About

Geoffrey Joyce, PhD, is a professor and chair of the Department of Pharmaceutical and Health Economics at the USC Alfred E. Mann School of Pharmacy and Pharmaceutical Sciences. He also serves as the director of health policy at the USC Schaeffer Center for Health Policy & Economics and is a research associate at the National Bureau of Economic Research. His research focuses on the costs of medical care and the role of insurance, with particular emphasis on drug benefit design, healthcare economics, Medicare Part D, aging and chronic disease, HIV/AIDS, and prescription drug pricing. Dr. Joyce has authored more than 125 peer-reviewed articles, reports, and book chapters, and his work has been published in leading medical, economic, health policy, and statistics journals. His research has been featured in prominent media outlets including the Wall Street Journal, the Washington Post, the Los Angeles Times, U.S. News & World Report, NBC Nightly News, CBS Evening News, and NPR. His studies examine various issues such as pharmacy benefit designs, the lifetime costs of chronic diseases, disparities in medication adherence, generic drug market dynamics, and out-of-pocket spending for Medicare beneficiaries. Dr. Joyce's work contributes to understanding the economic impacts of healthcare policies and pharmaceutical pricing, informing both academic discourse and public policy.

Research topics

• Medicine
• Internal medicine
• Psychology
• Family medicine
• Actuarial science
• Gerontology
• Pharmacology
• Finance
• Economic growth

Selected publications

• The changing Part D landscape

  Health Affairs Scholar · 2026-04-01

  articleOpen access1st authorCorresponding

  Introduction: Differences in how Medicare advantage prescription drug plans (MA-PDs) and stand-alone prescription drug plans (PDPs) are financed may contribute to fewer coverage options for traditional Medicare beneficiaries. The Inflation Reduction Act (IRA) capped annual out-of-pocket spending and reduced government reinsurance, thereby placing more cost responsibility on plans and manufacturers and potentially further limiting coverage. Methods: This study draws on 2020-2025 data from the CMS public use files to compare coverage options by cost-sharing categories (defined by mean premium and deductible) across plan types and plan sponsors. Results: A majority of MA-PDs maintained low-premium and low-deductible designs (2020 = 66.0%, 2025 = 62.4%), reflecting cross-subsidies from integrated medical benefits, while enhanced PDPs increasingly pursued low-premium and high-deductible structures (38.2%, 51.5%), and basic/actuarially equivalent (AE) PDPs maintained high-premium and high-deductible structures (66.0%, 65.1%). In 2025, more than half of brand-only drugs in unprotected classes were excluded from Part D formularies (enhanced MA-PDs = 51.8%, enhanced PDPs = 57.3%, basic/AE PDPs = 60.0%), especially in high-spending therapeutic areas. Conclusions: Results reflect increasing consolidation and reduced plan diversity, with potential consequences for access and affordability among traditional Medicare beneficiaries. Policymakers may need to reconsider subsidy and risk-adjustment mechanisms to preserve competition and equitable access across plan types in the evolving Part D marketplace.

  PublisherDOI

• Alzheimer's disease Stage 1 and 2: Biology, diagnostics, and treatment

  Alzheimer s & Dementia Translational Research & Clinical Interventions · 2026-01-01

  articleOpen access

  Abstract Advances in biomarker technology, digital cognitive assessments, and amyloid‐targeting therapies have redefined the opportunities for accurate and early diagnosis and care of Alzheimer's disease (AD). These advances also create new possibilities for intervention before the onset of cognitive impairment. This paradigm shift has increased the focus on Stages 1 and 2 of AD, in which individuals are cognitively unimpaired but exhibit biological evidence of disease. While early identification of AD offers an opportunity to intervene early to delay progression and preserve quality of life, it also presents complex challenges related to communicating diagnostic results to patients and their families, contextualizing the cost effectiveness of early diagnosis and treatment, and implementation of and equitable access to treatments. Recent, successfully enrolled, preclinical AD trials highlight the complex strategies required to identify asymptomatic, biomarker‐positive individuals on a large scale, and demonstrate critical knowledge gaps in inclusion, follow‐up, and long‐term outcome measurement. The Spring 2025 Alzheimer's Association Research Roundtable (AARR) meeting brought together academics, clinicians, industry, and regulatory leaders to exchange perspectives on current challenges, key learnings, and potential strategies for identifying and treating individuals in very early stages of AD, effectively and safely. This paper presents key takeaways from the Spring 2025 AARR meeting. Highlights New criteria for Alzheimer's disease (AD) enable early identification of AD pathology. Blood tests and digital cognitive assessment tools may facilitate early and personalized care. Plasma biomarkers may be scalable in clinical settings but need confirmation or follow‐up assessments in those individuals with equivocal results. Early therapy shows best results in people with lower initial amyloid and tau burden.

  PublisherDOI

• The cost of misaligned incentives in the pharmaceutical supply chain

  Health Affairs Scholar · 2025-06-25

  articleOpen access1st authorCorresponding

  Recent concerns over rising drug prices have focused on the role of pharmacy benefit managers or PBMs. While multiple players make up the pharmaceutical supply chain, PBMs are the conductors who effectively decide which drugs are covered and at what cost. Most PBM contracts tie their compensation to a percentage of a drug's list price, creating a financial incentive to favor high-cost, high-rebate drugs on plan formularies at the expense of lower-cost generics and biosimilars. Furthermore, the PBM industry is highly concentrated and vertically integrated with the country's largest health insurers, making it even harder to assess PBM performance and profitability. A simple analysis of annual drug spending at different reporting levels provides important insight into where the money goes and where savings could be achieved. We find that simply delinking compensation to the list price of a drug throughout the supply chain could reduce annual drug spending by more than $95b or nearly 15% of net spending without adversely affecting manufacturers' incentive to innovate.

  PublisherOA PDFDOI

• Disaggregating medical costs due to dementia

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: Medical care costs of high for persons living with dementia (PLWD) and are particularly high in the year of diagnosis, but remain higher than those without dementia thereafter. In part, this is due to differences in type and amount of health care used and this may vary by sociodemographic and economic background and/or may be due to differences in comorbidities and their management. There is a gap in understanding of the heterogeneity in medical costs due to dementia and the role of comorbid conditions in these costs. METHOD: In this study, we use Medicare Parts A, B, and D claims from 2006 to 2020 from 100% traditional Medicare beneficiaries to estimate medical costs among diverse PLWD. We used optimal propensity score matching to match PWLD to similar persons without dementia, using age, sex, race/ethnicity, six comorbid conditions, and Charlson Comorbidity Index. We quantified medical costs due to dementia, by type of medical care and across region, sex, race/ethnicity, and among PLWD with a high-cost comorbid condition of diabetes. We report costs for PLWD and matched persons without dementia. RESULT: In 2020, average medical costs across all settings were $35,828 among PLWD, compared to $29,752 among matched persons without dementia. Nearly half the $6,077 difference in overall costs was due to inpatient costs ($2,943). Cost differences attributable to dementia were largest among Asians ($8,931) and peaked at ages 65-70 ($7,992). Prevalence of diabetes among PLWD was 38.3% compared to 27.6% among persons who never developed dementia. Average medical costs for PLWD with diabetes were over 2 times higher for PLWD ($46,772), compared those without dementia ($21,140). CONCLUSION: This research will be the first to present nationally-representative estimates of medical costs attributable to dementia by region, sex, race and comorbid status. Our final set of analyses will further disentangle costs across additional patient subgroups, settings, and time to ascertain when, where, and for whom cost of dementia is highest. Results will inform comprehensive estimates of cost of dementia using the USCDM and may have clinical implications for management of comorbid conditions in PLWD.

  PublisherOA PDFDOI

• Costs of dementia: Comprehensive Estimates from U.S. Cost of Dementia Model (USCDM)

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: Dementia is associated with substantial economic burden for individuals, their care partners and family, communities, and society. There are myriad types of costs including those associated with medical care, long term care, supportive services, quality of life as well as others such as vulnerability to exploitation. There is a gap in annual, nationally representative, comprehensive cost estimates and variability across types of costs and persons. Tracking these costs over time as prevention, treatment and policy evolve will support efforts to reduce the economic impact of dementia. METHOD: We used nationally representative data from the Health and Retirement Study (HRS) of U.S. adults over age 50 and other nationally representative data sources. We employed the U.S. Cost of Dementia Model (USCDM) - a dynamic microsimulation model and estimated medical and long term care costs, costs of time and earnings loss associated with unpaid family caregiving and quality of life of persons living with dementia and their care partners. RESULT: In 2025, an estimated 5.6 million people are living with dementia in the United States. The medical and long-term care costs for dementia total $232 billion. Families and friends provide 6.8 billion hours of care, which is worth $233 billion. Many care partners also reduce their work hours or leave their jobs entirely, leading to a loss of $8.2 billion in annual earnings among care partners of parents with dementia. In addition to financial costs, people with dementia and their care partners experience a significant loss in quality of life, valued at $302 billion for those living with dementia and $5.9 billion among caregivers of parents or spouses with dementia. CONCLUSION: The USCDM, a comprehensive, well-validated dynamic microsimulation model, provides national, comprehensive estimates of total costs of dementia and by type of costs and who bears them. It is part of a larger effort to build capacity for researchers and policymakers to reduce the future impact of dementia.

  PublisherOA PDFDOI

• After Risk-Adjustment Change, Dementia Diagnoses Increased In Medicare Advantage Relative To Traditional Medicare

  Health Affairs · 2025-01-01 · 3 citations

  article

  In 2020, the Centers for Medicare and Medicaid Services reintroduced Alzheimer's disease and related dementias to its risk-adjustment payment model for Medicare Advantage (MA) plans. Using 2017-20 data for 100 percent of community-dwelling beneficiaries enrolled in Medicare, we evaluated how the reintroduction of dementia to the risk-adjustment model affected rates of new (incident) dementia diagnoses among beneficiaries enrolled in MA relative to those enrolled in traditional Medicare. In response to the payment change, annual incident dementia diagnosis rates in MA increased by 11.5 percent relative to traditional Medicare. This increase was concentrated among beneficiaries who were more likely to have undiagnosed dementia-specifically, beneficiaries who were Hispanic or Black, were ages eighty-five and older, or were dually eligible for Medicaid or received a Part D low-income subsidy. Only a third of the increase came through chart reviews. Financial incentives to detect dementia increased dementia diagnoses, particularly among beneficiaries at high risk for dementia and undetected dementia, but questions remain about potential overdiagnosis or upcoding.

  PublisherDOI

• Formulary Restrictions and Relapse Episodes in Persons With Relapsing-Remitting Multiple Sclerosis

  JAMA Network Open · 2025-08-01 · 2 citations

  articleOpen accessSenior authorCorresponding

  Importance: Narrow formularies can be used to increase rebates and manage the use of costly drug therapies in the US. Objective: To examine the association between the breadth of formulary coverage for multiple sclerosis (MS) disease-modifying therapies (DMTs) and MS relapse. Design, Setting, and Participants: This retrospective cohort study analyzed 100% Medicare administrative data from 2018 to 2022. The data analysis was conducted from August 1, 2024, to January 30, 2025. Beneficiaries with stand-alone prescription drug plans (PDPs) and Medicare Advantage Prescription Drug plans (MA-PDs) were followed for at least 5 quarters (baseline, quarters 1-4; follow-up, quarter 5). Calendar periods reflected formulary decision-making. Beneficiaries in the same Medicare Part D plan during baseline and follow-up, with relapsing-remitting MS, and with MS DMT use during baseline were included. Exposures: Formulary breadth was low coverage if the 4-quarter moving average of MS DMT drug or class coverage was below the median by plan type and quarter; it was considered high coverage if the average was above the median. Main Outcome and Measures: The primary outcome was MS relapse, including inpatient or outpatient treatment for MS. Multivariable logistic regressions were estimated separately for PDPs and MA-PDs, controlling for patient and plan characteristics and clustering for repeated observations by beneficiary. Results: The claims analysis included 50 162 unique beneficiaries in PDPs (mean [SD] age, 58.5 [12.1] years; 74.9% female) and 34 708 in MA-PDs (mean [SD] age, 58.2 [10.3] years; 77.2% female). Oral or injectable MS DMTs were frequently excluded from coverage (>50% excluded in 2022: PDPs, 11 of 15 DMTs; MA-PD, 9 of 15 DMTs). The MS relapse rate was greater for low- vs high-coverage PDPs (10.6% vs 9.5%; odds ratio [OR], 0.88 [95% CI, 0.84-0.92]) and MA-PDs (7.8% vs 6.9; OR, 0.88 [95% CI, 0.85-0.91]). In multivariable analyses, broader formulary coverage during baseline was associated with less MS relapse during follow-up (PDP: adjusted OR, 0.93 [95% CI, 0.90-0.96] for drugs and 0.94 [95% CI, 0.91-0.97] for classes; MA-PD: adjusted OR, 0.88 [95% CI, 0.83-0.94] for drugs and 0.92 [95% CI, 0.86-0.98] for classes). Conclusions and Relevance: In this cohort study of Medicare data, broader formulary coverage was associated with an 8% to 12% lower odds of MS relapse in MA-PDs and a 6% to 9% lower odds in PDPs. Formulary coverage and restrictions should be tailored for patient need, not just management of costs and use.

  PublisherOA PDFDOI

• Opportunities for using the USCDM for impact

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: The United States Cost of Dementia Model (USCDM) is a dynamic microsimulation model that generates nationally-representative estimates of comprehensive cost of dementia of individuals, families, payers and society. The model assesses cost across different stakeholders, and quantifies the impact on costs of changes in treatments, care, policies and population health for different populations. USCDM provides valuable insights for the research community, however its computational complexity may limit its full impact potential. METHOD: To increase access to broad audiences, the authors developed a prototype emulator that replicates the behavior of the full USCDM in a more computationally efficient and user friendly manner. The emulator used linear regression to predict key model outputs: number of people with any ADL limitations, the number of people living in nursing homes, diabetes prevalence, heart disease prevalence, dementia prevalence, Medicare costs, Medicaid costs, out-of-pocket medical costs, and total costs for all payers. We assessed impact on these outcomes from a reduction in all-cause mortality incidence, reduction in ADL limitations, and a delay in dementia onset. We compared regression model predictions to the actual USCDM output in terms of accuracy, interpretability, and efficiency. We performed these comparisons using a test data set of USCDM outputs that were not part of the sample data set used to estimate the emulator regression model. RESULT: For scenarios representing a change in all-cause mortality incidence, a reduction in the number of ADL limitations, or delayed dementia onset, we assessed accuracy by relative error - the percent difference between each USCDM output and the emulator prediction - and found the emulator outcome results were within 0.01% of USCDM outputs across all demographic population we assessed. The emulator yields these predictions in under 10 seconds while the USCDM would take 23 minutes in a 20-processor parallel computing environment. CONCLUSION: By improving accessibility to USCDM outputs, the emulator enhances usability for researchers conducting health economics analyses, policymakers evaluating the impact of interventions, and the layperson seeking a clearer understanding of dementia-related health and economic outcomes.

  PublisherOA PDFDOI

• Impact of dementia caregiving on the labor force participation and lifetime earning of middle age adults

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: Many persons living with dementia receive care from adult children. Providing care is mentally and physically demanding, and time intensive and may reduce caregivers' lifetime earnings and wealth. Population estimates of the financial impact of dementia caregiving has been limited by lack of nationally representative, longitudinal data on changes over time of middle-age adults' work and time spent caregiving and their parents' health, and of estimates using rigorous panel data methods. METHOD: We used nationally-representative data from 2012-2018 waves of the Health and Retirement Study on the demographic, economic and health characteristics of respondents ages 50 to 64 and of their parents/parents-in-law (hereafter "parents"). Primary outcomes were caregiving for a parent with dementia and labor force status. We first used fixed effect models to estimate the probability of providing care to a parent with dementia and quantify the association of dementia caregiving and labor supply independent of unobserved differences. We then employed the United States Cost of Dementia Model (USCDM), a validated dynamic microsimulation model, to quantify the impact of dementia caregiving on lifetime earnings and wealth accumulation for a cohort of middle age adults. RESULT: We used data from 7,522 respondents and their 12,010 living parents. Fifty four percent of respondents were female and mean age was 58. Sixty seven percent of respondents' parents were female and mean age was 82.7. About one-fifth (20.4%) of respondents had a least one parent with dementia. Among respondents who had a parent with dementia, 19.4% were providing care for an average 17.9 hours per week. Among dementia caregivers, 57.5% were in the labor force. A 10% increase in wage rate was associated with a 0.08% decline in the probability of being a dementia caregiver. The likelihood of working was reduced by 3.4% for dementia caregivers relative to non-caregivers. CONCLUSION: Dementia caregiving is common among middle age adult children and reduces their labor supply. For most workers, earnings levels peak at middle age and labor market exits are permanent. Ongoing work employs dynamic microsimulation to quantify the impact of dementia caregiving on lifetime earnings and household wealth accumulation and differences by sex, race, and education.

  PublisherOA PDFDOI

• Real-World Evidence of Brexpiprazole Use and 6-Month Mortality, Hospitalization, and Emergency Department Visits Among Persons With Dementia

  Neurology · 2025-07-03 · 1 citations

  articleOpen accessSenior author

  BACKGROUND AND OBJECTIVES: Alzheimer disease and other dementias are accompanied by depression and agitation and other behavioral and neuropsychiatric symptoms. In 2023, brexpiprazole became the first antipsychotic approved by the US Food and Drug Administration to treat agitation in persons with Alzheimer disease, but, like all atypical antipsychotics, it includes a black box warning of an increased risk of mortality among persons with dementia. This study provides real-world evidence of mortality in a heterogeneous sample of brexpiprazole users to understand effects in the population. METHODS: We used a 100% sample of Medicare claims data Parts A, B, and D from 2014 to 2023. Our sample was limited to beneficiaries with diagnosed dementia, who were continuously enrolled for at least 2 years and were new users of the atypical antipsychotics brexpiprazole or aripiprazole in a given year. We used matching and logistic regression to estimate the relationship between incident use of brexpiprazole, compared with aripiprazole, and mortality, emergency department (ED) visits, and hospitalization within 6 months. RESULTS: Among the 41,871 beneficiaries with dementia, 71.7% of brexpiprazole and 69.7% of aripiprazole users were women with a mean age of 75.7 and 78.0 years, respectively. Among persons living with dementia (PLWD), 6-month mortality was statistically lower among new users of brexpiprazole based on estimates from logistic regression and a matched sample of new users of brexpiprazole or aripiprazole (OR 0.49, [95% CI 0.37-0.65]).There was no statistical difference between the incident use of brexpiprazole and aripiprazole use for ED visits or hospitalization within 6 months of use initiation. Adjustment for potential unobserved confounding used two-stage least squares estimation and found no statistically significant differences in six-month mortality, ED visits, or hospitalizations between the 2 groups. DISCUSSION: Brexpiprazole use is not associated with differential mortality risk compared with aripiprazole use among PLWD. Brexpiprazole offers a treatment option which is important given the heterogeneity of effects of antipsychotics on persons. A two-stage least squares method is used to eliminate bias on estimates because of observed and unobserved differences between the 2 groups, but the small sample size is a limitation. CLASSIFICATION OF EVIDENCE: This study provides Class III evidence that brexpiprazole does not increase the risk of mortality at 6 months compared with aripiprazole in PLWD.

  PublisherOA PDFDOI

Recent grants

• NIH Grant R03HS013447

  NIH · $100k · 2004

• Evidence for Value-based Prescription Drug Plan for Diabetes, CVD and Chronic Illness

  NIH · $1.3M · 2019–2024

Frequent coauthors

• Dana P. Goldman

  University of Southern California

  96 shared

• Matt P. Wise

  28 shared

• Julie Zissimopoulos

  University of Southern California

  23 shared

• Anupam B. Jena

  Harvard University

  20 shared

• Samuel A. Bozzette

  18 shared

• Jennifer Pace

  Indiana University South Bend

  17 shared

• Christopher S. Saigal

  17 shared

• Erin Trish

  University of Southern California

  16 shared