Prognostic Indices for Hospitalized Older Adults with and without Alzheimer’s Disease and Related Dementias

Project Summary/Abstract We will develop and externally validate electronic health record (EHR)-based prognostic indices to improve healthcare quality for hospitalized older adults by identifying patients most likely to qualify for (and benefit from) specific health services. Our indices will predict: 1) 6-month mortality to guide hospice referral decisions; and 2) 2-year mortality to guide outpatient palliative care referrals and inform medication deprescribing. We will develop each of these indices for patients with and without Alzheimer’s disease and related dementias, since previous studies suggest that the trajectory of decline differs between older adults with and without Alzheimer’s disease and related dementias. Function is critically important to prediction in older adults, particularly those with Alzheimer’s disease and related dementias, and we have shown that the prognostic power of function increases with age. A major limitation that has hindered the effectiveness of predictive indices for hospitalized older adults has been the lack of functional data in the EHR. Recently, recognizing the critical role of function in providing high quality healthcare to older adults, hospital systems have started to routinely assess and document functional status. This development facilitates our proposal since we now have access to routine functional data from several hospitals, allowing us to incorporate these data into prognostic indices. With EHRs becoming ubiquitous in healthcare systems, our externally validated indices could be integrated into the EHR in most hospitals. To facilitate use in hospitals without EHR integration, we will develop parsimonious indices (also for ADRD and non-ADRD patients) available on ePrognosis, our free and widely used online compendium of geriatric prediction indices. We will set a new standard for equity-conscious prognostic model building by “baking equity” into model selection and incorporating neighborhood disadvantage as predictor representing social determinants of health. We have established a 4 site Collaboratory, which all routinely collect physical function data, clinical diagnoses, standardized delirium assessments, laboratory values, and physiologic measures: UCSF & Cleveland Clinic (development cohort); and Beth Israel Deaconess Medical Center & Johns Hopkins (external validation cohort). We propose to: (1) Develop full prognostic indices designed to be embedded in EHRs for 6-month and 2-year mortality for hospitalized older adults with and without Alzheimer’s disease and related dementias; (2) Develop parsimonious web-based prognostic indices designed to be accessed through ePrognosis for 6-month and 2-year mortality for older adults with and without Alzheimer’s disease and related dementias; (3) To internally validate the effectiveness of these prognostic indices; and (4) to externally validate these indices. These indices will help clinicians improve healthcare quality for older adults by prompting alignment of patient prognosis with health services (i.e. hospice and palliative care referrals) and deprescribing decisions.