Geographic Variation of Prevalence of Alzheimer’s Disease and Related Dementias in Central Appalachia

Abstract

Background:

Alzheimer’s disease and related dementias (ADRD) prevalence varies geographically in the United States.

Objective:

To assess whether the geographic variation of ADRD in Central Appalachia is explained by county-level sociodemographics or access to care.

Methods:

Centers for Medicare and Medicaid Services Public Use Files from 2015– 2018 were used to estimate county-level ADRD prevalence among all fee-for-service (FFS) beneficiaries with≥1 inpatient, skilled nursing facility, home health agency, hospital outpatient or Carrier claim with a valid ADRD ICD-9/10 code over three-years in Central Appalachia (Kentucky, North Carolina, Ohio, Tennessee, Virginia, and West Virginia). Negative binomial regression was used to estimate prevalence overall, by Appalachian/non-Appalachian designation, and by rural/urban classification. Models were then adjusted for county-level: 1) FFS demographics (age, gender, and Medicaid eligibility), comorbidities; 2) population sociodemographics (race/ethnicity, education, aging population distribution, and renter-occupied housing); and 3) diagnostic access (PCP visits, neurology visits, and imaging scans).

Results:

Across the 591 counties in the Central Appalachian region, the average prevalence of ADRD from 2015– 2018 was 11.8%. ADRD prevalence was modestly higher for Appalachian counties both overall (PR: 1.03; 95% CI: 1.02, 1.04) and after adjustment (PR: 1.02; 95% CI: 1.00, 1.03) compared to non-Appalachian counties. This difference was similar among rural and urban counties (p = 0.326) but varied by state (p = 0.004).

Conclusions:

The relative variation in ADRD prevalence in the Appalachian region was smaller than hypothesized. The case mixture of the dual eligible population, accuracy of the outcome measurement, and impact of educational attainment in this region may contribute to this observation.

Keywords

Alzheimer’s disease Appalachian region dementia rural population sociodemographic factors

INTRODUCTION

The burden of Alzheimer’s disease and related dementias (ADRD) is rapidly increasing alongside the growing population of older adults, where ADRD currently impacts an estimated 6.7 million adults aged 65 years and older in the United States (US). 1 However, this burden varies geographically,2 –5 possibly reflecting contextual influences, such as sociodemographic and environmental factors.1,6, 1,6 Specifically, known individual-level ADRD risk factors, such as chronic diseases, lifestyle factors, and education, may be influenced by contextual factors that are contributing to the disparate geographic patterning of ADRD prevalence.2,6–8 , 2,6–8 Additionally, the availability of health care resources likely influences health care access and, in turn, diagnosed prevalence of ADRD.6,9, 6,9 Adequate health care access is an ongoing concern in Appalachian regions of the US.10,11, 10,11

The prevalence of ADRD varies within US Appalachian regions and likely by rurality. In Ohio, Appalachian counties had a lower prevalence of ADRD compared to non-Appalachian counties from 2007 to 2017, and the burden varied by rural/urban designation. 2 Additionally, rural counties in Kentucky and West Virginia had a lower reported prevalence of ADRD compared to urban counties. 4 These observed differences persisted after accounting for demographics and comorbidities in both regions.2,4, 2,4 While trends in educational attainment have been improving over time in rural areas, the rural/urban education disparity still exists, where both rural and Appalachian regions are characterized by low rates of educational attainment compared to their urban and non-Appalachian counterparts.12,13, 12,13 Additionally, the prevalence of ADRD is likely underestimated in rural and/or Appalachian regions where access to adequate detection resources is limited. 6 Yet, the burden of ADRD has not yet been examined in larger geographic regions to help determine if the differences in prevalence between rural and urban regions can be attributed to disease detection access.

Individual characteristics and contextual characteristics alike can influence the development and/or the detection of ADRD. In regions with less access to diagnostic resources, it follows that there may be an underestimate of ADRD prevalence due to the lack of detection. The Appalachian region has a unique mix of higher levels of ADRD risk factors and limited diagnostic access.6,13, 6,13 Our objective was to assess ADRD prevalence in the Central Appalachian region stratifying by rural/urban status and Appalachian county designation from 2015 to 2018. We hypothesized that after accounting for individual demographics, county-level sociodemographics, and multiple measures of diagnostic access (i.e., numbers of primary care providers (PCPs), neurologists, and image scans) that ADRD prevalence would be higher in Appalachian and rural counties compared to non-Appalachian and urban counties.

MATERIALS AND METHODS

Study population and design

Using an ecologic study design, we utilized county-level summary data from the Centers for Medicare and Medicaid Services (CMS) for the period of 2015 to 2018. The Central Appalachian region is defined by the Appalachian Regional Commission (ARC) as Kentucky, North Carolina, Ohio, Tennessee, Virginia, and West Virginia and is the target population for the study.

Data source

We used existing data from multiple publicly availably population-level sources. Each of these sources represents aggregate data at the county level. The American Community Survey (ACS) is an ongoing survey that provides intercensal estimates of sample US addresses annually. 14 While annual estimates may not cover all geographic areas, the ACS pools across years (5 years) to provide estimates for smaller geographies. We used the 5-year estimates from 2015– 2018 for housing tenure (proportion renter-occupied housing), and sociodemographic measures (proportions 65 and older, 65 and older with at least a high school education, Black or African American (non-Hispanic), white (non-Hispanic), and Hispanic/Latinx).15 –19 We paired with ACS the CMS Geographic Variation Public Use Files (GVPUF). CMS summarize state- and county-level prevalence for ADRD and other chronic conditions.20,21, 20,21 These are publicly available and contain 100% of all Medicare claims for fee-for-service (FFS) enrollees and Medicare Advantage (MA) beneficiaries (including those under 65 years of age). These files are aggregated by calendar year and summarized by county Federal Information Processing System code. Estimates of county-year-level diagnostic access (counts of PCP visits, neurology visits, and imaging scans (e.g., MRI and CT scans) were calculated from 5% samples of the CMS Medicare Beneficiary Summary File and the CMS carrier base and line files during the same 5-year period (2015– 2018). 22 We also linked 2015– 2018 county-year level estimates of life expectancy at birth from the Institute for Health Metrics and Evaluation (IHME), which were derived from population and deaths data from the National Center for Health Statistics. 23 Due to the use of publicly available data, this study was deemed not to be human subject research and therefore exempt from Institutional Review Board review and approval in order for the study to be conducted and its results disseminated.

Outcome

The primary outcome of this study is the county level prevalence of ADRD. Cases of ADRD were identified through Medicare administrative claims using International Classification of Diseases (ICD) version 9 prior to October 2015, and version 10 after October 2015. FFS beneficiaries with at least one inpatient, skilled nursing facility, home health agency, hospital outpatient or Carrier claim with a valid ICD-9 code (331.0, 331.11, 331.19, 331.2, 331.7, 290.0, 290.10, 290.11, 290.12, 290.13, 290.20, 290.21, 290.3, 290.40, 290.41, 290.42, 290.43, 294.0, 294.10, 294.11, 294.20, 294.21, 294.8, 797) prior to October 2015 or ICD-10 code (F01.50, F01.51, F02.80, F02.81, F03.90, F03.91, F04, F05, F06.1, F06.8, G13.8, G30.0, G30.1, G30.8, G30.9, G31.01, G31.09, G31.1, G31.2, G94, R41.81, R54) after October 2015 were identified as a case. 24 This case definition has been previously validated; it correctly identifies 75– 80% of patients with ADRD. 25 CMS prevalence estimates were calculated taking the total number of beneficiaries with ADRD within that county and dividing by the total number of FFS beneficiaries in that county.

Primary independent variables

Primary exposures are Appalachian County and rural county designations. Counties in the Appalachian region were identified based on the Appalachian Regional Commission’s designation. 26 Urban versus rural designation was determined by the United States Department of Agriculture 2013 Economic Research Service Rural-Urban Continuum Codes (RUCCs). Rural counties were defined as counties with an urban population size of less than 20,000 (RUCC 6– 9) and urban counties as all other counties (RUCC 1– 5).

Covariates

Covariates include county-year level characteristics of the Medicare population (average age, gender, eligible for Medicaid, and other comorbid conditions (atrial fibrillation, chronic kidney disease, depression, diabetes, heart failure, hyperlipidemia, hypertension, ischemic heart disease, schizophrenia/other psychotic disorders, and stroke)), sociodemographic factors (proportions 65 and older, 65 and older with at least a high school education, Black or African American (non-Hispanic), white (non-Hispanic), Hispanic/Latinx, renter-occupied housing, below the federal poverty limit, and Parts A and B beneficiaries ever enrolled in MA, as well as average life expectancy at birth (i.e., <1 years) in years), and diagnostic access factors (counts of PCPs, neurologists, and imaging (i.e. MRI or CT) scans. Chronic conditions were identified similarly as ADRD using the appropriate ICD 9 and 10 codes from the CMS GVPUF. We defined PCP service using Healthcare Common Procedure Coding System (HCPCS) codes 99201-99205 and 99211-99215 who were rendered by “family practice”, “general practice”, and “internal medicine” provider specialty types. Neurology services were all services defined by all HCPCS codes of interest (i.e., 99201-99205 and 99211-99215) rendered under the “neurology” provider specialty type. CT and MRI services were defined by HCPCS codes: 70450, 70460, and 70470, and 70551 and 70533, respectively (for any specialty). Patient claims identified in the Carrier File were linked to the Master Beneficiary Summary File for aggregation at the county-year level.

Statistical analysis

County-level characteristics were summarized overall and by rural/urban designation. Mean (SD) proportions of the characteristics were calculated for the period of 2015– 2018 to describe demographic and comorbidity distributions of the Central Appalachian Medicare community. Generalized linear regression was used to assess the potential reasons for geographic variation in ADRD prevalence. Both Poisson regression and negative binomial regression were considered, and due to overdispersion, we applied negative binomial regression models. 27 Models were built in a pre-specified manner, with the base model (Model 1) including our geographic measures: indicators for rural counties and Appalachian counties, and a product-term for the interaction between rural and Appalachian counties. An offset term of the natural log of the eligible Medicare population was included in each model for the calculation of prevalence; Model 1 represents the crude prevalence estimates. Model 2 (adjusted geographic model) additionally adjusted for county-year-level demographics of the Medicare fee-for-service participants (age, gender, Medicaid eligibility, and comorbid conditions [atrial fibrillation, chronic kidney disease, depression, diabetes, heart failure, hyperlipidemia, hypertension, ischemic heart disease, schizophrenia/other psychotic disorders, and stroke]). Model 2 represents ADRD prevalence after adjustment for characteristics of the Medicare population. Model 3 (adjusted sociodemographic model) further adjusted for county-year-level proportions 65 and older, 65 and older with at least a high school education, Black or African American (non-Hispanic), white (non-Hispanic), Hispanic/Latinx, renter-occupied housing, and life expectancy at birth (i.e., <1 years) in years. Model 4 (adjusted diagnostic access model) further considered diagnostic access characteristics including county-year level counts of PCP visits, neurology visits, and imaging scans. As a post-hoc secondary analysis, Model 4 was further extended to determine if the association between Appalachian county designation and ADRD prevalence not only varied by rural/urban classification, but also by state (three-way interaction between Appalachian and rural county designations and state). Effect measure modification on the multiplicative scale by state was tested using the corresponding p-value of the three-way interaction. All analyses were conducted in SAS Version 9.4 (Cary, NC).

RESULTS

Crude average county-level ADRD prevalence among Medicare FFS beneficiaries ranged from 6.86 to 29.91% in the Central Appalachian region from 2015 to 2018 (Fig. 1). Overall, the county-level average age among the 65 years and older FFS population was 70.4 years (SD = 1.81) and counties had slightly fewer males than females (54% female; Table 1). Rural-Appalachian counties were slightly younger with an average age among the FFS population of 69 years compared to urban-Appalachian (70 years) and rural- and urban-non-Appalachian counties (71 years). Rural-Appalachian counties also had fewer females compared to rural-non-Appalachian and urban-Appalachian/non-Appalachian counties. However, the proportion of Medicaid eligible persons was highest in rural-Appalachian counties (31.2%), followed by urban-Appalachian (23.5%), rural-non-Appalachian (23.0%), and lowest in urban-Appalachian counties (17.5%). The proportions of older adults with diabetes, heart failure, and hypertension were highest among rural counties compared to urban counties. Educational attainment and renter-occupied housing was lowest among rural-Appalachian counties, followed by rural-non-Appalachian, urban-Appalachian, and highest among urban-non-Appalachian counties. Appalachian counties had lower proportions of Black and Hispanic older adults and a lower average life expectancy than that among non-Appalachian counties. The average number of yearly PCP visits, neurology visits, and imaging scans were highest among urban-non-Appalachian counties, followed by urban-Appalachian, rural-non-Appalachian, and rural-Appalachian counties.

Fig. 1

County-level distribution of mean ADRD prevalence from 2015 to 2018 in the Central Appalachian region of the US.

Table 1

Descriptive county-level characteristics of the Medicare fee-for-service population across counties within the Central Appalachian region overall and by Appalachian designation from 2015– 2018^*

	Overall	Rural-Appalachian^†	Rural-non-Appalachian^†	Urban-Appalachian	Urban-non-Appalachian
	(N = 591)^‡	(N = 149)^‡	(N = 118)^‡	(N = 106)^‡	(N = 218)^‡
	(n = 4,932,759)^‡	(n = 503,379)^‡	(n = 454,773)^‡	(n = 967,628)^‡	(n = 3,006,979)^‡
ADRD, Mean % (SD)	11.78 (1.76)	12.06 (1.67)	11.81 (1.75)	11.93 (1.38)	11.54 (1.95)
65 + Population, Median (Q1– Q3)	4,799 (2,666– 9628)	2,803 (1,842– 4,602)	3,545 (2,384– 5,148)	7,600 (3,961– 11,284)	9,264 (4,251– 15,973)
Age (y), Mean (SD)	70.42 (1.81)	69.18 (2.17)	70.90 (1.24)	70.15 (1.43)	71.13 (1.47)
Female, Mean % (SD)	53.73 (2.39)	51.82 (2.74)	54.34 (1.53)	54.61 (2.06)	54.78 (1.82)
Medicaid Eligible, Mean % (SD)	23.15 (9.43)	31.24 (9.51)	23.00 (6.98)	23.52 (6.78)	17.52 (7.36)
Comorbidities, Mean % (SD)
Atrial Fibrillation	9.61 (1.01)	9.52 (1.10)	9.49 (1.07)	9.87 (0.96)	9.59 (0.94)
Chronic Kidney Disease	24.19 (4.48)	24.40 (4.95)	24.48 (4.45)	24.57 (4.29)	23.70 (4.20)
Depression	16.15 (2.72)	16.89 (2.54)	14.97 (3.06)	17.55 (2.26)	15.60 (2.40)
Diabetes	29.98 (4.05)	31.27 (4.23)	30.92 (3.60)	29.81 (3.17)	28.67 (4.14)
Heart Failure	15.72 (3.07)	17.37 (3.42)	16.27 (2.75)	15.55 (2.47)	14.40 (2.61)
Hyperlipidemia	49.27 (6.84)	49.20 (7.72)	48.47 (6.77)	49.47 (6.14)	49.64 (6.54)
Hypertension	64.40 (5.23)	65.24 (5.66)	66.26 (4.50)	63.99 (4.38)	63.02 (5.26)
Ischemic Heart Disease	30.45 (5.03)	32.94 (5.51)	30.53 (4.36)	31.27 (4.48)	28.31 (4.32)
Schizophrenia/Other Disorders	2.28 (1.08)	2.58 (1.13)	2.22 (1.00)	2.44 (0.91)	2.05 (1.11)
Stroke	3.93 (0.66)	3.78 (0.70)	4.09 (0.74)	3.87 (0.61)	3.97 (0.60)
% 65+, Mean % (SD)	17.60 (3.79)	19.18 (3.51)	18.83 (3.48)	18.29 (2.89)	15.51 (3.57)
Education, Mean % (SD)^§>	75.20 (9.26)	68.02 (9.95)	73.80 (6.87)	76.21 (7.44)	80.38 (6.98)
Race/Ethnicity, Mean % (SD)
% White	84.93 (15.35)	95.06 (4.24)	78.39 (18.52)	92.50 (6.72)	77.86 (15.76)
% Black	10.43 (13.69)	2.02 (2.16)	17.76 (17.84)	4.05 (5.88)	15.30 (13.96)
% Hispanic	3.82 (3.95)	1.89 (2.08)	3.48 (3.28)	2.81 (2.45)	5.81 (4.86)
Renter-Occupied Housing, Mean % (SD)	28.68 (8.22)	25.55 (5.49)	28.02 (6.63)	28.47 (6.79)	31.29 (10.13)
Diagnostic Access, Median (Q1– Q3)
PCP Visits^#	11,120 (5,100– 23,370)	5,020 (2,660– 9,570)	7,380 (4,020– 11,650)	17,360 (8,730– 27,070)	22,960 (11,330– 40,370)
Neurology Visits^#	620 (280– 1,400)	260 (140– 540)	420 (220– 680)	1,030 (560– 1,720)	1,350 (600– 2,610)
Imaging Scans^**	1,100 (540– 2,170)	580 (260– 1,020)	780 (460– 1,160)	1,720 (930– 2,680)	2,020 (980– 3,830)
Percent Below FPL, Mean % (SD)^††	17.71 (6.27)	22.22 (6.17)	18.29 (5.42)	17.88 (4.59)	14.22 (5.31)
MA Participation Rate, Mean % (SD) ^‡‡	29.53 (10.91)	28.64 (8.27)	24.21 (7.42)	36.87 (11.33)	29.45 (11.92)
Life Expectancy, Mean (SD)^§§	76.01 (2.48)	74.48 (2.52)	75.71 (1.63)	75.64 (1.89)	77.41 (2.35)

^*Counties within Kentucky, North Carolina, Ohio, Tennessee, Virginia, and West Virginia comprise the county-level Central Appalachian region. ^†Appalachian designation was defined by Appalachian Regional Commission’s designation and rural counties were defined as counties with an urban population size of less than 20,000 (Rural-Urban Continuum Codes (RUCC) 6-9); Urban counties defined as all other counties (RUCC 1-5). ^‡‘N’ indicates number of counties; n indicates the total number of 65 + Medicare beneficiaries across counties overall and by county type in 2015. ^§Education defined as percent of the 65 + county-level population with a high school education. ^#Includes all outpatient primary care provider (PCP) and neurology visits. ^**Imaging scans include MRI, PET, and CT scans. ^††Population below 100% of the federal poverty limit (FPL). ^‡‡Percent of Part A and Part B beneficiaries who are ever enrolled in Medicare Advantage (MA). ^§§Life expectancy at birth (i.e., <1 years) in years.

Overall, small variation in ADRD prevalence between Appalachian (12.0%; 95% CI: 11.9%, 12.1%) and non-Appalachian (11.6%; 95% CI: 11.5%, 11.7%) counties was observed, where ADRD prevalence in Appalachian counties was 3% higher than ADRD prevalence in non-Appalachian counties before adjustment (95% confidence interval (CI): 1.02, 1.04; Table 2). Adjustment for demographic characteristics of the Medicare population in Model 2 attenuated this variation in ADRD prevalence overall (Prevalence ratio (PR): 1.00; 95% CI: 0.99, 1.01). However, adjustment for sociodemographic characteristics and housing tenure (PR: 1.02; 95% CI: 1.01, 1.03) and then diagnostic resources (PR: 1.02; 95% CI: 1.01, 1.03) reintroduced a marginal disparity. These associations were similar within rural and urban counties, however, the attenuation in the Appalachian/non-Appalachian disparity observed following demographic adjustment in Model 2 was only apparent in urban counties (PR: 0.99; 95% CI: 0.98, 1.00) relative to rural counties (PR: 1.02; 95% CI: 1.01, 1.04). Additionally, with each level of successive covariate adjustment, the estimated number of excess cases of ADRD per Appalachian county gets smaller. This is true on the population level as well; the excess number of expected cases overall would still see 3531 (95% CI: 1619, 5443) additional ADRD cases in Appalachian counties within Central Appalachia even after accounting for sociodemographic and access-related factors.

Table 2

Estimated ADRD prevalence and 95% confidence intervals (CIs) by Appalachian County designation, prevalence ratios (PRs) and 95% CIs, and excess number of ADRD cases in a given Appalachian county overall and by rural/urban county designation across sequential sets of covariate adjustment

	Appalachian^*		Non-Appalachian^*
	Prevalence	95% CI	Prevalence	95% CI	PR^†	95% CI	Excess^‡	95% CI
Overall
Model 1^$	0.120	(0.119, 0.121)	0.116	(0.115, 0.117)	1.03	(1.02, 1.04)	22	(14, 30)
Model 2^#	0.117	(0.116, 0.118)	0.117	(0.116, 0.118)	1.00	(0.99, 1.01)	N/A
Model 3^**	0.118	(0.117, 0.119)	0.116	(0.115, 0.116)	1.02	(1.01, 1.03)	16	(9, 23)
Model 4^††	0.118	(0.117, 0.119)	0.115	(0.115, 0.116)	1.02	(1.01, 1.03)	14	(7, 22)
Rural^‡‡
Model 1^$	0.121	(0.119, 0.122)	0.118	(0.116, 0.120)	1.02	(1.00, 1.04)	10	(3, 17)
Model 2^#	0.117	(0.115, 0.118)	0.114	(0.113, 0.115)	1.02	(1.01, 1.04)	6	(– 4, 13)
Model 3^**	0.118	(0.117, 0.119)	0.115	(0.114, 0.116)	1.03	(1.01, 1.04)	10	(1, 18)
Model 4^††	0.118	(0.116, 0.119)	0.115	(0.114, 0.116)	1.03	(1.01, 1.04)	11	(– 1, 20)
Urban^‡‡
Model 1^$	0.119	(0.118, 0.121)	0.116	(0.114, 0.117)	1.03	(1.02, 1.05)	35	(19, 50)
Model 2^#	0.117	(0.116, 0.118)	0.118	(0.117, 0.119)	0.99	(0.98, 1.00)	2	(– 11, 15)
Model 3^**	0.118	(0.117, 0.119)	0.116	(0.115, 0.117)	1.02	(1.01, 1.03)	28	(14, 43)
Model 4^††	0.118	(0.117, 0.119)	0.116	(0.115, 0.117)	1.02	(1.00, 1.03)	23	(10, 37)

^*Appalachian county designation based on Appalachian Regional Commission’s (ARC) designation. ^†Prevalence ratio (PR) comparing ADRD prevalence in Appalachian to non-Appalachian counties overall and by rural/urban county designation. ^‡Excess number of cases of ADRD in Appalachian counties in a given year overall and by rural/urban designation estimated by multiplying the prevalence difference by the average county-year level Appalachian (5,762), rural-Appalachian (3,358), and urban-Appalachian (9,142) FFS population size from 2015– 2018, respectively. ^$Model 1 includes Appalachian County designation, rural/urban county designation, and their interaction. P-value of the Appalachian/Rural County designation interaction term is 0.4809. ^#Model 2 adjusted for demographics of Medicare fee-for-service participants including age, gender, Medicaid eligibility, and comorbid conditions (i.e., atrial fibrillation, chronic kidney disease, depression, diabetes, heart failure, hyperlipidemia, hypertension, ischemic heart disease, schizophrenia/other psychotic disorders, and stroke). P-value of the Appalachian/Rural County designation interaction term is 0. 0004. ^**Model 3 additionally adjusted for race/ethnicity (i.e., county-level % white, % Black/African American, % Hispanic), housing tenure (i.e., % renter-occupied housing), % 65 and older, education (i.e., % 65 and older with a high school education), and life expectancy at birth. P-value of the Appalachian/Rural County designation interaction term is 0.3651. ^††Model 4 additionally adjusted for diagnostic access (i.e., PCP visits, neurology visits, and imaging scans). P-value of the Appalachian/Rural County designation interaction term is 0.3271. ^‡‡Rural counties defined as counties with an urban population size of less than 20,000 (Rural-Urban Continuum Codes (RUCC) 6-9); Urban counties defined as all other counties (RUCC 1-5).

The geographic variations in ADRD prevalence observed overall and by rural and urban county designation differed by state in the fully adjusted model (p-value of 3-way interaction between Appalachian, rural/urban, and state designation: 0.004; Table 3). Kentucky (PR: 1.06; 95% CI: 1.02, 1.10), Tennessee (PR: 1.08; 95% CI: 1.05, 1.11), and Virginia (PR: 1.05; 95% CI: 1.01, 1.10; Table 3) drove the Appalachian/non-Appalachian disparities in ADRD prevalence in rural counties. Subsequently, Appalachian/non-Appalachian variation in ADRD prevalence in urban counties was driven by North Carolina (PR: 1.05; 95% CI: 1.01, 1.08) and Tennessee (PR: 1.05; 95% CI: 1.03, 1.07). West Virginia is completely Appalachian, so within state comparisons were not estimable.

Table 3

Results of the negative binomial regression models of the estimated prevalence ratios (PRs) and 95% confidence intervals (CIs) comparing ADRD prevalence between Appalachian and non-Appalachian counties overall and by rural/urban designation across states in the fully adjusted model^*

	Overall		Rural^†		Urban^†
State	PR	95% CI	PR	95% CI	PR	95% CI	p ^‡
Kentucky	1.04	(1.01, 1.07)	1.06	(1.02, 1.11)	1.03	(0.99, 1.08)	0.244
North Carolina	1.04	(1.00, 1.08)	1.01	(0.96, 1.06)	1.05	(1.01, 1.08)	0.142
Ohio	0.99	(0.96, 1.01)	0.96	(0.93, 1.00)	0.99	(0.96, 1.01)	0.159
Tennessee	1.06	(1.04, 1.09)	1.08	(1.05, 1.12)	1.05	(1.02, 1.07)	0.069
Virginia	1.01	(0.98, 1.05)	1.05	(1.01, 1.10)	1.00	(0.97, 1.03)	0.026
West Virginia^$	–	–	–	–	–	–	–

^*Estimates from fully adjusted model (Model 4) stratified by rural/urban designation and adjusted for county-level demographics of Medicare fee-for-service participants including age, gender, Medicaid eligibility, and comorbid conditions (atrial fibrillation, chronic kidney disease, depression, diabetes, heart failure, hyperlipidemia, hypertension, ischemic heart disease, schizophrenia/other psychotic disorders, and stroke), race/ethnicity, renter-occupied housing, aging population distribution, education, life expectancy, and diagnostic access (PCP visits, neurology visits, and imaging scans). ^†Rural counties defined as counties with an urban population size of less than 20,000 (Rural-Urban Continuum Codes (RUCC) 6-9); Urban counties defined as all other counties (RUCC 1-5). ^‡P-value of the 3-way interaction term for Appalachian Regional Commission (ARC) designation, rural/urban designation, and state designation is 0.0035. ^$All West Virginia counties are Appalachian; thus, within-state PRs cannot be estimated.

DISCUSSION

In this ecologic study of the Central Appalachian region, we hypothesized that geographic variation in county-level ADRD prevalence would markedly differ between Appalachian and rural counties, however the data did not support this hypothesis. ADRD prevalence did vary slightly on the relative scale, but these differences between Appalachian and non-Appalachian counties were only of the magnitude of 2– 3%. While these associations were consistent with previous findings from Ohio only, 2 we hypothesized that the magnitude would have increased by accounting for differences in diagnostic access between counties in this region. Several potential reasons may account for these findings: the case mixture of the dual eligible population or MA populations in the region, the accuracy of measurement of the outcome, and the impact of educational attainment in the region.

Our analysis focused on the prevalence of ADRD among all FFS Medicare enrollees. However, the distribution of FFS compared to MA plans could contribute to the measurement of ADRD if there were systematic demographic differences between FFS and MA, and if this varied by area, we could be under- or overestimating the prevalence. Overall, there is small variation in the proportion of FFS Part A and B beneficiaries ever enrolled in MA across county types. A recent study comparing demographics and measures of health care access, use, and affordability among low-income MA and FFS beneficiaries did not find variation in access between enrollees of MA compared to FFS; low-income adults with MA were less likely to be from rural areas than low income adults with FFS. 28 Studies comparing key metrics of MA and FFS beneficiaries demonstrate more preventive care visits among MA than FFS. The lack of variation is likely not attributed to selection bias, as it seems FFS beneficiaries may be more prone to underdiagnosis. 29 Additionally, there is a varying proportion of persons in the population that are dual-eligible, where they have FFS Medicare, but are also Medicaid eligible. There was a higher proportion of Medicaid eligible persons in Rural-Appalachian counties (31.24%) than among all other county types, where the lowest proportion of Medicaid eligible persons is among urban non-Appalachian counties (17.52%) (Table 1). However, we accounted for Medicaid eligibility in our models and while it was individually associated with higher ADRD prevalence, its inclusion did not change the overall findings.

The use of administrative claims data is imperfect for measuring or capturing disease cases. Identifying a patient with ADRD in CMS uses the CCW case definition of over a three-year period, there must be at least 1 inpatient, skilled nursing facility, home health agency, hospital outpatient or Carrier claim with a valid ICD-9/10 code for an enrollee to be identified as having ADRD. 24 The CMS/CCW definition identifies 75– 80% of cases, but that means that there are 25% of cases that are potentially missed. 25 Zhu et al. reported that milder ADRD cases are often missed in claims data. 30 Additionally, claims algorithms underestimated ADRD among rural cases and cases with lower educational attainment, 31 indicating that misclassification may be differential by region type (e.g., lower sensitivity in Appalachian and rural counties). If only the more severe cases being detected among rural/Appalachian areas in claims data, such that PR is underestimated partially because shorter time to death among cases in rural areas (i.e., differences should be larger than observed).5,32, 5,32

In our analysis, we selected factors that are largely negative influences on cognition and ADRD, however, there are protective factors that may delay the onset of cognitive decline or prevent ADRD such as educational attainment and socialization. Educational attainment has grown in the US. Nearly 90% of adults 25 and older were high school graduates in 2017 compared to 82% in 1997, with bachelor’s degree attainment increasing from 16% to 21.3% in that same time period. 33 It has been thought that educational attainment influences cognitive reserve, and that education may be a mechanism to preserve cognition or slow cognitive decline.34 –36 Living in a rural community can influence the rate of cognitive decline. 32 Yet even when considering the changes in educational distribution, rural-urban differences persist in dementia. 3 Bachelor’s education has been growing in the Appalachian region, but not as fast as in the non-Appalachian regions of the same areas.37,38, 37,38 The US on average went up 3.3%, but Central Appalachia only increased 1.5% comparing 2012– 2016 to 2017– 2021. Additionally, education is related to health care access, as those with higher education may be best able to navigate through a dementia diagnosis. 39 The lag in education and the discrepancy in access means that differences in cognitive reserve would be more pronounced, so we would have anticipated more differences for ADRD rather than having the groups be more similar. Relatedly, socialization may also contribute to lower ADRD prevalence in rural and/or Appalachian regions. The tight-knit nature of rural communities have been found to be more socially cohesive than their urban counterparts, 40 which aligns with evidence supporting high frequencies of social interactions often reported among older rural compared to older urban residents. 41 With a known positive link between social engagement and cognition among older adults, higher levels of social cohesion that in turn positively influence social participation among rural residents may contribute to lower ADRD prevalence in rural areas. 42 Yet our analysis may have also been limited by the absence of other factors, such as air quality or built environment infrastructure, that could have had positive or negative influences on county-level ADRD prevalence. Future work should continue to explore the influence of protective health factors of the rural and Appalachian regions on ADRD prevalence.

PCP and neurologists are instrumental providers for the screening, detection, and diagnosis of dementia. While PCP availability is known to be limited in remote areas, 43 recent growth in the number of nurse practitioners (NPs) and Physician Assistants (PAs) poses opportunities for additional access to screening and initial diagnosis of dementia in underserved areas.44,45, 44,45 Telehealth is also emerging as a strategy for dementia detection in Appalachian or rural areas, however, differential availability of telehealth services by rurality and Appalachian designation still exists. 46 Higher rates of neurological disorders in Appalachian regions exist due to the poor sociodemographic makeup of these areas compared to their non-Appalachian counterparts. This variation in the burden of neurological disorders is increasing and, alternatively, supports a rising demand for specialty care in Appalachian areas over time. 47 The disparate need for neurologists exists even with consideration of rurality and area deprivation, where Appalachian census tracts were recently found to have between 25% and 35% lower spatial access to neurologists compared to non-Appalachian census tracts.46,48,49 , 46,48,49 We observed that having more neurologists was associated with higher prevalence similar to Lin et al., yet while supply of neurologists differs by geography, variation in demand is small and, for dementia specifically, use of specialist care rises with increasing supply of neurologists. 50 The vast majority of evidence that points to the declining availability of dementia support services and care providers rather demonstrates a rising need for additional diagnostic care access in Appalachian areas. The minimal relative variation in ADRD prevalence observed by Appalachian County designation may be indicative of that there is some other care-related factor that operates differentially for Appalachian areas such that the measurement of ADRD is offset to look equal to that in more resourced areas.

The magnitude of the overall differences in ADRD prevalence by Appalachian designation were smaller than originally hypothesized, however, some states demonstrated more pronounced variation than others. Overall variation in Appalachian compared to non-Appalachian counties existed in Kentucky, North Carolina, and Tennessee, where the difference was most pronounced in Tennessee. However, only Virginia showed variation by rurality as a driver of the Appalachian/non-Appalachian difference in rural but not urban counties. Our previous work demonstrated slightly lower ADRD prevalence among FFS beneficiaries in Appalachian compared to non-Appalachian counties from 2007 to 2017 in Ohio alone, where rural Appalachian counties had slightly higher ADRD prevalence and urban Appalachian counties had slightly lower prevalence than their non-Appalachian county counterparts from 2012 to 2017. 2 Conversely, we did not observe variation in ADRD prevalence by Appalachian designation overall or by rurality from 2015 to 2018 in Ohio specifically, even after accounting for differences in sociodemographics and diagnostic access. Our prior work motivated the current study’s evaluation of variation in ADRD prevalence across multiple Appalachian states likely characterized by poorer access to care than Ohio. For example, the lower ADRD prevalence observed in rural counties of Kentucky and West Virginia compared to urban counties in 2013 was hypothesized to be due to lack of diagnostic access corresponding with the disproportionate underdiagnosis of ADRD in rural areas. 4 However, we observed a higher ADRD prevalence in rural Appalachian counties of Kentucky compared to all other counties in the state across each level of adjustment, including PCP and neurologist counts. Our results may be reflecting recent increases in the burden of ADRD that disproportionately influences rural areas with large aging populations. The largely rural state’s sociodemographic and comorbid makeup may contribute to Tennessee being a main driver of the Appalachian/non-Appalachian variation in ADRD prevalence, 51 however, within state differences should be further explored.

Appalachian counties in Central Appalachia had relatively higher ADRD prevalence compared to non-Appalachian counties, but these differences may yet be understated. While accounting for socio-demographic and diagnostic access factors did not completely account for the relative difference, we hypothesize that the ascertainment of our outcome may be differentially misclassified which may mask the true magnitude of an Appalachian/non-Appalachian ADRD disparity. Future work should further evaluate if state level variation is a more appropriate scale than the entire Central Appalachian region.

AUTHOR CONTRIBUTIONS

Jeffrey J. Wing (Conceptualization; Funding acquisition; Methodology; Project administration; Supervision; Writing – original draft); Jenna I. Rajczyk (Data curation; Formal analysis; Writing – original draft); James F. Burke (Methodology; Supervision; Writing – review & editing).

Footnotes

ACKNOWLEDGMENTS

The authors have no acknowledgements to report.

FUNDING

This work was funded by the National Institute on Aging (R03AG078979) and by the Chronic Brain Injury Program, a university research center under the Discovery Themes Initiative and Enterprise for Research, Innovation & Knowledge at The Ohio State University.

CONFLICT OF INTEREST

The authors have no conflict of interest to report.

DATA AVAILABILITY

These data were derived from the following resources available in the public domain: Centers for Medicare and Medicaid Services – Medicare Geographic Variation (https://data.cms.gov/summary-statistics-on-use-and-payments/medicare-geographic-comparisons/medicare-geographic-variation-by-national-state-county); American Community Survey (https://data.census.gov/); Institute for Health Metrics and Evaluation – United States Mortality Rates and Life Expectancy by County, Race, and Ethnicity ().

References

2023 Alzheimer’s disease facts and figures. Alzheimers Dement 2023; 19: 1598–1695.

Wing

, Levine

, Ramamurthy

, et al. Alzheimer’s disease and related disorders prevalence differs by Appalachian residence in ohio. J Alzheimers Dis 2020; 76: 1309–1316.

Weden

, Shih

, Kabeto

, et al. Secular trends in dementia and cognitive impairment of U.S. rural and urban older adults. Am J Prev Med 2018; 54: 164–172.

Abner

, Jicha

, Christian

, et al. Rural-urban differences in Alzheimer’s disease and related disorders diagnostic prevalence in Kentucky and West Virginia. J Rural Health 2016; 32: 314–320.

Rahman

, White

, Mills

, et al. Rural-urban differences in diagnostic incidence and prevalence of Alzheimer’s disease and related dementias. Alzheimers Dement 2021; 17: 1213–1230.

Aranda

, Kremer

, Hinton

, et al. Impact of dementia: health disparities, population trends, care interventions, and economic costs. J Am Geriatr Soc 2021; 69: 1774–1783.

Diez Roux

, Mujahid

, Hirsch

, et al. The impact of neighborhoods on cardiovascular risk: the MESA Neighborhood Study. Glob Heart 2016; 11: 353–363.

Levasseur

, Généreux

, Bruneau

J-F

, et al. Importance of proximity to resources, social support, transportation and neighborhood security for mobility and social participation in older adults: results from a scoping study. BMC Public Health 2015; 15: 503.

Gauthier

, Leuzy

, Racine

, et al. Diagnosis and management of Alzheimer’s disease: Past, present and future ethical issues. Prog Neurobiol 2013; 110: 102–113.

10.

Morrone

, Cronin

, Schuller

, et al. Access to health care in Appalachia. J Appalach Health 2021; 3: 123–136.

11.

Donohoe

, Marshall

, Tan

, et al. Spatial access to primary care providers in Appalachia: evaluating current methodology. J Prim Care Community Health 2016; 7: 149–158.

12.

USDA ERS – Rural Education. USDA https://www.ers.usda.gov/topics/rural-economy-population/employment-education/rural-education/ (2021, accessed 13 February 2023).

13.

Pollard

and Jacobsen

. The Appalachian Region: A Data Overview from the 2012–2016 American Community Survey. PRB https://www.prb.org/resources/appalachias-aging-population-more-residents-ages-65-fewer-ages-25-to-64-signals-challenges-ahead/ (2018, accessed 10 November 2022).

14.

US Census Bureau. American Community Survey Information Guide.

15.

US Census Bureau. 2015–2018 American Community Survey 5-Year Estimates (S0101: Age and Sex), https://data.census.gov/table/ACSST5Y2018.S0101?t=Age%20and%20Sex&g=010XX00US$0500000.

16.

US Census Bureau. 2015–2018 American Community Survey 5-Year Estimates (S2502: Demographic Characteristics for Occupied Housing Units), https://data.census.gov/table/ACSST5Y2018.S2502?t=Owner/Renter%20(Householder)%20Characteristics&g=010XX00US$0500000.

17.

US Census Bureau. 2015–2018 American Community Survey 5-Year Estimates (S1501: Educational Attainment), https://data.census.gov/table/ACSST5Y2018.S1501?t=Education:Educational%20Attainment&g=010XX00US$0500000.

18.

US Census Bureau. 2015–2018 American Community Survey 5-Year Estimates (B03003: Hispanic or Latino Origin), https://data.census.gov/table/ACSDT1Y2022.B03002?t=Race%20and%20Ethnicity&g=010XX00US$0500000.

19.

US Census Bureau. 2015–2018 American Community Survey 5-Year Estimates (S1701: Poverty Status in the Past 12 Months), https://data.census.gov/table/ACSST5Y2018.S1701?t=Income%20and%20Poverty&g=010XX00US$0500000.

20.

Centers for Medicare and Medicaid Services. Medicare Geographic Variation – by National, State & County – Centers for Medicare & Medicaid Services Data. Data.CMS.gov, https://data.cms.gov/summary-statistics-on-use-and-payments/medicare-geographic-comparisons/medicare-geographic-variation-by-national-state-county (accessed 15 April 2024).

21.

Centers for Medicare and Medicaid Services. Specific Chronic Conditions, https://data.cms.gov/medicare-chronic-conditions/specific-chronic-conditions (2023, accessed 15 April 2024).

22.

Centers for Medicare and Medicaid Services. Master Beneficiary Summary File (MBSF) LDS. CMS.gov, https://www.cms.gov/data-research/files-for-order/limited-data-set-lds-files/master-beneficiary-summary-file-mbsf-lds (2023, accessed 15 April 2024).

23.

Institute for Health Metrics and Evaluation (IHME). United States Mortality Rates and Life Expectancy by County, Race, and Ethnicity 2000–2019. Seattle, United States of America: Institute for Health Metrics and Evaluation (IHME), 2022.

24.

Chronic Conditions Warehouse. 27 CCW Chronic Conditions Algorithms – Alzheimer’s Disease and Related Disorders or Senile Dementia–https://www2.ccwdata.org/web/guest/condition-categories-chronic (2022).

25.

Taylor

, Fillenbaum

and Ezell

. The accuracy of medicare claims data in identifying Alzheimer’s disease. J Clin Epidemiol 2002; 55: 929–937.

26.

Appalachian Counties Served by ARC. Appalachian Regional Commission, https://www.arc.gov/appalachian-counties-served-by-arc/ (accessed 30 March 2023).

27.

Hilbe

. Negative Binomial Regression . 2nd ed. New York: Cambridge University Press, 2012.

28.

Aggarwal

, Gondi

and Wadhera

. Comparison of Medicare Advantage vs traditional Medicare for health care access, affordability, and use of preventive services among adults with low income. JAMA Network Open 2022; 5: e2215227.

29.

Agarwal

, Connolly

, Gupta

, et al. Comparing Medicare Advantage and traditional Medicare: a systematic review. Health Affairs 2021; 40: 937–944.

30.

Zhu

, Ornstein

, Cosentino

, et al. Misidentification of dementia in Medicare claims and related costs. J Am Geriatr Soc 2019; 67: 269–276.

31.

McCarthy

, Chang

C-H

, Tilton

, et al. Validation of claims algorithms to identify Alzheimer’s disease and related dementias. J Gerontol A Biol Sci Med Sci 2021; 77: 1261–1271.

32.

Rahman

, White

, Thomas

, et al. Assessment of rural-urban differences in health care use and survival among Medicare beneficiaries with Alzheimer disease and related dementia. JAMA Network Open 2020; 3: e2022111.

33.

U.S. Census Bureau. Chapter 1: Population Trends and Educational Attainment . US Census Bureau, https://www.equityinhighered.org/resources/report-downloads/race-and-ethnicity-in-higher-education-a-status-report/ (2019).

34.

Meng

, D’Arcy

. Education and dementia in the context of the cognitive reserve hypothesis: a systematic review with meta-analyses and qualitative analyses. PLoS One 2012; 7: e38268.

35.

, Yu

J-T

, Tan

M-S

, et al. Cognitive reserve and Alzheimer’s disease. Mol Neurobiol 2015; 51: 187–208.

36.

Ciciora

, Vásquez

, Valachovic

, et al. Social and behavior factors of Alzheimer’s disease and related dementias: a national study in the U.S.. Am J Prev Med 2024; 66: 573–581.

37.

Pollard

, Srygley

and Jacobsen

. The Appalachian Region: A Data Overview from the 2017–2021 American Community Survey Chartbook, https://www.arc.gov/wp-content/uploads/2023/05/PRB_ARC_Chartbook_ACS_2017-2021_FINAL_2023-06.pdf#page=64 (2023).

38.

USDA. Educational attainment improved in rural America but educational gap with urban areas grew for bachelor’s degrees and higher. USDA Economic Research Service, http://www.ers.usda.gov/data-products/chart-gallery/gallery/chart-detail/?chartId=106147 (2023, accessed 17 April 2024).

39.

Park

, Jang

, Cho

, et al. Cognitive decline sensitivity by educational level and residential area: A descriptive study using long-term care insurance dementia registration data in South Korea. Medicine 2023; 102: e33003.

40.

Avery

, Hermsen

and Kuhl

. Toward a better understanding of perceptions of neighborhood social cohesion in rural and urban places. Soc Indic Res 2021; 157: 523–541.

41.

Carver

, Beamish

, Phillips

, et al. A scoping review: social participation as a cornerstone of successful aging in place among rural older adults. Geriatrics 2018; 3: 75.

42.

Douglas

, Georgiou

and Westbrook

. Social participation as an indicator of successful aging: an overview of concepts and their associations with health. Aust Health Rev 2017; 41: 455–462.

43.

Barth

, Nickel

and Kolominsky-Rabas

. Diagnosis of cognitive decline and dementia in rural areas — A scoping review. Int J Geriatr Psychiatry 2018; 33: 459–474.

44.

Streeter

, Zangaro

and Chattopadhyay

. Perspectives: using results from HRSA’s Health Workforce Simulation Model to examine the geography of primary care. Health Serv Res 2017; 52: 481–507.

45.

Poghosyan

, Brooks

, Hovsepian

, et al. The growing primary care nurse practitioner workforce: a solution for the aging population living with dementia. Am J Geriatr Psychiatry 2021; 29: 517–526.

46.

Buchalter

, Gentry

, Willis

, et al. Disparities in spatial access to neurological care in Appalachia: a cross-sectional health services analysis. Lancet Reg Health Am 2022; 18: 100415.

47.

Dall

, Storm

, Chakrabarti

, et al. Supply and demand analysis of the current and future US neurology workforce. Neurology 2013; 81: 470–478.

48.

Strowd

, Strauss

, Graham

, et al. Rapid implementation of outpatient teleneurology in rural Appalachia: barriers and disparities. Neurol Clin Pract 2021; 11: 232–241.

49.

Marra

, Hamlet

, Bauer

, et al. Validity of teleneuropsychology for older adults in response to COVID-19: a systematic and critical review. Clin Neuropsychol 2020; 34: 1411–1452.

50.

Lin

, Callaghan

, Burke

, et al. Geographic variation in neurologist density and neurologic care in the United States. Neurology 2021; 96: e309–e321.

51.

Luffman

, Joyner

, Tollefson

, et al. Rurality and COVID-19 in Tennessee: assessing and communicating pandemic emergence and transmission. Southeast Geogr 2021; 61: 203–221.