The study was approved by the Ethics Committee of the Affiliated Hospital of Xuzhou Medical University (XYFY2023-KL044-01) and registered in the Chinese Clinical Trials Registry (ChiCTR2300069382). We adhered to the reporting requirements established by the Strengthening the Reporting of Observational Studies in Epidemiology (15). Written informed consent was obtained from each participant prior to enrollment. Patients were recruited between March 2023 and March 2024.

Elderly (≥60 years) patients undergoing elective open heart surgery were eligible for enrollment. The exclusion criteria were as follows: Preoperative refusal to participate in this study or communication difficulties; inability to complete a frailty assessment due to absolute bed rest; preoperative comorbidities of severe hepatic and renal dysfunction; preoperative history of IABP (Intra-Aortic-Balloon-Pump), mechanical ventilation, or pacemaker implantation; those who are unable to be reached or refuse to cooperate with postoperative telephone follow-up.

During the preoperative assessment, in addition to routine examinations, all consenting participants underwent an assessment of the CAF and the WHO Disability Assessment Scale 2.0 (WHODAS) by a researcher who was unaware of the content of the postoperative follow-up; the methodology of the CAF assessment is described in the study by Sündermann et al. (16). The CAF can be broadly viewed as a composite of three frailty assessment scales [FP (Frailty Phenotype) (17), CFS (Canadian Clinical Frailty Scale) (18), and MPPT (Modified Physical Performance Test)], consisting primarily of biomarker assessments (serum albumin level, serum creatinine level, BMI, and FEV1), physical tests of fatigue, activity level, gait speed, grip strength, and balance stability, and including measures of the CFS. The total score of the CAF was 35 points, with patients scoring ≥11 being frail and those scoring <11 being non-frail.

The baseline assessment of preoperative disability was determined using the WHO Disability Assessment Scale 2.0 (WHODAS) (11–13). The WHODAS 2.0 consists of six main domains: mobility, self-care, getting by, cognition, and social participation. The 12 items of the WHODAS are scored as described in previous studies, with each item having a numerical value: none = 0; mild = 1; moderate = 2; severe = 3; and extreme = 4. The total score ranges from 0 (no disability) to 48 (total disability or death) and is then divided by 48 and multiplied by 100 to give a disability score of 0 (no disability or death). mild = 2; severe = 3; extreme = 4. The total score ranged from 0 (no disability) to 48 (total disability or death), then divided by 48 and multiplied by 100 to convert to a percentage of the disability score. Preoperative disability was defined as a WHODAS 2.0 total score percentage greater than or equal to 25%.

The primary outcome was the relationship between preoperative frailty and new disability or death at 3 months postoperatively. Postoperative patient follow-up was performed by an anesthesiologist who was unaware of the results of the preoperative evaluation. Telephone follow-up and electronic medical record documentation were the primary means of determining patients’ postoperative survival status. New disability at 90 days was defined as the occurrence of a WHODAS 2.0 ≥ 25% postoperatively in patients who were not disabled preoperatively; if the patient had a preoperative comorbid disability (WHODAS ≥ 25%), an increase in the percentage of WHODAS scores by 8 at 90 days postoperatively indicated the occurrence of new disability (11–13).

Secondary outcomes included (1) 90-day disability-free survival (DFS), (2) ICU length of stay, (3) postoperative non-hospital discharge, (4) postoperative length of stay, (5) major morbidity (19, 20) (Supplementary Table S1), (6) incidence of PPCs (21), (7) 90-day readmission rate, and (8) 90-day mortality.

We conducted a post hoc study to compare the validity of the CAF with the commonly used FP and CFS in predicting 90-day new disability or death. We asked whether the predictive validity of the traditional and rapidly assessable preoperative frailty scales (FP and CFS) would achieve the same categorical properties as the CAF, because the preoperative assessment of the CAF is time-consuming.

All patients were managed according to standard cardiac surgery protocols, with pre-operative perioperative risk assessment followed by anesthesia for surgery, post-operative transfer to the surgical intensive care unit, and after stabilization of vital signs, transfer to the general surgical ward for further management and routine post-operative rehabilitation for all patients.

Baseline clinical and demographic data were collected according to the protocol, including sex, age, smoking and alcohol consumption, comorbidities (diabetes, hypertension, myocardial infarction, etc.), ASA class, EuroSCORE II, pulmonary function (FEV1, FVC, and FEV1/FVC), type of surgery, and left ventricular ejection fraction (LVEF). Operating time, Cross-clamp time, bypass time, urine volume, blood loss, total fluid intake, and total fluid output were recorded during surgery. Data on ICU stay, postoperative hospital stay, non-hospital discharge, 90-day Major Morbidity, and readmission within 90 days were collected.

Data normality was tested by visual inspection of histograms and Shapiro–Wilk’s W test. All normally distributed and skewed continuous variables were expressed as mean (SD) or median (interquartile range [IQR]). Categorical variables were indicated as frequencies (%). Comparison of continuous variables among groups was performed with the use of the Student’s t-test or Mann–Whitney U-test, depending on the normality of the distribution. In contrast, the Fisher’s Exact test was used to compare categorical variables.

A Least Absolute Shrinkage and Selection Operator regression analysis was conducted with statistically significant risk factors included in the univariable study to remove non-zero characteristic components. After that, a multivariate logistic regression analysis (stepwise regression method) was used to identify the 90-day new disability or death risk variables. Internal validation was carried out using the bootstrap self-sampling technique (1,000 bootstrap samples repeatedly sampled), and the model’s discrimination was tested using the relatively adjusted C-index (concordance index). The calibration curve was drawn to evaluate the model’s consistency. In addition, the inverse probability treatment weighting (IPTW) approach was used for two groups to adjust for observed possible confounding factors. Multivariable logistic regression analyses were used to obtain the IPTW-adjusted odds ratio (OR) in the IPTW-adjusted cohort. The predictive validity was assessed using the area under the receiver operating characteristic (ROC) curve (AUC). An AUC of 0.5–0.7 implies poor prediction accuracy, whereas an AUC of 0.7–0.9 suggests high prediction performance. The DeLong test compared the AUC of different models.

p-value < 0.05 (two-sided) was considered statistically significant. R4.1.2 and MedCalc 20.0. Statistical software was used for analysis.

Based on the pre-test, the incidence of death or new disability at 90 days after cardiac surgery in elderly individuals is about 10%, the multivariate regression model includes at least 5 outcome events for each variable (22), a total sample size of 250 is required (5*5/10% = 250), and considering the 15% dropout rate, 250/(1–15%) = 295 patients are proposed to be included in this study.

We used complete cases for the initial analysis, and preoperative baseline data were complete for all participants. A total of 351 individuals were included in the study, and final data were complete for 325 individuals. Our overall proportion of missing values was small, our analysis was based on all available data without imputation.

An assessment of frailty was performed on 351 patients, and 325 patients were included in the final analysis (Figure 1). The prevalence of frailty was found to be 23.08%. Table 1 shows that age, proportion of female, NT-ProBNP, hsTnT, stroke/TIA, myocardial infarction, diabetes, proportion of preoperative disability, NYHA class, ASA class, EuroSCORE II, and duration of surgery were higher in patients in the frail group compared with those in the non-frail group. The rates of alcohol consumption, hemoglobin, albumin, FEV1, and FVC were lower (p < 0.05), while all other factors were not statistically different (p > 0.05).

New disability or death occurred within 90 days after surgery in 41 (12.6%) of our patients, including 23 (33.1%) in the frail group and 18 (7.2%) in the non-frail group of patients; there was a significant difference in new disability or death at 90 days between the groups (Table 2, p < 0.05). Table 2 shows that patients in the frail group had higher WHODAS scores, postoperative pulmonary complications, major morbidity at 90 days, ICU stay, postoperative hospitalization, in-hospital mortality, mortality within 90 days, and incidence of readmission within 90 days (p < 0.05), while disability-free survival (DFS) at 90 days postoperatively was lower (p < 0.05).

According to the Least Absolute Shrinkage and Selection Operator regression analysis (Figure 2), we selected eight non-zero characteristic variables including frailty, EuroSCORE II, diabetes, Creatinine, ASA class, BMI, time of operation and bypass time (Table 3). Then, taking these eight predictors and conducting a multifactor logistic regression using a stepwise regression method approach, five meaningful variables were finally identified (Figure 2). In multivariate analysis, frailty [OR, 3.31; 95% CI, 1.43–7.62], bypass time (OR, 1.01; 95% CI, 1.00–1.01), BMI (OR, 0.87; 95% CI, 0.77–0. 98), diabetes (OR, 3.47; 95% CI, 1.40–8.55), and EuroSCORE II (OR, 1.75; 95% CI, 1.20–2.57) were independently associated with 90-day new disability or death (Figure 3). Frailty (OR, 3.15; 95% CI, 1.55–6.43) was also independently associated with 90-day new disability or death in a multivariate analysis of the inverse probability weighted adjusted cohort (Table 4).

Model validation: We reported a bias-corrected concordance statistic to verify the internal validity of our primary model utilizing calibration and discrimination with a 1,000-sample bootstrapping approach. Our calibration curve revealed that our model was well-calibrated (Supplementary Figure S1). With a C-statistic of 0.824 and an optimism-corrected C-statistic of 0.810, the discriminative ability demonstrated strong model performance in predicting 90-day new disability or death.

Empirical ROC analysis showed that CAF (AUC = 0.762) predicted 90-day new disability or death postoperatively more reliably than the traditional risk assessment tools ASA + age (AUC = 0.656), EuroSCORE II (AUC = 0.643), and WHODAS (AUC = 0. 662). The difference in the area under the curve between the first three methods and CAF was significant (p values of 0.039, 0.019, and 0.029, corresponding to z values of 2.064, 2.350, and 2.190, respectively). It can also be seen that the CAF was also a better predictor of DFS among the four risk assessment tools (AUC = 0.799), but was a poorer predictor of PPCs and 90-day major morbidity for all four risk assessment tools (AUC < 0.700) (Figure 4).

Compared to the common clinical frailty assessment scales (CFS and FP), the CAF (AUC = 0.762) had superior predictive efficacy compared to CFS (AUC = 0.696) and FP (AUC = 0.686) (p-values of 0.036 and 0.005, corresponding to z-values of 2.096 and 2.819, respectively, Supplementary Figure S2).