Heart failure is a major public health problem that has been shown to increase with age, such that incidence rates rapidly increase after 80 years of age.1 Among older patients with heart failure, 18% to 54% show signs of frailty, a state of reduced physical and cognitive function that results in weakness.2 There are some overlapping symptoms between heart failure and frailty; they interact to accelerate the vicious cycle of frailty.3 One study showed that frail patients with cardiovascular disease had a 7-year survival rate of 12%, which was much lower than the survival rate in non-frail patients with cardiovascular disease (43%).4 Moreover, frailty among patients with heart failure has been associated with poor prognosis5 and reduced cardiac output capacity.6 Pre-frailty is the first step towards frailty; approximately 34.6% to 46.1% of individuals with pre-frailty progress to frailty in Japan.7 Therefore, improvements in frailty and pre-frailty are important considerations in the care of older patients with heart failure.

Age,8 nutrition,9 walking speed,9 heart function,8 and grip strength10 have been shown to influence frailty improvement among older patients with heart failure. However, the predictive accuracies of such factors remain unknown. Moreover, a combination of the predictors has been suggested to increase their predictive accuracy,11 although this hypothesis has not yet been tested. Additionally, nutrition,12 physical function,12 and quality of life12 have been reported to influence pre-frailty improvement among community-dwelling older individuals. To our knowledge, there are no reports of factors that influence pre-frailty improvement among patients with heart failure.

In this study, we used the classification and regression tree (CART) method,13 which facilitates the establishment of clinical prediction models that can identify the best combinations of medical signs, symptoms, and other findings to predict prognoses or treatment outcomes.13 Several clinical prediction models have been developed using the CART method to predict mortality in patients with heart failure.14 15 However, no clinical prediction models have been established to predict the prognosis of patients with heart failure complicated by pre-frailty and frailty.

Here, we aimed to use the CART method to develop models that could predict the prognosis of older patients with heart failure complicated by pre-frailty and frailty, then confirm the predictive accuracies of those models.

This pilot prospective cohort study followed the STROBE reporting guidelines. All included patients provided written informed consent to participate in the study. Identifying information was not collected to protect each patient’s privacy. This study was performed in accordance with the Declaration of Helsinki.

Recruitment, follow-up, and data collection were performed at two acute hospitals (Saiseikai Kure Hospital, Kure, Japan; Kure Kyosai Hospital, Kure, Japan) between July 2018 and December 2019. Potential participants were recruited by therapists at the rehabilitation department.

This study included patients who met the following inclusion criteria: age ≥65 years; hospitalisation for the treatment of heart failure; and presence of pre-frailty or frailty. The exclusion criteria were complications during hospitalisation and/or severe dementia (defined as a revised Hasegawa’s Dementia Scale score ≤9).

Cardiac rehabilitation was performed by physical or occupational therapists to improve physical condition, restore walking ability during hospitalisation, and expand the activities of daily living. Rehabilitation programmes were established by physical or occupational therapists in accordance with physicians’ orders. Initially, aerobic exercises and resistance training programmes were provided according to each patient’s physical condition. Exercise intensity was determined using multiple indices, including target heart rate (convenient method: resting heart rate + 20 bpm), talk test, and Borg scale (11-13) for the chest and lower limbs. The type of exercise was modified (ie, duration extended and load increased) with consideration of each patient’s symptoms and haemodynamics. If necessary, functional exercises (eg, neuromuscular facilitation, joint range of motion, and muscle strengthening exercises), exercises for activities of daily living, and psychological support were implemented for patients and their families. Exercises for activities of daily living were customised according to the functions that each patient needed for discharge. Overall, the duration and frequency of intervention were 30 minutes to 1 hour per day and 5 days per week, respectively.

Variables included demographic and clinical characteristics, frailty assessment results, and physical function. Demographic characteristics included age, sex, body mass index, living arrangement (with family members or alone), New York Heart Association class, medical history (eg, heart failure, coronary artery disease, valvular disease, hypertension, diabetes mellitus, dyslipidaemia, atrial fibrillation, chronic renal dysfunction, and/or stroke), cognitive function assessed using the revised Hasegawa’s Dementia Scale, Life-Space Assessment score, interval from admission to initiation of cardiac rehabilitation, interval from admission to rehabilitation room entry, and length of hospital stay. Hasegawa’s Dementia Scale scores of 21-30, 15-20, 10-14, and ≤9 were regarded as normal, suspected dementia, mild to moderate dementia, and severe dementia, respectively.16 The Life-Space Assessment developed by Baker et al17 was used to evaluate life-space mobility. Up to 120 points were assigned based on the degree of independence in each life-space level during the month prior to the assessment; higher scores were considered indicative of broader life-space and/or greater independence.

Clinical characteristics included blood data, cardiac function, and pharmacotherapy. Blood data included the Geriatric Nutritional Risk Index, brain natriuretic peptide level, estimated glomerular filtration rate (eGFR), and haemoglobin (Hb) level. Cardiac function was evaluated using left ventricular ejection fraction (LVEF), as determined by echocardiography. Pharmacotherapy data included whether patients were receiving dopamine, dobutamine, noradrenaline, phosphodiesterase III inhibitor, or diuretics.

Frailty was assessed using the following five conditions based on the Cardiovascular Health Study (CHS) Index: slow gait speed, weakness, exhaustion, low activity, and weight loss.18 Slow gait speed was defined as <1.0 m/s. Weakness was assessed using maximum grip strength according to sex-specific cut-offs (<26 and <18 kg for men and women, respectively). Exhaustion was assessed using the question “During the past 2 weeks, have you felt tired without a specific reason?” A positive response to this question (ie, “yes”) was considered indicative of exhaustion. Physical activity was evaluated using the question “Do you engage in low levels of physical activity to improve your health?” A negative response to this question (ie, “no”) was considered indicative of a low activity level. Weight loss was assessed using the question “Have you lost 2 kg or more in the past 6 months?” A positive answer to this question was considered indicative of weight loss. There are various criteria for assessing frailty; Fried’s frailty phenotype model18 and the accumulated deficit model established by Mitnitski et al19 are well known. The CHS criteria and the Frailty Index were developed based on the above frailty phenotype model and accumulated deficit model, respectively. Furthermore, a Japanese version of the CHS criteria (J-CHS) has been established,20 and its validity has been confirmed.21 Thus, we selected the J-CHS to assess frailty. Patients with none of the above conditions were considered non-frail (robust), patients with one to two conditions were considered pre-frail, and patients with three conditions or more were considered frail.22

Physical function was assessed using the Short Physical Performance Battery score, 10-metre walk time, single-leg standing time, and hand grip strength. Patients performed the Short Physical Performance Battery test in the following sequence, in accordance with the National Institute on Aging protocol: standing balance tests, gait test (4 m), and chair stand test (five repetitions). The sum of the three test components comprised the final Short Physical Performance Battery score, which ranged from 0 to 12; a score of 12 indicated optimal lower extremity function.23 24 Moreover, 10-m walk time was measured at a comfortable walking speed to assess walking ability. The 10-m walk test has demonstrated high validity and reliability in multiple populations, including healthy older individuals and patients with stroke, neurological disorders, or orthopaedic dysfunction.25 26 Measurements were performed twice at an interval of 30 s; the smaller value was used to indicate walking ability. Standing balance was evaluated by measuring single-leg standing time, which reflects the ability to maintain the body’s centre of gravity within its base of support. A stopwatch was used to measure the duration that a patient could stand on one leg with their eyes open and hands on their waist, without any assistance or falling. A second trial was performed if the result of the first trial was <60 s.27 Hand grip strength (kg) was measured using a digital hand grip strength dynamometer (TKK-5101; Takei Scientific Instruments, Tokyo, Japan or 12B3X00030; Tsutsumi Works, Chiba, Japan). Accordingly, patients were asked to squeeze the dynamometer with maximum effort during two trials for each hand. The maximum value (rounded to the nearest 0.1 kg) for either the left or right hand was used for subsequent analyses.28

Demographic and clinical characteristics were assessed upon admission or each time, whereas cognitive function and Life-Space Assessment were assessed during the first physical therapy session. Frailty and physical function were assessed upon initial entry into the rehabilitation room and at discharge. The amount of change in physical function (discharge−initial) was calculated.

To reduce selection bias, outcomes were selected based on the methods of previous studies.16 17 18 19 20 21 22 23 24 25 26 27 28 To reduce measurement bias, the first author (T Umehara) was not involved in participant enrolment or data collection. Patients received explanations about the purpose of the study but did not receive information concerning the hypothesis tested.

Classification and regression tree analysis29 was used to predict the primary outcomes. Patients with pre-frailty at admission were categorised into an improved group and a non-improved group: patients who were non-frail at discharge were assigned to the improved group, whereas patients who were pre-frail or frail were assigned to the non-improved group. Similarly, patients with frailty at admission were categorised into an improved group and a non-improved group: patients who were non-frail or pre-frail at discharge were assigned to the improved group, whereas patients who were frail were assigned to the non-improved group. Binary trees were used to recursively split predictor variables based on answers to yes/no questions for each variable. All statistical distributions were considered without limitation to linear relationships between outcome variables and predictor variables. These algorithms have been used to develop prediction models in various fields.30 31 32 The CART method with the Gini index was used for the following models: A, which predicted the presence or absence of pre-frailty improvement during hospitalisation; and B, which predicted the presence or absence of frailty improvement during hospitalisation. Pre-frailty or frailty improvement was the dependent variable, whereas demographic and clinical characteristics, pharmacotherapy, and amount of change in physical function were the independent variables. The accuracies of the CART models were evaluated using the area under the receiver operating characteristic curve (AUROC). The method proposed by Delong et al33 was used to identify optimal cut-off points. Subsequently, the sensitivity, specificity, positive likelihood ratio (PLR), and negative likelihood ratio (NLR) were calculated. To assess model validity, cross-validation was performed as follows: the sample was divided into 10 subgroups and the model developed from nine subgroups was used to test the 10th subgroup; this was repeated for all 10 combinations and the rates of misclassification were averaged. Model validity was considered high when the misclassification rates were similar before and after cross-validation. The accuracies of the CART models were evaluated using the AUROC developed from each method. The maximum Youden index (sensitivity + specificity − 1) was defined as the optimal cut-off point. All statistical analyses were performed using SPSS (Windows version 23.0; IBM Corp, Armonk [NY], United States) and the significance level was set at 5%.

Sample size calculations were conducted using MedCalc statistical software, version 19.2 (MedCalc Software bvba, Ostend, Belgium). Before plotting an AUROC, the following values were established: statistical significance (P<0.05), alpha (0.05), statistical power (0.80), and the AUROC value to be included in the null hypothesis (0.5). The AUROC could distinguish between non-predictive (AUROC<0.5), less predictive (0.5<AUROC^lt;0.7), moderately predictive (0.7<AUROC<0.9), highly predictive (0.9<AUROC<1), and perfectly predictive (AUROC=1).34 In this study, an AUROC value of 0.7 was considered indicative of superior statistical discrimination. The frailty improvement ratio (ie, positive/negative ratio) considerably varied among previous studies.35 36 37 Therefore, the positive/negative ratio used here was set at 1:1-5. Moreover, a large sample size was needed to allow for the possibility of stratified analysis. Thus, 62 to 120 patients with frailty were required: 20-31 and 31-100 in the improved and non-improved groups, respectively.

Figure 1 shows the flowchart of patient recruitment. Among the 30 patients with pre-frailty, two with severe dementia were excluded; 28 patients were included in the analysis. Among the 161 patients with frailty, five with severe dementia were excluded; 156 patients were included in the analysis. The patient characteristics are summarised in Tables 1 and 2 (patients with pre-frailty and patients with frailty, respectively).

Figure 2 shows CART model A, which predicted the presence or absence of pre-frailty improvement during hospitalisation. Among the 28 patients with pre-frailty, seven experienced improvements. Single-leg standing time at admission was identified as the best single discriminator for pre-frailty improvement (≤4.4 or >4.4 s). Among patients with single-leg standing time of >4.4 s at admission, the next predictor was LVEF (≤36.8% or >36.8%). Our CART analysis resulted in the establishment of three terminal nodes. The terminal node with the highest probability of a favourable outcome (pre-frailty improvement) was defined as rank 1, whereas the terminal node with the lowest probability of a favourable outcome was defined as rank 3. Based on the AUROC (95% confidence interval [CI]), this CART model had an accuracy of 0.96 (95% CI=0.89-1.00), with an optimal cut-off point of rank 1 (sensitivity: 85.7%, specificity: 95.2%, PLR: 18.0, and NLR: 0.15). The misclassification rates before and after cross-validation were 7.1% and 28.6%, respectively.

Figure 3 shows CART model B, which predicted the presence or absence of frailty improvement during hospitalisation. Among the 156 patients with frailty, 57 experienced improvements. Hand grip strength at admission was identified as the best single discriminator for frailty improvement (≤16.8 or >16.8 kg). Among patients with hand grip strength of >16.8 kg at admission, the next predictor was eGFR (≤27.0 or >27.0 mL/min/1.73 m²). Among patients with hand grip strength of ≤16.8 kg at admission, the next predictor was eGFR (≤83.5 or >83.5 mL/min/1.73 m²). Among patients with eGFR ≤83.5 mL/min/1.73 m², the next predictor was Hb level (≤14.3 or >14.3 g/dL). Among patients with Hb level ≤14.3 g/dL, the next predictor was change in single-leg standing time (≤4.9 or >4.9 s). Our CART analysis resulted in the establishment of six terminal nodes. The terminal node with the highest probability of a favourable outcome (frailty improvement) was defined as rank 1, whereas the terminal node with the lowest probability of a favourable outcome was defined as rank 6. Based on the AUROC (95% CI), this CART model had an accuracy of 0.84 (95% CI=0.78-0.91), with an optimal cut-off point of rank 4 (sensitivity: 70.1%, specificity: 86.9%, PLR: 5.3, and NLR: 0.3). The misclassification rates before and after cross-validation were 19.2% and 35.3%, respectively.

Model A predicted pre-frailty improvement with high accuracy; it identified single-leg standing time at admission as the best predictor, followed by LVEF. Notably, changes in physical function during hospitalisation were not identified as predictors. These results suggest that conditions at admission strongly influence pre-frailty improvement during hospitalisation; moreover, improvement can be expected among patients with good physical function (single-leg standing time >4.4 s) and cardiac function (LVEF >36.8%) at admission. To our knowledge, no study has examined the factors that influence pre-frailty improvement among patients with heart failure. However, one study found that physical function, nutrition, and quality of life were factors that influenced pre-frailty improvement among community-dwelling older individuals.12 The above findings suggest that although physical function is a common factor that influences pre-frailty improvement among older patients with heart failure and community-dwelling older individuals, cardiac function specifically influences pre-frailty improvement among patients with heart failure.

Model B predicted frailty improvement with moderate accuracy; it identified hand grip strength at admission as the best predictor, followed by eGFR, Hb level, and change in single-leg standing time during hospitalisation. Thus, frailty improvement can be expected among patients with good hand grip strength and/or renal function at admission. However, the cut-off values for eGFR, an index of renal function, considerably differed between patients with good (>16.8 kg) and poor (≤16.8 kg) hand grip strength (27 and 83.5 mL/min/1.73 m², respectively). A previous study also showed that older patients with heart failure who had higher hand grip strength were more likely to experience frailty improvement.10 Although no association has been identified between renal function and frailty, renal function is known to influence improvements in exercise capacity among patients with heart failure.38 Moreover, our results showed that, despite the presence of poor physical and renal function at admission, patients with a high Hb level (>14.3 g/dL) were likely to improve from frailty. Although there is no published literature concerning an association between Hb level and frailty, a low Hb level has been shown to cause fatigability,39 40 which is one of the criteria for assessing frailty using the CHS. Furthermore, the Hb level reportedly influences improvements in exercise capacity among patients with heart failure.38

Despite poor hand grip strength, weak renal function, and a low Hb level at admission, frailty improvement was observed in most patients whose single-leg standing time during hospitalisation was >4.9 s. In previous studies that sought to predict frailty improvement among older patients with heart failure, investigators mostly focused on conditions at admission without considering changes during hospitalisation.41 The present results indicate that, despite the presence of poor conditions at admission, patients can recover from frailty by improving physical function during hospitalisation; therefore, rehabilitation is essential during hospitalisation. Resistance training is known to improve single-leg standing time among older individuals.42 Thus, we recommend cardiac rehabilitation, including resistance training, to improve frailty among older patients with heart failure.

The clinical implications of our findings are as follows. Thus far, no algorithms have been established concerning pre-frailty and frailty improvements in older patients with heart failure. Using the CART method, we developed models that could predict the prognosis of older patients with heart failure complicated by pre-frailty and frailty. These models will provide useful information for patients and caregivers. Many of the factors extracted in this study were only assessed at the time of initial rehabilitation. Thus, the prognoses of patients with pre-frailty and patients with frailty can be inferred (to some extent) at the time of initial rehabilitation. Additionally, an increase in single-leg standing time during hospitalisation was associated with frailty improvement. For older patients with heart failure who show signs of frailty, interventions to increase single-leg standing time may help to improve frailty. The PLR and NLR were used to assess the diagnostic performances of the CART models; these parameters revealed that both model A (pre-frailty) and model B (frailty) had good performance. However, our data should be interpreted cautiously because of the small number of patients with pre-frailty.

There were several notable weaknesses and limitations in this study. First, the sample size was limited; the numbers of patients with pre-frailty and frailty were 28 and 156, respectively. Therefore, this study should be regarded as a pilot prospective cohort study. Despite the moderate to high accuracies of models A and B, more large-scale studies are needed to enhance the generalisability of our results. Second, three aspects of frailty exist, namely physical, social, and mental frailty; mental frailty was not considered in this study. A previous study reported that patients with multifaceted frailty, including physical, mental, and social frailty, had worse prognoses compared with patients who had physical frailty alone.5 Therefore, additional studies are needed to develop models that predict improvements in multifaceted frailty, including mental frailty. Third, interventions were customised for each patient; they were not uniform. However, physical therapists in both study hospitals received 2 weeks of training in cardiac rehabilitation, and the methods of cardiac rehabilitation were standardised as much as possible. Fourth, we used the J-CHS to measure frailty; the use of this tool to assess patients with heart failure is potentially controversial. The J-CHS was developed for older adults and has been used in multiple studies.7 21 Additionally, the reliability and validity of this tool have been confirmed.21 We thus consider this use of the J-CHS to be appropriate. Fifth, this study was performed in an unblinded manner, and we could not completely rule out the potential for bias. However, to reduce the measurement bias, the first author (T Umehara) was not involved in participant enrolment or data collection. Sixth, a selection bias might have been present because patients were only recruited at two hospitals in Japan. Caution is needed when generalising our results, particularly to patients in other countries.

By using the CART method, we developed moderately to highly accurate prediction models for pre-frailty and frailty improvement among older patients with heart failure. Model A, which predicted pre-frailty improvement, showed that patients with good single-leg standing time and cardiac function at admission are likely to experience pre-frailty improvement. Furthermore, Model B, which predicted frailty improvement, showed that patients with good hand grip strength, excellent renal function, and/or a high Hb level at admission are likely to experience frailty improvement. Notably, despite the presence of poor conditions at admission, frailty improvement may occur in patients who show improvement in single-leg standing time during hospitalisation. Overall, our results suggest that cardiac rehabilitation to prolong single-leg standing time is necessary to improve frailty, particularly when conditions at admission are poor.

Concept or design: All authors.
Acquisition of data: T Umehara and N Katayama.
Analysis or interpretation of data: T Umehara, M Tsunematsu, M Kakehashi.
Drafting of the manuscript: T Umehara, A Kaneguchi and Y Iwamoto.
Critical revision of the manuscript for important intellectual content: T Umehara, A Kaneguchi, Y Iwamoto.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

All authors have disclosed no conflicts of interest.

We thank Y Yamamoto, M Ota, Y Nakashima, W Kawakami, and M Iwanaga (Kure Kyosai Hospital, Kure, Japan), as well as S Nagao, Y Kawabata, and D Tomiyama (Saiseikai Kure Hospital, Kure, Japan) for their support in data collection.

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

The study was approved by the Ethics Committees of Saiseikai Kure Hospital (Ref 127) and Kure Kyosai Hospital (Ref 31-17). Patients provided written informed consent to participate.

1. Bleumink GS, Knetsch AM, Sturkenboom MC, et al. Quantifying the heart failure epidemic: prevalence, incidence rate, lifetime risk and prognosis of heart failure The Rotterdam Study. Eur Heart J 2004;25:1614-9. Crossref

2. Smart N, Marwick TH. Exercise training for patients with heart failure: a systematic review of factors that improve mortality and morbidity. Am J Med 2004;116:693-706. Crossref

3. Joyce E. Frailty in advanced heart failure. Heart Fail Clin 2016;12:363-74. Crossref

4. Newman AB, Gottdiener JS, Mcburnie MA, et al. Associations of subclinical cardiovascular disease with frailty. J Gerontol A Biol Sci Med Sci 2001;56:158-66. Crossref

5. Rodríguez-Pascual C, Paredes-Galán E, Ferrero-Martínez AI, et al. The frailty syndrome is associated with adverse health outcomes in very old patients with stable heart failure: A prospective study in six Spanish hospitals. Int J Cardiol 2017;236:296-303. Crossref

6. Denfeld QE, Winters-Stone K, Mudd JO, Hiatt SO, Chien CV, Lee CS. Frequency of and significance of physical frailty in patients with heart failure. Am J Cardiol 2017;8:1243-9. Crossref

7. Kojima G, Taniguchi Y, Iliffe S, Jivraj S, Walters K. Transitions between frailty states among community-dwelling older people: A systematic review and meta-analysis. Ageing Res Rev 2019;50:81-8. Crossref

8. Takabayashi K, Ikuta A, Okazaki Y, et al. Clinical characteristics and social frailty of super-elderly patients with heart failure—the Kitakawachi clinical background and outcome of heart failure registry. Circ J 2016;81:69-76. Crossref

9. Sze S, Zhang J, Pellicori P, Morgan D, Hoye A, Clark AL. Prognostic value of simple frailty and malnutrition screening tools in patients with acute heart failure due to left ventricular systolic dysfunction. Clin Res Cardiol 2017;106:533-41. Crossref

10. Bekfani T, Pellicori P, Morris DA, et al. Sarcopenia in patients with heart failure with preserved ejection fraction: impact on muscle strength, exercise capacity and quality of life. Int J Cardiol 2016;222:41-6. Crossref

11. Umehara T, Tanaka R, Tsunematsu M, et al. Can the amount of interventions during the convalescent phase predict the achievement of independence in activities of daily living in patients with stroke? A retrospective cohort study. J Stroke Cerebrovasc Dis 2018; 27:2436-44. Crossref

12. Gené Huguet L, Navarro González M, Kostov B, et al. Pre Frail 80: multifactorial intervention to prevent progression of pre-frailty to frailty in the elderly. J Nutr Health Aging 2018;22:1266-74. Crossref

13. McGinn TG, Guyatt GH, Wyer PC, Naylor CD, Stiell IG, Richardson WS. Users’ guides to the medical literature: XXII: how to use articles about clinical decision rules. Evidence-Based Medicine Working Group. JAMA 2000;284:79-84. Crossref

14. Fonarow GC, Adams KF Jr, Abraham WT, Yancy CW, Boscardin WJ, ADHERE Scientific Advisory Committee, Study Group, and Investigators. Risk stratification for in-hospital mortality in acutely decompensated heart failure: classification and regression tree analysis. JAMA 2005;293:572-80. Crossref

15. Arenja N, Breidthardt T, Socrates T, et al. Risk stratification for 1-year mortality in acute heart failure: classification and regression tree analysis. Swiss Med Wkly 2011;141:w13259. Crossref

16. Imai Y, Hasegawa K. The revised Hasegawa’s Dementia Scale (HDS-R)—Evaluation of its usefulness as a screening test for dementia. J Hong Kong Coll Psychiatr 1994;4:20-4.

17. Baker PS, Bodner EV, Allman RM. Measuring life-space mobility in community-dwelling older adults. J Am Geriatr Soc 2003;51:1610-4. Crossref

18. Fried LP, Tangen CM, Walston J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci 2001;56:146-56. Crossref

19. Mitnitski AB, Mogilner AJ, Rockwood K. Accumulation of deficits as a proxy measure of aging. ScientificWorldJournal 2001;1:323-36. Crossref

20. Satake S, Arai H. The revised Japanese version of the Cardiovascular Health Study criteria (revised J-CHS criteria). Geriatr Gerontol Int 2020;20:992-3. Crossref

21. Makizako H, Shimada H, Doi T, Tsutsumimoto K, Suzuki T. Impact of physical frailty on disability in community-dwelling older adults: a prospective cohort study. BMJ Open 2015;5:e008462. Crossref

22. Satake S, Shimada H, Yamada M, et al. Prevalence of frailty among community-dwellers and outpatients in Japan as defined by the Japanese version of the Cardiovascular Health Study criteria. Geriatr Gerontol Int 2017;17:2629-34. Crossref

23. Bernabeu-Mora R, Medina-Mirapeix F, Llamazares-Herrán E, García-Guillamón G, Giménez-Giménez LM, Sánchez-Nieto JM. The Short Physical Performance Battery is a discriminative tool for identifying patients with COPD at risk of disability. Int J Chron Obstruct Pulmon Dis 2015;10:2619-26. Crossref

24. Guralnik JM, Simonsick EM, Ferrucci L, et al. A short physical performance battery assessing lower extremity function: association with self-reported disability and prediction of mortality and nursing home admission. J Gerontol 1994;49:85-94. Crossref

25. Bohannon RW. Comfortable and maximum walking speed of adults aged 20-79 years: reference values and determinants. Age Ageing 1997;26:15-9. Crossref

26. van Hedel HJ, Wirz M, Dietz V. Assessing walking ability in subjects with spinal cord injury: validity and reliability of 3 walking tests. Arch Phys Med Rehabil 2005;86:190-6. Crossref

27. Michikawa T, Nishiwaki Y, Takebayashi T, Toyama Y. One-leg standing test for elderly populations. J Orthop Sci 2009;14:675-85. Crossref

28. Huang C, Niu K, Kobayashi Y et al. An inverted J-shaped association of serum uric acid with muscle strength among Japanese adult men: a cross-sectional study. BMC Musculoskelet Disord 2013;14:258. Crossref

29. Breiman L, Friedman J, Olshen R, Stone C. Classification and Regression Tree Analysis. Boston University; 2014: 1-16.

30. Aguiar FS, Almeida LL, Ruffino-Netto A, Kritski AL, Mello FC, Werneck GL. Classification and regression tree (CART) model to predict pulmonary tuberculosis in hospitalized patients. BMC Pulm Med 2012;12:40.Crossref

31. Takahashi O, Cook EF, Nakamura T, Saito J, Ikawa F, Fukui T. Risk stratification for in-hospital mortality in spontaneous intracerebral haemorrhage: a Classification and Regression Tree analysis. QJM 2006;99:743-50. Crossref

32. Westreich D, Lessler J, Funk MJ. Propensity score estimation: neural networks, support vector machines, decision trees (CART), and meta-classifiers as alternatives to logistic regression. J Clin Epidemiol 2010;63:826-33. Crossref

33. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44:837-45. Crossref

34. Swets JA. Measuring the accuracy of diagnostic systems. Science 1988;240:1285-93. Crossref

35. Espinoza SE, Jung I, Hazuda H. Lower frailty incidence in older Mexican Americans than in older European Americans: the San Antonio Longitudinal Study of Aging. J Am Geriatr Soc 2010;58:2142-8. Crossref

36. Schuurmans H, Steverink N, Lindenberg S, Frieswijk N, Slaets JP. Old or frail: what tells us more? J Gerontol A Biol Sci Med Sci 2004;59:962-5. Crossref

37. Collard RM, Boter H, Schoevers RA, Oude Voshaar RC. Prevalence of frailty in community-dwelling older persons: a systematic review. J Am Geriatr Soc 2012;60:1487-92. Crossref

38. Silverberg DS, Wexler D, Blum M, Iaina A. The cardio renal anemia syndrome: correcting anemia in patients with resistant congestive heart failure can improve both cardiac and renal function and reduce hospitalizations. Clin Nephrol 2003;60 Suppl 1:S93-102.

39. Roy CN. Anemia in frailty. Clin Geriatr Med 2011;27:67-78. Crossref

40. Chaves PH, Ashar B, Guralnik JM, Fried LP. Looking at the relationship between hemoglobin concentration and prevalent mobility difficulty in older women. Should the criteria currently used to define anemia in older people be reevaluated? J Am Geriatr Soc 2002;50:1257-64. Crossref

41. Reeves GR, Whellan DJ, O’Connor CM, et al. A novel rehabilitation intervention for older patients with acute decompensated heart failure: The REHAB-HF pilot study. JACC Heart Fail 2017;5:359-66. Crossref

42. Gonzalez AM, Mangine GT, Fragala MS, et al. Resistance training improves single leg stance performance in older adults. Aging Clin Exp Res 2014;26:89-92. Crossref