STRUCTURAL SENSITIVITY ANALYSES OF MORTALITY AND LOCATION OF CARE IN A SIMULATION OF ALZHEIMER'S DISEASE
Author(s)
Kansal A, DosSantos R, Tafazzoli A
Evidera, Bethesda, MD, USA
OBJECTIVES: Simulating outcomes of patients with AD requires modeling the natural history and management of AD over decades. A particular challenge of AD simulation is that a large fraction of the cost of management is driven by institutional care and thus is concentrated at the end of patients’ lives. This may result in model outcomes that are highly sensitivity to the interaction between predictors of mortality and location of care (LoC) may have a substantial impact on the predicted economic outcomes. This study quantifies the impact of using different published approaches to mortality and location of care in a simulation of AD. METHODS: Disease progression was simulated over 10 years using the AD Archimedes Condition-Event (ACE) simulator in a cohort of patient profiles with Mini-Mental State Exam (MMSE) scores ≥25. Three published sets of predictors of mortality were included: one where mortality was determined by current disease stage, one where it was determined by baseline stage, and one that was independent of disease stage. Two published predictors for LoC were also included: one where LoC was based on current disease stage and a second where LoC was determined by time spent at current stage. Combinations of these approaches were simulated to predict patients outcomes, with a focus on time spent in institutional care. RESULTS: The fraction of patients predicted to require institutional care varied between 10% and 15%, with time in institutional care for these patients ranging from 1.7 to 3.1 years depending on the combination of mortality and LoC predictors selected. Overall, there was a nearly 2.5-fold difference in total time of institutional care depending of the specific published predictors incorporated. CONCLUSIONS: Time spent in institutional care varies markedly when using published analyses in a simulation of AD, supporting a need for structural sensitivity analyses in AD simulation.
Conference/Value in Health Info
2016-05, ISPOR 2016, Washington DC, USA
Value in Health, Vol. 19, No. 3 (May 2016)
Code
PRM84
Topic
Methodological & Statistical Research
Topic Subcategory
Modeling and simulation
Disease
Mental Health, Neurological Disorders