Frailty is increasingly recognized as a leading indicator of poor health outcomes, even death, as well as a barometer of how well patients respond to treatment. To truly provide patient-centered care, providers should be aware of each patient's frailty status and incorporate it into clinical decision making. Providers can now offer a number of invasive and aggressive procedures for cardiovascular disease, which involve risk, and can be painful. The treatment intensity need to match the expected patient outcome, yet providers do not have a reliable method to estimate prognosis for frail patients. In the study proposed here, we will use a novel approach that leverages the electronic health record (EHR) in identifying patient frailty status, with the goal of supporting retrospective clinical studies and prospective clinical decision making. Our preliminary studies have demonstrated the availability of frailty-related findings in EHR, the feasibility of extracting frailty findings, and the feasibility of using EHR-extracted frailty for outcome prediction.
The specific aims of the project are to 1) Create a frailty ontology building on existing functional status and quality of life measurements; 2) Develop ontology guided, natural language processing (NLP) methods for extracting frailty descriptions and measurements; 3) Develop a model to aggregate NLP-extracted frailty findings to generate a patient-level frailty score; 4) Examine the all-cause mortality and all-cause hospital readmission one year after major cardiac procedures in heart failure patients with different frailty scores and assess the impact of this information on surgical decision making.
Frailty is an important, but often overlooked, determinant of health outcomes in older adults. Providers can now offer invasive and aggressive treatments for various conditions, but these interventions involve risk and careful patient selection that balances risk and benefit is imperative. We will develop automated methods to extract frailty information from clinical records and generate an aggregated frailty score, which will enable retrospective analyses of EHR data for risk prediction and support prospective clinical decisional making.
| Cheng, Yan; Shao, Yijun; Weir, Charlene R et al. (2017) Predicting Adverse Outcomes in Heart Failure Patients Using Different Frailty Status Measures. Stud Health Technol Inform 245:327-331 |
