Millions of Americans have Alzheimer's Disease and Related Dementias (?dementia?), with many millions more expected to develop it over the next few decades; dementia is a life-altering condition with high levels of associated morbidity. Most of these patients are Medicare beneficiaries. With new treatments on the horizon, there is both a promise for future improvements and risk because of the potential cost of such treatments combined with growing numbers of patients, especially risk of fiscal strain for patients, families, and the Medicare program (total spending is estimated to reach $1 trillion dollars/year by 2050). Despite the promise and risk, there is limited information about current dementia care within the traditional fee-for-service Medicare program (aka TM), the larger of Medicare's two components (the other component is Medicare Advantage (MA), which is administered by private plans). One major barrier to examining care nationally within the Medicare program is the uncertain validity of dementia diagnoses in claims data. To address these issues, we first will use a novel dataset with individual-level linkages of longitudinal data from a dementia registry, electronic health record, and Medicare claims (2006-17) to predict which patients with dementia-related diagnosis codes have true disease. We then will apply this approach to a national dataset of all TM beneficiaries to identify beneficiaries with dementia (2006-21), assess care patterns, and examine the impact of Medicare policy changes on beneficiaries' receipt of guideline-concordant dementia care. We will exploit a natural experiment in which policy changes shift the distribution of patients in TM vs. MA within each county, i.e., changes in MA penetration because of mandated MA benchmark changes. Prior work has found that MA has better process quality compared to TM for some chronic conditions, e.g., diabetes, and that MA penetration favorably impacts guideline adherence and care in TM for such conditions. We will investigate the effect of MA penetration on dementia care within TM. We have three aims:
Aim 1) Validation of a claims-based dementia definition among those who have a dementia diagnosis;
Aim 2) Examination of the impact of MA penetration on guideline-concordant diagnostic evaluation for TM dementia patients;
and Aim 3) Examination of the impact of MA penetration on guideline-concordant treatments for TM dementia patients. In summary, we will apply modern data science approaches to identify the patients in the Medicare program who have dementia, then examine how changes in the Medicare program affects dementia care within each county in the United States. These Medicare policy changes both help generate evidence and could lend themselves to future interventions to improve dementia care, e.g., through adjustments in Medicare quality incentives. Moreover, these data could help inform patients, families, clinicians, and policy makers about how we can improve care for this rapidly expanding population of patients.
Dementia is a life-altering condition that is common and growing more so among older Americans. Most dementia patients have Medicare, but little is known about the care that they receive. One challenge is correctly identifying the patients who have dementia. We will use novel information and modern approaches for identifying who has dementia, and then examine care for these patients across the national Medicare program.
