Racial/Ethnic disparities in Initiation and Persistent Use of Anti-dementia Medications

Abstract

BACKGROUND:

Racial/ethnic disparities in anti-dementia medications use in longitudinally followed research participants are unclear.

METHODS:

Study included initially untreated participants followed in NACC-UDS who were ≥65 at baseline with AD dementia.

Outcomes:

(1) any new Acetylcholinesterase (AChEI) treatment during follow-up, (2) persistence of treatment during follow-up categorized into: intermittent treatment (<50% follow-ups reporting treatment), persistent (≥50% follow-ups), and always treated. Outcomes for memantine treatment were similarly constructed.

RESULTS:

Controlling for participant characteristics, Black/African American and Hispanic participants remained less likely than White participants to report any new AChEI or memantine treatment during follow-up.

Among those who reported new treatment during follow-up, both Black/African American and Hispanic participants were less likely than White participants to be persistently treated with AChEI and memantine.

DISCUSSION:

Substantial racial/ethnic treatment disparities remain in controlled settings of longitudinal research where participants have access to dementia experts, suggesting wider disparities in the larger community.

INTRODUCTION

Underserved/under-represented populations are disproportionally affected in the growing number of older adults living with Alzheimer’s disease (AD) and other dementias in the United States.^1–6 Despite being a primary objective of the National Alzheimer’s Plan (NAPA), racial/ethnic disparities in the prevalence of dementia have not been narrowing over time.⁷ A recent white paper from the Alzheimer’s Association noted substantial gaps in the scientific literature regarding the impact of racial/ethnic disparities in AD and other dementias.⁸ A recent study reported less than half of individuals with dementia were informed by a physician about their dementia. Unrecognized or delayed diagnoses of AD and other dementias are more common in racial/ethnic minorities.^9–13 After receiving a diagnosis, racial/ethnic minorities are less likely to access dementia care services and research trials.¹⁰

Several medications are approved by the Food and Drug Administration (FDA) for AD,^14–17 including acetylcholinesterase inhibitors (AChEIs: donepezil, rivastigmine, and galantamine) for all stages of AD, and memantine, an N-methyl-D-aspartate (NMDA) receptor antagonist, for moderate to severe disease. A number of studies have examined disparities in the use of AD medications in various settings.^{9, 10, 18–26} A study using 2001–2003 data from Medicare Current Beneficiary Survey (MCBS) reported lower utilization rates of anti-dementia medication in Blacks/African Americans and Hispanics/Latinos, although disparities remained statistically significant only for Blacks/African Americans after adjusting for demographic and health characteristics.²² Another study using 2009–2010 Medicare claims data showed similar rates of initiation of AChEI or memantine treatment between Blacks/African American and White beneficiaries, but Blacks/African American and Hispanics/Latino beneficiaries discontinued medication treatment at a faster rate.²⁵

Many of these studies used large administrative data to identify dementia. However, administrative data lack cognitive assessments and have been shown to mis-identify cases when compared to reference standard such as a clinical diagnosis. These difficulties have been shown to be higher in racial/ethnic minority groups. Participants in the National Alzheimer’s Coordinating Center Uniform Data Set (NACC-UDS) have been followed since 2005 in National Institute of Aging (NIA) funded Alzheimer’s Disease Centers (ADCs) located throughout the US with careful clinical diagnosis. In this select group of individuals who have access to dementia experts, disparities in AChEI and memantine use in those with possible or probable AD at study enrollment were reported as early as 2007.²⁴ Individuals who were referred by clinicians or clinics to participate in research were more likely to take an AChEI and memantine than those who did not.²⁴ A recent study showed in participants who were diagnosed with AD dementia at baseline, AChEI and memantine treatment rates increased in both White and Blacks/African American participants at the first follow up compared to baseline, but in both years treatment rates were lower in Blacks/African American than White participants.²⁶ Although individuals are followed in NACC-UDS for long periods of time, it is unclear if such treatment disparities observed at study enrollment and at the first follow-up year remained or may have changed over time.

In this study, we aim to extend our understanding of racial/ethnic differences in anti-dementia medication treatment in several ways. We hypothesized that individuals who participated in longitudinal AD research may have better access to dementia experts and possibly better access to treatment. Unlike studies using administrative data where clinical diagnosis of dementia are lacking, we were able to focus our analysis on individuals who have been clinically diagnosed with AD dementia, for whom anti-dementia medication have been approved by the FDA. We took advantage of the longitudinal nature of the NACC-UDS data to examine uptake and persistence of medication treatment over time. We improved on study methodology by modelling uptake and persistence of AChEI and memantine treatment together.

METHODS

Data Source and Sample Derivation

Data are drawn from the National Alzheimer’s Coordinating Center Uniform Data Set (NACC-UDS) Recruitment, participant evaluation, and diagnostic criteria have been detailed elsewhere.^{27, 28} Briefly, beginning in September 2005, participants have been followed prospectively at approximately 12-month intervals from 39 past and present NIA funded ADCs located throughout the US using a standardized protocol. Informed consent was provided by all participants and their study partners.

The current study considered all participants enrolled in NACC-UDS between September 2005 (start date of the UDS) and November 2020 (N=43,342) who had at least one follow-up visit (n=29,817). To be included in the current analysis, participants had to be 65 or older at baseline with a primary etiologic diagnosis of AD, regardless of clinical diagnosis of mild cognitive impairment (MCI) or AD. Because it is unusual for AChEIs to be prescribed de novo in severe disease, and in fact in end stage dementia, medications are often discontinued, individuals with baseline Clinical Dementia Rating (CDR)=3 were excluded from the analysis sample (n=290, 3.7%). A small proportion of participants who reported being Asian (n=170, 2.2%) or of other race/ethnicities (n=72, 0.9%) also were excluded from the analysis. To examine uptake and persistence of anti-dementia medication treatment after study enrollment, analysis focused on the 3,276 participants who reported being untreated with any anti-dementia medication at baseline. Characteristics of the participants who were already treated with anti-dementia medications at baseline (excluded sample) are shown in Supplemental Table 1. Because average follow-up was 4 years, analysis included all available data for up to 5 UDS visits.

Measures

Cholinesterase Inhibitor and Memantine Use

Medication use was determined by clinician or ADC staff, based on participant/co-participant report, medical records, and/or observation, and included all medications a participant reported taking at each annual visit.²⁹ Up to 40 medications can be reported at each visit.

Outcomes for AChEI treatment were (1) any new treatment during follow-up, and (2) among those who had a new treatment, persistence of treatment during follow-up, measured by the proportion of follow-up visits reporting being treated, further categorized into three mutually exclusive and ordered groups of increasing persistence of treatment: (1) intermittent treatment (treated in <50% of follow-up visits), (2) persistent treatment (treated in ≥50% of follow-up visits but not always), and (3) always treated in all follow-up visits. Outcomes for memantine treatment were constructed similarly.

Demographic Characteristics

Our main independent variable was self-reported race/ethnicity, grouped into (1) non-Hispanic White (abbreviated below as White), (2) non-Hispanic Black/African American (abbreviated below as Black/African American), and (3) Hispanic/Latinx. Other demographic characteristics included age, sex, years of education, marital status (married/living as domestic partner/companion, widowed, vs. other), living alone (yes/no), co-participant relationship (spouse/partner/companion, child, sibling, vs. other), type of referral (through professional contact with clinician, nurse, healthcare provider, other ADC or non-ADC staff, through non-professional contact with relative or friend, vs. other unspecified contact), and total years of follow-up.

Clinical Characteristics

Clinical diagnoses were made using research criteria for AD and MCI as outlined by the UDS coding guidebook³⁰. The CDR³¹ was used to measure dementia severity. Participants’ cognitive status was measured by Mini Mental Status Exam (MMSE)³² and Montreal Cognitive Assessment (MoCA)³³, harmonized based on a crosswalk study conducted by NACC.³⁴ Participants’ function was measured using the Functional Assessment Questionnaire (FAQ)³⁵ reported from interviews with study partners. Neuropsychiatric behaviors were measured using Neuropsychiatric Inventory Questionnaire (NPI-Q).³⁶ Depressive symptoms were measured using the 15-item Geriatric Depression Scale (GDS-15).³⁷ Participants’ medical history was obtained by clinician interview and review of medical records as reported to NACC-UDS. Indicators for current use of 16 classes of medications and total number of medications a participant was taking, excluding anti-dementia medications, was used as a proxy for overall health status. Cardiovascular risk factors included history of diabetes, hypertension, and hypercholesterolemia, ascertained by self-report or clinician assessment.

Statistical Analyses

We examined baseline characteristics of individuals who were untreated with any anti-dementia medication by race/ethnic group. We compared uptake of new AChEI or memantine treatment during follow-up. Among those who reported a new treatment during follow-up, we further examined persistence of treatment by race/ethnic group.

Because treatment decisions are likely made jointly rather than independently, estimation of AChEI and memantine treatment were modelled jointly using bivariate models. Specifically, any new AChEI treatment (yes/no) and any new memantine treatment (yes/no) were modelled using bivariate probit regressions models. Because of their ordered nature, outcomes for persistence of AChEI or memantine treatment (intermittent, persistent, and always treated) were estimated using bivariate ordered probit regression models. In these models, inter-relationship between outcomes is estimated by the correlation coefficient ρ. If ρ is statistically insignificant, then outcomes can be estimated independently. If ρ is statistically significant, then outcomes should not be estimated independently. To facilitate interpretation of coefficient estimates, results were presented in terms of average marginal effects, which estimates the change in probability of the outcome falling into a specific category when an independent variable increases by one unit. All models controlled for demographic and clinical characteristics described above and accounted for clustering within ADCs. Statistical significance was set a priori at p=0.05. Analyses were performed using Stata 16.0.³⁸

RESULTS

Baseline Sample Characteristics

The 3,276 participants who were not treated with an anti-dementia medication at baseline included 2,454 (74.9%) White, 532 (16.2%) Black/African American, and 290 (8.9%) Hispanic participants (Table 1). Average age was 77.2±7.0, 45% was male, with 14.8±3.8 years of education. 62% of the participants were married or had a partner, 21.8% were widowed. 54% had a spouse/partner, 28.5% a child, and 6.6% a friend/neighbor as co-participant. 52.4% were diagnosed with MCI, 47.6% with AD. 69.4%, 22.2%, and 5.1% of the participants had CDR=0.5, 1, 2, respectively. Average MMSE was 24.7±4.6, GDS was 2.3±2.4, FAQ was 7.4±7.9, and NPI-Q was 3.2±3.9. More than 80% of follow-up visits had the same etiologic diagnosis as in baseline, suggesting anti-dementia medication remained appropriate in these visits (Supplemental Table 2).

Table 1.

Characteristics of Participants Who Reported No Anti-dementia Treatment at Baseline.

Variable	All Sample	Non-Hispanic White	Non-Hispanic Black/African American	Hispanic	p-value
N (%)	3,276 (100%)	2,454 (74.9%)	532 (16.2%)	290 (8.9%)
Age, mean (SD)	77.2	77.2	77.2	76.7	0.4831
	(7.0)	(7.0)	(7.1)	(6.9)
Male (%)	45.0	49.7	28.0	36.2	<0.001
Years of education, mean (SD)	14.8	15.6	13.5	10.0	<0.001
	(3.8)	(3.1)	(3.6)	(5.4)
Years since symptom onset, mean (SD)	3.5	3.5	3.1	3.7
	(2.8)	(2.8)	(2.8)	(2.8)	0.011
Lives alone (%)	23.4	21.0	34.8	23.1	<0.001
Co-participant relationship to participant (%)
Spouse/partner	54.0	62.8	26.9	29.3	<0.001
Adult child	28.5	23.7	40.2	47.2
Friend	6.6	5.6	12.4	4.5
Other	11.0	7.9	20.5	19.0
Marital status (%)
Married/partnered	62.0	69.9	36.3	42.4	<0.001
Widowed	21.8	18.2	34.2	29.7
Other	16.2	11.9	29.5	27.9
Years of follow-up, mean (SD)	3.9	4.0	3.6	3.7	<0.001
	(2.2)	(2.2)	(2.0)	(1.9)
Referral Source (%)					<0.001
Non-professional (self, relative, friend)	29.7	29.8	33.3	22.8
Professional (clinician, nurse, other healthcare professionals)	46.1	48.0	34.1	52.6
Other/Unknown	24.2	22.3	32.6	24.6
Diagnosis (%)					<0.001
Mild cognitive impairment, MCI	52.4	54.8	47.6	40.7
Dementia	47.6	45.2	52.4	59.3
Geriatric Depression Scale (GDS), mean (SD)	2.3	2.1	2.5	2.9	<0.001
	(2.4)	(2.3)	(2.5)	(2.8)
Functional Assessment Questionnaire (FAQ), mean (SD)	7.4	7.0	7.9	10.5	<0.001
	(7.9)	(7.5)	(8.6)	(9.3)
Neuropsychiatric Inventory Questionnaire (NPIQ), mean (SD)	3.2	2.9	3.9	4.4	<0.001
	(3.9)	(3.4)	(4.8)	(5.1)
Mini Mental Status Exam (MMSE), mean (SD)	24.7	25.3	23.2	21.4	<0.001
	(4.6)	(4.1)	(5.3)	(5.4)
Clinical Dementia Rating (CDR), (%)					<0.001
0.5	69.4	73.1	63.5	48.6
1	22.2	20.5	23.9	34.1
2	5.1	3.4	8.6	12.8
Hypertension (%)	54.3	49.3	73.5	62.1	<0.001
Hypercholesterolemia (%)	53.5	53.3	53.2	55.5	0.771
Diabetes (%)	15.1	11.1	26.1	28.6	<0.001
Depression (%)	28.9	28.9	25.2	35.9	0.006
Number of medications taken, mean (SD)	6.0	6.0	6.1	5.5	0.1220
	(4.5)	(4.5)	(5.0)	(3.9)
Class of Medication (%)
Lipid lowering medication	45.8	46.0	44.7	46.2	0.861
Anticoagulant/Antiplatelet agent	42.4	43.9	39.7	34.1	0.003
NSAID	41.5	42.5	40.5	34.8	0.037
Beta-Blocker	23.4	22.8	25.5	24.1	0.397
Antidepressant	23.0	24.2	16.3	24.8	0.001
Calcium channel blocker	16.3	13.3	30.3	15.5	<0.001
Diuretic	16.3	13.5	28.2	17.6	<0.001
Angiotensin II inhibitor	12.1	11.1	14.8	14.8	0.021
Diabetes medication	11.6	8.6	20.0	21.4	<0.001
Anxiolytic, sedative, hypnotic agent	10.4	11.3	6.9	8.6	0.007
Antiadrenergic agent	9.8	10.4	6.9	10.3	0.049
Antipsychotic	2.2	1.6	2.7	6.2	<0.001
Antiparkinson agent	2.2	2.5	1.2	1.4	0.096
Estrogen hormone therapy	2.2	2.4	1.2	2.1	0.211
Vasodilator	2.1	1.6	3.6	2.8	0.009
Estrogen + progestin therapy	0.1	0.2	0.0	0.0	0.513

Participants reported taking an average of 6.0±4.5 medications. The five most commonly reported classes of medication included lipid lowering medications (45%), anticoagulant/antiplatelet agents (42%), NSAIDs (42%), beta-blockers (23%), and antidepressants (23%). Hypertension (54.3%), hypercholesterolemia (53.5%), depression (28.9%), and diabetes (15.1%) were common. 46.1% of the participants were referred through a professional contact, 29.7% through a non-professional contact, and 24.2% from other non-specified sources. Average follow-up was 3.9±2.2 years.

Participants’ characteristics differed by race/ethnic groups for all variables except for age and hypercholesterolemia. Specifically, Black/African American and Hispanic participants were more likely to be female, had fewer years of education, and were more impaired with higher CDR and other clinical characteristics. A higher proportion of Black/African Americans lived alone and had an adult child as study partner than Hispanics or Whites. Although total number of medications reported were similar across race/ethnic groups, use of several individual classes of medications were statistically different.

New Treatment during Follow-up

During the follow-up period, 45.6% of White, 32.1% Black/African American, and 41.4% Hispanic participants reported a new AChEI treatment. 24.9% of White, 16.2% Black/African American, and 21.4% Hispanic participants reported a new memantine treatment (both p<0.001, Table 2).

Table 2.

Proportions of Participants with New and Persistence of Treatment during Follow-up

Variable	All Sample	Non-Hispanic White	Non-Hispanic Black/African American	Hispanic	p-value
N	3,276	2,454	532	290
New treatment during follow-up (%)
Any new AChEI	43.1	45.6	32.1	41.4	<0.001
Any new Memantine	23.2	24.9	16.2	21.4	<0.001
No new AChEI and no new Memantine	52.0	49.4	63.4	53.5
New treatment of AChEI only	24.9	25.8	20.5	25.2
New treatment of Memantine Only	5.0	5.0	4.5	5.2
New treatment of both AChEI and Memantine	18.2	19.9	11.7	16.2
Persistence of treatment among participants who had a new treatment during follow-up (%)
AChEI
<50% visits (intermittent treatment)	13.5	12.9	16.4	15.0	0.734
≥50% visits (persistent treatment)	28.8	29.2	26.3	29.2
Always treated	57.6	57.9	57.3	55.8
Memantine
<50% visits (intermittent treatment)	22.4	22.6	19.8	24.2	0.545
≥50% visits (persistent treatment)	37.3	38.5	33.7	30.6
Always treated	40.3	39.0	46.5	45.2

Bivariate probit regression analysis showed that after controlling for participant’s demographic and clinical characteristics, Black/African American participants were 9.9% and Hispanic participants were 7.4% less likely than White participants to report any new AChEI treatment during follow-up (p=0.002 and 0.027, respectively, Panel A, Table 3). Black/African American participants were 8.2% and Hispanic participants were 8.5% less likely than White participants to report any new memantine treatment during follow-up (both p=0.002).

Table 3.

Average Marginal Effects of Race/Ethnicity on Any New Treatment during Follow-up, Bivariate Probit Estimation Results

	Non-Hispanic Black/African American (reference= Non-Hispanic White)		Hispanic (reference= Non-Hispanic White)
	Marginal effect	p	Marginal effect	p
	(SE)		(SE)
	[95%CI)		[95%CI)
Any new AChEI	−0.099	0.002	−0.074	0.027
	(0.032)		(0.034)
	[−0.161, −0.037]		[−0.140, −0.008]
Any new Memantine	−0.082	0.002	−0.085	0.002
	(0.026)		(0.028)
	[−0.134, −0.031]		[−0.139, −0.031]
No treatment (no new AChEI, no new Memantine)	0.128	0.000	0.105	0.009
	(0.036)		(0.040)
	[0.058,0.199]		[0.026,0.183]
Memantine Only	−0.014	0.156	−0.019	0.026
	(0.010)		(0.009)
	[−0.034,0.005]		[−0.036,−0.002]
AChEI only	−0.043	0.160	−0.016	0.567
	(0.030)		(0.028)
	[−0.102,0.017]		[−0.071,0.039]
Both AChEI and Memantine	−0.071	0.000	−0.069	0.001
	(0.019)		(0.021)
	[−0.109,−0.034]		[−0.110,−0.028]

Notes:

Average marginal effects estimates the change in probability in the outcome when an independent variable increases by one unit. Here, the main independent variable is race/ethnicity (reference group=non-Hispanic white). For example, Panel A, estimate on any new AChEI treatment shows that holding all other covariate the same, and regardless of memantine treatment, the change in the probability of initiating any new AChEI treatment for a Black/American participant compared to a White participant is 9.9% or −0.099 lower, and is statistically significant at p=0.002. Panel B, estimate on no treatment (no new AChEI, no new memantine) shows that holding all other covariate the same, the change in the probability of remaining untreated for a Black/American participant compared to a White participant is 12.8% or 0.128 higher, and is statistically significant at p<0.001. In Panel B, because the outcomes of any new treatment can only fall into one of the 4 categories, average marginal effects sum to 0.

Full estimation results are in Supplemental Tables 3.

More specifically, compared to White participants, Black/African American participants on average were 12.8% more likely to remain untreated with any new anti-dementia medication, i.e., no new AChEI and no new memantine treatment, and 7.1% less likely to have initiated both new AChEI and new memantine treatments (both p<0.001, Panel B, Table 3). Hispanic participants were 10.5% more likely to remain untreated with any new anti-dementia medication, and 6.9% less likely to have initiated both new AChEI and memantine treatment (both p<0.01).

Persistence of Treatment during Follow-up

Among those who had a new anti-dementia medication treatment during follow-up, 13.6%, 28.8%, and 57.6% reported AChEI intermittently (<50% of follow-up visits), persistently (≥50% of follow-up visits), and always being treated (Table 2). For memantine treatment, these rates were 22.4%, 37.3%, and 40.3%, respectively. Unadjusted rates of persistent treatment did not differ significantly between race/ethnic groups.

Bivariate ordered probit regression analysis showed that for persistence of AChEI treatment, on average, Black/African American participants were 9.6% less likely than White participants to always be treated with AChEI during follow-up (Table 4). They were more likely than White participants to fall into never treated (by 2.7%), intermittent (by 3.1%), and persistently treated (by 3.9%) group. For memantine, Black/African American participants were on average 11.6% more likely than White participants to report never been treated during follow-up. They were less likely than White participants to fall into intermittent (by 1%), persistent (by 4%), and always treated group (by 6.6%). Patterns of differential persistence of treatment between Hispanic and White participants were similar to those between Black and White participants.

Table 4.

Average Marginal Effects of Race/Ethnicity on Persistence AChEI and Memantine Treatment during Follow-up, Bivariate Ordered Probit Estimation Results

	Non-Hispanic Black/African American (reference=Non-Hispanic White)		Hispanic (reference=Non-Hispanic White)
	Marginal effect	p	Marginal effect	p
	(SE)		(SE)
	[95% CI]		[95% CI]
Persistence of AChEI Treatment
Never treated	0.027	0.024	0.032	0.040
	(0.012)		(0.015)
	[0.004,0.050]		[0.001,0.062]
<50% visits (intermittent treatment)	0.031	0.024	0.036	0.040
	(0.014)		(0.018)
	[0.004,0.057]		[0.002,0.071]
≥50% visits (persistent treatment)	0.039	0.025	0.046	0.042
	(0.017)		(0.023)
	[0.005,0.073]		[0.002,0.090]
Always treated	−0.096	0.023	−0.114	0.039
	(0.042)		(0.055)
	[−0.179,−0.013]		[−0.221,−0.006]
Persistence of Memantine Treatment
Never treated	0.116	0.008	0.117	0.042
	(0.044)		(0.058)
	[0.030,0.201]		[0.004,0.230]
<50% visits (intermittent treatment)	−0.01	0.016	−0.01	0.056
	(0.004)		(0.005)
	[−0.018,−0.002]		[−0.020,<0.001]
≥50% visits (persistent treatment)	−0.04	0.009	−0.04	0.044
	(0.015)		(0.020)
	[−0.070,−0.010]		[−0.080,−0.001]
Always treated	−0.066	0.008	−0.067	0.042
	(0.025)		(0.033)
	[−0.115,−0.017]		[−0.132,−0.003]

Notes:

Average marginal effects estimates the change in probability in the outcome when an independent variable increases by one unit. Here, the main independent variable is race/ethnicity (reference group=non-Hispanic white). For example, under Persistence of AChEI treatment, estimate on Always treated shows that holding all other covariate the same, the change in the probability of being always treated with AChEI for a Black/American participant compared to a White participant is 9.6% or −0.096 lower, and is statistically significant at p<0.023. Because the outcomes on Persistence of AChEI treatment can only fall into one of the 4 categories (from never treated to always treated), average marginal effects sum to 0.

Full estimation results are in Supplemental Tables 4.

Full estimation results on new and persistent treatment are shown in Supplemental Tables 3 and 4. There were positive correlations between new AChEI and memantine treatment (ρ=0.54, p<0.001) and between persistent AChEI and memantine use (ρ=0.24, p<0.001), suggesting that AChEI and memantine treatment should be modeled simultaneously.

DISCUSSION

In this study, we examined racial/ethnic differences in the uptake and persistence of anti-dementia medications among previously untreated participants who have been formally diagnosed with AD dementia at study enrollment and followed in longitudinal AD research in NIH funded Alzheimer’s Disease Centers. We hypothesized that research participants who have received a formal diagnosis of AD dementia within the context of a longitudinal study that requires a study partner would have better access to dementia experts, have increased awareness of their disease, can better adhere to treatment recommendations, thereby decreasing racial/ethnic differences in the uptake and persistence of dementia medication treatment. However, we found that both Black/African American and Hispanic participants were less likely than White participants to report any new AChEI and memantine treatment during follow-up. Among those who reported a new treatment during follow-up, both Black/African American and Hispanic participants were less likely than White participants to be persistently treated. Results suggest that racial/ethnic disparities in the uptake and persistence of treatment remain even in controlled settings of longitudinal research studies.

In the current study, non-Hispanic Whites were less functionally impaired at baseline with 73% of White participants presenting with CDR of 0.5, vs. 64% of Black/African American and 49% of Hispanic participants. These observations are consistent with existing evidence showing racial/ethnically under-represented populations to be more impaired at first dementia diagnosis and that diagnosis occurs at a later stage of the disease for these groups.^{39, 40} These baseline differences in dementia severity suggest that the more symptomatic racial/ethnic minority cohorts should have higher treatment rates, instead, White participants were still more likely to be treated at any severity, magnifying disparities in treatment.

To put our results in context, two studies that were also based on the UDS are particularly of note. An early study using baseline data from 2005–2007 found Blacks were less likely to be treated with ACHEI (odds ratio, OR=0.59) and memantine than non-Hispanic Whites (OR=0.43), respectively, and Hispanics were less likely to be treated with memantine than non-Hispanic Whites (OR=0.60).²⁴ A more recent study using data from baseline and first follow-up visit between 2005–2019 reported minority groups as a whole (combining Black, Hispanic, and other minority groups as one) was less likely to be treated with AChEI (OR=0.50), and memantine (OR=0.49) than White participants.²⁶ Odds ratios on initiation of AChEI treatment in Black and Hispanic participants compared to White participants in the current study were 0.71 and 0.81, and odds ratios on initiation of memantine in Black and Hispanic participants compared to White participants were 0.65 and 0.74, respectively. It is unclear whether our results are inferentially distinct from earlier results. However, the higher estimates in the current study raise the possibility of potential benefit of longitudinal research participation and highlight the importance of examining how participation in longitudinal research may affect treatment disparities in future studies.

Participants in the current study reported taking an average of 6 medications. Of note, while total number of medications were similar across racial/ethnic groups, utilization in several classes of medications were different between groups. Models that controlled for individual classes of medications showed better fit than those controlled for total number of medications, suggesting the importance of examining individual classes of medications. Future research should further examine individual medications to help clarify appropriateness of treatment and potential drug-drug interactions; however they are beyond the scope of the current study. Our estimation models showed similar estimated coefficients on racial/ethnic groups, boosting our confidence in these results.

Our estimates that having a family member such as a spouse or child as a study partner were associated with higher likelihood of uptake and persistence of both AChEI and memantine highlight the importance of having family members in the treatment and management of disease. Interestingly, Whites were more likely to have a spouse/partner as their study partner, whereas Black/African Americans and Hispanics were more likely to have an adult child as their study partner. Black/African Americans were also more likely to live alone and to have a friend as their study partner. Whether differences in the dynamics of a spousal relationship vs. a child/friend relationship in disease management and support may explain, at least in part, racial differences in medication uptake should be examined in future studies.^{11, 41}

This study is one of the few studies that examined both AChEI and memantine treatment. The strong correlations between AChEI and memantine treatment suggest that methodologically these outcomes should be estimated together. A small proportion of participants was treated with memantine alone or treated with memantine in mild dementia. Although these treatment decisions are not indicated, results are consistent with wide off-label use of memantine in mild AD reported in earlier studies.^{18, 42}

This study has several limitations. First, data on medication use is limited to two weeks before each visit. Unlike studies that use administrative data, medication use between visits is unknown. Information on reasons for treatment or non-treatment is not reported. It is unclear whether participants may have had prior treatment before study enrollment, and if there was prior treatment, why treatment were discontinued. Second, NACC-UDS is not a representative sample of the general population. The US Census Bureau reports education level for White, Black/African American, and Hispanic older adults at 14.0, 12.7, and 10.7 years, respectively.⁴³ In comparison to the general population, White participants in our sample had 1.6 more years of schooling, and Black/African American and Hispanic participants had 0.8 and 0.7 more years of schooling. The specialized nature of NIA-supported ADCs and the cohort of participants who volunteered to participate in research may not generalize to patients seen in primary care in the community. The sample also has relatively few medical and psychiatric illnesses than the general population of the same age. Taken together, it is likely that the disparities reported in this study underestimate true disparities in the community. Third, while we controlled for many demographic and clinical characteristics in the estimation models, socioeconomic variables such as income or insurance coverage that may be sources of disparities are not available in the data and therefore cannot be accounted. For example, dually-eligible older adults struggle with access barriers such as having a usual source of care. It is unclear how strongly these characteristics may affect disparities observed in this study. Several studies of veterans receiving care from VA medical centers in which access to care and costs of medications should not be strong sources of disparities showed worse outcomes in Black veterans.^{23, 44} Some of the reasons for lower dementia treatment rates in under-represented populations include lower rates of recognition of dementia in community clinicians^{11–13, 39}, less medicalized view of dementia and higher likelihood to believe memory loss and dementia as part of normal aging⁴⁵, beliefs and attitudes about medicine and alternative treatment or practices, lack of trust in medical doctors, and socio-economic concerns⁴⁶. Systemic barriers related to racism and cultural recognition of dementia also may play a role.⁴⁷ Future studies with representative samples are needed to test these causal pathways and elucidate sources of disparities in dementia medication usage. Because of small sample sizes, we did not include Asian, American Indian or Alaska Native, Native Hawaiian or other Pacific Islanders. Experiences of these under-represented groups need to be examined in future studies.

In summary, in this large national study of well-characterized individuals with a formal diagnosis of AD dementia and known dementia severity who have been followed in longitudinal AD research in NIH funded Alzheimer’s Disease Centers, we found disparities remained in the uptake and persistent treatment of anti-dementia medications in under-represented populations. Smaller magnitude of treatment disparities estimated in this longitudinal study compared to those reported using baseline data suggest a possible benefit of longitudinal research participation that has not been examined before. Strengths of this study include large population of study, longitudinal follow-up, careful characterization of sample, and inclusion of both AChEI and memantine in the analyses. Disparities found in a well-controlled study such as ours where study participants have access to experts have implications for the larger community, where disparities are likely to be even wider. In addition to individual level factors explored in this study, future studies should explore systemic and provider level factors that may contribute to these disparities. Targeted effort to improve treatment of dementia in under-represented populations are urgently needed.

Research In Context.

Systematic review:

Individuals with dementia in racial/ethnically under-represented groups are less likely to use prescriptive anti-dementia medications even among those engaged in research. It is unclear if treatment disparities remain in those who are followed in longitudinal research.

Interpretation:

After controlling for individual’s demographic and clinical characteristics, substantial racial/ethnic treatment disparities remained in controlled settings of longitudinal research where participants have access to dementia experts, suggesting wider disparities in the larger community. Smaller magnitude of treatment disparities estimated in this longitudinal study compared to those reported when only baseline data were used suggest possible benefit of longitudinal research participation that has not been examined before.

Future directions:

In addition to individual level factors explored in this study, future studies should explore systemic and provider level factors that may contributed to these disparities. Targeted effort to improve treatment of dementia in racial/ethnic under-represented groups are urgently needed.