Contribution of individual and cumulative frailty-related health deficits on cardiac rehabilitation completion

Patients were eligible for CR if they experienced an acute cardiovascular event, including coronary artery disease, myocardial infarction, percutaneous coronary intervention, coronary artery bypass graft or valve surgery, heart failure or a combination of other diagnoses with few referrals (e.g., arrhythmia, heart transplant). Patients were excluded if they had a critical illness (e.g., for renal failure, cancer) requiring other treatments or another condition that would preclude them from participating in the program, such as severe stroke. Patients were also excluded if a frailty index (FI) could not be calculated. A total of 4,004 patients enrolled in a 12-week exercise and education-based CR program between 2005–2015 were eligible for the current study [25].

Routinely collected CR data from the nurse practitioner, dietician, and physiotherapist staff were used in this study. Clinical data were subsequently entered into a research database by a research assistant that were used for the present study. Participant characteristics included admission demographics consisting of age in years, sex (% female), smoking status (non-smoker, ex-smoker, current smoker), marital status (Divorced/separated, widowed, single, married/living with partner), education (< grade 12, grade 12/GED, community college, bachelor’s degree, post secondary education), employment (disability, unemployed, part-time, full-time, retired, other), beta-blocker and lipid lowering medication use, and referring diagnosis (see above list for eligibility into CR).

Frailty was measured with a 25-item FI at admission using a guidelines-based approach and previously used in the CR cohort [25, 26]. The following items were included in the FI: 1) cardiovascular risk factors including triglycerides, total cholesterol, low density lipoprotein, and high-density lipoprotein, fasting blood glucose, systolic and diastolic blood pressure, resting pulse rate, pulse pressure, and mean arterial pressure; 2) cardiovascular symptoms with the New York Heart Association functional class system; 3) cardiovascular fitness from an exercise stress test; 4) quality of life with the SF-36 questionnaire to capture physical, mental, social and general health domains; 5) body composition with body mass index, waist circumference, and bioelectrical impedance (to measure percent fat and fat free mass); and 6) diet quality with the Food Frequency Questionnaire. The FI is a ratio of the total accumulated health deficits across multiple physiologic systems (scores range from 0 to 1; a higher score indicates a higher frailty level). A valid FI was determined if participants had less than 30% missing data for the FI constituents [27]. An in depth description of the FI and its constituents was previously published [25]. FI distribution by program completion can be found in Table 1. The primary outcome measure for the current study was CR program completion; defined as attendance of at least 65% of sessions, coupled with the completion of a discharge assessment [25].

The CR program was a phase two traditional center- and group-based 12-week behaviour change intervention, with sessions focused on education (1 session/week) and exercise (up to 2 sessions/week), each lasting approximately 60 min. Exercise sessions were structured as follows: a 10-min warm-up, 40 min of aerobic training, and 10 min of resistance training. The goals of the program were to reduce CVD risk factors by incorporating exercise, improving diet, and optimizing medical management. Patients were given an exercise prescription for both group-based sessions and for their home-based program. These prescriptions were individually tailored to the patient [25].

The 12-week CR program provided patients with a multidisciplinary healthcare team, consisting of a CR program lead, nurse practitioner, nurses, physiotherapist, and dietician, guided by the cardiac medical director [25]. Upon admission to CR, personal CR goals were established using the SMART-framework [28] and were facilitated with assistance from CR staff [25]. A graded exercise test using a cycle ergometer or treadmill was used to determine a patients’ maximum metabolic equivalent (MET) level achieved to aid in exercise prescription. Examples of patient goals in CR include weight loss (i.e. lowering BMI, reducing waist circumference), improving physical fitness (i.e. returning to previous sport/hobbies, improving physical strength), and improving cardiovascular profile (i.e. lowering cholesterol, control/manage diabetes, smoking cessation, and hypertension management) [25].

Demographic variables are presented as frequency (percent) or mean (standard deviation) for categorical and continuous variables, respectively. Demographics, and program completion across levels of frailty were compared using logistic regression. Individual variables contributing to the FI were entered into multivariable models controlling for age, sex, education, marital status, employment, smoking status, and referring diagnosis. Based on individual deficits, odds ratios for program completion and 95% confidence intervals were estimated. All variables were standardized such that the odds ratios presented represent a 10% change in each of the individual FI deficits. The cumulative effect of deficits in domains related to cardiovascular health, body composition, and quality of life categories were also investigated for eligible participants (Supplemental Table S1). The categorizations were as follows, cardiovascular (systolic blood pressure, diastolic blood pressure, mean arterial pressure, pulse pressure, pulse rate, cholesterol level, triglyceride level, low-density lipoproteins, high-density lipoproteins, blood glucose, and NYHA functional category); body composition (BMI (body mass index), waist circumference, fat-free mass, food frequency questionnaire); quality of life (METs, SF-36 physical function, SF-36 mental health, SF-36 role physical, SF-36 general health, SF-36 energy, SF-36 role physical, SF-36 change in health past year, SF-36 bodily pain). The cut-offs indicating a deficit for this analysis were published previously [25] (Supplemental Table S2); however, for non-binary categorizations used in the previous paper, categories were appropriately combined for analysis. That is, if a variable consisted of 4 categories, the lower two were combined to signify the presence of a deficit and the higher two combined to signify the absence of a deficit. For example, NYHA categories three and four were combined to represent the presence of a deficit for that variable. Where necessary, categories were collapsed such that no individual category contained less than 5% of the overall sample. Additionally, cumulative traditional cardiovascular (cardiovascular plus METs) and non-traditional cardiovascular (SF-36 variables and body composition variables) deficits were investigated. In these analyses, cumulative deficits was defined as the sum of the number of deficits an individual had in each previously described category. Covariate selection for final models was performed using backward stepwise selection, with a liberal p-value (0.15) [29]. Effect modification of individual frailty deficits with program non-completion was investigated by age, sex, referring diagnosis, and baseline frailty. All analyses were completed using SAS 9.4 (SAS Institute, Cary, NC, USA).

This study was an analysis of data collected as part of a CR program. Nova Scotia Health Authority privacy policies were followed during the conduction of the study. Thus, the risk for harming individual patients was minimal. No known privacy breaches occurred during this study.

There are limitations to be considered with the present study. Regarding selection bias, individuals who did not have sufficient information to calculate an FI could have differed significantly from the remainder of the population, therefore influencing our results. However, these individuals did not differ significantly based on demographic variables, and represented a small minority of the overall sample. A second limitation includes the method by which the accumulation of deficits analysis was completed. If patients had missing values for any of the constituent variables, the maximum number of deficits that could have been identified would be one less than the total number of deficits available. However, all participants considered had enough valid variables to construct a representative FI based on guideline recommendations, which should mean that the risk of bias in this case is low. It is also important to consider that participants in this study reside in a large urban center, which may impact the generalizability of these results. This study also does not capture other reasons why people dropped out of CR; for example, younger people may be less likely to complete CR due to work or family commitments. The current study could also have issues related to multiple comparisons; however, false discovery rate testing did not detect any significant changes in p-values for the individual variable analysis (Supplemental Table S4). Additionally, the FI used in the current study did not contain deficits related to other chronic conditions which are known to impact CR completion. Lastly, we could not determine the effect of changes in FI variables with program completion rates as participants did not attend their formal outtake assessment appointment where routine data used in the FI are collected.

Despite these limitations, our investigation advances the knowledge on how frailty contributes to completing CR. This information may be useful for program directors regarding which participants should be targeted with specific interventions for retention. That is, if administrators are aware of the most relevant frailty-related deficits, they can work to retain individuals who would otherwise be likely to drop out of the program. To identify these participants, data would have to be entered at program initiation, with a code written that could identify patients with the deficits discussed in this paper. Program administrators could then target individuals with the highest risk of drop out at program intake, based on their current deficits – both traditional and non-traditional – by offering them additional sessions or learning materials relevant to their current situation. Following this, a matched quasi-experimental study could be undertaken to compare completion rates before and after the addition of this new intervention, with the goal of identifying whether or not it was efficacious.