Ignoring life expectancy can lead to poor clinical decisions. Healthy older adults who could benefit from screening have low screening rates. Older adults with dementia or metastatic cancer are screened for slow-growing cancers that are unlikely to cause symptoms but may lead to distress from false-positive results, invasive work-ups and treatments. Life expectancy calculators offer the possibility of moving beyond arbitrary age-based cutoffs to a more rational life expectancy based decision-making, so that screening and prevention can be targeted to those Veterans most likely to benefit. However, current life expectancy calculators are time consuming for clinical staff to use and have not focused on the long time frames most relevant for screening decisions.
The objectives of the study are to (a) develop and validate a 10-year life expectancy calculator using VA electronic data; (b) develop and validate an Augmented life expectancy calculator that adds functional assessments (Activities of Daily Living or ADL and Instrumental Activities of Daily Living or IADL assessments) Medicare data to the other risk factors and determine whether the addition of these additional data elements lead to improved prediction; (c) apply the life expectancy calculator to a national VA cohort to determine the proportion of colorectal cancer (CRC) screening that occurs in veterans with limited life expectancy (<25% likelihood of life expectancy of 10+ years) and extended life expectancy (>75% likelihood of life expectancy of 10+ years) in both veterans within the recommended age (50-75) for CRC screening as well as veterans at more advanced age (76-85) when screening is not routinely recommended.
To achieve these objectives, we will leverage the wealth of electronic data available within VA (demographics, comorbidities, laboratory results and pharmacy records) and use state-of-the-art prediction modeling techniques to develop life expectancy calculators. Specifically, we will use LASSO (Least Absolute Shrinkage and Selection Operator) regression to simultaneously select important factors for predicting life expectancy as well as estimate coefficients for those factors. We will develop several models to determine whether the addition of specific data elements (functional status and non-VA CMS data) improves life expectancy prediction. Finally, we will utilize life expectancy prediction to determine how much colorectal cancer (CRC) screening is occurring in Veterans with a limited life expectancy who are unlikely to benefit and how much CRC screening is occurring in Veterans with an extended life expectancy who are more likely to benefit.
Aim 1: We have obtained, linked and started to clean and our analytic dataset of 3.9 million veterans with outpatient visits in 2005 with potential predictor variables during a 1 year lookback period (~180 comorbidity categories, ~380 drug classes, laboratory values, vital signs, utilization data).
Aim 2: We are awaiting final approval of VA-Medicare data.
Aim 3: none so far.
By developing and validating a VA electronic data driven life expectancy calculator, this project will determine the rates of potentially inappropriate CRC overscreening in age-appropriate veterans (age 50-75) with a limited life expectancy (<25% likelihood of life expectancy 10+ years). Conversely, this project will also identify potentially inappropriate CRC underscreening in veterans beyond the recommend age for screening (age 76-85) with an extended life expectancy (>75% likelihood of life expectancy of 10+ years). This work will provide the critical foundation for an intervention to improve the targeting of CRC screening will estimate an individual veterans' life expectancy to 1) suppress CRC screening clinical reminders for patients with limited life expectancy (age 50-75) or 2) trigger CRC screening clinical reminders for patients with extended life expectancy (age 76-85).
External Links for this Project
Grant Number: I01HX002135-01A2
- Lee SJ, Larson EB, Dublin S, Walker R, Marcum Z, Barnes D. A Cohort Study of Healthcare Utilization in Older Adults with Undiagnosed Dementia. Journal of general internal medicine. 2018 Jan 1; 33(1):13-15. [view]
- Lee SJ, Smith AK, Diaz-Ramirez LG, Covinsky KE, Gan S, Chen CL, Boscardin WJ. A Novel Metric for Developing Easy-to-Use and Accurate Clinical Prediction Models: The Time-cost Information Criterion. Medical care. 2021 May 1; 59(5):418-424. [view]
- Brown RT, Diaz-Ramirez LG, Boscardin WJ, Cappola AR, Lee SJ, Steinman MA. Changes in the Hierarchy of Functional Impairment From Middle Age to Older Age. The journals of gerontology. Series A, Biological sciences and medical sciences. 2022 Aug 12; 77(8):1577-1584. [view]
- Jing B, Boscardin WJ, Deardorff WJ, Jeon SY, Lee AK, Donovan AL, Lee SJ. Comparing Machine Learning to Regression Methods for Mortality Prediction Using Veterans Affairs Electronic Health Record Clinical Data. Medical care. 2022 Jun 1; 60(6):470-479. [view]
- Odden MC, Lee SJ, Steinman MA, Rubinsky AD, Graham L, Jing B, Fung K, Marcum ZA, Peralta CA. Deprescribing Blood Pressure Treatment in Long-Term Care Residents. Journal of The American Medical Directors Association. 2021 Dec 1; 22(12):2540-2546.e2. [view]
- Deardorff WJ, Jing B, Jeon SY, Boscardin WJ, Lee AK, Fung KZ, Lee SJ. Do functional status and Medicare claims data improve the predictive accuracy of an electronic health record mortality index? Findings from a national Veterans Affairs cohort. BMC geriatrics. 2022 May 18; 22(1):434. [view]
- Kotwal AA, Lee SJ, Dale W, Boscardin WJ, Waite LJ, Smith AK. Integration of an Objective Cognitive Assessment Into a Prognostic Index for 5-Year Mortality Prediction. Journal of the American Geriatrics Society. 2020 Aug 1; 68(8):1796-1802. [view]
- Nicosia FM, Spar MJ, Steinman MA, Lee SJ, Brown RT. Making Function Part of the Conversation: Clinician Perspectives on Measuring Functional Status in Primary Care. Journal of the American Geriatrics Society. 2019 Mar 1; 67(3):493-502. [view]
- Lee AK, Jing B, Jeon SY, Boscardin WJ, Lee SJ. Predicting Life Expectancy to Target Cancer Screening Using Electronic Health Record Clinical Data. Journal of general internal medicine. 2022 Feb 1; 37(3):499-506. [view]
Treatment - Observational, Prognosis, TRL - Applied/Translational
Cancer, Clinical Diagnosis and Screening, Decision Support, Predictive Modeling