Comparing the Performance of Different Instruments for Diagnosing Frailty and Predicting Adverse Outcomes among Elderly Patients with Gastric Cancer

OBJECTIVES: To examine the diagnostic performance of the Tilburg Frailty Indicator (TFI), 11-factor modified frailty index (mFI-11), and 5-factor modified frailty index (mFI-5) for frailty defined by Frailty Phenotype (FP), as well as to compare the predictive ability of TFI, mFI-11, and mFI-5 for adverse outcomes in hospital among elderly patients undergoing gastric cancer surgery. DESIGN: A prospective cohort study. SETTING: Hospitalization setting, Nanjing, China. PARTICIPANTS: We recruited 259 elderly patients undergoing gastric cancer surgery from a tertiary hospital. MEASUREMENTS: Frailty was assessed by the FP, TFI, mFI-11, and mFI-5 before surgery, respectively. The receiver operating characteristic (ROC) curves were plotted to compared the diagnostic performance of TFI, mFI-11, and mFI-5 using FP as the reference. ROC curves were used to examine the performance of TFI, mFI-11, and mFI-5 in predicting adverse outcomes. The area under the curve (AUC)>0.70 was regarded as an indicator of good performance. RESULTS: The prevalence of frailty ranged from 8.5% (mFI-11) to 45.9% (TFI). The AUCs of TFI (AUC: 0.764, p<0.001) was significantly greater than that of mFI-11 (AUC: 0.600, p=0.033) and mFI-5 (AUC: 0.600, p=0.0311) in the detection of frailty defined by FP, with quite different sensitivity and specificity at their original cutoffs. TFI and mFI-11 both had statistically significant but similarly inadequate predictive accuracy for adverse outcomes in hospital, including total complications (AUCs: 0.618; 0.621), PLOS (AUCs: 0.593; 0.639), increased hospital costs (AUCs: 0.594; 0.624), and hypoproteinemia (AUCs: 0.573; 0.600). For the mFI-5, only the predictive ability for hypoproteinemia was statistically significant, with poor accuracy (AUC: 0.592, p<0.0055). CONCLUSION: The TFI performed slightly better than mFI-11 and mFI-5 in our study. Moreover, future studies are needed to further determine an optimal frailty instrument with great diagnostic and predictive accuracy.