The primary purpose of this study was to create a CFS for use in children and adolescents with heart disease and evaluate the relationship between frailty and cross-sectional and longitudinal outcomes. A CFS was produced using the five frailty domains introduced by Fried et al. using modified domain-specific measures for this unique patient population using the rubric that we have previously published [5, 15]. Only two studies of frailty in children and adolescents are available for us to compare our findings. Sgambat et al. measured frailty in n = 557 children and adolescents aged 6–19 years with chronic kidney disease. The authors found that participants with ≥ 3 frailty indicators (13% of the sample) one year after study entry were 3.16 times more likely to develop frequent infections, including any reported bacterial, viral, or fungal infection, and 2.81 times more likely to have hospitalization over the subsequent 3-years [32]. A multicenter study by Lurz et al. measured frailty in two groups of children and adolescents aged 5–17 years with chronic liver disease, including patients listed for liver transplantation (n = 35) and patients with compensated chronic liver disease without evidence of decompensation (n = 36). Patients listed for liver transplantation had a significantly higher CFS compared to patients with compensated liver disease. However, they found no significant correlations between CFS and either the Pediatric End-Stage Liver Disease (PELD) or Model of End-Stage Liver Disease (MELD) scores in the cohort listed for liver transplantation [16].
Developing the CFS
Our first objective was to develop a frailty scoring structure specific for children and adolescents with heart disease, using the studies by Sgambat et al. and Lurz et al. as a model. Sgambat and colleagues combined the Fried frailty model with the adult chronic kidney disease frailty model from McAdams-Demarco et al., which includes markers of inflammation in addition to a measure of fatigue and two measures of body composition [32, 33]. The researchers assessed frailty domains at study entry and 1 year later, where frailty was indicated by the degree of change in scores between the two time periods [32]. Our study design mimicked that of Lurz and colleagues where frailty was assessed at a single time point, relied primarily on the Fried frailty model, and scores were derived for each frailty domain by converting outcomes to age and sex-specific z-scores [16]. However, in our CFS development, we found that utilizing z-scores was only feasible for three of the five frailty domains (6-MWT, handgrip strength dynamometry, and TSfT), as age and sex-specific normative values are not available for measures of exhaustion/fatigue via the PedsQL MFS or physical activity via the PAQ-C/A. Lurz and colleagues also used the PedsQL MFS and PAQ-C/A; however, the researchers derived z-scores from a sample of youth with sickle-cell disease for PedsQL MFS [34] and from a study validating the PAQ-C/A for a Dutch population [35]. We believed these studies may not serve as an appropriate comparison for our sample of children and adolescents with heart disease from the U.S. Thus, we chose to use the raw questionnaire scores to produce frailty points for these domains (Table 1).
We made two other modifications to the frailty scoring system developed by Lurz et al. First, points for the body composition domain were assigned using a bidirectional method where patients who were either overweight/obese or underweight were awarded higher frailty points, designating them as frailer. This approach was based on a study of 18,337 patients aged 10–35 years with CHD from the Society of Thoracic Surgeons Database published by O’Byrne and colleagues. The authors in this study observed that either underweight or obese weight status was independently associated with an increased risk of adverse outcomes, including unplanned cardiac operation, reoperation for bleeding, prolonged hospitalization, and wound infection [26]. Secondly, we expanded the CFS range from 0–5 (Fried et al.) and 0–10 (Lurz et al.) to a range of 0–25 [5, 16]. Allowing for greater variability and finer distinctions in CFS scores may provide greater sensitivity to detect change (i.e., worsening of frailty associated with disease progression or improvement as the result of an intervention).
Frailty and Outcomes
The cross-sectional analysis found noteworthy correlations between individual frailty domains and/or CFS for both medical record-derived and self-report outcomes (Table 4). We were heartened by the distribution of correlations between individual frailty domains and outcomes. Fried and colleagues’ five-component frailty phenotype was designed as a multi-dimensional assessment of physical function, with independent constructs ranging from strength, endurance, body composition, and fatigue, together representing factors associated with declines across multiple physiologic systems [6]. In our cross-sectional analyses, each outcome was associated with one to three individual frailty domains, highlighting the benefits of the comprehensive approach, and illustrating the relationships between domains and various, distinct outcomes.
Eight outcomes had notable correlations with the CFS, 7 of which had correlations > 0.4. Chronotropic index and VO2peak are both well-established predictors of mortality in congenital heart disease populations [36–39]. Taking a high number of medicines in a normal day was positively correlated with CFS. A 2019 study by Woudstra et al. examined data from 14,138 adults with CHD from the CONCOR registry and found the use of multiple medications (≥ 5) was present in 30% of the sample and was related to four-fold higher risk of all-cause mortality and five-fold higher risk of hospitalization [40]. The CFS was correlated with two well-established subjective markers of a patient’s functional state (NYHA classification) and quality of life (Child PedsQL and parent-proxy PedsQL). A systematic review by Latal and colleagues noted a higher likelihood of child-reported impairment in quality of life and parent-reported psychological maladjustment in children with more severe cardiac disease [41]. Lastly, the number of ancillary medical specialists was positively correlated with CFS suggesting patients with greater medical needs are frailer, or vice versa; although this may be speculative as we found no studies that explored the number of medical specialists on morbidities. Overall, our findings suggest that the CFS may have clinical utility in estimating markers of global clinical morbidities cross-sectionally.
Our longitudinal analysis identified six significant correlations between outcome measures and individual frailty domains and/or CFS. Although there are longitudinal studies in older adolescents and adults with CHD, few prospective longitudinal studies spanning childhood and adolescence are available for us to compare our findings. These studies in youth with heart disease are generally limited in scope to examination of procedural techniques [42, 43], different CHD diagnoses/complexity [44, 45], or specific predictors such as exercise capacity [46–48] or weight status [49]. Our study’s small sample size, resulting in part from large loss-to-follow-up, likely contributed to our limited findings. However, our analysis did identify a notable correlation between an increase in the number of heart failure medications prescribed and the CFS. Heart failure medications have been used as an indicator of increased acuity in several heart failure models, such as the New York University Pediatric Heart Failure Index [50] and the Seattle Heart Failure Model [51] and have been linked to mortality [52].
Although research suggests that body composition is a valuable predictor of short- and long-term outcomes in youth with heart disease [26, 49], we did not observe any correlations between body composition domain and outcomes that met our criteria for highlighting at either baseline or follow-up time points. We did observe a 0.38 correlation between parent-proxy-reported PEDsQL at baseline and body composition frailty score, but the estimate was too imprecise (CI -0.19, 0.77) to rule out an effect in either direction. It is possible that the TSfT method to assess body composition was not sensitive enough to capture variability within the cohort. The methods used to assess each frailty domain were chosen as they are low-cost and require minimal training, allowing frailty assessments to be performed in environments with limited resources. Measures of body composition with greater sensitivity should be explored in future research.
Limitations and Future Directions
Our study has several weaknesses that limit the impact of our results. The most notable limitation is the small sample size, particularly at follow-up. We were able to acquire follow-up outcomes from the medical records and questionnaires from 82% and 44% of our original sample, respectively. Several methods were used to reapproach participants at follow-up, including telephone and email communication, in-person communication while participants were in the hospital for clinical visits, and mailings to participants’ homes. Participants followed standard clinical care from their primary cardiologists following the baseline frailty assessment, resulting in inconsistencies in clinical follow-up frequency, clinical testing, and management.
Future studies with larger and more generalizable samples are needed to optimize the frailty phenotype for specific for children and adolescents with heart disease. This work should include the exploration of alternate cost-effective methods with greater sensitivity for each domain. In calculating the CFS, future studies should explore analysis where domains are weighted to provide greater predictive strength for children and adolescents with heart disease. Methods of reducing frailty should be studied using CFS as this could benefit patient outcomes.