Specialist Availability and Drug Adherence in Older Adults with Dementia Across Regions of the United States

Abstract

Background:

Access to specialists facilitates appropriate Alzheimer’s disease and related dementia (ADRD) medication use and adherence. However, there is little information on the impact of specialists’ availability on ADRD medication adherence, especially in regions of the United States (US) where specialists are scarce, e.g., the Deep South (DS).

Objective:

To ascertain whether the availability of specialty physicians in the DS and other US regions predicts ADRD medication adherence among community-dwelling older adults

Methods:

We conducted secondary analyses of claims data for 54,194 Medicare beneficiaries with ADRD in 2013–2015. Medication adherence was measured using the proportion of days covered (PDC). Multivariable-adjusted Modified Poisson regression was used to examine associations of adherence with physicians’ availability by region.

Results:

The race/ethnicity distribution was 81.44% white, 9.17% black, 6.24% Hispanic, 2.25% Asian, and 1% other; 71.81% were female, and 42.36% were older than 85 years. Beneficiaries across regions differed in all individual and contextual characteristics except sex and comorbidities. Neurologists and psychiatrists’ availability was not significantly associated with adherence (DS = 1.00, 0.97–1.03 & non-DS = 1.01, 1.00–1.01). Race and having ≥1 specialist visits were associated with a lower risk of adherence in both regions (p < 0.0001). Advanced age, dual Medicare/Medicaid eligibility, and living in non-large metropolitan areas, were associated with adherence in the non-DS region.

Conclusion:

Among older Americans with ADRD, a context defined by specialist availability does not affect adherence, but other context characteristics related to socioeconomic status may. Research should further examine the influence of individual and contextual factors on ADRD treatment among older adults.

Keywords

Alzheimer’s disease medication adherence Medicare physician availability specialist use socioeconomic factors

1 INTRODUCTION

Alzheimer’s disease and related dementias (ADRD) affect nearly 5 million Americans [1]. In 2050, approximately 13.2 million older Americans are projected to have a diagnosis of ADRD, representing a public health priority [1]. Providing adequate pharmacological treatment is a critical element of care for people with dementia. Prior to 2021, all treatments for Alzheimer’s disease (AD) addressed only the symptoms. Their effects are not sustained after discontinuation, so a person receiving the benefits of AD medications depends upon their continuous use [2]. Although these pharmacologic therapies for AD do not reverse the disease pathology, medication adherence optimizes symptom management, delays institutionalization, and reduces health care expenditures [3–5]. While there is limited evidence of the clinical benefits in the medium term [6], clinical and cost-effectiveness data suggested that treatment continuation with cholinesterase inhibitors (ChEIs) was cost-effective considering cognitive and functional benefits and health-related quality of life of individuals diagnosed with ADRD [7, 8]. Among older adults, the group most at risk for AD, poor medication adherence was associated with worsening of neuropsychiatric symptoms, avoidable hospitalizations, and increased emergency department visits [9]. More specifically, health economic analyses for Medicare beneficiaries with AD have shown a significant reduction in health care expenditures associated with adherence to ChEIs [10]. Therefore, understanding what affects medication adherence in older adults with AD is warranted.

Medication adherence is influenced by individual, community, and health care system-level factors [11, 12]. At the individual and community levels, previous studies described associations between patterns of adherence to AD medications and characteristics such as race and geographic location [2, 13, 14]. For example, African American and Asian Medicare beneficiaries were less likely to use antidementia medication than White beneficiaries [13], and patients in the South of the United States (US) were less likely to be adherent to oral AD medications than those from the West [2]. The reasons for such geographic differences remain unclear. One factor that may play a role is the availability of ADRD specialists which differs across geographic regions [15, 16]. Having access to physicians and to a specific type of clinician (e.g., neurologist, psychiatrist) near the time of diagnosis was essential for AD medication initiation [17, 18], while a higher frequency of physician visits was associated with AD medication initiation, lower AD medication discontinuation rate, and increased ChEI therapy persistence [18, 19]. The specialty of the physician is also important for AD treatment: visits to psychiatrists and neurologists were related to a higher likelihood of receiving this treatment than visits to geriatricians [18]. The availability of these specialists is likely to also impact adherence to AD medications. According to the Behavioral Model of Access to Care, higher availability could be considered a contextual enabling factor influencing patients’ behaviors through proximity to providers and care facilities [20]. A context with an appropriate supply of specialists may facilitate access, utilization, and physician involvement in patient care contributing to a context of overall patient engagement and positive beliefs about the benefits of medication adherence [21, 22]. Prior research exploring geographic variations in neurologist density, neurologic conditions, and involvement in neurologic care reported that a higher neurologist supply increases access to neurologist care for conditions such as dementia and stroke [23], but it is not known if such higher supply may affect health behaviors, such as medication adherence for people with dementia. This is particularly important for some regions of the US, specifically for states of the Deep South such as Alabama, Mississippi, and South Carolina, which are characterized by lower socioeconomic status, rising cases of ADRD, higher presence of underrepresented populations, and identified as “dementia neurology deserts” based on current and projected shortages of neurologists. The neurology workforce is currently expected to remain inadequate to meet the health needs of people with AD and other related dementias in 2025 [15].

The objective of this study was to address the knowledge gap regarding the role of specialists in maintaining appropriate dementia pharmacotherapy by ascertaining whether the availability of neurologists and psychiatrists was associated with AD medication adherence in community-dwelling older adults with dementia. We evaluated this question in the context of individual and contextual factors that influence medication adherence. Moreover, we analyzed the relationship of individual and community factors and medication adherence by geographic region, namely for the Deep South versus the rest of the US (non-Deep South). Our results will provide knowledge to elucidate factors to intervene to optimize adherence to AD treatment and ultimately improve outcomes of older adults with AD.

2 METHODS

This study was a secondary analysis of 2013–2015 administrative claims data for a 5% random sample of Medicare beneficiaries. The UAB Institutional Review Board deemed this exempt study research. Data included Medicare Part A (hospital), B (outpatient) claims for physician visits, outpatient departments, inpatient and skilled nursing facility admissions, and Medicare Part D Drug Event file for filled prescriptions. Beneficiaries’ county of residence was linked to county-level data on demographic community characteristics and healthcare resource availability from the Area Health Resources File (AHRF) [24].

2.1 Study design

We identified ADRD cases in a three-year period (2013–2015), a time span that has been found optimal to identify ADRD cases using administrative claims [25, 26]. ADRD cases were beneficiaries who were 66 years old or older in 2013 and had claims with International Classification of Disease (ICD) diagnosis codes for Alzheimer’s disease (331.0, G30.0, G30.1, G30.8, G30.9), frontotemporal dementia (331.11, 331.19, G31.01, G31.09), vascular dementia (290.40–290.43, F01.50, F01.51), and other forms of dementia (Cerebral Degeneration, Presenile/Senile Dementia, Amnestic Disorder In Conditions Classified Elsewhere, Dementia Not Classified Elsewhere, Dementia/Unspecified, Other Persistent Mental Disorders Due To Conditions Classified Elsewhere, Senility With No Mention Of Psychosis; 331.2, 331.7, 290.0, 290.20–290.21, 290.3, 290.10–290.13, 294.0, 294.10, 294.11, 294.20, 294.21, 294.80, 797, G31.1, G94, G13.2, G13.8, G31.2, G91.4, F02.80, F02.81, F03, F03.90, F03.91, F04, F05, F06.8, F06.1,R41.81, R54). This algorithm has been found to have high sensitivity and specificity (≥0.85) [27]. For each case, we defined follow-up as the period after the index event (first claim that contributed to the case identification) and until institutionalization in a skilled nursing facility or hospice for 30 days or more, or death, or one year, whichever came first, with a minimum of 7 days.

We then restricted the sample to Medicare beneficiaries continuously enrolled in fee for service Medicare Part A (hospital), B (outpatient), and D (prescription drugs), who had at least two prescription refills in the Part D event file (at least 7-day supply) of an ADRD medication in the follow-up period. ADRD medications were donepezil, galantamine, rivastigmine, and memantine. To examine existing comorbid conditions, we further restricted the sample to beneficiaries with continuously enrolled in fee for service Medicare Part A (hospital), B (outpatient) in a look-back period of 365 days before the index event (N = 54,194)

2.2 Measures

2.2.1 Outcome

The primary outcome of interest was a binary variable equal to 1 if the beneficiary was adherent to ADRD medications and 0 otherwise. To obtain this variable, ADRD medication adherence was measured using the proportion of days covered (PDC). PDC was calculated as the ratio of total days with AD drugs at hand over the period from the start of the first prescription (or from the index date if it started before then) to the end of the last prescription refill (or the end of follow-up if it came first). The number of total days with at least one AD drug at hand was the sum of days of supply for all refills (including gaps between refills of <60 days). We used a conservative approach and did not account for higher adherence during hospital stays. The main outcome of being adherent was equal to 1 if the PDC was ≥0.80 or 80% days covered.

2.2.2 Independent variable

The main independent variable was the availability of neurologists and psychiatrists at the county level. The supply of neurologists and psychiatrists (number of neurologists and psychiatrists’ office-based practices as identified in AHRF) was calculated as per 1,000 persons in the county.

2.3 Conceptual framework

The Andersen Behavioral Model of Health Services Use guided predisposing, enabling, and need factors selection at the individual and context levels [20]. Individual-level predisposing factors included demographic characteristics such as sex, race, ethnicity, and age. For race and ethnicity, we used the reclassification based on the race code used by the Social Security Administration and an algorithm from the Research Triangle Institute to better identify Hispanic and Asian beneficiaries [28]. The percentage of other race population in the regression analysis included Asian, Hispanics and other races due to small sizes at the county level by region. Individual-level enabling factors included an indicator of dual eligibility for Medicare/Medicaid and the Area Deprivation Index at the 9-level ZIP code (a proxy for personal socioeconomic status) [29]. Individual-level medical need factors included clinical characteristics such as comorbidities and having at least one specialist visit (specialty code in physicians’ claims equal to 13 for neurology, 26 and 27 for psychiatry and general psychiatry). Comorbid conditions were identified in the look-back period using the Charlson comorbidities algorithm [30]. The use of specialists was included as the individual perception of health can be influenced by access to programs or healthcare providers [31]. Context-level predisposing factors included education (2012–2016 mean proportion of county population 25 and older with less than high school diploma determined by American Community Survey). Context-level enabling factors together with the main independent variable, county-level availability of specialists, included a county indicator for persistent poverty (defined as 20 percent or more of county residents were poor, measured by the 1980, 1990, 2000 censuses, and the American Community Survey 5-year average data for 2007-11) [24], and residence in large or small metro, micropolitan or rural areas (based on the Urban Influence Code) [32]. Then, we stratified by geographic region (Deep South versus non-Deep South). The Deep South region was defined as the states of Alabama, Georgia, Louisiana, Mississippi, and South Carolina, and non-Deep South as the rest of the US states.

2.4 Analysis

We first conducted descriptive statistics of predisposing, enabling, and need factors and the outcome of interest. We then ascertained differences across geographic regions (Deep South versus non-Deep South). Chi-square for categorical variables and T-test for continuous variables were used to identify statistically significant differences.

Crude and multivariable-adjusted Modified Poisson regression were used to examine the associations of being adherent with physicians’ availability adjusting for predisposing, enabling, and need factors by region. Analyses were performed using SAS, version 9.4. (SAS Institute, Inc., Cary, NC). Results were considered statistically significant and practically meaningful when the accompanying statistical test yields a two-tailed probability of 0.0001.

3 RESULTS

3.1 Sample characteristics

Among the 54,194 Medicare beneficiaries who were users of antidementia medications, 81.44% were White, 9.17% Black, 6.24% Hispanic, 2.25% Asian, and 1% of other race and ethnicity; 71.81% were female, and 42.36% were older than 85 years (Table 1). Concerning individual-level medical need factors, 39.30% of the beneficiaries had three or more comorbidities, and 51.98% had at least one specialist visit. With respect to individual level enabling factors, 32.76% were dually eligible for Medicaid, and the mean ADI was 48.97 (SD = 27.96). When ascertaining context-level predisposing factors, 11.88% lived in areas with lower education. Regarding context-level enabling factors, 46.30% of the beneficiaries lived in large metropolitan areas, and 6.86% lived in areas with high persistent poverty. The mean availability of neurologists per 1K was 0.23 (SD = 0.22), and psychiatrists per 1K was 0.52 (SD = 0.56).

Table 1
Characteristics of Medicare beneficiaries* with Alzheimer’s disease and associated dementia in 2013–2015 (N = 54,194) by geography (Deep South [AL, GA, LA, MS, SC] versus not)

Factors ALL Non-Deep South¹ Deep South²

N % N % N % p

54,194 100 48,338 89.20 5,856 10.80

Adherence to AD medication Yes 43,741 80.71 39,138 80.97 4,603 78.60 <0.0001

No 10,453 19.29 9,200 19.03 1,253 21.40

Length of follow up

Mean 336.24 336.34 335.49 0.4886

(SD) (78.11) (77.94) (79.50)

Predisposing

Gender Male 15,276 28.19 13,685 28.31 1,591 27.17 0.0665

Female 38,918 71.81 34,653 71.69 4,265 72.83

Race/Ethnicity White 44,138 81.44 39,558 81.84 4,580 78.21 <0.0001

Black 4,967 9.17 3,791 7.84 1,176 20.08

Hispanic 3,380 6.24 3,325 6.87 55 0.94

Asian 1,223 2.25 1,203 2.49 20 0.34

Other 486 0.90 461 0.95 25 0.43

Age 66–74 8,010 14.78 6,963 14.4 1,047 17.88 <0.0001

75–84 23,229 42.86 20,555 42.52 2,674 45.66

85+ 22,955 42.36 20,820 43.07 2,135 36.46

Need

# Comorbidities CCI³ 0 8,808 16.25 7,787 16.11 1,021 17.43 0.0064

1-2 24,092 44.45 21,465 44.41 2,627 44.86

3+ 21,294 39.30 19,086 39.48 2,208 37.71

Visits to specialists

≥1 visit with a specialist No 26,024 48.02 22,987 47.55 3,037 51.86 <0.0001

Yes 28,170 51.98 25,351 52.45 2,819 48.14

≥1 visit with a neurologist No 35,484 65.48 31,525 65.22 3,959 67.60 0.0003

Yes 18,710 34.52 16,813 34.78 1,897 32.40

≥1 visit with a psychiatrist No 42,840 79.05 38,075 78.77 4,765 81.37 <0.0001

Yes 11,354 20.95 10,263 21.18 1,091 18.63

Enabling

M/M dually eligible⁴ No 36,440 67.24 32,939 68.14 3,501 59.79 <0.0001

Yes 17,754 32.76 15,399 31.86 2,355 40.22

ADI ⁵ Mean (SD) 48.97 47.429 61.661 <0.0001

(27.96) 27.95 24.62

Context-Level Predisposing

Low Education Code 2015 0 47,557 87.75 42,918 88.78 4,639 79.22 <0.0001

1 6,437 11.88 5,220 10.79 1,217 20.78

Context-Level Enabling

Neurologists per 1K Mean 0.23 0.24 0.19 <0.0001

(SD) (0.22) (0.22) (0.20)

Psychiatrists per 1K Mean 0.52 0.54 0.36 <0.0001

(SD) (0.56) 0.57 0.41

Persistent Poverty-year 0 50,280 92.78 45,653 94.45 4,627 79.01 <0.0001

1 3,714 6.86 2,485 5.14 1,229 20.99

Urban/rural residence Large Metro 25,094 46.30 23,886 49.41 1,256 21.45 <0.0001

Small Metro 17,145 31.64 14,509 30.02 2,670 45.59

Micropolitan 9,748 17.99 8,092 16.74 1,656 28.28

Rural 2,007 3.70 1,730 3.57 277 4.73

Missing 200 0.37 200 0.42

Factors	ALL	Non-Deep South¹	Deep South²
	54,194	100	48,338	89.20	5,856	10.80
Adherence to AD medication	Yes	43,741	80.71	39,138	80.97	4,603	78.60	<0.0001
	No	10,453	19.29	9,200	19.03	1,253	21.40
Length of follow up
	Mean	336.24		336.34		335.49		0.4886
	(SD)	(78.11)		(77.94)		(79.50)
Predisposing
Gender	Male	15,276	28.19	13,685	28.31	1,591	27.17	0.0665
	Female	38,918	71.81	34,653	71.69	4,265	72.83
Race/Ethnicity	White	44,138	81.44	39,558	81.84	4,580	78.21	<0.0001
	Black	4,967	9.17	3,791	7.84	1,176	20.08
	Hispanic	3,380	6.24	3,325	6.87	55	0.94
	Asian	1,223	2.25	1,203	2.49	20	0.34
	Other	486	0.90	461	0.95	25	0.43
Age	66–74	8,010	14.78	6,963	14.4	1,047	17.88	<0.0001
	75–84	23,229	42.86	20,555	42.52	2,674	45.66
	85+	22,955	42.36	20,820	43.07	2,135	36.46
Need
# Comorbidities CCI³	0	8,808	16.25	7,787	16.11	1,021	17.43	0.0064
	1-2	24,092	44.45	21,465	44.41	2,627	44.86
	3+	21,294	39.30	19,086	39.48	2,208	37.71
Visits to specialists
≥1 visit with a specialist	No	26,024	48.02	22,987	47.55	3,037	51.86	<0.0001
	Yes	28,170	51.98	25,351	52.45	2,819	48.14
≥1 visit with a neurologist	No	35,484	65.48	31,525	65.22	3,959	67.60	0.0003
	Yes	18,710	34.52	16,813	34.78	1,897	32.40
≥1 visit with a psychiatrist	No	42,840	79.05	38,075	78.77	4,765	81.37	<0.0001
	Yes	11,354	20.95	10,263	21.18	1,091	18.63
Enabling
M/M dually eligible⁴	No	36,440	67.24	32,939	68.14	3,501	59.79	<0.0001
	Yes	17,754	32.76	15,399	31.86	2,355	40.22
ADI ⁵	Mean (SD)	48.97		47.429		61.661		<0.0001
		(27.96)		27.95		24.62
Context-Level Predisposing
Low Education Code 2015	0	47,557	87.75	42,918	88.78	4,639	79.22	<0.0001
	1	6,437	11.88	5,220	10.79	1,217	20.78
Context-Level Enabling
Neurologists per 1K	Mean	0.23		0.24		0.19		<0.0001
	(SD)	(0.22)		(0.22)		(0.20)
Psychiatrists per 1K	Mean	0.52		0.54		0.36		<0.0001
	(SD)	(0.56)		0.57		0.41
Persistent Poverty-year	0	50,280	92.78	45,653	94.45	4,627	79.01	<0.0001
	1	3,714	6.86	2,485	5.14	1,229	20.99
Urban/rural residence	Large Metro	25,094	46.30	23,886	49.41	1,256	21.45	<0.0001
Small Metro	17,145	31.64	14,509	30.02	2,670	45.59
Micropolitan	9,748	17.99	8,092	16.74	1,656	28.28
Rural	2,007	3.70	1,730	3.57	277	4.73
Missing	200	0.37	200	0.42

^*Medicare beneficiaries with Part A, B, D, and no HMO in follow-up period, i.e., one year post index event or until death or institutionalization, whichever comes first, and A, B, no HMO in one year before index event. ¹Non-Deep South, US states not in the Deep South; ²Deep South, States of Alabama, Georgia, Louisiana, Mississippi, and South Carolina; ³CCI, Charlson Comorbidity Index; ⁴MM, Medicare/Medicaid; ⁵ADI, Area Deprivation Index.

Bivariate analyses revealed significant differences between beneficiaries in the Deep South and those in the rest of the country (p < 0.0001) in all characteristics except gender and number of comorbidities (Table 1). When examining predisposing factors, compared to beneficiaries in the non-Deep South region, beneficiaries from the Deep South were less likely to be 85 or older (36.46% versus 43.07%), more likely to be black (20.08% versus 7.84%), and to live in lower education areas (20.78% versus 10.79%). Regarding need factors, beneficiaries in the Deep South were less likely to have at least one specialist visit during the follow-up period (48.14% versus 52.45%). About enabling factors, beneficiaries in the Deep South were more likely to be Medicare/Medicaid dually eligible (40.22% versus 31.86%), live in areas with higher ADI mean (mean scores 61.66 versus 47.43), live in counties with high persistent poverty (21% versus 5.14%) live in small metropolitan areas (45.59% versus 30.02%) and with fewer neurologists (means 0.19 versus 0.24 per 1k) and psychiatrists (means 0.36 versus 0.54 per 1k) than those in the non-Deep South region.

3.2 Adherence analysis

The period from the first to the last prescription refill covered on average 86% of the follow-up period (±23%). Over this period, 80.71% of beneficiaries were considered adherent to AD medication (PDC ≥0.80). There were no differences in adherence by quintiles of providers’ availability (range 80.19% –81.15, p = 0.34). Beneficiaries in the Deep South (78.60%) were less likely to be adherent than beneficiaries in the non-Deep South (80.97%) (p < 0.0001). Length of follow up was 336.24 (±78.11). There were no differences in the length of follow up across regions (p = 0.4886).

Multivariable-adjusted risk ratios (RR) and 95% CIs are presented in Table 2. In analyses adjusted for individual and context level factors, neurologists and psychiatrists’ availability was not significantly associated with adherence (non-Deep South = 1.01, 1.00–1.01 versus Deep South = 1.00, 097–1.03).

Table 2
Regression analysis sociodemographic, clinical and provider factors and being adherent to ADRD medication by region

Characteristic Deep South¹n = 5,856 Non-Deep South²n = 48,338

Relative Risk 95% Confidence Interval Relative Risk 95% Confidence Interval

Neurologists and psychiatrists per 1k 1.00 (0.97–1.03) 1.01 (1.00–1.01)

Individual-level Predisposing Factors Sex Male (ref) 1.00 1.00

Female 0.99 (0.96–1.02) 0.99 (0.99–1.01)

Age 66–74 (ref) 1.00 1.00

75–84 0.99 (0.96–1.04) 1.01 (0.99–1.02)

85+ 1.03 (0.99–1.07) 1.03 (1.02–1.05)^

Race/Ethnicity White (ref) 1.00 1.00

Black 0.90 (0.86–0.94)^ 0.90 (0.85–0.92)^

Other³ 0.85 (0.74–0.97) 0.92 (0.91–0.94)^

Individual-level Need Factors CCI⁴ 0 (ref) 1.00 1.00

1-2 1.01 (0.97–1.05) 1.00 (0.99–1.02)

>3 0.98 (0.94–1.02) 0.99 (0.98–1.00)

Visits to specialists No 1.00 1.00

Yes 0.94 (0.91–0.97)^ 0.97 (0.96–0.98)^

Individual-level Enabling Factors MM⁵ Dually Eligible Yes 1.04 (1.01–1.07) 1.02 (1.01–1.03)^

No 1.00 1.00

ADI⁶ Bottom quintile 1.00 1.00

Second 1.01 (0.95–1.07) 0.99 (0.98–1.01)

Third 0.97 (0.91–1.03) 0.98 (0.97–0.99)

Fourth 0.97 (0.91–1.03) 0.97 (0.95–0.98)^

Top quintile (most deprived) 0.93 (0.88–0.99) 0.96 (0.94–0.98)^

Context-level Predisposing factors Low Education -2015 0 1.00 1.00

1 0.99 (0.95–1.03) 0.96 (0.95–0.98)^

Context-Level Enabling Factors Persistent Poverty 2014 0 1.00 1.00

1 1.02 (0.98–1.06) 0.97 (0.94–0.99)

Area of Residence Large metropolitan 1.00 1.00

Small Metropolitan 1.01 (0.93–1.09) 1.06 (1.04–1.09)^

Micropolitan 1.02 (0.99–1.06) 1.02 (1.01–1.03)^

Non-core rural 1.03 (0.97–1.08) 1.03 (1.02–1.05)^**

Characteristic	Deep South¹n = 5,856	Non-Deep South²n = 48,338
	Neurologists and psychiatrists per 1k	1.00	(0.97–1.03)	1.01	(1.00–1.01)
Individual-level Predisposing Factors	Sex	Male (ref)	1.00		1.00
		Female	0.99	(0.96–1.02)	0.99	(0.99–1.01)
	Age	66–74 (ref)	1.00		1.00
		75–84	0.99	(0.96–1.04)	1.01	(0.99–1.02)
		85+	1.03	(0.99–1.07)	1.03	(1.02–1.05)^**
	Race/Ethnicity	White (ref)	1.00		1.00
		Black	0.90	(0.86–0.94)^**	0.90	(0.85–0.92)^**
		Other³	0.85	(0.74–0.97)	0.92	(0.91–0.94)^**
Individual-level Need Factors	CCI⁴	0 (ref)	1.00		1.00
		1-2	1.01	(0.97–1.05)	1.00	(0.99–1.02)
		>3	0.98	(0.94–1.02)	0.99	(0.98–1.00)
	Visits to specialists	No	1.00		1.00
		Yes	0.94	(0.91–0.97)^**	0.97	(0.96–0.98)^**
Individual-level Enabling Factors	MM⁵ Dually Eligible	Yes	1.04	(1.01–1.07)	1.02	(1.01–1.03)^**
		No	1.00		1.00
	ADI⁶	Bottom quintile	1.00		1.00
		Second	1.01	(0.95–1.07)	0.99	(0.98–1.01)
		Third	0.97	(0.91–1.03)	0.98	(0.97–0.99)
		Fourth	0.97	(0.91–1.03)	0.97	(0.95–0.98)^**
		Top quintile (most deprived)	0.93	(0.88–0.99)	0.96	(0.94–0.98)^**
Context-level Predisposing factors	Low Education -2015	0	1.00		1.00
		1	0.99	(0.95–1.03)	0.96	(0.95–0.98)^**
Context-Level Enabling Factors	Persistent Poverty 2014	0	1.00		1.00
		1	1.02	(0.98–1.06)	0.97	(0.94–0.99)
	Area of Residence	Large metropolitan	1.00		1.00
		Small Metropolitan	1.01	(0.93–1.09)	1.06	(1.04–1.09)^**
		Micropolitan	1.02	(0.99–1.06)	1.02	(1.01–1.03)^**
		Non-core rural	1.03	(0.97–1.08)	1.03	(1.02–1.05)^**

^**p < 0.0001. ¹Deep South, States of Alabama, Georgia, Louisiana, Mississippi, and South Carolina; ²Non-Deep South, US states not in the Deep South; ³Other, Asian, Hispanic, and other races; ⁴CCI, Charlson Comorbidity Index; ⁵MM, Medicare/Medicaid; ⁶ADI, Area Deprivation Index.

In the Deep South region, only belonging to minority racial and ethnic groups (Blacks RR = 0.90, 0.86–0.94) and having at least one specialist visit (RR 0.94, 0.91–0.97) were associated with a lower risk of being adherent (p < 0.0001) (Table 2).

In the non-Deep South region individual-level predisposing factors as age and race and ethnicity were significantly associated with adherence (Table 2). Beneficiaries aged 85 and older were more likely to be adherent than those 66–74 (RR = 1.03, 1.02–1.05). Differences across race and ethnic groups were statistically significant (Blacks RR = 090, (0.85–0.92), other race RR = 0.92, 0.91–0.94) compared to Whites, p < 0.0001). Individual-level need factors measured as having at least one specialist visit was associated with a lower probability of being adherent (RR = 0.97, 0.96–0.98). As for individual level enabling factors, Medicare/Medicaid dually eligible beneficiaries were more likely to be adherent than those not eligible (RR = 1.02, 1.01–1.03). Living in the most deprived areas was associated with lower medication adherence. (RR = 0.96, 0.94–0.98).

Regarding context-level predisposing and enabling factors, beneficiaries living in counties with low education had lower medication adherence (RR = 0.96, 0.95–0.98), while living in non-large metropolitan areas (Small Metropolitan, RR = 1.06, 1.04–1.09; micropolitan, RR = 1.02, 1.01–1.03; non-core rural, RR = 1.03, 1.02–1.05) was associated with a higher probability of being adherent (p < 0.0001).

4 DISCUSSION

In this sample of Medicare beneficiaries with ADRD, most (8 out of 10) were considered adherent to ADRD medication. Our results using PDC somewhat differed from previous reports that found ADRD drug medication use in about 60% of Commercial and Medicare Advantage Beneficiaries with ADRD [2]. These differences in adherence may be explained by the use of different methods to measure treatment adherence (e.g., Proportion of days covered (PDC) versus Medication possession ratio (MPR)) and different samples (Medicare fee for services beneficiaries versus Managed healthcare plan members). However, the availability of neurologists and psychiatrists was not a significant predictor of adherence to AD drugs. This finding was consistent across the Deep South, where the availability of these specialists is traditionally low, and in the rest of the country. After adjusting for individual and contextual-level factors, Black beneficiaries had lower medication adherence compared to White beneficiaries, regardless of the region of the country beneficiaries were from. Similarly, regardless of region, we also found that having at least one specialist visit was associated with poor medication adherence. Further, community-level factors in the non-Deep south region like living in the most deprived neighborhoods and areas with low education were associated with lower medication adherence.

Substantial geographic variations in the availability of neurologists have been described [23, 34]. Our study found that the availability of neurologists and psychiatrists was higher in the non-Deep South region than in the Deep South region. These findings are consistent with earlier data reporting that regions in the Deep South had a low average density of neurologists practicing in clinical settings [23]. It is unclear how the variation in specialists’ availability affects the care of older adults with dementia. Earlier research found a positive association between neurologist per square mile and the probability of seeing a neurologist for neurological conditions such as epilepsy, multiple sclerosis, and Parkinson’s disease, but not for dementia [34]. One reason may be because most of the ambulatory visits for people with dementia are done by primary care providers [34]. This may explain why the availability of specialists was not associated with adherence in older adults in this Medicare beneficiaries’ sample. Interestingly, we found that individual need factors such as having at least one specialist visit was associated with lower medication adherence. This finding is somewhat in contrast to the study conducted by Suh et al. that found a lower risk of discontinuation was associated with provider involvement measured as frequent visits to physicians’ offices for AD-related reasons [35]. Our findings may be explained by the fact that specialists may be seeing more severe cases with no obvious response to therapy or adverse effects [36] or that specialists are more likely to diagnose non-ADRD conditions as the source of impairment and discontinue medications that are not indicated. In our analysis, we did not have information regarding the dementia stage and thus could not adjust for this variable: therefore, the reason for the association between specialist visits and adherence remains unclear. Future studies should seek to better understand factors related to affordability and social and cultural drivers of acceptability of medical care as potential contributors to adherence differences to help to develop interventions.

Lower AD medication adherence was associated with individual-level factors as race and ethnicity. These results are consistent with emerging research focused on treatment disparities in AD for underrepresented populations [14, 37]. Data suggest that Black, Non- Hispanic other race, and Hispanic community-dwelling Medicare beneficiaries with a reported diagnosis of dementia had fewer office visits than Non-Hispanic White community-dwelling Medicare beneficiaries which may result in fewer prescriptions [18]. Among underrepresented populations, African American patients and people receiving low-income subsidies were less likely to receive a prescription and be treated for dementia [17]. Also, Medicare beneficiaries who belong to underrepresented populations were less likely to be treated with ChEIs than white beneficiaries [38]. In line with this, our previous analysis found that across racial and ethnic groups, Black Medicare beneficiaries were the least likely to have at least one ADRD prescription [39]. These disparities are present not only in medication initiation but also in medication adherence. Similar to our study, others have found that racial differences in adherence persisted among white and non-white individuals with ADRD after controlling for individual (i.e., age, gender, income, health status) and community-level factors (e.g., geographic region, urban residence) [13, 18]. African American patients were less likely to use and be adherent to dementia medication than non-Hispanic white patients [14, 17, 40]. Adding to the existing literature, in our study, racial and ethnic disparities in medication adherence remained when adjusting for contextual factors such as neighborhood deprivation and the community’s level of education. However, racial differences were not fully explained by the enabling, predisposing, and need factors included in the analyses. Further study of health beliefs and literacy within racial and ethnic groups may be worthwhile to explain factors that may impact AD treatment among underrepresented groups.

In this analysis, we found contextual factors to be associated with adherence, at least in the non-Deep South, specifically, living in non-large metropolitan areas, neighborhood deprivation and community level of education. Beneficiaries living in small metropolitan, micropolitan, and non-core rural areas were more likely to be adherent. Area of residence has been linked with discontinuation rates. Specifically, previous research found that beneficiaries living in a small metropolitan county had slower discontinuation rates [33]. In addition, beneficiaries from high deprivation and low education areas were less likely to adhere to medications. However, in our previous work, similar factors did not affect whether beneficiaries had at least one ADRD drug prescription [39]. Further studies should explore whether drug affordability and health literacy are the underlying mechanisms of these findings, and to understand what impacts initiation and appropriate use of ADRD drug medications. Furthermore, examining the interaction of personal and context level factors may be worthwhile, particularly to explain how socioeconomic status at the individual and context level may impact AD medication adherence.

4.1 Limitations

This study has some limitations. We included Medicare beneficiaries continuously enrolled in fee for service Medicare Part A, B, and D, based on having an ADRD diagnosis and dementia medications prescriptions. Results are not generalizable to beneficiaries on managed care or with fee-for-service Medicare but without Part D coverage or individuals without a diagnosis or mild stage (i.e., MCI) users. Evaluation of adherence is based on drug prescription and not actual intake by the beneficiaries. We did account for hospital stays when beneficiaries may be taken off medications by providers, and, thus, the PDC may be underestimated. The disease severity or caregiver involvement cannot be established in these claims data: These factors have been previously related to medication adherence [41]. In our study, the geographic level used was constrained by the data available. For instance, the availability of providers is only available from AHRF at the county level. Furthermore, availability is only one component of access to care—even if available, providers may not be easily accessible in the county. Lastly, ADI scores were obtained at the ZIP code level, a level that may be too large to capture the true person’s neighborhood disadvantage [42]. However, we found some associations between ADI and medication adherence.

4.2 Conclusions

Among older Americans with ADRD, a context defined by the availability of specialists does not affect adherence, but other context characteristics related to socioeconomic status may. Our results add to the growing literature highlighting the role of community-level social determinants of health, such as education and neighborhood deprivation in ADRD medication adherence. The impact of individual and contextual factors on aspects of AD treatment beyond medication adherence among people with ADRD needs further investigation. Future research with more diverse samples may help elucidate differences in AD pharmacological treatment and inform public health and policy efforts to intervene in social and contextual factors, considering that AD disproportionately affects underrepresented populations and socially at-risk people.

Footnotes

ACKNOWLEDGMENTS

The authors have no acknowledgments to report.

FUNDING

Research reported in this publication was supported by National Institutes of Health through a supplement from the National Institute of Aging to the Obesity Health Disparities Research Center and the National Institute of Aging through the Alzheimer Disease Center at the University of Alabama at Birmingham under the awards numbers U54MD000502-17S2 and P20AG068024.

CONFLICT OF INTEREST

The authors have no conflict of interest to report.

DATA AVAILABILITY

The data that support the findings of this study were obtained from The Centers for Medicare and Medicaid Services (CMS) and were accessed by the principal investigator under a data use agreement (DUA) with CMS, which does not allow data sharing. Thus, the data are not available from investigators, but can be obtained under a new DUA with CMS.

References

Office of the Assistant Secretary for Planning and Evaluation. (2021) What is Alzheimer’s Disease and Related Dementias | ASPE (hhs.gov), Accessed on April 1, 2021.

Borah

, Sacco

, Zarotsky

(2010) Predictors of adherence among Alzheimer’s disease patients receiving oral therapy. Curr Med Res Opin 26, 1957–1965.

Arlt

S LR

, Rösler

, von Renteln-Kruse

(2008) Adherence to medication in patients with dementia: Predictors and strategies for improvement. Drugs Aging 25, 1033–1047.

Geldmacher

, Kirson

, Birnbaum

, Eapen

, Kantor

, Cummings

, Joish

(2014) Implications of early treatment among Medicaid patients with Alzheimer’s disease. Alzheimers Dement 10, 214–224.

Campbell

, Dexter

, Perkins

, Gao

, Li

, Skaar

, Frame

, Hendrie

, Callahan

, Boustani

(2013) Medication adherence and tolerability of Alzheimer’s disease medications: Study protocol for a randomized controlled trial. Trials 14, 125.

Parsons

, Lim

, Loy

, McGuinness

, Passmore

, Ward

, Hughes

(2021) Withdrawal or continuation of cholinesterase inhibitors or memantine or both, in people with dementia. Cochrane Database Syst Rev 2, CD009081.

Bond

, Rogers

, Peters

, Anderson

, Hoyle

, Miners

, Moxham

, Davis

, Thokala

, Wailoo

, Jeffreys

, Hyde

(2012) The effectiveness and cost-effectiveness of donepezil, galantamine, rivastigmine and memantine for the treatment of Alzheimer’s disease (review of Technology Appraisal No. 111): A systematic review and economic model. Health Technol Assess 16, 1–470.

Knapp

, King

, Romeo

, Adams

, Baldwin

, Ballard

, Banerjee

, Barber

, Bentham

, Brown

, Burns

, Dening

, Findlay

, Holmes

, Johnson

, Jones

, Katona

, Lindesay

, Macharouthu

, McKeith

, McShane

, O’Brien

, Phillips

PPJ

, Sheehan

, Howard

(2017) Cost-effectiveness of donepezil and memantine in moderate to severe Alzheimer’s disease (the DOMINO-AD trial). Int J Geriatr Psychiatry 32, 1205–1216.

, Han

, Choi

, Park

(2019) Effect of drug compliance on health care costs in newly-diagnosed dementia: Analysis of nationwide population-based data. J Psychiatr Res 118, 31–37.

10.

Mucha

, Shaohung

, Cuffel

, McRae

, Mark

, del

Valle M

(2008) Comparison of cholinesterase inhibitor utilization patterns and associated health care costs in Alzheimer’s disease. J Manag Care Pharm 14, 451–461.

11.

Maxwell

, Stock

, Seitz

, Herrmann

(2014) Persistence and adherence with dementia pharmacotherapy: Relevance of patient, provider, and system factors. Can J Psychiatry 59, 624–631.

12.

Brady

, Weinman

(2013) Adherence to cholinesterase inhibitors in Alzheimer’s disease: A review. Dement Geriatr Cogn Disord 35, 351–363.

13.

Barthold

, Joyce

, Ferido

, Drabo

, Marcum

, Gray

, Zissimopoulos

(2020) Pharmaceutical treatment for Alzheimer’s disease and related dementias: Utilization and disparities. J Alzheimers Dis 76, 579–589.

14.

Hernandez

, McClendon

, Zhou

, Sachs

, Lerner

(2010) Pharmacological treatment of Alzheimer’s disease: Effect of race and demographic variables. J Alzheimers Dis 19, 665–672.

15.

Rao

, Manteau-Rao

, Aggarwal

(2017) [P1-561]: Dementia neurology deserts: What are they and where are they located in the U.S.? Alzheimers Dement 13, P509.

16.

University of Michigan Behavioral Health Workforce Research Center (2018) Estimating the Distribution of the U.S. Psychiatric Subspecialist Workforce. UMSPH, Ann Arbor, MI.

17.

Koller

, Hua

, Bynum

JPW

(2016) Treatment patterns with antidementia drugs in the United States: Medicare Cohort Study. J Am Geriatr Soc 64, 1540–1548.

18.

Zuckerman

, Ryder

, Simoni-Wastila

, Shaffer

, Sato

, Zhao

, Stuart

(2008) Racial and ethnic disparities in the treatment of dementia among Medicare beneficiaries. J Gerontol B Psychol Sci Soc Sci 63, 328–333.

19.

Amuah

, Hogan

, Eliasziw

, Supina

, Beck

, Downey

, Maxwell

(2010) Persistence with cholinesterase inhibitor therapy in a populationbased cohort of patients with Alzheimer’s disease. Pharmacoepidemiol Drug Saf 19, 670–679.

20.

Andersen

(1995) Revisiting the behavioral model and access to medical care: Does it matter? J Health Soc Behav 36, 1–10.

21.

Amico

, Mugavero

, Krousel-Wood

, Bosworth

, Merlin

(2018) Advantages to using social-behavioral models of medication adherence in research and practice. J Gen Intern Med 33, 207–215.

22.

Jaglal

, Guilcher

SJT

, Bereket

, Kwan

, Munce

, Conklin

, Versnel

, Packer

, Verrier

, Marras

, Pitzul

, Riopelle

(2014) Development of a chronic care model for neurological conditions (CCM-NC). BMC Health Serv Res 14, 409.

23.

Lin

, Callaghan

, Burke

, Skolarus

, Hill

, Magliocco

, Esper

, Kerber

(2021) Geographic variation in neurologist density and neurologic care in the United States. Neurology 96, e309–e321.

24.

Area Health Resource File. https://data.hrsa.gov/topics/health-workforce/ahrf

25.

Goodman

, Lochner

, Thambisetty

, Wingo

, Posner

SF LS

(2017) Prevalence of dementia subtypes in United States Medicare fee-for-service beneficiaries, 2011-2013. Alzheimers Dement 13, 28–37.

26.

Taylor

DH Jr

, Fillenbaum

GG Ezzel ME

(2002) The accuracy of Medicare claims data in identifying Alzheimer’s disease. J Clin Epidemiol 55, 929–937.

27.

Taylor

DH Jr

, Ostbye

, Langa

, Weir

, Plassman

(2009) The accuracy of Medicare claims as an epidemiological tool: The case of dementia revisited. J Alzheimers Dis 17, 807–815.

28.

Jarrín

, Nyandege

, Grafova

, Dong

, Lin

(2020) Validity of race and ethnicity codes in Medicare administrative datacompared with gold-standard self-reported race collected duringroutine home health care visits. Med Care 58, E1–E8.

29.

Singh

(2003) Area deprivation and widening inequalities in US mortality, 1969-1998. Am J Public Health 93, 1137–1143.

30.

Sharabiani

Aylin

, Bottle

(2012) Systematic review of comorbidity indices for administrative data. Med Care 50, 1109–1118.

31.

Travers

, Hirschman

, Naylor

(2020) Adapting Andersen’s expanded behavioral model of health services use to include older adults receiving long-term services and supports. BMC Geriatr 20, 58.

32.

United States Department of Agriculture, Economic Research Service. Urban Influence Code In.

33.

Thorpe

, Fowler

, Harrigan

, Zhao

, Kang

, Hanlon

, Gellad

, Schleiden

, Thorpe

(2016) Racial and ethnic differences in initiation and discontinuation of antidementia drugs by Medicare beneficiaries. J Am Geriatr Soc 64, 1806–1814.

34.

Van der Goes

, Ney

, Garrison

(2019) Determinants of specialist physician ambulatory visits: A neurology example. J Med Econ 22, 830–839.

35.

Suh

, Thomas

, Valiyeva

, Arcona

S VL

(2005) Drug persistency of two cholinesterase inhibitors: Rivastigmine versus donepezil in elderly patients with Alzheimer’s disease. Drugs Aging 22, 695–707.

36.

Hogan

, Bailey

, Black

, Carswell

, Chertkow

, Clarke

, Cohen

, Fisk

, Forbes

, Man-Son-Hing

, Lanctôt

, Morgan

, Thorpe

(2008) Diagnosis and treatment of dementia: 5. Nonpharmacologic and pharmacologic therapy for mild to moderate dementia. CMAJ 179, 1019–1026.

37.

Gilligan

, Malone

, Warholak

, Armstrong

(2012) Racial and ethnic disparities in alzheimer’s disease pharmacotherapy exposure: An analysis across four state medicaid populations. Am J Geriatr Pharmacother 10, 303–312.

38.

Gruber-Baldini

, Stuart

, Zuckerman

, Simoni-Wastila

, Miller

(2007) Treatment of dementia in community-dwelling and institutionalized Medicare beneficiaries. J Am Geriatr Soc 55, 1508–1516.

39.

Pisu

, Martin

, Shan

, Pilonieta

, Kennedy

, Oates

, Kim

, Geldmacher

(2021) Dementia care in diverse older adults in the U.S. Deep South and the rest of the United States. J Alzheimers Dis 83, 1753–1765.

40.

Poon

, Lal

, Ford

, Braun

(2009) Racial/ethnic disparities in medication use among veterans with hypertension and dementia: A national cohort study. Ann Pharmacother 43, 185–193.

41.

Ownby

(2006) Medication adherence and cognition: Medical, personal and economic factors influence level of adherence in older adults. Geriatrics 61, 30–35.

42.

Durfey

SNM

, Kind

AJH GR

et al. (2018) Impact of risk adjustment for socioeconomic status on Medicare advantage plan quality rankings. Health Aff (Millwood) 37, 1065–1072.