Abstract
Emerging evidence demonstrates that frailty measures can predict adverse outcomes after cardiac procedures. Our objectives were to examine whether the inclusion of frailty measures adds incremental predictive value to existing surgical risk prediction models in patients undergoing cardiac surgery and to evaluate the reporting and methods of studies that investigated the prediction of frailty measures in cardiology. The inclusion of frailty measures adds incremental predictive value on existing perioperative risk‐scoring systems. We systematically searched the EMBASE, MEDLINE, and Web of Science databases for relevant studies. Studies were included according to predefined inclusion criteria. The quality of included studies was appraised using the QUADAS‐2 tool. Data were extracted and synthesized according to predefined methods. Twelve studies were included in the analysis. Included studies demonstrated the incremental predictive value of frailty measures on existing surgical risk models for mortality, but the predictive value of frailty measures alone was not consistent across literature. Few studies that investigated the predictive ability of frailty measures reported all important model performance measures. When comparing the predictive value of frailty measures with existing models, few studies reported if the frailty measurement was separately performed from the existing perioperative risk assessment. The addition of frailty measures to the existing perioperative risk models improved the prediction performance for mortality, but the incorporation of frailty assessment into perioperative risk assessment requires further evidence before making health policy recommendations.
Keywords: cardiac surgery, frailty, perioperative, predictive value
1. INTRODUCTION
The contemporary aging of society leads to an increasing proportion of elderly patients undergoing cardiac surgery.1 While elderly patients may benefit from cardiac surgical procedures, some suffer from adverse outcomes such as death and major complications.2 Risk scores are important tools for predicting morbidity and mortality following cardiac surgery.3 The European System for Cardiac Operative Risk Evaluation (Euro SCORE) and the Society of Thoracic Surgeons (STS) score are two commonly used risk scores for predicting mortality and morbidity. Although both risk scores have been exclusively validated, recent evidence suggests that these scoring systems have poor accuracy as these risk scores mainly focus on medical diagnoses and comorbidities and do not reflect the true biological status of the patient.3
Identifying patients who are the most or least likely to benefit from cardiac surgical procedures is a critical step toward better clinical pathways aimed at prevention.4 The addition of frailty to existing perioperative risk scores may allow for the improved prediction of the postoperative outcomes.4 Frailty is defined as a state of decreased physiological reserve and vulnerability to stressors such as acute or chronic illness and iatrogenic stressors.5 Because surgery is an iatrogenic physiological stressor, cardiac surgery is an inherently relevant setting for frailty.6 Accumulating evidence has demonstrated that frailty is a predictor of adverse outcomes following cardiac surgery, and frailty assessment in patients undergoing cardiac surgery might improve the determination of operative risk.1
Previous studies have highlighted the predictive value of frailty and suggested that significant frailty measurements could potentially be incorporated into existing surgical risk prediction models. However, whether frailty adds incremental predictive value above existing risk models and how frailty measures could be added to existing risk models remain to be determined.7, 8 Although there have been an increasing number of studies assessing the predictive ability of frailty measures, some studies do not follow methodological recommendations, limiting their reliability and applicability.9 The objectives of this study were to examine whether the inclusion of frailty measures adds incremental predictive value to existing surgical risk models, with a specific focus on the Euro SCORE and the STS score, and to evaluate the reporting and methods of studies that have investigated the predictive ability of frailty measures in cardiac surgery. We aimed to provide evidence to guide health policy decision‐making regarding the incorporation of frailty in the perioperative risk assessment of patients undergoing cardiac surgery.
2. METHOD
2.1. Literature search
Using the keywords “cardiac surgery,” “frailty,” and “predictive value,” we used a systematic search strategy to identify studies that were eligible for inclusion, beginning with EMBASE, followed by Medline and Web of Science. References from reviews were manually searched for further relevant studies. No language limitations were applied.
2.2. Inclusion criteria
Inclusion criteria were as follows: (1) the study aimed to compare the predictive ability of frailty measures and commonly used surgical risk models (STS score and Euro SCORE); (2) the study population comprised patients undergoing cardiac surgery, including conventional surgeries and minimally invasive procedures, and we acknowledge that cardiac surgery has moved toward less‐invasive approaches and that interventions such as percutaneous valve repair or replacement may have an impact on the prognosis of eligible patients;19 (3) the primary endpoints were postoperative outcomes (including mortality and morbidity); and (4) the study reported model performance measures, such as calibration, discrimination, classification, overall model performance, and global model fit.9 Calibration and discrimination are recommended to be reported in prediction research.9, 20 Calibration is defined as the agreement between the predicted outcome and the observed outcome. Common measures include calibration plots, calibration intercepts and slopes, and the Hosmer‐Lemeshow statistic.9, 20 Discrimination is defined as the ability to separate participants with and without the outcome. The c statistic or area under the receiver operating characteristic curve (AUC) can be used to assess discrimination.20, 21 The predictive ability of the model is not random when the c statistic or AUC is greater than 0.5. The predictive value is considered high, medium, and low if the AUC is greater than 0.9, 0.7 to 0.9, and 0.5‐0.7, respectively.22 Net reclassification improvement (NRI) and integrated discrimination improvement (IDI) have recently been proposed to assess the performance of a prediction model when the key interest is whether a new variable or marker improves model performance.23 NRI and IDI focus on reclassification tables, which demonstrate how many participants are reclassified by adding a new variable to the model.24 NRI quantifies correct improvements, and IDI integrates NRI over all possible cut‐off values for the probability of the outcome.24 Overall performance measures, such as the R2 and the Brier scores, are traditional measures used to quantify how close predictions are to the actual outcome.23 The Akaike information criterion (AIC) and Bayesian information criterion (BIC) may also be reported as a measure of global model fit.17
2.3. Study selection
Two reviewers independently reviewed all abstracts identified in the initial search based on the inclusion criteria, and studies meeting the inclusion criteria were included for full‐text review. Full‐text review of articles was performed independently by two reviewers.
2.4. Data extraction and synthesis
Information on the literature source, patient characteristics, reference risk score model, model with frailty measures added, outcomes, and model performance measures was extracted by one reviewer and independently audited by another reviewer. As we focused on the incremental predictive value of frailty measures above the STS score and the Euro SCORE, studies were synthesized and presented separately by different risk‐scoring systems. Although we acknowledged the heterogeneity of the populations, given limited data in the included studies, we did not categorize data by different heart diseases or types of cardiac surgery.
2.5. Quality assessment of the included studies
The quality of the included studies was independently appraised by two reviewers using QUADAS‐2, a tool for the systematic review of diagnostic accuracy studies.25 QUADAS‐2 contains four domains: patient selection, index test, reference standard, and flow and timing. Each domain is appraised in terms of the risk of bias. Patient selection, index test, and reference standard are also appraised in terms of concerns regarding applicability.25 Signaling questions are included to help judge the risk of bias.25 Disagreements were resolved by consensus.
3. RESULTS
3.1. Literature selection
Figure 1 shows the results of the literature search. A total of 3141 published studies were included for screening. After title and abstract screening, 101 studies were included for full‐text screening. Finally, 12 studies met the inclusion criteria for analysis. Seven studies investigated a multidimensional frailty index, 3, 8, 10, 11, 12, 15, 18 two studies focused on a single‐item measure, 13, 17 and three studies included both.4, 14, 16
Figure 1.
Flowchart showing number of articles included during each stage of the systematic review process
Left, the number of articles included during each stage.
Right, the number of articles excluded during each stage.
AUC, area under the receiver operating characteristic curve; NRI, net reclassification index; IDI, integrated discrimination index; n, sample size; Black dots (•), reasons to exclude articles
3.2. Quality assessment of the included literature
The quality of the included studies is summarized in Table 1. In terms of the risk of bias, only one study was judged as “low” for all key domains.10 Due to inappropriate exclusions, four studies were judged as “medium risk” for patient selection.4, 8, 16, 17 Nine studies were judged as “unclear” for the index test and reference standard because insufficient data were reported to answer the signaling question “Were the index test results interpreted without knowledge of the results of the reference standard?” or “Were the reference standard results interpreted without knowledge of the results of the index test?”3, 4, 8, 11, 12, 13, 14, 15, 18 Four studies were “medium risk” for flow and timing because of a high rate of loss to follow up.3, 4, 13, 16 Concerns regarding applicability were rated as “low” for all studies.
Table 1.
Assessment of methodological quality
| Risk of bias | Applicability concerns | ||||||
|---|---|---|---|---|---|---|---|
| Study | Patient selection | Index test | Reference standard | Flow and timing | Patient selection | Index test | Reference standard |
| Kovacs et al10 | Low | Low | Low | Low | Low | Low | Low |
| Forcillo et al11 | Low | Unclear | Unclear | Low | Low | Low | Low |
| Afilalo et al8 | Medium | Unclear | Unclear | Low | Low | Low | Low |
| Sündermann et al3 | Low | Unclear | Unclear | Medium | Low | Low | Low |
| Jung et al4 | Medium | Unclear | Unclear | Medium | Low | Low | Low |
| Kobe et al12 | Low | Unclear | Unclear | Low | Low | Low | Low |
| Grossman et al13 | Low | Unclear | Unclear | Medium | Low | Low | Low |
| Green et al14 | Low | Unclear | Unclear | Low | Low | Low | Low |
| Sündermann et al15 | Low | Unclear | Unclear | Low | Low | Low | Low |
| Afilalo et al16 | Medium | Low | Low | Medium | Low | Low | Low |
| Afilalo et al17 | Medium | Low | Low | Low | Low | Low | Low |
| Shimura et al18 | Low | Unclear | Unclear | Low | Low | Low | Low |
Risk of bias is considered: “low” if all signaling questions for a domain are answered “yes”; “medium” if one to two signaling questions for a domain are answered “no”; “high” if all signaling questions for a domain are answered “no”; “unclear” if insufficient data are reported to permit judgment.
3.3. Model performance measures reported in the included literature
All studies reported discrimination measures. Five studies reported classification measures such as NRI and IDI.4, 8, 13, 16, 17 Two studies analyzed the AIC and/or BIC as a measure of model fit.8, 17 Only one study reported calibration measures.17 No studies reported overall model performance, such as the Brier score or R2 (Table S1, Supporting information).
3.4. Incremental predictive value of frailty measures above the STS score
Five studies compared the predictive ability between frailty measures and STS score (Table 2).8, 11, 12, 14, 17 The sample size ranged from 131 to 1020, and the study populations were patients aged 75 or older. All studies investigated patients undergoing transcatheter aortic valve implantation or surgical aortic valve replacement. One study also enrolled patients undergoing coronary artery bypass grafting (CABG).17 The AUC for the STS score ranged from 0.579 to 0.727. The addition of frailty increased the AUC to 0.69 to 0.772, depending on the frailty measures added. All studies reported mortality, including 30‐day mortality, in‐hospital mortality, and 1‐year mortality, as a study outcome. Five studies reported AUCs, and the results demonstrated that, regardless of the definition of frailty, adding frailty measures to the STS score resulted in a higher AUC for predicting mortality. Two studies reported IDI. One study showed that the estimated IDI was statistically significant, regardless of the frailty measurement tools used and the outcomes reported.8 Another study reported IDI but did not provide P values.17 No studies reported NRI.
Table 2.
Incremental predictive value of frailty measures above STS score
| Patient characteristics | Reference standard | Frailty measures | Predictive value of frailty measures | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Author | N | Age | Procedure | Outcomes | Model | AUC | Model | AUC | NRI | IDI |
| Forcillo et al11 | 361 | 82 (76‐86) | TAVI | 30‐d mortality | STS | 0.579 | Albumin + grip strength + ADL + walk test | 0.727 | ||
| Reference standard + Albumin + grip strength + ADL + walk test | 0.699 | |||||||||
| Afilalo et al8 | 1020 | 82 (77‐86) | TAVI/SAVR | 1‐year mortality | STS + type of procedure | 0.713 | Reference standard + Fried scale | 0.724 | 0.012** | |
| Reference standard + Fried scale + cognition | 0.734 | 0.020** | ||||||||
| Reference standard + Rockwood CFS | 0.743 | 0.027** | ||||||||
| Reference standard + SPPB | 0.734 | 0.023** | ||||||||
| Reference standard + Bern Scale | 0.753 | 0.031** | ||||||||
| Reference standard + Columbia scale | 0.752 | 0.031** | ||||||||
| Reference standard + Essential Frailty Toolset | 0.784 | 0.067** | ||||||||
| 30‐d mortality | STS + type of procedure | Fried scale | 0.008** | |||||||
| Fried scale + cognition | 0.013** | |||||||||
| Rockwood Clinical frailty scale | 0.007* | |||||||||
| SPPB | 0.013** | |||||||||
| Bern Scale | 0.007* | |||||||||
| Columbia scale | 0.007* | |||||||||
| Essential frailty toolset | 0.026** | |||||||||
| Kobe et al12 | 130 | 83.3 ± 4.8 | TAVI | In‐hospital mortality | STS Score | chair rise + weakness + stairs + creatinine + CFS | 0.73 | |||
| chair rise + weakness + stairs + creatinine + CFS + STS score | 0.77 | |||||||||
| 30‐d mortality | chair rise + weakness + stairs + creatinine + CFS | 0.68 | ||||||||
| chair rise + weakness + stairs + creatinine + CFS + STS score | 0.69 | |||||||||
| Green et al14 | 159 | 86.2 ± 7.7 | TAVI | 1‐year mortality | Clinical model (with STS score) | 0.727 | Reference standard + albumin | 0.734 | ||
| Reference standard + albumin + ADL | 0.749 | |||||||||
| Reference standard + albumin + ADL + gait speed | 0.767 | |||||||||
| Reference standard + albumin + ADL + gait speed + grip strength | 0.772 | |||||||||
| Afilalo et al17 | 131 | 75.8 ± 4.4 | CABG/valve replacement | In‐hospital postoperative mortality or major morbidity | STS score | 0.70 | STS score + gait speed | 0.74 | 5%a | |
Abbreviations: ADL, activity of daily living; AUC, area under the receiver operating characteristic curve; CABG, coronary artery bypass grafting; CFS, clinical frailty scale; IDI, integrated discrimination index; N, sample size; NRI, net reclassification index; SAVR, surgical aortic valve replacement; SPPB, short physical performance battery; TAVI, transcatheter aortic valve implantation.
P < 0.05;
P < 0.01.
P value not reported.
3.5. Incremental predictive value of frailty measures above the Euro SCORE
Two studies compared the predictive ability between frailty measures and the Euro SCORE (Table 3). The first study enrolled 57 patients, with an average age of 70.2 years, who underwent open heart surgery.10 The study demonstrated that, compared with the Euro SCORE II, the Edmonton Frailty Scale and the Clinical Frailty Scale had lower AUCs for postoperative complications and mortality. For predicting postoperative mortality, the study showed an AUC of 0.816 for the Euro SCORE. This was the highest AUC among the included studies.10 The second study included 133 patients undergoing elective CABG or valve procedures.4 The study showed that, compared with the Euro SCORE II, frailty measures, such as the modified Fried Scale, the Short Physical Performance Battery (SPPB), and the Frailty Index, had higher AUCs for postoperative delirium. The estimated NRI and IDI were statistically significant (P < 0.05).4 While the modified Frailty Scale demonstrated better model performance, the predictive ability of each domain was different. Only weight loss, grip strength, and gait speed showed higher AUCs for postoperative delirium, and only weight loss demonstrated statistically significant model improvement.4
Table 3.
Incremental predictive value of frailty measures above the Euro SCORE
| Patient characteristics | Reference standard | Frailty measures | Predictive value of frailty measures | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Author | N | Age | Procedure | Outcome | Model | AUC | Model | AUC | NRI | IDI |
| Kovacs et al10 | 57 | 70.2 ± 4.3 | Open heart surgery | Postoperative complications | Euro SCORE II | 0.665 | Edmonton Frail Scale | 0.652 | ||
| Clinical Frailty Scale | 0.636 | |||||||||
| Postoperative deaths | Euro SCORE II | 0.816 | Edmonton Frail Scale | 0.738 | ||||||
| Clinical Frailty Scale | 0.778 | |||||||||
| Jung et al4 | 133 | 71.0 | Elective CABG/valve procedures | Postoperative delirium | Euro SCORE II | 0.695 | Modified Fried Scale (≥3 of 7) | 0.745 | 74.9%** | 6.5%** |
| Weight loss | 0.709 | 44.6%* | 5.0%* | |||||||
| Grip strength | 0.700 | 41.1% | 2.3% | |||||||
| Low physical activity | 0.674 | 27.1% | 1.1% | |||||||
| Exhaustion | 0.668 | 28.6% | 0.9% | |||||||
| Depression | 0.683 | 20.0% | 0.5% | |||||||
| Cognitive impairment | 0.677 | 16.9% | 0.1% | |||||||
| Gait speed | 0.701 | −8.8% | 0.0% | |||||||
| SPPB (score ≤9) | 0.699 | 49.9%* | 2.5%* | |||||||
| SPPB (score 4‐6) | 0.732 | 53.8%* | 10.1%** | |||||||
| Frailty Index ≥0.2 | 0.722 | 55.7%* | 3.7%** | |||||||
| Frailty Index ≥0.25 | 0.716 | 64.6%** | 4.4%* | |||||||
| Frailty Index ≥0.3 | 0.730 | 70.0%** | 5.7%** | |||||||
Abbreviations: AUC, area under the receiver operating characteristic curve; CABG, coronary artery bypass grafting; IDI, integrated discrimination index; N, sample size; NRI, net reclassification index; SPPB, short physical performance battery; TAVI, transcatheter aortic valve implantation.
P < 0.01;
P < 0.05.
3.6. Predictive value of frailty measures compared with the STS score and the Euro SCORE
Five studies compared the predictive ability of frailty measures with the STS score and the Euro SCORE, with a sample size ranging from 152 to 1215 (Table 4).3, 13, 15, 16, 18 All studies investigated elderly patients and reported mortality as an outcome. In terms of the AUC for mortality, three studies showed that the Euro SCORE was better,13, 15, 18 while one study demonstrated that the STS score was better,16 and one study showed that the two risk models were equal.3 Comparing frailty measures with the two risk models, two studies showed that the Euro SCORE and the STS score were better,15, 16 and one study showed that the frailty measures were better.3 Two studies demonstrated that adding frailty measures to the two risk models improved model performance in predicting mortality.13, 18
Table 4.
Predictive value of frailty measures compared with Euro SCORE and STS score
| Patient characteristics | Reference standard | Frailty measures | Predictive value of frailty measures | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Author | N | Mean age | Procedure | Outcome | Model | AUC | Model | AUC | NRI | IDI |
| Sundermann et al3 | 400 | ≥74 | Cardiac surgery | 1‐year mortality | Euro SCORE II | 0.67 | chair rise + weakness + stairs + creatinine + CFS | 0.76 | ||
| STS Score | 0.67 | |||||||||
| Grossman et al13 | 411 | 83.8 | TAVI | 1‐year mortality | STS Score | 0.65 | STS Score + albumin | 0.71 | 0.58** | |
| Euro SCORE II | 0.66 | Euro SCORE II + albumin | 0.72 | 0.64** | ||||||
| Sundermann et al15 | 400 | ≥74 | Cardiac surgery | 30‐d mortality | Euro SCORE II | 0.79 | Grip strength + walking speed + balance + rise up from chair + pick up a pen + put on and remove a jacket | 0.71 | ||
| STS Score | 0.76 | |||||||||
| Afilalo et al16 | 152 | 75.9 ± 4.4 | CABG/valve surgery | In‐hospital mortality or major morbidity | Euro SCORE | 0.65 | Five‐item CHS scale (gait speed + grip strength + inactivity + exhaustion + weight loss) | 0.60 | ||
| STS Score | 0.67 | Seven‐item CHS scale (five‐Item CHS scale + cognitive impairment + depressed mood) | 0.58 | |||||||
| MSSA frailty scale (gait speed + grip strength + inactivity + cognitive impairment) | 0.56 | |||||||||
| Gait speed | 0.64 | |||||||||
| Shimura et al18 | 1215 | 84.4 ± 5.0 | TAVR | Mortality | Euro SCORE | 0.63 | Euro SCORE + CFS | 0.66 | ||
| STS Score | 0.61 | STS Score + CFS | 0.64 | |||||||
Abbreviations: AUC, area under the receiver operating characteristic curve; CABG, coronary artery bypass grafting; CFS, clinical frailty scale; IDI, integrated discrimination index; N, sample size; MSSA, MacArthur study of successful aging; NRI, net reclassification index; SPPB, short physical performance battery; TAVI, transcatheter aortic valve implantation.
P < 0.01;
P < 0.05.
4. DISCUSSION
Our study suggests that the AUC of frailty for predicting mortality in patients undergoing cardiac surgery varied with the definition of frailty. Compared with the two commonly used surgical risk models (STS score and Euro SCORE), frailty measures could result in either higher or lower AUCs, depending on the frailty measures used. Adding frailty to the two risk models improved the AUC for predicting mortality, including 30‐day mortality and 1‐year mortality, regardless of the frailty measured added. The statistically significant NRI and IDI further supported that the addition of frailty can improve the ability of the Euro SCORE and the STS score to predict 30‐day mortality and 1‐year mortality in patients undergoing cardiac surgery.
The true biological status of a patient is not captured well by current commonly used risk models, making the incorporation of factors that describe the biological status into existing perioperative risk assessments important.3 Examining the AUC values, our study found that the addition of frailty to the Euro SCORE and the STS score improved the discrimination of the model with respect to predicting mortality. The improvements were also demonstrated via the NRI and IDI statistics, which are considered more intuitive measures of the performance of prediction models.4 These results suggest the possibility and feasibility of incorporating frailty measures into standard preoperative risk prediction models in terms of predictive ability. Because there are many frailty measurements that can be short, fast, and crude or sophisticated and time consuming, it is difficult to determine which frailty measurement to use.26 Although the wide variations in frailty measurement tools may result in discrepancies in the definition and operationalization of frailty, our study demonstrates that the addition of frailty measures to existing risk‐scoring systems improves the prediction performance, regardless of the frailty measures used. By providing additional information not captured by existing perioperative risk models, the incorporation of frailty measurements into clinical practice may help clinicians better identify and manage patients’ health conditions.26
Most studies reported predictive effects derived from the multivariate and/or univariate analyses.9 Reviews assessing the predictive ability of frailty also mainly presented odds ratios or hazard ratios.27 Measures of model performance, such as the c statistics (or AUC) for discriminative ability and goodness‐of‐fit statistics for calibration, were reported only occasionally.27 In prediction research, calibration and discrimination are considered two main types of prediction performance measures.9 When adding a new variable to an established model, improvements in the model are more intuitively demonstrated via NRI and IDI.24 By reviewing studies that aimed to understand the predictive value of frailty measures compared with existing risk models, our study found that the majority of prediction studies only measured discrimination (c statistics). This finding is similar to the results of Bouwmeester et al, who found the AUC to be the most frequently reported discrimination statistics.9 When assessing whether the predicted outcome indeed matches the observed outcome, calibration and overall performance measures are essential.9 Therefore, our study suggests that, in further studies examining the predictive value of frailty, more comprehensive reporting of model performance measures is needed.
Several studies reviewing frailty assessment have offered valuable information on how frailty was measured in patients undergoing cardiac surgery.2, 27, 28, 29 Consistent with these studies, our study found that gait speed (as a measure of mobility) was an accurate single‐item frailty measure used to predict prognosis. The Clinically Frailty Scale (CFS), a well‐validated frailty measurement tool that is frequently reported in reviews, was used in 5 of the 12 included studies. However, while previous reviews have mainly investigated the association between frailty and prognosis, our study focused on a core clinical question: “Is the addition of frailty measure to existing perioperative risk models feasible or valuable?” We believe that the findings of our study will complement those of previously published studies evaluating frailty assessments, investigating the predictive value of frailty measures, and assessing the feasibility of the incorporation of frailty into perioperative risk assessments.
Our study has several important limitations. First, we included all patients who underwent cardiac surgery and did not separate the types of cardiac procedures. Because different frailty measures may be suited for different patients, the incremental predictive values of frailty in patients undergoing open heart surgery may not be translated directly to patients undergoing minimally invasive procedures.26 Thus, further studies may need to separate the types of cardiac procedures and address heterogeneity. Second, given the limited data, our study failed to perform a meta‐analysis, in which data from multiple studies would be combined and the effect estimated; thus, it is difficult to quantify the overall incremental predictive value of frailty measures above the STS score and Euro SCORE. Third, the methodological quality was appraised using QUADAS‐2. Although it is considered a desirable tool that allows for a more transparent judgment of bias and applicability of primary diagnostic accuracy studies, the reporting of sample size considerations was not assessed in our study.9, 25 Thus, we could not assess statistical power or effective sample size for the included studies.9 Fourth, our study focused on the STS score and Euro SCORE, two commonly used models for perioperative risk assessments. Other risk stratification scoring systems, such as Synergy between Percutaneous Coronary Intervention with Taxus and Cardiac Surgery (SYNTAX) and the Charlson comorbidity index, may also be valuable in cardiac surgery.30 Fifth, our study did not capture qualitative or quantitative details about changes in the decision to operate, the choice of a specific procedure, or the anesthetic plan that resulted from the frailty diagnosis.31 Meanwhile, evidence suggests that frailty is a dynamic process in which interventions could worsen or improve frailty status.32 These details may complicate comparisons of existing predictive models alone and models combined with frailty measurements. Finally, our study only suggests that, in general, the addition of frailty measures to the STS score or/and the Euro SCORE has incremental predictive value. Issues such as which frailty measures should be added in different populations were beyond the scope of our study.
5. CONCLUSION
Despite the recognition of frailty as a powerful predictor in patients undergoing cardiac surgery, whether to adopt frailty assessments in routine clinical practice remains unclear. Our study suggests that the addition of frailty measures to two commonly used risk score models could improve model performance with respect to predicting mortality, but more research that addresses the heterogeneity of populations and frailty assessment tools is needed to make more definitive health policy recommendations.
Supporting information
TABLE S1. Model performance measures reported in the included studies
ACKNOWLEDGMENT
No funding source to disclose
Conflict of interest
The authors declare no potential conflict of interests.
Li Z, Ding X. The incremental predictive value of frailty measures in elderly patients undergoing cardiac surgery: A systematic review. Clin Cardiol. 2018;41:1103–1110. 10.1002/clc.23021
REFERENCES
- 1. Nicolini F, Agostinelli A, Vezzani A, et al. The evolution of cardiovascular surgery in elderly patient: a review of current options and outcomes. Biomed Res Int. 2014;2014:1‐10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Kim DH, Kim CA, Placide S, Lipsitz LA, Marcantonio ER. Preoperative frailty assessment and outcomes at 6 months or later in older adults undergoing cardiac surgical procedures. Ann Intern Med. 2016;165:650‐660. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Sündermann S, Dademasch A, Rastan A, et al. One‐year follow‐up of patients undergoing elective cardiac surgery assessed with the comprehensive assessment of frailty test and its simplified form. Interact Cardiovasc Thorac Surg. 2011;13:119‐123. [DOI] [PubMed] [Google Scholar]
- 4. Jung P, Pereira MA, Hiebert B, et al. The impact of frailty on postoperative delirium in cardiac surgery patients. J Thorac Cardiovasc Surg. 2015;149:869‐875. [DOI] [PubMed] [Google Scholar]
- 5. Bergman H, Ferrucci L, Guralnik J, et al. Frailty: an emerging research and clinical paradigm—issues and controversies. J Gerontol Ser A Biol Med Sci. 2007;62:731‐737. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Afilalo J, Alexander KP, Mack MJ, et al. Frailty assessment in the cardiovascular care of older adults. J Am Coll Cardiol. 2014;63:747‐762. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Puri R, Iung B, Cohen DJ, Rodes‐Cabau J. TAVI or No TAVI: identifying patients unlikely to benefit from transcatheter aortic valve implantation. Eur Heart J. 2016;37:2217‐2225. [DOI] [PubMed] [Google Scholar]
- 8. Afilalo J, Lauck S, Kim DH, et al. Frailty in older adults undergoing aortic valve replacement: the FRAILTY‐AVR study. J Am Coll Cardiol. 2017;70:689‐700. [DOI] [PubMed] [Google Scholar]
- 9. Bouwmeester W, Zuithoff NP, Mallett S, et al. Reporting and methods in clinical prediction research: a systematic review. PLoS Med. 2012;9:1‐12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Kovacs J, Moraru L, Antal K, Cioc A, Voidazan S, Szabo A. Are frailty scales better than anesthesia or surgical scales to determine risk in cardiac surgery? Korean J Anesthesiol. 2017;70:157‐162. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Forcillo J, Condado JF, Ko YA, et al. Assessment of commonly used frailty markers for high‐and extreme‐risk patients undergoing transcatheter aortic valve replacement. Ann Thorac Surg. 2017;104:1939‐1946. [DOI] [PubMed] [Google Scholar]
- 12. Kobe AR, Meyer A, Elmubarak H, et al. Frailty assessed by the FORECAST is a valid tool to predict short‐term outcome after transcatheter aortic valve replacement. Innovations. 2016;11:407‐413. [DOI] [PubMed] [Google Scholar]
- 13. Grossman Y, Barbash IM, Fefer P, et al. Addition of albumin to traditional risk score improved prediction of mortality in individuals undergoing transcatheter aortic valve replacement. J Am Geriatr Soc. 2017;65:2413‐2417. [DOI] [PubMed] [Google Scholar]
- 14. Green P, Woglom AE, Genereux P, et al. The impact of frailty status on survival after transcatheter aortic valve replacement in older adults with severe aortic stenosis: a single‐center experience. J Am Coll Cardiol Intv. 2012;5:974‐981. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Sündermann S, Dademasch A, Praetorius J, et al. Comprehensive assessment of frailty for elderly high‐risk patients undergoing cardiac surgery. Eur J Cardiothorac Surg. 2011;39:33‐37. [DOI] [PubMed] [Google Scholar]
- 16. Afilalo J, Mottillo S, Eisenberg MJ, et al. Addition of frailty and disability to cardiac surgery risk scores identifies elderly patients at high risk of mortality or major morbidity. Circ Cardiovasc Qual Outcomes. 2012;5:222‐228. [DOI] [PubMed] [Google Scholar]
- 17. Afilalo J, Eisenberg MJ, Morin JF, et al. Gait speed as an incremental predictor of mortality and major morbidity in elderly patients undergoing cardiac surgery. J Am Coll Cardiol. 2010;56:1668‐1676. [DOI] [PubMed] [Google Scholar]
- 18. Shimura T, Yamamoto M, Kano S, et al. Impact of the clinical frailty scale on outcomes after transcatheter aortic valve replacement. Circulation. 2017;16:116. [DOI] [PubMed] [Google Scholar]
- 19. Chiu KM. Minimally invasive cardiac surgery. Formosan J Surg. 2013;46:183‐188. [Google Scholar]
- 20. Harrell FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15:361‐387. [DOI] [PubMed] [Google Scholar]
- 21. Vergouwe Y, Steyerberg EW. Validity of prognostic models: when is a model clinically useful? Semin Urologic Oncol. 2002;20:96‐107. [DOI] [PubMed] [Google Scholar]
- 22.Chen JY, Zhang AD, Lu HY, Guo J, Wang FF, Li ZC. CHADS2 versus CHA2DS2‐VASc score in assessing the stroke and thromboembolism risk stratification in patients with atrial fibrillation: a systematic review and meta‐analysis. J Geriatr Cardiol. 2013;10:258‐266. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Steyerberg EW, Vickers AJ, Cook NR, et al. Assessing the performance of prediction models: a framework for some traditional and novel measures. Epidemiology. 2010;21:128‐138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Pencina MJ, D'Agostino RB, Vasan RS. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med. 2008;27:157‐172. [DOI] [PubMed] [Google Scholar]
- 25. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS‐2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med. 2011;155:529‐536. [DOI] [PubMed] [Google Scholar]
- 26. Dent E, Kowal P, Hoogendijk EO. Frailty measurement in research and clinical practice: a review. Eur J Intern Med. 2016;31:3‐10. [DOI] [PubMed] [Google Scholar]
- 27. Abdullahi YS, Athanasopoulos LV, Casula RP, et al. Systematic review on the predictive ability of frailty assessment measures in cardiac surgery. Interact Cardiovasc Thorac Surg. 2017;24:619‐624. [DOI] [PubMed] [Google Scholar]
- 28. Sepehri A, Beggs T, Hassan A, et al. The impact of frailty on outcomes after cardiac surgery: a systematic review. J Thorac Cardiovasc Surg. 2014;148:3110‐3117. [DOI] [PubMed] [Google Scholar]
- 29. Partridge JS, Harari D, Dhesi JK. Frailty in the older surgical patient: a review. Age Ageing. 2012;41:142‐147. [DOI] [PubMed] [Google Scholar]
- 30.Díez‐Delhoyo F, Cebada FS, Cressa LM, Rivera‐Juárez A, Elízaga J, Fernández‐Avilés F. Prognostic value of the residual SYNTAX score in octogenarian patients with non–ST‐elevation acute coronary syndrome. Rev Esp Cardiol. 1970;1:217‐219. [DOI] [PubMed] [Google Scholar]
- 31. Hall DE, Arya S, Schmid KK, et al. Association of a frailty screening initiative with postoperative survival at 30, 180, and 365 days. JAMA Surg. 2017;152:233‐240. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Ayesta A, Astiz MT, Masa MJ, Segovia J, Cosío MD, Martínez‐Sellés M. Rationale and design of the FELICITAR registry (Frailty Evaluation After List Inclusion, Characteristics and Influence on Transplantation and Results). Clin Cardiol. 2018;41:293‐299. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
TABLE S1. Model performance measures reported in the included studies