A Comparison of Dementia-free Life Expectancy Estimates Across Competing Algorithmic Classifications: New Knowledge and Considerations from the Health and Retirement Study - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice J Gerontol B Psychol Sci Soc Sci . Author manuscript; available in PMC: 2026 Apr 17. Published in final edited form as: J Gerontol B Psychol Sci Soc Sci. 2026 May 7;81(6):gbag051. doi: 10.1093/geronb/gbag051 Search in PMC Search in PubMed View in NLM Catalog Add to search A Comparison of Dementia-free Life Expectancy Estimates Across Competing Algorithmic Classifications: New Knowledge and Considerations from the Health and Retirement Study Marc A Garcia Marc A Garcia , PhD 1. Lerner Center for Public Health Promotion, Aging Studies Institute, & Department of Sociology, Maxwell School of Citizenship & Public Affairs, Syracuse University, Syracuse, New York, United States Find articles by Marc A Garcia 1 , Wassim Tarraf Wassim Tarraf , PhD 2. Wayne State University, Institute of Gerontology & Department of Healthcare Sciences, Detroit, Michigan, United States Find articles by Wassim Tarraf 2 , Chi-Tsun Chiu Chi-Tsun Chiu , PhD 3. Institute of European and American Studies, Academia Sinica, Taipei, Taiwan Find articles by Chi-Tsun Chiu 3 , Amy D Thierry Amy D Thierry , PhD, MPH 4. Department of Public Health Sciences, Xavier University of Louisiana, New Orleans, Louisiana, United States Find articles by Amy D Thierry 4 , Joseph L Saenz Joseph L Saenz , PhD 5. Edson College of Nursing and Health Innovation, Arizona State University, Phoenix, Arizona, United States Find articles by Joseph L Saenz 5 , Adriana M Reyes Adriana M Reyes , PhD 6. Brooks School of Public Policy, Cornell University, Ithaca, New York, United States Find articles by Adriana M Reyes 6 , Roland J Thorpe Jr Roland J Thorpe Jr , PhD 7. Johns Hopkins Alzheimer’s Disease Resource Center for Minority Aging Research, Johns Hopkins Bloomberg School of Public Health, Johns Hopkins University, Baltimore, Maryland, United States 8. Program for Research on Men’s Health, Hopkins Center for Health Disparities Solutions, Johns Hopkins Bloomberg School of Public Health, Johns Hopkins University, Baltimore, Maryland, United States Find articles by Roland J Thorpe Jr 7, 8 Author information Copyright and License information 1. Lerner Center for Public Health Promotion, Aging Studies Institute, & Department of Sociology, Maxwell School of Citizenship & Public Affairs, Syracuse University, Syracuse, New York, United States 2. Wayne State University, Institute of Gerontology & Department of Healthcare Sciences, Detroit, Michigan, United States 3. Institute of European and American Studies, Academia Sinica, Taipei, Taiwan 4. Department of Public Health Sciences, Xavier University of Louisiana, New Orleans, Louisiana, United States 5. Edson College of Nursing and Health Innovation, Arizona State University, Phoenix, Arizona, United States 6. Brooks School of Public Policy, Cornell University, Ithaca, New York, United States 7. Johns Hopkins Alzheimer’s Disease Resource Center for Minority Aging Research, Johns Hopkins Bloomberg School of Public Health, Johns Hopkins University, Baltimore, Maryland, United States 8. Program for Research on Men’s Health, Hopkins Center for Health Disparities Solutions, Johns Hopkins Bloomberg School of Public Health, Johns Hopkins University, Baltimore, Maryland, United States * Corresponding author: Marc A. Garcia, PhD. [email protected] PMC Copyright notice PMCID: PMC13085996 NIHMSID: NIHMS2159044 PMID: 41912419 The publisher's version of this article is available at J Gerontol B Psychol Sci Soc Sci Abstract Objectives: To create new dementia-free life expectancy estimates based on three competing algorithmic classifications of dementia and compare these estimates with dementia-free life expectancy estimates based on the well-established Langa-Weir (LW) dementia classification to ascertain the consistency of competing algorithmic methods on the overall burden of quality of life measured by the number of years lived dementia-free and the proportion of late-life years dementia-free to total life expectancy. Methods: Data from the Health and Retirement Study (1998-2016) are used to estimate Sullivan-based life tables of dementia-free life expectancies by race/ethnicity, nativity, and gender for adults aged 70 and older. Dementia ascertainment was operationalized using the Langa-Weir (LW), Expert, Hurd, and LASSO classifications. We test for both within-group differences (i.e., nativity for Latinos) and gender differences to comprehensively document the overall burden of dementia among at-risk populations. Results: Expert, LASSO, and Hurd competing algorithm estimates were generally equal to the LW estimates for dementia prevalence after age 70, with a few exceptions. White men and women, and Black women exhibit better cognitive functioning in the LW model, whereas the results for other race/ethnic and nativity groups are more consistent across models. Discussion: Small differences in population-level estimates across algorithmic classification are observed; however, these small differences may be important for understanding disparities across groups. Therefore, more work is needed 1) to understand for whom and which models are most accurately identifying dementia and 2) to place these models relative to gold-standard adjudicated dementia assessments. Keywords: Minority aging (race/ethnicity), Measurement, Alzheimer's disease Introduction Alzheimer's disease (AD) and AD-related dementias (ADRDs) are growing public health concerns that impact individuals, families, communities, and healthcare systems. The growing dementia crisis in the United States, driven by population aging, affects approximately 7.2 million adults aged 65 and older, with the total economic burden of AD/ADRD estimated to be $781 billion in 2025 ( Alzheimer's Association, 2025 ). Yet, estimates of AD/ADRD prevalence across and within population-based studies are highly variable, in large part, due to inconsistencies in accurately assessing AD/ADRD prevalence across diverse population subgroups, and the time and financial costs associated with competing methodological approaches and diagnostic measures used to ascertain dementia status ( Gianattasio et al., 2020 ; Gianattasio, Prather, et al., 2019 ; Gianattasio, Wu, et al., 2019 ; Hudomiet et al., 2022 ). Thus, identifying cost-effective and accurate diagnostic strategies for specific population subgroups can be particularly useful, as the care needs and access to services at various stages of AD/ADRD onset and development vary significantly among at-risk groups by race/ethnicity, nativity, and gender ( Garcia et al., 2023 ; Lin & Liu, 2024 ; Park & Chen, 2020 ; Saadi et al., 2017 ). Emerging research comparing new and existing algorithms for dementia assessment shows substantial variation in the prevalence of dementia by race/ethnicity and gender ( Gianattasio et al., 2020 ; Gianattasio, Wu, et al., 2019 ), and nativity status measured separately (i.e., subgroups of U.S.-born vs. foreign-born, independent of race/ethnicity and gender) ( Hudomiet et al., 2022 ). These findings show that older non-Latino Black and Latino adults have higher prevalence rates of dementia compared to non-Latino White adults with traditional screening tools (i.e., Langa-Weir algorithm) exhibiting lower reliability among racial/ethnic subgroups, leading to significant differences in diagnosis rates, with older Black adults having a higher risk of underdiagnosis compared to their White counterparts. Differences in the derived algorithms are based on methods developed independently across a series of research groups and operationalized recently for testing and validation ( Crimmins et al., 2011 ; Gianattasio et al., 2020 ; Hurd et al., 2013 ; Wu et al., 2013 ). Briefly, the Langa–Weir (LW) algorithm applies cutoffs to cognitive tests (TICS-M for self-report) and to proxy and interviewer assessments of memory, cognitive status, and IADLs (for proxy interviews), and derives scores summarizing cognitive and/or functional data from the HRS interview to identify persons with severe cognitive impairment or dementia. Cutoffs are validated against external clinical data from the ADAMS study, a subset of the HRS, as described elsewhere ( Langa et al., 2005 ). The Expert, LASSO, and Hurd algorithms apply regression and machine learning techniques to HRS survey data to predict cognitive classifications in the ADAMS Wave A and to use the estimated model parameters to generate post hoc classifications for the entire HRS sample. Importantly, challenges with metrics for assessing AD/ADRD prevent accurate comparisons across studies, making it difficult to assess disease progression and the impact of public health interventions, which hinder progress in population health. Two critical aspects for advancing AD/ADRD in public health are: 1) the development of culturally appropriate, robust, and detailed national estimates to highlight the distribution and impact of the disease across different at-risk populations; and 2) comparing competing algorithmic classifications of dementia ascertainment that facilitate the assessment of the average burden of disease over the life course measured by both the estimated number of years (quantity) lived dementia-free and the proportion (quality) of late-life years dementia-free to total life expectancy. Building on the above findings, we use the Health and Retirement Study to examine between-group and within-group differences in estimates of dementia prevalence and dementia-free life expectancy, as well as the consistency of these estimates across competing algorithmic classifications of dementia ascertainment. Methods Data This study used data from 1) the Health and Retirement Study (HRS), a nationally representative study of individuals over age 50 and their spouses in the United States sponsored by the National Institute on Aging (grant number NIA U01AG009740), and is conducted by the University of Michigan ( Bugliari et al., 2020 ); 2) the public-use National Health Interview Survey Linked Mortality Files (NHIS-LMF, 1999-2015) ( Blewett et al., 2019 ); 3) the Langa-Weir (LW) classification of cognitive function data file ( Langa, 2020 ) for 1995-2016 HRS respondents; and 4) the Gianattasio-Power Predicted Dementia Probability Scores and Dementia Classifications data file for 2000-2016 ( Gianattasio et al., 2020 ) that includes three newly developed algorithms: the (modified) Hurd Model, the Expert Model, and the LASSO Model. We merged the HRS, LW cognitive function, and Gianattasio-Power files and restricted our analytic sample to 2000-2016 HRS respondents aged 70 and older for estimating dementia prevalence. This analytical sample was used to estimate the age-, race/ethnicity-, nativity-, and gender-specific prevalence of dementia, defined using the Langa-Weir, Expert, Hurd, and LASSO classifications separately. We used the public-use NHIS-LMF mortality follow-up from the National Death Index through December 31, 2015, to estimate age-, race/ethnicity-, nativity-, and gender-specific mortality rates. Given our focus on Latino ethnicity, nativity status, and gender, we used the NHIS-LMF to estimate mortality due to the small number of deaths among Latinos in the HRS ( N =572). Notably, the NHIS-LMF contains 2,378 deaths for foreign-born Latinos (1,158 for men and 1,220 for women) and 1,850 deaths for U.S.-born Latinos (910 for men and 940 for women) compared to 305 deaths for foreign-born Latinos (135 for men and 170 for women) and 267 deaths for U.S.-born Latinos (124 for men and 143 for women) in the HRS. Prior research shows that mortality linkage quality in the NHIS-LMF is highly accurate among older Latinos (Lariscy et al., 2015) and that life expectancy estimates based on this data are comparable to U.S. vital statistics data ( Bauldry et al., 2023 ; Garcia et al., 2019 ; Hayward et al., 2014 ). We then integrated information on age-, race/ethnicity-, nativity-, and gender-specific dementia prevalence with age-, race/ethnicity-, nativity-, and gender-specific mortality rates to construct Sullivan life table models of dementia-free life expectancy. All models incorporated sampling weights provided by HRS and NHIS-LMF. At-Risk groups Race/ethnicity, nativity, and gender are used to identify key at-risk groups for our analysis. Race/ethnicity, nativity, and gender individually have been shown to be important factors for understanding dementia prevalence ( Chen & Zissimopoulos, 2018 ; Crimmins et al., 2018 ; Farina et al., 2022 ; Hudomiet et al., 2022 ; Langa et al., 2017 ; Rocca et al., 2011 ) and dementia-free life expectancy ( Farina et al., 2020 ; Hale et al., 2020 ). The HRS obtains information on race/ethnicity separately from participants self-reporting their race (White, Black, and Other) and whether they consider themselves Hispanic/Latino at first observation. Information on nativity is based on place of birth and obtained at the first interview. Following prior research ( Bauldry et al., 2023 ; Garcia et al., 2019 ), responses are combined with respondent gender to create eight at-risk groups: non-Latino White (hereafter, White) men, non-Latino Black (hereafter, Black) men, U.S.-born Latinos, foreign-born Latinos, and identical groups for women. Due to small sample sizes, we exclude foreign-born White and Black respondents. We also exclude respondents categorized as “other” race due to ambiguity in racial identification. Analytic plan First, we provide descriptive statistics to characterize the target population by race/ethnicity, nativity status (Latino-specific), and gender (see Supplementary Table 1 ). Second, we generate dementia prevalence estimates across competing algorithmic classifications of dementia ascertainment for 2000 and 2016 for the target populations (see Figure 1 and Supplementary Table 2 ). Third, we generate prevalence-based Sullivan dementia-free life expectancy estimates for racial/ethnic and nativity groups of interest by gender (see Table 1 and Figure 2 ). As with previous work, we follow a multistep process. First, we estimate logistic regression models to calculate gender-specific dementia prevalence rates for each race/ethnic and nativity group. Second, we use Gompertz hazard models to calculate age-specific (e.g., at age 70) total life expectancies, along with dementia prevalence. We also calculate life expectancies with dementia and dementia-free life expectancies for each group. As such, life expectancy estimates (total, with dementia, and dementia-free) can be interpreted as the gender-specific total expected number of years lived after age 70, as well as the average number of years lived with dementia and dementia-free for each race/ethnic and nativity group member, assuming shared constant risk of disease and mortality. Fifth, to calculate inferential statistics (i.e., standard errors and 95% confidence intervals) for the estimated averages and test differences in estimates, we use bootstrapping methods following standard procedures detailed in prior work ( Garcia et al., 2021 ; Garcia et al., 2019 ). For comparative purposes, we estimate dementia prevalence and dementia-free life expectancy among pan-ethnic Latinos across competing algorithmic classifications of dementia ascertainment ( Supplementary Figures 1 - 2 and Supplementary Tables 2 - 3 ), given our interest in nativity status as an overlooked axis of stratification in the development of algorithmic methods for dementia classification among older Latinos. Figure 1. Open in a new tab Dementia Prevalence Estimates by Race/Ethnicity, Nativity, and Gender (2000 and 2016). Table 1: A Comparison of Sullivan Dementia-free Life Expectancy Estimates at Age 70 Across Competing Algorithmic Classifications by Race/Ethnicity, Nativity, and Gender. Dementia Algorithmic Classification Differences By Classification Variables LW Expert Hurd LASSO LW-Expert LW-Hurd LW- LASSO Female White Dementia-free 15.0 13.2 * 13.5 * 13.2 * 1.8 * 1.5 * 1.8 * Dementia 2.2 4.1 * 3.7 * 4.0 * −1.9 * −1.5 * −1.8 * Total Life Expectancy 17.3 17.3 17.3 17.3 - - - Ratio Healthy 0.87 0.76 * 0.78 * 0.76 * −0.11 * 0.09 * −0.11 * Black Dementia-free 11.4 9.9 * 10.7 * 9.7 * 1.5 * 0.7 * 1.7 * Dementia 4.9 6.5 * 5.6 * 6.6 * −1.6 * −0.7 * −1.7 * Total Life Expectancy 16.3 16.3 16.3 16.3 - - - Ratio Healthy 0.70 0.61 * 0.66 * 0.60 * 0.09 * 0.04 * 0.10 * U.S.-Latino Dementia-free 12.3 12.4 * 12.0 * 11.9 * −0.1 * 0.3 * 0.4 * Dementia 5.4 5.4 * 5.7 * 5.8 * 0.0 * −0.3 * −0.4 * Total Life Expectancy 17.7 17.7 17.7 17.7 - - - Ratio Healthy 0.70 0.70 * 0.68 * 0.67 * 0.00 * 0.02 * 0.03 * Foreign-Born Latino Dementia-free 13.6 13.4 * 12.5 * 12.2 * 0.2 * 1.1 * 1.4 * Dementia 6.4 6.5 * 7.4 * 7.7 * −0.1 * −1.0 * −1.3 * Total Life Expectancy 20.0 20.0 20.0 20.0 - - - Ratio Healthy 0.68 0.67 * 0.63 * 0.61 * 0.01 * 0.05 * 0.07 * Male White Dementia-free 13.3 12.0 * 12.2 * 12.1 * 1.3 * 1.1 * 1.2 * Dementia 1.4 2.7 * 2.6 * 2.6 * −1.3 * −1.2 * −1.2 * Total Life Expectancy 14.7 14.7 14.7 14.7 - - Ratio Healthy 0.91 0.82 * 0.83 * 0.82 * 0.09 * 0.08 * 0.09 * Black Dementia-free 9.8 9.3 * 9.7 * 9.2 * 0.5 * 0.1 * 0.6 * Dementia 3.7 4.2 * 3.8 * 4.3 * −0.5 * −0.1 * −0.6 * Total Life Expectancy 13.5 13.5 13.5 13.5 - - - Ratio Healthy 0.73 0.69 * 0.72 * 0.68 * 0.04 * 0.01 * 0.05 * U.S.-Latino Dementia-free 11.4 12.3 * 10.8 * 11.2 * −0.9 * 0.6 * 0.2 * Dementia 4.5 3.6 * 5.2 * 4.7 * 0.9 * −0.7 * −0.2 * Total Life Expectancy 15.9 15.9 15.9 15.9 - - - Ratio Healthy 0.72 0.77 * 0.68 * 0.71 * −0.05 * 0.04 * 0.01 * Foreign-Born Latino Dementia-free 13.2 13.7 * 12.2 * 12.6 * −0.5 * 1.0 * 0.6 * Dementia 3.8 3.3 * 4.8 * 4.5 * 0.5 * −1.0 * −0.7 * Total Life Expectancy 17.0 17.0 17.0 17.0 - - - Ratio Healthy 0.77 0.81 * 0.72 * 0.74 * −0.04 * 0.05 * 0.03 * Open in a new tab Notes. LW = Langa-Weir. Data from the U.S. Health and Retirement Study, 2000-2016. * p ≤ .05. Figure 2. Open in a new tab Dementia-free Life Expectancy by Race/Ethnicity, Nativity, and Gender. Results Prevalence across competing algorithms. Overall, our estimates of dementia prevalence were largely concordant across algorithms, consistently pointing to a decline in rates between 2000 and 2016 (see Figure 1 ). In line with Langa et al. (2017) and Farina and colleagues (2022) , these results point to a general downward trend in the prevalence of dementia after age 70 in 2016 compared to 2000 across the competing algorithms ( Chen et al., 2019 ; Farina et al., 2022 ; Langa et al., 2017 ; Zhu et al., 2021 ). Group-specific estimates, however, suggest meaningful heterogeneity across algorithms. When stratified by race/ethnicity and nativity, our results indicate that the estimated dementia prevalence rates for the LW algorithm are lower than those of competing algorithms for older White adults, regardless of gender. In contrast, the prevalence estimates for older Black men were generally similar across algorithms. Notably, the Expert algorithm prevalence estimates for U.S.-born and foreign-born Latinos were substantially lower than those of competing algorithms. These trends were largely similar across the 2 considered periods of 2000 and 2016. The similarity in patterns over time suggests persistent differences rather than sampling or wave-specific artifacts. Specific to gender, the estimated dementia prevalence among Black women respondents was lower in the LW and Hurd algorithms than in other estimates. However, all algorithms generated largely equivalent estimates of dementia for both U.S.-born and foreign-born Latinas. Dementia-free life expectancy estimates across competing algorithms at age 70 Overall, we found modest but potentially meaningful group variations in dementia-free life expectancy across algorithms. Furthermore, the relative ranking of estimates from the considered subgroups differed by algorithm. Dementia-free life expectancy estimates were relatively stable across competing algorithms (see Table 1 , Figure 2 ), ranging from 0.1 years (U.S.-born Latinas) to 1.8 years (White women). The estimated number of dementia-free years (at age 70) was lowest for Black men in the LASSO model, U.S.-born and foreign-born Latinos in the Hurd model, and White men in the Expert model, and highest for Black and White men in the LW model, and U.S.-born and foreign-born Latinos in the Expert model. Overall, women had consistently larger estimates of dementia-free years lived (at age 70) compared to men across all algorithms (except for foreign-born Latinos in Expert and LASSO models), independent of race/ethnicity or nativity status. Estimates of life expectancy with dementia showed more pronounced algorithmic variability. The estimated number of years (at age 70) with dementia was highest for Black women (6.6 years) and both U.S.-born (5.8 years) and foreign-born Latinas (7.7 years) in the LASSO model and White women (4.1 years) in the Expert model. In contrast, the estimates for years with dementia were lowest for women in the LW model across all race/ethnic and nativity groups, particularly for White women (2.2 years). The size of the within-group disparities among all at-risk groups for both dementia-free life expectancy and life expectancy with dementia was the largest in the LW and LASSO models for women. When expressed as proportions of later life spent dementia-free (cognitively healthy), White men and women had the highest estimates, Black and U.S.-born Latino men and women exhibited broadly similar intermediate proportions, and foreign-born Latinas had the lowest. Overall, the LW algorithm tended to yield higher proportions of dementia-free life expectancy than alternative algorithms, highlighting that algorithmic choice can potentially influence group comparisons in cognitively healthy longevity and dementia burden. Specifically, our estimates indicate that under the LW algorithm, White men, followed by White women, have the highest proportions of dementia-free life expectancy beyond age 70 (0.91 and 0.87, respectively). Foreign-born Latino men had a relatively higher proportion of dementia-free life expectancy (0.77) compared to other Latino groups and Black men and women. Black and U.S.-born Latino men and women had largely equivalent proportions of dementia-free life expectancy, estimated at 0.70 for Black and Latina women and 0.73 and 0.72 for Black and U.S.-born Latino men, respectively. Foreign-born Latinas had the lowest estimated proportion of dementia-free life expectancy (0.68) compared to all other racial/ethnic and nativity groups considered. When comparing across competing algorithms, with a few exceptions (Expert model for U.S.-born Latino men and women, and foreign-born Latino men), we found that the LW algorithm estimated higher proportions of dementia-free life expectancy than the Expert, Hurd, and LASSO models, and consistently so across the considered at-risk groups. The relative difference in proportions of dementia-free life expectancy, however, varied across racial/ethnic and nativity groups, ranging from 11% for White women (0.76-0.87) and 9% among White men (0.82-0.91) to 10% for Black women (0.60-0.70) and 5% for Black men (0.68-0.73). Among Latinos, the range was 3% for U.S.-born Latinas (0.67-0.70), 9% for U.S.-born Latino men (0.68-0.77), 7% for foreign-born Latinas (0.61-0.68), and 9% for foreign-born Latino men (0.72-0.81). Discussion Algorithms for dementia classification are helpful for population-wide inference (e.g., the United States) and for examining temporal trends in rates and burden assessment. However, caution is warranted when treating these algorithms as interchangeable for subgroup-specific estimation. Although our analyses were limited to a subset of dementia classification algorithms available in the HRS, the observed heterogeneity in prevalence, dementia-free life expectancy, and dementia years burden suggests that similar issues are likely to arise with other dementia algorithms used in population health research. Stated differently, the inputs underlying algorithmic classification matter, and may need to be tailored, or at least carefully evaluated, when applied to specific sex, race/ethnic, and nativity subgroups, particularly in the absence of a uniformly applied research classification framework. In light of this, two sub-recommendations merit emphasis. First, algorithm choice should be recognized as a source of uncertainty in disparity-focused research, underscoring the need for continued methodological work toward convergence on a recognizable "gold standard" for research-based dementia classification. Second, researchers should explicitly describe the advantages and limitations of the algorithms they employ and clearly articulate how algorithm selection aligns with their research questions and the population under study. Saito, Robine, and Crimmins (2014) note that the proportion of healthy (i.e., dementia-free) life expectancy to total life expectancy is an important indicator of population health that should not be overlooked. Specifically, both length (i.e., the number of years) and proportion of dementia-free life expectancy should be used together to inform health policy because increases in both dementia-free life expectancy and total life expectancy (i.e., public health goals) could result in a decrease in the proportion of late-life years lived dementia-free, not indicative of actual improvements in AD/ADRD in the population ( Saito et al., 2014 ). Specific to our study, the proportion of healthy life expectancy is an important indicator to consider when comparing racial/ethnic, nativity, and gender disparities and the overall burden of AD/ADRD across groups. For example, although older White men and foreign-born Latinos spend similar lengths of time dementia-free (LW: 13.3 years vs. 13.2 years), White men (0.91) spend a greater proportion of their remaining years after age 70 dementia-free compared to foreign-born Latino men (0.77) despite having a lower total life expectancy (14.7 years for White men vs. 17.0 years for foreign-born Latino men). Therefore, we argue that in addition to examining subgroup patterns of AD/ADRD, health researchers should also go beyond using dementia-free years as a sole indicator of the overall burden of dementia among older at-risk adults, given differences in overall life expectancy between groups. Considering Latinos as a pan-ethnic group (i.e., not examining variation by nativity status) leads to a masking of lower dementia-free life expectancy for U.S.-born women and men, whereas models indicate that foreign-born older adults live more years dementia-free compared to their U.S.-born counterparts. However, across all modeling classifications, foreign-born women exhibit greater estimates of years lived with dementia and, thus, lower cognitively healthy ratios than U.S.-born women. This pattern is reversed for men. The disparity between U.S.-born and foreign-born Latinos would have been missed if nativity status were not explicitly assessed as a differentiating social status indicator of dementia. Thus, future research should aim to further understand the role of nativity status in cognitive health among Latinos, particularly identifying which biopsychosocial determinants may help explain the nuanced differences in dementia-free expectancies and cognitively healthy ratios for U.S.-born and foreign-born women. Furthermore, Latino heritage and country/region of origin should be examined to better contextualize how dementia status may vary across Latino population subgroups. While the algorithms compared in this study overall provide similar estimates of dementia life expectancies across racial/ethnic, gender, and nativity groups, some discrepancies require further interrogation. Specifically, compared to the Langa-Weir algorithm, other algorithms estimate living fewer years dementia-free at age 70, particularly for White, Black, and foreign-born Latino women. This corresponds to estimates that depict a worse state of dementia life expectancy than the Langa-Weir estimates. Therefore, the algorithm selected for assessing dementia status at the population level has implications for measuring racial/ethnic, nativity, and gender disparities as well as evaluating progress made by public health and clinical interventions. Limitations First, although these estimates help contextualize and guide future use of these competing algorithms in dementia disparities, our findings should be interpreted as descriptive, and future work should validate them by providing tests that allow inferential statements across competing algorithms (i.e., tests that allow comparisons across groups and algorithms). Second, algorithmic classifications of dementia are not clinical diagnoses. Future studies should incorporate clinical diagnoses of dementia, along with measures used in the HRS, to determine whether competing algorithmic approaches differ in their ability to accurately estimate dementia prevalence and dementia-free life expectancy, and to examine racial/ethnic and gender disparities in these outcomes. Last, the Sullivan method relies on prevalence rather than transition probabilities and therefore cannot capture the dynamic interplay between dementia onset and mortality. If subgroup sample sizes allow, incidence-based multi-state life tables are ideal for estimating health expectancy from a longitudinal survey. Conclusion Future research on dementia algorithms should expand to incorporate 1) data for other under-researched racial/ethnic groups, such as Asian and Middle Eastern and North African (MENA) populations, and 2) data from low- and middle-income countries, as many of these countries experience a greater burden of disease due to rapid population aging, with dementia prevalence projected to increase rapidly over the coming decades. Advancing the application of such algorithms is crucial to achieving accurate, population subgroup-specific estimates of dementia necessary for determining the cognitive status and needs of population subgroups. Supplementary Material 1 NIHMS2159044-supplement-1.pdf (600.1KB, pdf) Funding This work was supported by the National Institute on Aging of the National Institutes of Health [grant numbers L30AG074461, P30AG059298, and K02AG059140]. Footnotes Conflict of Interest The authors declare no conflict of interest. Data Availability This study uses data from the Health and Retirement Study (HRS) and contributed projects based on HRS public data provided by researchers who want to share their work with the research community. These data are publicly available and can be accessed in accordance with the access rules established by HRS at https://hrs.isr.umich.edu/data-products . This study also used the NHIS Linked Mortality File (LMF) Public Use Data available at https://nhis.ipums.org/nhis/aboutIPUMSNHIS.shtml . No preregistration was conducted for this study. References Alzheimer's Association. (2025). 2025 Alzheimer's disease facts and figures. Alzheimer's & Dementia, 21(4). 10.1002/alz.70235 [ DOI ] [ Google Scholar ] Bauldry S, Thomas PA, Sauerteig-Rolston MR, & Ferraro KF (2023). Racial–ethnic disparities in dual-function life expectancy. The Journals of Gerontology: Series A, 78(7), 1269–1275. 10.1093/gerona/glad059 [ DOI ] [ Google Scholar ] Blewett LA, Rivera Drew JA, King ML, & Williams KC (2019). IPUMS health surveys: National health interview survey. Minneapolis: IPUMS. [ Google Scholar ] Bugliari D, Carroll J, Hayden O, Hayes J, Hurd M, Karabatakis A, Main R, Marks J, McCullough C, & Meijer E (2020). RAND HRS longitudinal file 2016 (V2) documentation. Santa Monica, CA: RAND Center for the Study of Aging. [ Google Scholar ] Chen C, & Zissimopoulos JM (2018). Racial and ethnic differences in trends in dementia prevalence and risk factors in the United States. Alzheimer's & Dementia: Translational Research & Clinical Interventions, 4, 510–520. 10.1016/j.trci.2018.08.009 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Chen Y, Tysinger B, Crimmins E, & Zissimopoulos JM (2019). Analysis of dementia in the US population using Medicare claims: insights from linked survey and administrative claims data. Alzheimer's & Dementia: Translational Research & Clinical Interventions, 5, 197–207. 10.1016/j.trci.2019.04.003 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Crimmins EM, Kim JK, Langa KM, & Weir DR (2011). Assessment of cognition using surveys and neuropsychological assessment: the Health and Retirement Study and the Aging, Demographics, and Memory Study. Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 66(suppl_1), i162–i171. 10.1093/geronb/gbr048 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Crimmins EM, Saito Y, Kim JK, Zhang YS, Sasson I, & Hayward MD (2018). Educational differences in the prevalence of dementia and life expectancy with dementia: Changes from 2000 to 2010. The Journals of Gerontology: Series B, 73(suppl_1), S20–S28. 10.1093/geronb/gbx135 [ DOI ] [ Google Scholar ] Farina MP, Hayward MD, Kim JK, & Crimmins EM (2020). Racial and educational disparities in dementia and dementia-free life expectancy. The Journals of Gerontology: Series B, 75(7), e105–e112. 10.1093/geronb/gbz046 [ DOI ] [ Google Scholar ] Farina MP, Zhang YS, Kim JK, Hayward MD, & Crimmins EM (2022). Trends in dementia prevalence, incidence, and mortality in the United States (2000–2016). Journal of aging and health, 34(1), 100–108. 10.1177/08982643211029716 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Garcia MA, Diminich ED, Lu P, Arévalo SP, Sayed L, Abdelrahim R, & Ajrouch KJ (2023). Caregiving for foreign-born older adults with dementia. The Journals of Gerontology: Series B, 78(Supplement_1), S4–S14. 10.1093/geronb/gbac153 [ DOI ] [ Google Scholar ] Garcia MA, Downer B, Chiu C-T, Saenz JL, Ortiz K, & Wong R (2021). Educational benefits and cognitive health life expectancies: Racial/ethnic, nativity, and gender disparities. The Gerontologist, 61(3), 330–340. https://doi.org/Educational benefits and cognitive health life expectancies: Racial/ethnic, nativity, and gender disparities [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Garcia MA, Downer B, Chiu C-T, Saenz JL, Rote S, & Wong R (2019). Racial/ethnic and nativity differences in cognitive life expectancies among older adults in the United States. The Gerontologist, 59(2), 281–289. 10.1093/geront/gnx142 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Gianattasio KZ, Ciarleglio A, & Power MC (2020). Development of algorithmic dementia ascertainment for racial/ethnic disparities research in the US Health and Retirement Study. Epidemiology, 31(1), 126–133. DOI: 10.1097/EDE.0000000000001101 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Gianattasio KZ, Prather C, Glymour MM, Ciarleglio A, & Power MC (2019). Racial disparities and temporal trends in dementia misdiagnosis risk in the United States. Alzheimer's & Dementia: Translational Research & Clinical Interventions, 5, 891–898. 10.1016/j.trci.2019.11.008 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Gianattasio KZ, Wu Q, Glymour MM, & Power MC (2019). Comparison of methods for algorithmic classification of dementia status in the health and retirement study. Epidemiology, 30(2), 291–302. DOI: 10.1097/EDE.0000000000000945 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Hale JM, Schneider DC, Mehta NK, & Myrskylä M (2020). Cognitive impairment in the US: Lifetime risk, age at onset, and years impaired. SSM-Population Health, 11, 100577. 10.1016/j.ssmph.2020.100577 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Hayward MD, Hummer RA, Chiu C-T, González-González C, & Wong R (2014). Does the Hispanic Paradox in U.S. Adult Mortality Extend to Disability? Population Research and Policy Review, 33(1), 81–96. 10.1007/s11113-013-9312-7 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Hudomiet P, Hurd MD, & Rohwedder S (2022). Trends in inequalities in the prevalence of dementia in the United States. Proceedings of the National Academy of Sciences, 119(46), e2212205119. 10.1073/pnas.2212205119 [ DOI ] [ Google Scholar ] Hurd MD, Martorell P, Delavande A, Mullen KJ, & Langa KM (2013). Monetary costs of dementia in the United States. New England Journal of Medicine, 368(14), 1326–1334. DOI: 10.1056/NEJMsa1204629 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Langa KM (2020). Langa-Weir classification of cognitive function (1995 Onward). Survey Research Center Institute for Social Research, University of Michigan. [ Google Scholar ] Langa KM, Larson EB, Crimmins EM, Faul JD, Levine DA, Kabeto MU, & Weir DR (2017). A comparison of the prevalence of dementia in the United States in 2000 and 2012. JAMA internal medicine, 177(1), 51–58. doi: 10.1001/jamainternmed.2016.6807 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Langa KM, Plassman BL, Wallace RB, Herzog AR, Heeringa SG, Ofstedal MB, Burke JR, Fisher GG, Fultz NH, & Hurd MD (2005). The Aging, Demographics, and Memory Study: study design and methods. Neuroepidemiology, 25(4), 181–191. 10.1159/000087448 [ DOI ] [ PubMed ] [ Google Scholar ] Lin Z, & Liu H (2024). Race/ethnicity, nativity, and gender disparities in unmet care needs among older adults in the United States. The Gerontologist, 64(4), gnad094. 10.1093/geront/gnad094 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Park S, & Chen J (2020). Racial and ethnic patterns and differences in health care expenditures among Medicare beneficiaries with and without cognitive deficits or Alzheimer’s disease and related dementias. BMC geriatrics, 20(1), 482. 10.1186/s12877-020-01888-y [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Rocca WA, Petersen RC, Knopman DS, Hebert LE, Evans DA, Hall KS, Gao S, Unverzagt FW, Langa KM, & Larson EB (2011). Trends in the incidence and prevalence of Alzheimer’s disease, dementia, and cognitive impairment in the United States. Alzheimer's & Dementia, 7(1), 80–93. 10.1016/j.jalz.2010.11.002 [ DOI ] [ Google Scholar ] Saadi A, Himmelstein DU, Woolhandler S, & Mejia NI (2017). Racial disparities in neurologic health care access and utilization in the United States. Neurology, 88(24), 2268–2275. 10.1212/WNL.0000000000004025 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Saito Y, Robine J-M, & Crimmins EM (2014). The methods and materials of health expectancy. Statistical Journal of the IAOS, 30(3), 209–223. DOI: 10.3233/SJI-140840 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Wu Q, Tchetgen EJT, Osypuk TL, White K, Mujahid M, & Glymour MM (2013). Combining direct and proxy assessments to reduce attrition bias in a longitudinal study. Alzheimer Disease & Associated Disorders, 27(3), 207–212. DOI: 10.1097/WAD.0b013e31826cfe90 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Zhu Y, Chen Y, Crimmins EM, & Zissimopoulos JM (2021). Sex, race, and age differences in prevalence of dementia in Medicare claims and survey data. The Journals of Gerontology: Series B, 76(3), 596–606. 10.1093/geronb/gbaa083 [ DOI ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials 1 NIHMS2159044-supplement-1.pdf (600.1KB, pdf) Data Availability Statement This study uses data from the Health and Retirement Study (HRS) and contributed projects based on HRS public data provided by researchers who want to share their work with the research community. These data are publicly available and can be accessed in accordance with the access rules established by HRS at https://hrs.isr.umich.edu/data-products . This study also used the NHIS Linked Mortality File (LMF) Public Use Data available at https://nhis.ipums.org/nhis/aboutIPUMSNHIS.shtml . No preregistration was conducted for this study. ACTIONS View on publisher site PDF (745.3 KB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top