CR sessions attended, and there were no co-pays (Vermont policy), and although patients were responsible for transportation costs, parking was free.
Participation was defined as attendance at scheduled sessions and completion of recommended exercise sessions and other scheduled activities. Program staff verified participation. Incentives were distributed as cash immediately upon session completion. Participation in an introductory group meeting and the initial exercise session earned participants $20 each. Participation in subsequent exercise sessions was incentivized by using an escalating payment schedule. Participation in the second exercise session earned $4, and each subsequent session completed increased incentive value by $2, up to a maximum of $50 per session (total possible incentive earnings was $1,238). Failure to attend a session (unless advanced notice was given) resulted in no earnings for that session, and the amount possible to be earned in the next scheduled session was reset to $4. Successful participation in 2 consecutive sessions following a reset returned the incentive value to the amount it was prior to the reset. This escalating incentive schedule combined with a reset contingency has been experimentally demonstrated to promote periods of continuous adherence (30).
Participants were recruited from April 2014 to January 2017 (Figure 1). Study participants included 130 patients enrolled in Medicaid who had been hospitalized for CR-qualifying events (acute myocardial infarction, coronary artery bypass grafting surgery, heart valve replacement or repair, percutaneous coronary intervention [with or without stent placement], or stable outpatient chronic systolic heart failure). Eligibility criteria included enrollment in Medicaid or other state-supported insurance policy and residence in the catchment area of the University of Vermont Medical Center Cardiac Rehabilitation Program without plans to leave the area. Patients were ineligible if they had participated in CR in the previous 5 years (defined as completion of ≥5 sessions), were non-English speaking, had dementia or current untreated axis 1 psychiatric disorder other than nicotine dependence, had longevity-limiting systemic disease that would preclude participation (e.g., advanced cancer), or had comorbidities that would preclude participation in supervised exercise (e.g., angina at rest, uncontrolled ventricular arrhythmia, severe arthritis or lung disease, or New York Heart Association functional class IV chronic heart failure).
Recruitment and randomization
Potentially eligible patients were approached by research staff following the patient's qualifying event. Upon confirming eligibility and providing written consent, participants were allocated by using sealed envelopes, based on 1:1 randomization, to intervention or control condition. Following consent, all participants were referred to CR. Staff collecting outcome data were blind to condition assignment. The study was approved by the University of Vermont Institutional Review Board.
Study participants were not separated from other patients during CR sessions. However, they comprised a minority of the CR population at any one time and did not tend to interact with each other. Research staff interacted with all CR participants during sessions, regardless of study participation and condition. To reduce potential negative bias, the general benefits of being in the study (e.g., assessment earnings, thorough health checks during assessments) were emphasized. Participants did not report negative feelings toward other participants, although some did report mild disappointment in being assigned to the control group.
The primary study outcomes were the number of CR sessions completed and completion of CR, defined as completion of ≥30 exercise sessions. Secondary outcomes included changes that occurred between baseline and 4-month assessments in fitness (peak VO2 was measured directly by expired gas analysis or was estimated by metabolic equivalents), body composition (body mass index [BMI], waist circumference), sociocognitive measurements (31), Behavior Rating Inventory of Executive Function score (32), Stop Signal Task reaction time (SSRT) (33), and quality of life (MacNew Cardiac Health Status Questionnaire) (34). Basic clinical and demographic characteristics (age, sex, educational attainment, race/ethnicity, smoking status, BMI, CR-qualifying diagnosis) were collected at the time of consent. Hospital contacts (ED visits and inpatient hospitalizations) were obtained for each patient over the course of 1 year beginning 1 month after hospital discharge. Contacts were extracted from the electronic health record by medical center staff otherwise unaffiliated with the study.
Characteristics of individuals randomized to 1 of the 2 conditions were compared by using the chi-square goodness-of-fit test or Fisher exact test for categorical variables and the Student t-test or Wilcoxon rank sum test for continuous variables. The number of sessions completed was analyzed as a continuous variable by using the Wilcoxon rank sum test or converted to a binary categorical variable (<30 vs. ≥30 sessions) and examined using the chi-square test. Logistic regression was used to examine predictors of CR completion. Univariate logistic regression was conducted with 7 possible predictors (treatment condition, sex, qualifying diagnosis, smoking status prior to hospitalization, age at consent, educational attainment, and BMI). Variables that contributed to the outcome at a p value of ≥0.25 were included in an initial model. This was winnowed to predictors achieving significance at a p level of <0.05. All variables that were either initially excluded or dropped were tested again, one-by-one, using a model with only significant predictors. Interactions between significant predictors that remained in a tentative final model were tested.
Changes over time were assessed by using paired differences in scores from intake to 4 months after intake for BMI, waist circumference, fitness (peak VO2), MacNew Cardiac Health Status Questionnaire responses (34), self-reported global executive composite executive function (ExecF), SSRT, and the Achenbach system of empirically based assessment (ASEBA) instrument of anxiety and depression. Analyses were conducted using the entire sample by treatment condition and by completion status (i.e., <30 vs. ≥30 sessions). Due to non-normal distributions, the Wilcoxon signed rank test was used. Contributions of other variables (treatment condition, sex, surgical status, current smoker, BMI, age) to changes in secondary outcomes were examined by using analyses of covariance (ANCOVA).
Given the proportion of participants with zero hospital contacts (39.2%), hospitalizations and ED visits combined were analyzed by using simple negative binomial regression models. Two models were used to predict the number of hospital contacts, 1 model with treatment condition and 1 with completer status (<30 vs. ≥30 sessions) as the sole predictor.
The expected discounted life years gained (EDLYG) per CR participant was derived by using propensity-matched results (16) based on sessions completed (n), as EDLYG = a [(1 − exp (-b-cn)], where a = 9.9873 years (additional years per additional 5-year survivor), exp = expected, b = 0.0459 (the 5-year mortality reduction from initiating CR), and c = 0.0023 (the 5-year mortality reduction from each CR session). The cost of each CR session to the health system included incentive earnings (where applicable) plus 30% for incentive administration based on a comparably complex trial (35), as well as the payment allowed by Medicare ($102). The preliminary incremental cost-effectiveness ratio (ICER) equals EDLYG divided by the program’s incremental cost. Confidence intervals (CIs) were derived from 1,000 bootstrap replications.
The study was designed to have >80% power to detect a difference in CR attendance participation rates of 20%. Across all tests, statistical significance was defined as a p value of <0.05 (2-tailed) and 95% CI.
Demographic and clinical characteristics were collected at consent from all 130 participants and did not differ between treatment conditions (Table 1). Measurements gathered at the intake assessment differed only by SSRT (a component of ExecF), which was higher (indicating more impulsivity) in the incentive condition (p = 0.010). Participants were representative of a high-risk population. Education levels varied widely, and patients had considerable psychiatric and other medical comorbidities. Elevated depression or symptoms of anxiety were present in 59% of participants; 40% had elevated problem scores on self-reported ExecF; average BMI was in the obese range; and 42% were current smokers. Eight of 130 patients (6%) carried a primary diagnosis of systolic heart failure. Left ventricular ejection fraction was ≤45% for 24 of 130 patients; and for this group, mean ejection fraction was 34.5 ± 10.2% (range 17% to 45%). Of these 24 patients, 23 were taking evidence-based medical therapy consisting of beta-blockers and angiotensin inhibitors or receptor blockers.
Primary outcomes (i.e., number of CR sessions completed and proportion completing ≥30 sessions) and hospital contacts (hospitalizations and ED visits) were collected for all 130 participants. The trial ended after 1 year's hospital data collection was completed. Of the 130 patients randomized, 112 patients (86%) attended the initial assessment (55 in the incentive condition; 57 in the controls). Of those who completed an intake assessment, 103 participants (92%) completed the 4-month follow-up assessment. Assessment attendance did not differ by condition.
CR participation and completion
Participants in the incentive condition completed more CR sessions (22.4 vs. 14.7, respectively; p = 0.013) and were almost twice as likely (55.4% vs. 29.2%, respectively; p = 0.002) to complete CR than those in the control condition (Figure 2). The percentages of patients who attended at least 1 session did not differ by condition. Only 2 variables predicted CR completion: being assigned to the incentive condition (odds ratio [OR]: 3.38; 95% CI: 1.55 to 7.38) and not being a current smoker at the time of hospitalization (OR: 4.55; 95% CI: 2.02 to 10.25). There were no significant interactions between these 2 variables.
Fitness, quality of life, and body composition
Measurements of fitness (peak VO2) and cardiac event-specific quality of life (MacNew Cardiac Health Status Questionnaire responses) improved in all patients between intake and follow-up (p < 0.001) but did not differ by treatment condition or completion status (Table 2). The only significant predictor of change in peak VO2 was smoking status, with current smokers showing less improvement (p = 0.028). Measurements of body composition (waist circumference and BMI) did not change significantly, although there were trends toward a decrease in waist circumference in both the incentive group versus the control group (p = 0.054) as well as those in program completers versus dropouts (p = 0.077) (Table 2). No other variable significantly predicted changes in BMI or waist circumference.
There were no significant changes in the combined measurement of anxiety and depression symptoms (using the ASEBA scale for depression and anxiety), although there was a trend toward symptom reduction in the incentive condition (p = 0.083) (Table 2). The objective measurement of ExecF (SSRT) did not change significantly in this sample. However, ExecF scores, which were self-reported using the Behavior Rating Inventory of Executive Function questionnaire, improved significantly in patients in the incentive condition (p < 0.001). No other variable significantly predicted changes in cognitive outcomes.
The number of times patients contacted the hospital system (hospitalization or ED visit) ranged from 0 to 21, with a median of 1 and a mean of 2.05 (Figure 3). More than one-half of the sample (60.8%) entered the hospital system at least once during the year following their initial event. The number of hospital contacts was lower in those who completed CR (1.36 vs. 2.55, respectively; p = 0.012), with a trend for fewer contacts in patients in the incentive condition (1.62 vs. 2.48, respectively; p = 0.079). No other variable significantly predicted hospital contacts.
Mean earnings in the incentive group were $716 per patient (all in U.S. dollars) (95% CI: $694 to $740). Additional CR sessions completed by incentive patients cost the health care system $1,730 (95% CI: $1,057 to $2,388) or $222 (95% CI: $183 to $421) per additional session completed. The added EDLYG by incentive patients was 0.1809 (95% CI: 0.0243 to 0.3324). The preliminary ICER was $9,336 (95% CI: $6,471 to $27,085) per life-year added.
Financial incentives increased the mean number of CR sessions completed and almost doubled completion rates among lower-SES patients for whom outpatient CR was medically indicated. Given the paucity of research of successful interventions to increase sessions of CR attended generally and the nearly nonexistent studies undertaken among lower-SES patients in particular, these results are promising. Additionally, given the high-risk nature of this population, as demonstrated by high rates of current smoking, psychiatric comorbidities, and hospital contact, improving CR attendance has the potential to make significant differences in longer-term health outcomes.
Completion of CR is an especially important outcome as completers showed multiple health benefits compared with noncompleters. In one study of more than 13,000 patients with diabetes mellitus, those who completed CR had reduced mortality rates (hazard ratio [HR]: 0.55), hospitalizations (HR: 0.91), and cardiac event-specific hospitalizations (HR: 0.78) compared to noncompleters, even after adjusting for demographics and presence and severity of medical conditions (20). Other studies have found CR completion was associated with a lower mortality risk (adjusted HR: 0.59), all-cause hospitalization (adjusted HR: 0.77), and cardiac-related hospitalization (adjusted HR: 0.68) (36). Those findings are reflected in the current trial, where CR completion was significantly associated with reduced hospitalization usage. CR completion is also associated with improvements in quality of life (37) and cognitive improvements (38), areas where positive trends were seen in the current study.
Several aspects of CR can help explain effects on hospital contacts (1). First, upon starting CR, patients were assigned to a case manager. Second, at each session, patients were weighed, vital signs were assessed; and patients were queried about symptoms, risk-related behaviors (e.g., smoking), and medication concerns. This repeated clinical contact allows for behavior intervention, symptom monitoring, and early referral to a treating physician, as necessary, which can reduce hospitalizations by keeping symptoms in check, halting escalation of symptoms, and preventing unnecessary ED visits.
Preliminary cost analyses excluded 2 offsetting effects: CR savings in hospitalizations and medical care of added life years. Nevertheless, the results suggest that the incentive intervention was inexpensive in relation to its expected health gains. Because the intervention's ICER ($9,336) was only one-sixth of the 2017 U.S. per capita gross national income ($58,270) (39), it was “very” cost effective (40) and almost 1 order of magnitude more cost effective than the increasingly used implantable cardioverter-defibrillator (ICER: $37,031 to $138,458) (41).
Characteristics of the population may help explain why incentives were necessary and successful. ExecF, the ability to plan and undertake complex tasks and other higher order functions, is a reliable predictor of accessing and completing health care regimens (42). Lower-SES populations are more likely to have ExecF deficits (43). Indeed, 40% of participants in the present study reported ExecF difficulties. One aspect of ExecF difficulties that is more common in lower-SES patients is the bias for the present, wherein individuals disproportionately weigh the relatively immediate and tangible costs of adhering to medical regimens relative to the delayed and relatively intangible but more substantive benefits of adherence (44). Incentives can harness these present-biased preferences by providing immediate, tangible, positive outcomes (earning incentives) for adhering to medical regimens (e.g., attending CR sessions) (45). In the present trial, participants in the incentive condition also reported improvements in ExecF, a finding that suggests they may experience improvements in self-care that extend beyond the intervention.
This study was conducted at a single clinical site with a fairly racially homogenous patient population. Baseline assessments were conducted in the CR program, which may have inflated CR participation rates in the control condition. Clinical staff could not be fully blinded to patient assignment, which might have affected staff/patient interactions, although staff followed standardized protocols for promoting attendance independent of condition assignment. Hospitalizations outside the University of Vermont Medical Center system might have been missed. Finally, sample size was determined based on the power needed to demonstrate effects of incentives on CR attendance. Studying effects of incentives on health outcomes such as fitness, psychocognitive parameters, hospital use, and more comprehensive cost effectiveness would have required a larger sample. For example, post hoc analyses suggest a study would require 180 participants per condition to be powered for fitness effects and 335 participants for cardiac event-specific quality of life assessment.
Financial incentives improve CR participation and completion in lower-SES, high-risk cardiac patients. Increasing access to CR in such high-risk cardiac populations is critically important to decreasing health disparities. Examining the reliability and generalizability of this effect appears warranted.
COMPETENCY IN MEDICAL KNOWLEDGE 1: Financial incentives improve CR participation among lower-SES patients.
COMPETENCY IN MEDICAL KNOWLEDGE 2: Lower-SES patients are at high risk for recurrent events yet are much less likely to attend and complete CR. Therefore, intense intervention may be required to increase lower-SES patient participation in CR.
TRANSLATIONAL OUTLOOK 1: Although lower-SES patients who completed CR had fewer ED visits and hospitalizations, larger trials with a longer follow-up period will be required to experimentally demonstrate the health and care service use benefits of increased CR participation.
TRANSLATIONAL OUTLOOK 2: Attendance in CR is unsatisfactory in lower-SES patients. As incentives have been shown to substantially increase CR participation in this high-risk population, the acceptability and feasibility of integrating an incentives-based approach to clinical care should be explored.
Supported by U.S. National Institutes of Health Center of Biomedical Research Excellence award P20GM103644 from the National Institute of General Medical Sciences. The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- body mass index
- cardiac rehabilitation
- emergency department
- expected discounted life years gained
- executive function
- incremental cost-effectiveness ratio
- oxygen uptake
- socioeconomic status
- Received October 18, 2018.
- Revision received November 25, 2018.
- Accepted December 9, 2018.
- 2019 American College of Cardiology Foundation
- Suaya J.A.,
- Stason W.B.,
- Ades P.A.,
- Normand S.-L.T.,
- Shepard D.S.
- Ades P.A.,
- Keteyian S.J.,
- Balady G.J.,
- et al.
- Rengo J.L.,
- Savage P.D.,
- Barrett T.,
- Ades P.A.
- Mackenbach J.P.,
- Kulhánová I.,
- Menvielle G.,
- et al.
- Díaz-Toro F.,
- Verdejo H.E.,
- Castro P.F.
- Fang J.
- Beatty A.L.,
- Truong M.,
- Schopfer D.W.,
- et al.
- Suaya J.A.,
- Shepard D.S.,
- Normand S.-L.T.,
- et al.
- Sun E.Y.,
- Jadotte Y.T.,
- Halperin W.
- Castellanos L.R.,
- Viramontes O.,
- Bains N.K.,
- Zepeda I.A.
- Pack Q.R.,
- Squires R.W.,
- Lopez-Jimenez F.,
- et al.
- Armstrong M.J.,
- Sigal R.J.,
- Arena R.,
- et al.
- Gaalema D.E.,
- Elliott R.J.,
- Monford Z.H.,
- Ades P.A.,
- Higgins S.T.
- Hammill B.G.,
- Curtis L.H.,
- Schulman K.A.,
- Whellan D.J.
- Gaalema D.E.,
- Savage P.D.,
- Rengo J.L.,
- et al.
- McMahon S.R.,
- Ades P.A.,
- Thompson P.D.
- Achenbach T.M.,
- Newhouse P.A.,
- Rescorla L.
- Guy S.C.,
- Isquith P.K.,
- Gioia G.A.
- Garner B.R.,
- Lwin A.K.,
- Strickler G.K.,
- Hunter B.D.,
- Shepard D.S.
- Kardis P.,
- Bruce A.,
- Michaels J.,
- Barnett S.D.
- Milani R.V.,
- Lavie C.J.
- World Bank
- Alosco M.L.,
- Spitznagel M.B.,
- van Dulmen M.,
- et al.
- Ades P.A.,
- Gaalema D.E.