Approaches to optimally target frailty screening among people with HIV in clinical care: findings from the Centers for AIDS Research Network of Integrated Clinical Systems (CNICS)

Background: Frailty screening in HIV care has been recommended, however, screening adds burden to busy clinics. We compared criteria that predict concurrent frailty to identify approaches to optimally target frailty screening. Methods: The development cohort included people with HIV (PWH) at six U.S. sites. Frailty was assessed based on four components of a modified Fried phenotype: fatigue, unintentional weight loss, low mobility, and poor physical activity. We evaluated demographic and clinical characteristics, comorbidities, and substance use as predictors of who should be screened using selection approaches for simple and complex frailty screening tools, including machine learning approaches. We compared discrimination and calibration including area under the receiver operator characteristic (ROC) curve (AUC), sensitivity, and specificity in a validation cohort (7 th site). Results: Among the 9,592 PWH in the development cohort, 11% were frail. AUC ranged from 0.52 for simple screening approaches such as age-based to 0.84 for complex approaches in the development and validation cohorts. Using an age cutoff >50 years reduced the percentage of PWH needing screening by over half but also reduced the sensitivity to 58% in the validation cohort. Complex approaches required 47% to be screened and had a sensitivity of 89%. Conclusions: Age-based frailty screening approaches (e.g., >50 years) miss many frail PWH. Complex tools had marginally better testing characteristics but would be more difficult to implement in clinical care. Simple targeted screening based on three characteristics (age, gender, and depressive symptoms) identified 89% of frail PWH and reduced the number who needed screened by 52%.