The challenges of using the Hospital Frailty Risk Score

Machine learning for predicting readmission risk among the frail: Explainable AI for healthcare

“…Using existing methods to detect frailty is challenging since frailty-related diagnosis codes are subject to under-coding, coding may over-represent frailty due to comorbidities such as dementia, whereas frail patients with other comorbidities such as cardiac conditions or cancer might be grouped with non-frail patients. 17 The FRS-26-ICD used in the present study is at risk for Here, we show the top two observations with the lowest number of readmissions and the highest number of readmissions similar issues since coding for ambiguous syndromes such as weakness, fatigue, and dysphagia, which are indicators in the FRS-26-ICD, may not be a priority for the healthcare provider when considering the primary reason for admission when the number of codes possible for data entry is limited. However, in future research, capture of these important patient conditions that are salient in frailty and can be used in modeling frailty will make it possible to represent the patient's health status with higher accuracy.…”

“…Using existing methods to detect frailty in the acute care setting is challenging since frailty-related diagnosis codes are subject to under-coding, whereas frail patients with other comorbidities, such as cardiac conditions or cancer, might be grouped with non-frail patients. 17 Our study utilized a proxy measure for frailty (FRS-26-ICD) 18 drawn from ICD-10, Clinical Modification (ICD-10-CM) disease diagnosis codes that encompass, common geriatric syndromes, psycho-social factors, and blood biomarkers. The FRS-26-ICD defines frailty as a clinical syndrome resulting from multi-system physiologic impairments and failed integrative responses with diminished capacity to resist and recover from stressors.…”

“…Using existing methods to detect frailty in the acute care setting is challenging since frailty-related diagnosis codes are subject to under-coding, whereas frail patients with other comorbidities, such as cardiac conditions or cancer, might be grouped with non-frail patients. 17 …”

Measuring Frailty in Health Care Databases for Clinical Care and Research

“…Moreover, frail individuals may not be classified exclusively into a single group (e.g., frail people with cancer and frail people with heart disease may be classified into different groups despite similar levels of frailty). 38) Data-Driven Selection with a Reference Standard If a population-based dataset containing information on a reference standard frailty measure (e.g., frailty phenotype or deficit-accumulation frailty index) and administrative claims data is avaiable, specific codes can be selected against the reference frailty measure (also known as supervised learning). Several variable selection algorithms have been applied-e.g., stepwise regression, 39,40) penalized regression, 39,41) or tree-based algorithms.…”

“…Moreover, frail individuals may not be classified exclusively into a single group (e.g., frail people with cancer and frail people with heart disease may be classified into different groups despite similar levels of frailty). 38 )…”