A user‐friendly calculator to estimate length of time from initial presentation to need for full‐time care or death in individual patients with Alzheimer’s disease
Yaakov Stern, Eric Stallard, Hyunnam Ryu, Anton J Kociolek, Seonjoo Lee, Stephanie Cosentino, Carolyn W Zhu, Michelle Hernandez, Bruce Kinosian, Yian Gu- Psychiatry and Mental health
- Cellular and Molecular Neuroscience
- Geriatrics and Gerontology
- Neurology (clinical)
- Developmental Neuroscience
- Health Policy
- Epidemiology
Abstract
Background
Alzheimer’s disease (AD) exhibits substantial variability in its initial presentation and rate of progression. Our goal was to understand AD progression well enough to create a calculator that uses presented clinical features to estimate the length of time from initial presentation to major disease outcomes in individual patients with AD.
Method
The variability in rate of AD progression is strongly related to the variability in specific disease signs and symptoms identified at initial presentation. We developed a longitudinal Grade of Membership prediction algorithm and demonstrated its ability, based on initial presentation of six fixed covariates and 73 time‐varying clinical signs and/or symptoms, to accurately predict time to need for full‐time care (FTC) and death in two groups: a mainly white, clinic‐based group of AD patients (N = 229), and a statistically representative community‐based sample of Hispanic (N = 211) and non‐Hispanic (N = 62) older adults with MCI or AD. We then created an online calculator for use by physicians, relatives, or patients to generate these predictions.
Result
The calculator is implemented as an online, freely available R Shiny web app: https://cnd‐columbia.shinyapps.io/AD_Calc/. The user is prompted for current clinical information about the patient. Most of this information can be gleaned from patient observation, healthcare providers, or medical records. Unknown items can be skipped or sensibly approximated. The calculator generates individual‐specific survival‐probability curves, including median survival time, over the next 10 years. It also generates conditional probabilities of need for FTC in a way that allows the overall survival time to be decomposed into time free of need for FTC and time in need of FTC. The calculator does not predict which specific individuals would die before the median versus after the median; that prediction cannot be done.
Conclusion
Given the substantial variability over individuals in rate of disease progression, knowing the statistics of the individual‐specific survival times as well as the times to need for FTC can be of great value to patients with AD, their caregivers, and health care providers. The new calculator meets this need. In the future, the calculator also can be extended to predict time to other important disease milestones.