What You Should Know:
– AHRQ conducted a study to address the operational gap between CFIs and EFIs. This project focused on validating an established CFI using linked claims-EHR databases of multiple large health systems. The project provides a systematic approach that health systems can use to examine the quality of the EHR data and prepare it for the application of EFI measures.
– The findings demonstrated that structured EHR data can be used by healthcare providers to identify frail patients using validated EFIs; however, claims data can identify additional frailty cases compared to EHR data. EFIs can also be used to improve the prediction of various healthcare utilization outcomes. Risk stratification developers may integrate EFI in their model development process, and population health managers may incorporate EFI in disease management efforts.
Insights into Frailty and Functional Disabilities in EHRs
Frailty is conceptualized as decreased physiologic reserve and inability to withstand physical and psychological stressors. The goal of frailty measurement is to identify high-risk older adults and to target interventions to prevent adverse health outcomes. Despite the utility of frailty in identifying older adults at risk, and an abundance of frailty measures in the literature, healthcare providers continue to lack pragmatic tools to cost-effectively screen large patient populations for frailty.
Screening tools for frailty may identify individuals in need of further evaluation at the point of care, but such tools still require the availability of or collection of new data that is specific to the score (e.g., gait speed, chair rise, grip strength) and cannot be automatically calculated from information already in a patient’s chart. Healthcare providers and health insurance plans are actively seeking ways to measure frailty using insurance claims, electronic health records (EHRs), and on a more limited scale, health risk assessments. Applying and scaling frailty indexes across adult populations enable providers and plans to identify frail individuals at high risk for mortality, disability, and healthcare utilization. Multiple claims-based frailty indexes (CFIs) have been developed and validated over the past few years; however, healthcare providers often do not have access to the insurance claims records of their entire population of patients, thus necessitating the development of reliable EHR- based frailty indexes (EFI). Nonetheless, a challenge with developing EFI measures is the lack of frailty variables captured as structured codes within EHRs.
To address the operational gap between CFIs and EFIs, this project focused on validating an established CFI using linked claims-EHR databases of multiple large health systems: Johns Hopkins Medical Institute (JHMI); Optum Labs Data Warehouse (OLDW), which includes data from 55 health systems; and Kaiser Permanente Mid-Atlantic States (KPMAS). Task 2 of this project assessed and compared the EHR and claims data of these data sources to ensure sufficient data quality for frailty analysis. Task 3 of the project compared the EFI and CFI using EHR and claims data of each data source. Tasks 1 and 4 focused on administrative and dissemination efforts (e.g., data use agreements, scientific publications) and are not covered in this report.
The project provides a systematic approach to healthcare providers to examine the quality of the EHR data and prepare it for the application of EFI measures (Task 2). The EFI showed to be a valid measure of frailty when compared to a custom patient survey at KPMAS, and when compared to CFI measures of the same population across all data sources. An acceptable concordance of EFI and CFI was found and shown to be stable across multiple health systems.. The concordance of EFI and CFI was also acceptable across different patient groupings such as age, sex, and race. Finally, the EFI were found to be predictive of current and future healthcare utilization outcomes, such as inpatient hospitalization, emergency department admission, and nursing home admission.
In conclusion, the project findings demonstrated that structured EHR data can be used by healthcare providers to identify frail patients using validated EFIs; however, claims data can identify additional frailty cases compared to EHR data. Further research is needed to evaluate the role of unique EHR features, such as unstructured data in physician notes, in developing EFIs that have a higher sensitivity and specificity in identifying patients with frailty.
EFIs can also be used to improve the prediction of various healthcare utilization outcomes. Risk stratification developers may integrate EFI in their model development process, and population health managers may incorporate EFI in disease management efforts. Future studies should evaluate the interaction of comorbidity indexes with EFIs in predicting healthcare utilization outcomes and adjusting total healthcare costs.
