Development and Validation of a Predictive Risk Score for Screening Frailty in Veteran Health Administration Hospitalized Patients
19 Pages Posted: 12 Apr 2019More...
Background: The use of administrative data could provide an efficient method to identify hospitalized patients at risk of complications due to frailty. This is shown in a recent study which developed a frailty risk score using administrative data from United Kingdom hospital encounters. However, its applicability to United States (US) patients is unknown. We sought to develop a frailty risk score using healthcare administrative data from the Veterans Health Administration (VA), the largest integrated health care system in the US.
Methods: We developed a frailty risk score using hierarchical clustering analysis to identify VA patients (n=127,350) with diagnostic codes associated with frailty and high resource use. The score was validated in two ways. First, we determined how well the score predicted adverse outcomes in 75,453 VA patients. Second, we determined frailty status based on clinical observations and assessed how well the frailty score predicted frailty status in 9,565 VA patients.
Findings: In the developmental cohort, those with the highest frailty scores were older (mean=71*9 versus 67·3 years) and had higher mortality (22·7% versus 11·8%). Upon validation, patients in the highest frailty risk group had a higher adjusted odds of hospital stays lasting >10 days (odds ratio(OR) =2·39; 95%confidence interval(CI)=2·15-2·66), 30-day mortality (OR=2·14;95%CI=1·84-2·49), and readmission (OR=2·60;95%CI=2·34- 2·89), versus the lowest frailty group. Validation with clinician assessed frailty status revealed high predictability, as each unit increase in frailty score was associated with 20% higher odds of being clinically assessed as frail (95%CI=1·19-1·21,c-statistic=0·82).
Interpretation: We produced a predictive model to identify frail hospitalized United States patients using administrative data. The score's performance was strong in predicting clinician-assessed frailty. Applications include use by clinicians to triage patients, by hospitals for resource allocation, and by health systems for risk adjustment.
Funding Statement: This study was supported by a VHSR&D funded grant.
Declaration of Interests: The authors report no potential conflicts of interest relevant to this article.
Ethics Approval Statement: The study is approved by the institutional ethical review board of the Charleston and National VA.
Suggested Citation: Suggested Citation