key: cord-0831334-uxtj7ye2 authors: Bennett, Erin E.; Kwan, Abraham; Gianattasio, Kan Z.; Engelman, Brittany; Dowling, N. Maritza; Power, Melinda C. title: Estimation of dementia prevalence at the local level in the United States date: 2021-12-31 journal: Alzheimers Dement (N Y) DOI: 10.1002/trc2.12237 sha: 47b3fc38a3c4092d008ba29a6520d0246b89f04b doc_id: 831334 cord_uid: uxtj7ye2 INTRODUCTION: Ensuring adequate and equitable distribution of resources to support persons living with dementia relies on understanding the burden and distribution of dementia in a population. Our goal was to develop an approach to estimate dementia prevalence at the local level in the United States using publicly available data. METHODS: Our approach combines publicly available data on dementia prevalence and demographic data from the US Census to estimate dementia prevalence. We illustrate this approach by estimating dementia prevalence in persons aged 65 and older in Philadelphia, PA; Chicago, IL; and Atlanta, GA. RESULTS: Overall, we estimate the prevalence of dementia among those 65 and older to be 11.9% in Philadelphia, 11.8% Chicago, and 12.3% in Atlanta. Estimates across Philadelphia localities vary from 9.3% to 15.9%. DISCUSSION: Our approach provides a cost‐effective method to generate estimates of dementia prevalence at the local level. HIGHLIGHTS: Brain health needs assessments require understanding of local dementia prevalence. Our approach can be used to estimate dementia prevalence in individual communities. This information can inform decisions about distribution of resources. allocation of resources, whether the availability of programs and services to support those living with dementia and their care partners meets community needs, and whether subpopulations most impacted by dementia have access to a proportional share of services. As such, municipalities could benefit from an understanding of the burden of dementia in their communities. In the United States, estimating dementia prevalence in a given municipality can be challenging. Presently, there is no national surveillance system for dementia, and given the costs of developing and maintaining successful surveillance systems, 1 it is unclear if and when we can expect one in the United States. Until such a system is implemented, we must rely on other approaches. Other administrative sources of information, such as Medicare claims, are unlikely to provide adequate information, as many people with dementia remain undiagnosed, and the sensitivity and specificity of a Medicare claim for identifying people living with dementia varies across sociodemographic subgroups. [2] [3] [4] [5] Existing epidemiologic studies that estimate and forecast the prevalence of dementia have focused on the national population, [6] [7] [8] and so have limited utility for public health planning at the local level. Our goal was to develop an approach for estimation of dementia prevalence at the local level using publicly available data. Broadly, this approach uses available, if limited, information on dementia prevalence from US-based cohort studies to generate prevalence estimates for predetermined demographic subgroups, which are then combined with census data to estimate dementia prevalence in specific geographic areas. Here we describe and illustrate our approach by generating estimates of dementia prevalence for three major US cities-Philadelphia, PA; Chicago, IL; and Atlanta, GA-as well as in smaller geographic units (public use microdata areas, or PUMAs) within Philadelphia. To begin, we reviewed the published literature and cohort websites for reports of dementia prevalence or incidence, stratified by age, sex, and/or race-ethnicity. We then selected a subset of these reportsbased on availability of stratified risk estimates, racial/ethnic diversity of the sample, and calendar period for the reported risk estimatesto inform our approach. Using data from these reports, we generated estimates of dementia prevalence within defined age-, sex-, and race/ethnicity categories. Next, we combined these stratified estimates with demographic information from the US Census Bureau to estimate dementia prevalence in persons aged 65 and older in populations of interest ( Figure 1 ). As we rely on published or publicly-available statistics, this work is not human subjects research. We searched PubMed for US population-based studies published after 2000 using the search terms "dementia," "prevalence," "incidence," and Hub on June 11, 2020. 12 We also used data from two additional samples ( [17] [18] [19] fit to the reported data on age group-specific prevalence, to derive estimates for alternate age categories. 20 MAP/MARS reported sex-and race-specific dementia prevalence estimates for the 65 to 84, 85 to 89, and 90+ age groups. Within each sex-and race-specific subgroup, we estimated the prevalence of dementia in age groups, 65 to 74, 75 to 84, and 85+, using an exponential best-fit line, assuming that the midpoint of each desired proportion of males and females in each race-age group, and the ratios of dementia prevalence of women to men in each age category. We used the estimated prevalence of dementia for females and for males from ADAMS 13 to derive the women-men prevalence ratios. (In the absence of more specific data, we chose ADAMS because it was designed to allow recovery of nationally representative estimates and allowed us to keep estimates across the four samples independent, but we acknowledge that using a single ratio may not fully represent changes in sex difference by age and race.) For deriving the proportion of males and females in each age-race category, we used the number of males and females in each age-race group as reported in the 2012 US Census 5-year population estimates for the three Chicago neighborhoods from which CHAP participants were recruited. 21 Assuming the dementia prevalence in each age-race category is a weighted average of the prevalence of dementia in men and the prevalence of dementia in women in each age-race category, we were thus able to solve for age-, sex-, and race-specific dementia prevalence. Using the steps above, we were able to generate estimates of dementia prevalence within subcategories defined by age (65-74, 75-84, and 85+), sex (male and female), and two racial categories (Black, White) in all four primary samples (ARIC, CHAP, Kaiser, MAP/MARS). For non-Hispanic Whites and Blacks, we averaged the race-, age-, and sex-specific dementia prevalence estimates from ARIC, Kaiser, CHAP, and MAP/MARS to produce an estimated dementia prevalence for each race-age-sex group. As an initial check, we confirmed these were reasonably similar to reported statistics from other samples. 22 To estimate the age-and sex-specific dementia prevalence in Asian and Hispanic groups, we multiplied the final age-and sex-specific dementia prevalence estimates in non-Hispanic Whites by the estimated prevalence ratios comparing dementia prevalence in Hispanics or Asians to non-Hispanic Whites. For Asians, this prevalence ratio was based on data from Kaiser, 10 which was the only large, recent study with the relevant data. For Hispanics, this prevalence ratio was based on data from both Kaiser 10 and WHICAP-I, 14 which include different Hispanic/Latinx subpopulations. Similar to CHAP, WHICAP-I only published AD dementia incidence rates cross-classified by age and race. Therefore, we first calculated age-and race-specific prevalence rates assuming a 7.4-year duration of dementia. We then approximated sex-specific dementia prevalence rates within each stratum using the Street and 181th Street in the Washington Heights and Inwood neighborhoods in New York in 2010, 21 using a process identical to that described above for CHAP. Finally, we averaged race-specific prevalence rates from the two cohorts for Whites and Hispanics, then calculated a prevalence ratio comparing the prevalence of dementia in Hispanics to that in Whites. Estimation of dementia prevalence based on demographic information The steps above ultimately result in a single set of dementia prevalence estimates for subgroups defined by age (65-74, 75-84, and 85+), sex (male and female), and race/ethnicity (non-Hispanic Black, non-Hispanic White, Asian, and Hispanic). For persons who are not White, Black, Hispanic, or Asian, we assumed the risk of dementia in each age/sex category to be equivalent to that observed for non-Hispanic Whites. We then used information on the age, sex, and racial/ethnic make-up of a given population provided by the US Census Bureau to derive estimates of dementia prevalence for that population. Specifically, the number of individuals of a given population in each age, sex, and racial/ethnic group is multiplied by the estimated dementia prevalence for that group, producing the expected number of persons in each group with dementia. The total number of expected cases of dementia in the given population is then divided by the population size over 65 to estimate the overall prevalence of dementia in older adults. To validate our dementia prevalence calculator, we predicted dementia prevalence in ADAMS and in the United States and compared these val- We chose to illustrate our approach by estimating the prevalence of dementia in three US cities: Philadelphia, PA; Chicago, IL; and Atlanta, GA. Though two of the cohorts used for prevalence estimation were based in Chicago (CHAP and MAP/MARS), neither was intended to be representative of the entire city. To estimate dementia prevalence in Chicago and Atlanta we used ACS 1-year estimates for the 2019 population. 21 All of our cities of interest were defined as census places. To demonstrate that our approach can capture local variation in dementia prevalence, we leveraged data from Philadelphia's 11 PUMAs. For Philadelphia specifically, PUMAs are roughly spatially aligned with planning districts, and therefore may be useful for local planning purposes. For Philadelphia and the 11 Philadelphia PUMAs, we used the ACS 5-year estimates for the 2018 population, as 1-year estimates were incomplete for specific PUMAs. 21 Our validation exercises produced estimates of dementia prevalence consistent with those reported elsewhere. Because of growing Overall, we estimate the dementia prevalence among those 65 and older to be 11.9% in Philadelphia, 11.8% in Chicago, and 12.3% in Atlanta ( Table 2 ). As expected, prevalence was higher in older age groups, and in Black or Hispanic subpopulations. Dementia prevalence was largely consistent within demographic subgroups across cities. However, the prevalence of dementia among Asians differed across cities, reflecting the different age distribution of Asians aged 65 and older in Chicago, Atlanta, and Philadelphia. Estimates of dementia prevalence varied from 9.3% to 15.9% across Philadelphia PUMAs (Table 3 and Figure 3 ). Here we demonstrate a method that uses publicly available data to generate actionable estimates of dementia prevalence at the local level. This approach is cost effective and can be used to quantify expected differences in dementia prevalence within and across municipalities. Given that there is no cure for dementia, much of the public health discussion and guidance from the US Centers for Disease Control focuses on risk factor prevention and early identification of cognitive decline. 26 Even with these preventive steps, the country will see an increase in dementia prevalence in the coming years due to the changing age structure of the US population. To confront this growing public health concern, public health departments need tools to estimate the local burden of dementia and identify disparities within their coverage area to properly and equitably allocate resources. This is especially true in the absence of formal national and local surveillance systems. While a dementia surveillance system would enhance our public institutions' ability to respond to the needs of those living with dementia, there are concerns about the cost and accuracy of dementia surveillance, 1 especially given that half of all persons living with dementia may be undiagnosed. 27 Additionally, these surveillance programs may take years to build, and public health practitioners will require tools to estimate dementia prevalence in local communities in the meantime. Our approach serves this need. Our approach has several strengths. Our validation exercises show that our approach can produce reasonable estimates of dementia prevalence in settings where a dementia surveillance study is infeasible given time or resource constraints. We were able to identify several data sources with dementia prevalence or incidence estimates from Figure 2 ), we believe this assumption is justifiable. We make similar assumptions about the relative prevalence of dementia among men and women across cohorts. Additionally, with the exception of the WHICAP cohort, 28 we lacked information on the proportion of participants in each cohort residing in a nursing home, a population that may be of particular interest for local public health departments. While our approach makes use of the available data, and is rational and well considered, other choices may also be justifiable and would likely result in slightly different dementia prevalence estimates. While other methods, such as small area estimation, 29, 30 could be used to quantify disease prevalence in smaller geographic areas and may have relative advantages, our approach remains a practical option for obtaining dementia prevalence data for cities and their subunits. Our approach precludes calculating measures of uncertainty, including prediction or confidence intervals. Rather, we prioritize communicating actionable information on the burden of dementia, while acknowledging that these are estimates. Although diagnostic criteria for dementia varied across studies, we assumed that they are measuring the same quantity. However, differences in criteria may contribute to differences across study-specific estimates of dementia prevalence. 31, 32 We also assume that dementia prevalence estimates in regions from which our cohorts of interest are drawn will generalize to other regions in the United States after accounting for differences in demographic distributions. Specifically, our approach assumes that, across the United States, equivalent age-sex-race/ethnic groups (such as non-Hispanic White females aged 65-74) have similar distributions of other dementia-related predictors, such as education. This assumption may not be appropriate in some settings. Given lack of information, we assumed the prevalence of dementia for persons identifying as a racial/ethnic group other than non-Hispanic White, non-Hispanic Black, Hispanic, and Asian was equivalent to that of non-Hispanic Whites. We are also unable to explore potential heterogeneity in dementia prevalence within our defined subgroups, despite evidence that dementia risk varies across Hispanic and Asian subpopulations. 10 Here, we demonstrate that it is possible to use publicly available data on dementia risk to derive local estimates of dementia prevalence. This approach is cost effective and can provide valuable information to municipalities, states, and local governments as they plan to support persons living with dementia and their care partners. This work was funded by a grant from DC Health to MCP (CHA2020-000024). The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. Melinda C. Power reports grants from the United States National Institutes of Health (NIH), Department of Defense (DOD), and DC Health. N Maritza Dowling reports grants from the United States National Institutes of Health (NIH). All authors were supported by a grant from DC Health. Dementia registries around the globe and their applications: a systematic review Racial disparities and temporal trends in dementia misdiagnosis risk in the United States The accuracy of Medicare claims as an epidemiological tool: the case of dementia revisited Prevalence and determinants of undetected dementia in the community: a systematic literature review and a meta-analysis Implications of the use of algorithmic diagnoses or Medicare claims to ascertain dementia The impact of changes in population health and mortality on future prevalence of Alzheimer's disease and other dementias in the United States Racial and ethnic estimates of Alzheimer's disease and related dementias in the United States (2015-2060) in adults aged >/= 65 years Alzheimer disease in the United States (2010-2050) estimated using the 2010 census Mild cognitive impairment and dementia prevalence: the atherosclerosis risk in communities neurocognitive study (ARIC-NCS) Inequalities in dementia incidence between six racial and ethnic groups over 14 years Prevalence and incidence of clinically diagnosed Alzheimer's disease dementia from 1994 to 2012 in a population study RADC Research Resource Sharing Hub. Rush Alzheimer's Disease Center (RADC) Prevalence of dementia in the United States: the aging, demographics, and memory study. Neuroepidemiology Incidence of AD in African-Americans, Caribbean Hispanics, and Caucasians in Northern Manhattan Overview and findings from the rush memory and aging project The Minority Aging Research Study: ongoing efforts to obtain brain donation in African Americans without dementia Risk of dementia hospitalisation associated with cardiovascular risk factors in midlife and older age: the Atherosclerosis Risk in Communities (ARIC) study The incidence of dementia: a meta-analysis Incidence and prevalence of dementia in the cardiovascular health study Rates and risk factors for progression to incident dementia vary by age in a population cohort Dementia and Alzheimer disease incidence: a prospective cohort study Effect of socioeconomic disparities on incidence of dementia among biracial older adults: prospective study Alzheimer's Disease Facts and Figures. Alzheimer's Association Twenty-seven-year time trends in dementia incidence in Europe and the United States The public health road map to respond to the growing dementia crisis Defeating Alzheimer's disease and other dementias: a priority for European science and society Hospitalization and Alzheimer's disease: results from a community-based study A review of small area estimation problems and methodolgical developments Multilevel regression and post stratification for small-area estimation of population health outcomes: a case study of chronic obstructive pulmonary disease prevalence using the behavioral risk factor surveillance system National estimates of the prevalence of Alzheimer's disease in the United States Sources of variability in estimates of the prevalence of Alzheimer's disease in the United States Systematic review of dementia prevalence and incidence in US race/ethnic populations Heterogeneity in 14-year dementia incidence between Asian American subgroups When to Use 1-year, 3-year, or 5-year Estimates Understanding and Using American Community Survey Data: What all Data Users Need to Know Government Printing Office Disparities in incidence of COVID-19 among underrepresented racial/ethnic groups in counties identified as hotspots during COVID-19 exacerbating inequalities in the US COVID-19 and African Americans COVID-19 in people with diabetes: understanding the reasons for worse outcomes Characteristics of persons who died with COVID-19-United States Misidentification of dementia in Medicare claims and related costs