Wisconsin Healthcare Utilization Cost Among American Indians/Alaska Natives with and without Alzheimer’s Disease and Related Dementias

INTRODUCTION

Approximately 6.2 million Americans have Alzheimer’s disease and related dementias (ADRD), resulting in an annual direct cost of $355 billion [1]. By 2050, the prevalence of Americans ages 65 and older is predicted to increase from 58 to 88 million [1], the prevalence of dementia is expected to double, and the economic burden is expected to triple to $1.1 trillion annually [1]. The incremental cost of ADRD is greater than any other medical condition [2] and remains in the presence of co-existing medical conditions [3, 4]. ADRD-related healthcare utilizations costs vary as a function of time since diagnosis, with the year after diagnosis incurring the greatest incremental cost [5, 6]. A recent Medicare review found that incremental incident cost of care for the year following ADRD diagnosis ranged from $8,938 to $39,794 [7], but research examining the economic impact of ADRD among groups historically marginalized in the U.S. is limited, despite vast health inequities among marginalized racial and ethnic groups (particularly in terms of ADRD) [1, 8]. One woefully under-included population [1] with a rapidly increasing incidence of ADRD is American Indians and Alaska Native populations(AI/AN) [9].

AI/AN are among the fastest growing racial groups in terms of ADRD [9]. Researchers predict that one in three AI/AN adults ages 65 years or older will develop ADRD in the next two decades [9]. One can speculate why this group has shown such rapid growth in ADRD rates: 1) increased awareness of ADRD in this population due to more research and public health campaigns [1], which has led to increased patient sought care; 2) greater life expectancy of AI/AN individuals over the past centuries has resulted in more people aging to the time of dementia diagnosis [10]. However, for a clearer understanding of this phenomenon, we must wait for the results of ongoing research aimed at determining why incidence rates among AI/AN individuals are growing, while the rest of the U.S. population decreases in incidence (but increases ADRD prevalencewith aging).

After adjusting for risk factors known to disproportionately affect AI/AN individuals (i.e., diabetes, stroke, hyperlipidemia, hypertension, and commercialized tobacco use), the relative risk of ADRD in the AI/AN population is 49% higher than in the Non-Hispanic White population [11]. Unfortunately, prevalence estimates for ADRD among AI/AN adults vary substantially, ranging from 4.2% to 9.1% [8, 12]. These prevalence estimates are from three different time periods and locations, using the only known existing sources [8, 12]. This prevalence rate presumably underestimates the true number of AI/AN adults with an ADRD diagnoses, as less than half of all ADRD cases are diagnosed in the general U.S. population, and less than one-third of AI/AN adults ages 65 years or older experiencing memory loss discussed this symptom with their healthcare provider [13]. Underreporting may be due to the longstanding and significant mistrust of the healthcare services outside the tribe [1], which limits accurate and appropriate treatment for ADRD. Additionally, multiple barriers inhibit enrollment in Medicare and Medicaid programs for older AI/AN adults, including transportation, cultural and language barriers, limited awareness of program eligibility and access routes, and differing definitions of “elders” for tribal communities (55 years old) versus Medicaid and the Older Americans Act (60 years old), which can result in a gap in care [14].

For those AI/AN individuals who do obtain an ADRD diagnosis, AI/AN individuals’ utilization of healthcare services significantly increases, particularly within the first year [15]. Utilization increase spans across all types of healthcare and settings and is significantly higher than utilization rates of individuals without ADRD [15]. Despite growing interest and research in the impact of ADRD on AI/AN individuals, only one incremental cost estimate of healthcare utilization associated with ADRD exists among this population. A recent article found that when looking at Indian Health Service (IHS)/Tribal treatment costs for AI/AN individuals ages 65 and older, unadjusted mean total costs for those with ADRD was $5400 more than those without ADRD [16]. However, no study to date has examined the cost of ADRD among AI/AN individuals in other state-fundedprograms.

This study examined the incremental 5-year health care utilization costs of ADRD for Medicaid and state program eligible AI/ANs ages 40 and older. We hypothesized that AI/ANs with ADRD would incur significant incremental costs, even after matching cases for co-existing medical diagnoses common in the AI/AN population [17, 18]. We also hypothesized that by using a more comprehensive definition of ADRD, to highlight the likely underdiagnoses of this disease in the AI/AN population [13], an even greater incremental cost would be observed among AI/AN adults with versus without ADRD.

METHODS

Paid electronic health record data from Wisconsin Department of Health Services were retrospectively analyzed to compare non-Tribal and IHS utilization costs among Medicaid and state program eligible AI/AN adults, ages 40 and older, based on the presence or absence of ADRD throughout data review (7/1/2015 through 7/1/2020) using R (Version 4.1.0) [19]. Controls (without ADRD) were matched to cases (with ADRD) using exact matching (Matchit package) [20] across subclasses defined by crossings of subjects’ age group, sex, county type (rural, urban, or reservation land), diabetes status, chronic obstructive pulmonary disease (COPD)/emphysema status, and myocardial infarction (MI) status; all available controls in a subclass were matched. Percentile-based bootstrapped 95% confidence intervals (CIs) of weighted mean difference (cases’ mean minus controls’ mean), over 2000 bootstrap iterations, were used to compare billed costs for these participants, totaled over 5 years of claims and encounter data for each subject; tests accounted for both the subclass sampling strata and weights produced through the matching process. Due to the heavy right tail in cost distributions, and the relatively small size of cases in some instances, the bootstrapped CIs were the most appropriate method of analysis, as opposed to weighted t-tests which might have assumption issues, or weighted non-parametric methods that deal with medians instead of the means (primary statistic of interest). In each bootstrap iteration, case and control groups before matching were each resampled with replacement back to the original size of the group, then matching was performed and the 5-year weighted mean cost difference calculated. Age groups for analyses were defined as subject age on July 1st, 2020; groups were formed in 5-year intervals starting at 40 years of age (i.e., 40 to <45, 45 to <50, etc.). Health conditions (i.e., ADRD, diabetes, COPD/emphysema, and MI) were defined as presence of the health condition’s diagnoses as of July 1st, 2020. Weighted Chi-square tests of association (using survey package) [21] were also employed to confirm balance was achieved across age groups, sex, diabetes mellitus status, COPD/emphysema status, and MI status in the matching process; no weighted imbalances occurred, as expected. Analyses were performed across all payor programs of a combined Fee-for-service (FFS), HMO, and managed care organizations/long-term care (MCO/LTC) dataset. Due to the structure of Medicaid state plans, including home and community-based services (HCBS) waivers, individuals can co-occur across benefit programs, with those individuals receiving Medicaid card only services under either FFS or HMO plan structures. Wisconsin’s state Medicaid plan offers both an FFS and HMO program option that members may choose between, including changing from one plan to the other within the same benefit year. To capture these nuances over the 5-year data collection period, unique participant IDs were used to account for and aggregate billed data to ensure against data duplication.

In secondary analyses, the case/control covariate of dementia status was modified to be “ADRD and possible misdiagnoses” (ADRD APMD), which was a diagnosis of ADRD, malnutrition, or adult failure to thrive (AFTT) anywhere within the 5-year period of encounters and/or claims. All other aspects of the matching and analyses were unchanged.

RESULTS

Confidence intervals on the weighted 5-year mean cost difference revealed that AI/ANs with ADRD accrued significantly more costs (Mean = $278,186.73) than those without ADRD (Mean = $128,945.21; $149,241.52 weighted mean difference; 95% CI on mean difference: $77,888 to $245,126; Table 1). Secondary analyses found that AI/ANs with ADRD APMD accrued significantly more costs (Mean = $308,529.30) than those without ADRD APMD (Mean = $110,015.67; a $198,513.63 weighted mean difference; 95% CI on mean difference: $160,307 to $237,917).

DISCUSSION

Consistent with hypotheses, ADRD was associated with a 2- to 3-fold increase in total billed dollars, depending on the type of payor program and disease definition. We found an incremental cost of $149,241.52 over 5 years (approximately $29,848.30/year) for AI/ANs with ADRD. Visual inspection of the distribution of incremental costs of ADRD by age group within both payor programs did not reveal a clear pattern, suggesting the reported disparity in costs exist across all age groups. Secondary analyses revealed that a broader definition of ADRD, including malnutrition and AFTT, resulted in mean incremental 5-year cost for ADRD of $198,513.63 ($39,702.73/year).

Notable strengths and limitations of these analyses include incomplete data from Indian Health Service, Tribal, and Urban Indian Health Program (ITU; the first line of medical defense for many AI/ANs), and no data from Veterans Health Affairs, likely underestimating the economic burden of ADRD in the AI/AN population. Second, participant mortality data was unavailable, prohibiting evaluating impact of death on cost of care. Competing risk is particularly important in studies of ADRD and older adults [22] as death during the study may obfuscate the impact of ADRD on healthcare utilization cost. Although death rates between demented and non-demented groups could be different and confounded with cost, this is unlikely unless the occurrence of death rates are unexpectedly higher in the non-demented versus demented group [23]. Thus, any confounding that may occur would underestimate the 5-year cost differences observed in this study. Authors made the a priori decision to limit the number of matched conditions to COPD/emphysema, MI, and diabetes to allow for adequate matching of participants, but future analyses would benefit from including additional relevant health and psychiatric conditions (i.e., stroke, hypertension, kidney disease, depression, etc.). The greatest incremental cost was observed when using a more comprehensive definition of ADRD (ADRD APMD), but this definition did not include altered mental status (delirium), which commonly co-occurs with and/or is misdiagnosed as ADRD [24]. This could affect the accuracy of our analyses in that we may have missed individuals whose delirium is the first sign of an underlying neurodegenerative process. An important strength of this study is that data were obtained directly from the Wisconsin Department of Health Services and results remained consistent when repeating analyses using several other methods (non-parametric/median-based methods without bootstrapping, 1-to-1 matching without weighting, and regression analyses). Dollar amounts found in this study were higher than annual combined Medicare/Medicaid costs reported elsewhere [1] and may highlight a potential elevated cost of healthcare utilization for ADRD among AI/AN populations, but further research is needed to directly examine this point. The data used in this study did not provide cost component data (e.g., hospitalization, outpatient visits, examinations, medications, etc.), which should also be analyzed in future work to examine exactly where the reported costs are stemming from. Lastly, zip-code level data was unavailable, precluding analyses specific to the eleven Tribal Nations of Wisconsin. Each tribal nation is unique in its culture and members and may have differing costs of care related to dementia; future work should engage with tribes to hopefully include tribal data and highlight these possible differences.

Based on the national prevalence estimate of 702,325 AI/AN adults ages 65 or older [25] and the ADRD prevalence range of 4.2% to 9.1% [8, 12], we estimate ADRD in AI/AN costs an additional $880.45 million to $1.91 billion annually. Moreover, we estimate ADRD APMD in AI/AN costs an additional $1.17 billion to $2.54 billion annually. Importantly, we used the prevalence estimate of AI/AN adults 65 and older (rather than 40 and older) as two of the existing estimates of ADRD prevalence in AI/AN individuals are calculated for adults ages 65 and older [8, 12]. This, along with the factors of AI/AN individuals being underreported in the 2020 census [26] and significant underdiagnoses of ADRD in AI/AN individuals [13] suggests an even greater economic impact than the dollar amounts listed above. Taken together, this work exemplifies the potential economic benefit of allocating funds to prevent, accurately diagnose, and treat ADRD in AI/AN adults. These funds should be allocated with the input and collaboration of tribal leaders, who have already reported that a chief unmet need is dementia-related care and awareness [27].