ABSTRACT Predicting life expectancy in older adults with advanced (stage IV or incurable stage III) cancers is challenging due to heterogeneity in the presence of chronic conditions and geriatric syndromes, which independently increase the risk of death. Accurate prediction and communication of life expectancy is a key component of shared decision-making for older adults with advanced cancers and their providers. However, evidence shows that oncologists tend to overestimate life expectancy in older adults with advanced cancer. Standard tools currently available for predicting life expectancy perform poorly in this population. The geriatric assessment is a set of validated, patient-reported and objective measures to assess comorbidity, functional status, physical performance, cognitive status, psychological status, nutritional status, polypharmacy, and social support, and has potential to improve prediction of life expectancy for older adults with advanced cancer. Yet, no studies have explicitly sought to harness information captured from the geriatric assessment to develop a clinical prediction model for life expectancy in older adults with advanced cancers. This study will leverage detailed information from geriatric assessment and advanced statistical methods to develop and validate a new life expectancy prediction model for older adults (age 70 and over) with advanced cancer. We will then evaluate the extent of improvement in predictive performance of this new GA model compared with oncologists’ self- reported estimates of life expectancy and a model based on routinely collected clinical data (age, comorbidities, and performance status). To achieve these aims, we will link the most up-to-date vital status data from the Medicare program with detailed geriatric assessments for >800 patients aged 70 years and older with advanced cancer and their oncologists participating in two community-based, cluster-randomized trials. Our multidisciplinary study team has the requisite expertise in data linkage, predictive modeling, geriatric oncology, and palliative care to conduct this research and translate our findings into meaningful advances for this vulnerable population. Our work will lead to the develop of a clinical prediction tool that can be used in future experimental work to test whether the life expectancy prediction model based on geriatric assessment can improve patient-provider communication about prognosis and the delivery of high-quality, patient-centered cancer care.