Development and validation of a clinical model to estimate the probability of frailty in older emergency patients

Abstract

Abstract Objectives Most instruments to measure frailty are impractical to operationalize in emergency department (ED), due to the busy work of the staff and the acute presentations of the patients. The current work aimed to develop and validate a clinical model by combining clinical data and biomarkers available in the ED to evaluate the probability of frailty among older emergency patients. Design: A cross-sectional research. Setting and participants: A total of 273 adults aged 65 years and older who visited the ED of a university-affiliated hospital in Beijing, China during the period from January 2021 to September 2021 were included. Measurements: Fried's frailty phenotype was employed for assessing frailty. Clinical data and blood biomarkers were collected within 24 hours of admission. Clinical data included demographic characteristics, chronic diseases, main causes of admission and nutrition risk screening (NRS) 2002 score. Blood biomarkers included hallmarks of inflammation, nutrition and metabolism, coagulation and other biomarkers such as alanine transaminase (ALT), aspartate transaminase (AST), N-terminal pro-brain natriuretic peptide (NT-pro-BNP), and creatinine. Results Totally 151 patients were frail. Eight variables were included in the diagnosis model: nutritional risk, prior ischemic heart disease (IHD), age > 77 years, ALT ≤ 9.2 U/L, albumin ≤ 31.9 g/L, homocysteine > 18 umol/L, 25-OH vitamin D ≤ 10.89 ng/mL, and NT-pro-BNP > 2070 ng/L. Through the area under receiver operating characteristics curve (AUC) (0.855), leave-one-out cross-validation (AUC = 0.828), as well as Hosmer-Lemeshow test (P = 0.615), the validity of the diagnosis model was verified. The clinical diagnosis model was converted into a scale score with a total score of 10. The AUC of the diagnostic score for predicting frailty was 0.851 (95% confidence interval: 0.806–0.895, P< 0.001) and the cutoff value was 4.5 points with the Youden index of 0.579. Following adjustment for gender and main causes of admission, the diagnostic score was an independent predictor for 28-day mortality, with a hazard ratio of 1.486 and a 95% confidence interval of 1.224–1.804 (P < 0.001). Conclusion In the current work, a clinical diagnosis model to estimate the probability of frailty in older emergency patients was developed using clinical data and biomarkers generally readily available in the ED. This diagnosis model may facilitate targeted intervention and risk stratification in the EDs.