Background: Coffee intake has been linked to improved cardiometabolic (CM) health, yet underlying microbial and metabolic pathways remain unclear. Methods: At HCHS/SOL Visit 1 (2008-11), usual coffee intake was estimated from two 24-hour dietary recalls and food propensity questionnaires among 13634 adults. Serum metabolomics (653 metabolites) were profiled by LC-MS at Visit 1 (n=6180) and Visit 2 (2014-17; n=718). Gut microbiome was characterized by shotgun metagenomics at Visit 2 (n=2758). CM traits, including adiposity, glycemic, lipid, blood pressure, kidney, and liver markers, were measured at both visits. Among 10730 participants free of diabetes, CVD, and cancer at baseline, 2493 diabetes cases occurred during a median follow-up of 10.2 years. Results: Higher coffee intake was correlated with lower levels of waist-to-hip ratio, liver enzymes (AST, ALT, GGT), glucose/insulin indices (2-h OGTT glucose, fasting insulin, HOMA-IR, HOMA-b), TG, and blood pressure, but higher LDLC levels (all P <.05). Each additional serving/d of coffee intake was related to an HR of 0.89 (95% CI 0.80-0.98; Fig1 ) for diabetes. Coffee intake was positively related to gut microbial α-diversity and four species ( e.g., Clostridium phoceensis ), while inversely related to 27 species, including pro-inflammatory Ruminococcus gnavus , Clostridium glycyrrhizinilyticum , and Blautia hansenii , and opportunistic Klebsiella oxytoca ( q <.05), which were linked to unfavorable CM traits ( Fig2 ). We identified 165 coffee-related and 154 microbiota-related metabolites, with 55 overlapping. These 264 metabolites were clustered into 18 modules using WGCNA, one of which was mainly composed of caffeine metabolites and positively correlated with coffee intake and Clostridium phoceensis . A sphingomyelin module was positively related to coffee intake and lower diabetes risk (HR=0.88 [0.82-0.95] per SD increase in the module; Fig3 ). Modules of branched-chain amino acid (BCAA) and microbial bile acid metabolites were inversely related to coffee intake and higher diabetes risk, with HRs of 1.34 (1.24-1.45) and 1.07 (1.00-1.14), respectively. Glycerophospholipid module was positively related to Ruminococcus gnavus (inversely related to coffee intake) and an HR of 1.17 (1.09-1.26) for diabetes. Conclusions: Coffee intake was associated with gut microbial and metabolomics signatures converging on lipid, BCAA, and bile acid metabolism, and these pathways may illuminate the biological basis of coffee’s CM benefits.

The identification of novel blood-based biomarkers of small vessel disease of the brain (SVD) may improve pathophysiologic understanding and inform the development of new therapeutic strategies for prevention. We evaluated plasma proteomic associations of white matter fractional anisotropy (WMFA), white matter hyperintensity (WMH) volume, enlarged perivascular space (ePVS) volume, and the presence of microbleeds (MB) on brain magnetic resonance imaging (MRI) in the population-based Multi-Ethnic Study of Atherosclerosis (MESA). Eligible MESA participants had 2941 plasma proteins measured from stored blood samples (collected in 2016-2018) using the antibody-based Olink proteomics platform, and completed brain MRI scans in 2018-2019. Participants with quality control exclusion of protein measurements, missing covariate data, and poor quality or missing MRI outcome variables were excluded. The cross-sectional association between the abundance of each plasma protein (normalized protein expression – a relative protein quantification unit measured on a log 2 scale) was modeled separately with WMFA, WMH volume, total ePVS volume, and the presence of MBs using multivariable linear or modified Poisson regression, adjusting for demographic variables, estimated glomerular filtration rate (eGFR), and SVD risk factors. The Benjamini-Hochberg procedure was used to control the false discovery rate (FDR) <0.05 to account for multiple hypothesis testing. For proteins independently associated with the SVD markers on MRI, penalized regression with least absolute shrinkage and selection operator (LASSO) was used to create a parsimonious proteomic model. Eligible participants (total N=709) had a mean age of 73 years, 53% were women, 25% were Black, 17% were Chinese, 19% were Hispanic or Latino, and 39% were White (Table 1). After adjustment for demographics, eGFR, and SVD risk factors, 769 plasma proteins were associated with WMFA (FDR <0.05) (Figure). LASSO regression identified a 37-protein model predictive of WMFA (Table 2). We did not find plasma proteins to be independently associated with WMH volume, ePVS volume, or the presence of MBs. Multiple circulating proteins – implicated in central nervous system myelination, lipid metabolism, angiogenesis, coagulation, cellular adhesion and migration, appetite regulation, energy homeostasis, systemic inflammation, and immune regulation – were independently associated with WMFA in a multi-ethnic cohort of older adults.

Background: Plant-based diets are associated with a lower risk of cardiovascular disease (CVD). Proteomics may improve our understanding of the biological pathways underlying these associations. Objectives: Using large-scale proteomics, we aimed to examine if plant-based diet-related proteins, which have been previously identified, are associated with incident CVD and subtypes of CVD in the Atherosclerosis Risk in Communities (ARIC) Study and Framingham Heart Study (FHS) Offspring cohort. Methods: Discovery analyses were based on 9078 participants free of CVD at ARIC visit 3 (1993-1995). Cox proportional hazards regression was used to evaluate the associations between plant-based diet-related proteins and incident CVD, coronary artery disease, heart failure, and stroke. Replication analyses were based on 1279 participants without CVD in the FHS Offspring cohort. Results: In the ARIC study, over a median follow-up of 21 y, there were 3167 CVD events. At a false discovery rate <0.05, 26 of 73 plant-based diet-related proteins were significantly associated with incident CVD, after adjusting for important confounders. Eighteen, 1, and 0 proteins were associated with heart failure, stroke, and coronary artery disease, respectively. Three and 2 additional proteins were associated with CVD and heart failure risk in the FHS Offspring cohort at the nominal threshold (P < 0.05). In the ARIC Study and FHS Offspring cohort, soluble advanced glycosylation end product-specific receptor was inversely associated with incident CVD whereas thrombospondin-2 (THBS2) and N-terminal pro-BNP was positively associated with incident CVD. THBS2 was positively associated with incident heart failure, whereas neuronal growth factor regulator 1 and insulin-like growth factor-binding protein 1 was inversely associated. Conclusions: These proteins highlight several pathways that could explain plant-based diets-CVD associations.

BACKGROUND:Limited data exist on associations of immune cell subsets with longitudinal changes in subclinical coronary artery disease. METHODS:In the MESA (Multi-Ethnic Study of Atherosclerosis) study, we used a case-cohort approach to explore associations of 28 immune cell subsets measured at baseline (2000-2002) with longitudinal changes in coronary artery calcium (CAC). We examined incident CAC from examination 2 (2002-2004) through examination 5 (2010-2012) in participants with 0 CAC at baseline using multivariable-adjusted Cox regression. In participants with CAC >0 at baseline, we analyzed changes in CAC through examination 5 using multivariable-adjusted linear mixed models. Because no studies have investigated immune cells and longitudinal CAC changes, analyses were considered exploratory, with P<0.05 as the threshold for possible significance. RESULTS:Of 975 participants with immune cells subsets and CAC measurements at baseline, 378 had CAC 0 at baseline (mean age, 58.4; 37.0% men) and 597 had CAC >0 at baseline (mean age, 65.7; 57.6% men). Natural killer cells were associated with higher incident CAC (hazard ratio [HR], 1.26 per SD higher natural killer cell proportion; P=0.03), whereas T helper type cells were associated with lower incident CAC (HR, 0.81; P=0.04). B cells were associated with CAC progression (β=53.1 Agatston units per SD higher B-cell proportion, P=0.04), whereas CD14+CD16+ monocytes (β=-71.6; P=0.03) and T regulatory cells (β=-61.9; P=0.03) were associated with lower CAC progression. CONCLUSIONS:Natural killer cells may be associated with incident CAC and T regulatory cells may be associated with attenuated CAC progression, among other findings. These warrant replication and experimental investigation.

Background Lipoprotein(a) [Lp(a)] is associated with atherosclerotic cardiovascular disease. An Lp(a) threshold of ≥125 nmol/L is commonly used to identify individuals at higher risk for events, but there is a paucity of data on individuals of Hispanic/Latino descent. Objectives The purpose of this study was to provide a comprehensive evaluation of Lp(a) and its association with 10-year cardiovascular disease risk and mortality among Hispanic/Latino adults in the United States. Methods We evaluated the association between Lp(a) and myocardial infarction (MI), ischemic stroke, and all-cause mortality among 16,117 Hispanic Community Health Study/Study of Latinos individuals. Event rates were compared across Lp(a) quintiles. Multivariable Cox proportional hazards models assessed the relationship between events and Lp(a) across increasing quintiles, log-transformed Lp(a), and ≥125 nmol/L vs <125 nmol/L. Sampling weights and survey methods were used to account for the stratified probability sampling of the cohort. Results Among the Hispanic Community Health Study/Study of Latinos target population (median age 41.1 years, 52.4% women), the median Lp(a) was 19.7 nmol/L (Q1-Q3: 7.3-60.6 nmol/L), with 11.4% having Lp(a) ≥125 nmol/L, and the highest Lp(a) quintile defined as >77 nmol/L. Over a median follow-up of 9.8 years, 883 events (135 MI, 99 stroke, 649 all-cause mortality) occurred. The age-adjusted incidence rate of the composite events (MI, stroke, and all-cause mortality) was 505.2 per 100,000 person-years. After multivariable adjustment, each 1-SD increase in log-transformed Lp(a) was associated with a higher risk of MI (HR: 1.47; 95% CI: 1.14-1.89). Compared with Lp(a) <125 nmol/L, elevated Lp(a) ≥125 nmol/L conferred an increased risk of MI (HR: 2.29; 95% CI: 1.45-3.63), all-cause mortality (HR: 1.43; 95% CI: 1.05-1.93), and composite events (HR: 1.56; 95% CI: 1.22-2.01), but not stroke. Findings were consistent when comparing the highest Lp(a) quintile to the lower 4 quintiles, but the elevated risk was observed only for MI and composite events. Conclusions Hispanic/Latino individuals with elevated Lp(a) are at an increased risk of MI and all-cause mortality. Although Lp(a) ≥125 nmol/L is a valid risk threshold, Hispanics/Latinos show a continuous relationship between increasing Lp(a) levels and MI risk.

Introduction: Abnormalities in the function and structure of the left atrium, called atrial cardiopathy, are a precursor of atrial fibrillation (AF) and an important risk factor for several other cardiovascular outcomes, yet detecting abnormalities is challenging due to the cost and limited accessibility of high-quality cardiac imaging. Objective: To develop and validate a deep learning model (ECG-AI) that estimates left atrial (LA) structure and function from the resting 12-lead electrocardiogram and reliably predicts cardiovascular outcomes. Methods: We trained ECG-AI models on cardiac magnetic resonance imaging data from the UK Biobank (n=21,749) to estimate LA minimum volumes, maximum volumes and ejection fraction. Volumes were indexed to body surface area. Cox regression models adjusting for clinical risk factors evaluated associations of LA with incident AF, ischemic stroke and cardioembolic stroke in an external cohort of adults aged ≥ 65, the Cardiovascular Health Study. We compared prediction models for incident AF and cardioembolic stroke that included [1] age and sex only (base), [2] base + ECG-AI LA measures (ECG-AI), [3] CHARGE-AF risk score alone, and [4] CHARGE-AF + ECG-AI (combined). Results: ECG-AI estimates were moderately correlated with direct imaging measures (r=0.40-0.50) but demonstrated strong independent associations with outcomes in the external cohort and outperformed a conventional ECG measure of atrial cardiopathy, P terminal force in V1. In the Cardiovascular Health Study, each standard deviation increase in ECG-AI LA minimum volume yielded HRs of 1.44 (95% CI 1.35-1.47) for AF, 1.45 (95% CI 1.37-1.53) for ischemic stroke, and 1.66 (95% CI 1.47-1.86) for cardioembolic stroke, the hallmark complication of AF and atrial cardiopathy (Figure 1). In contrast, none of the ECG-AI measures was associated with large-artery atherosclerotic or small vessel stroke. In 5-year prediction models (Figure 2), the ECG-AI measures outperformed the CHARGE-AF risk prediction tool for both AF (delta AUC 0.03, 95% CI 0.01-0.05) and cardioembolic stroke (delta AUC 0.01, 95% CI -0.05-0.08). Conclusion: Our fully-trained ECG-AI tool estimates LA function and identifies individuals at elevated-risk for cardiovascular disease using only standard data from the inexpensive and widely-available 12-lead ECG.

OBJECTIVE:Personalized approaches to ischemic stroke diagnosis are needed. We determine differences in proteomic signatures of incident embolic (EIS) and thrombotic stroke (TIS) by age and resultant pathways using large-scale proteomics. METHODS:Participants in the Atherosclerosis Risk in Communities Study (ARIC) from visit 2 (V2, 1990-1992) until 2020 without prevalent stroke with available SomaScan data (4,955 protein targets) at V2 (mid-life, n = 10,929), and then again at visit 5 (V5, 2011-2013, n = 4,463) were included. Covariate adjusted Cox hazard models determined the association between proteins, and adjudicated incident EIS or TIS from V2 to V5 and from V5 to 2020. RESULTS:Among 10,929 participants (56% female, 23% Black, follow-up ~20 years), 20 proteins measured in mid-life were associated with either EIS (n = 168) or TIS (n = 459) in mid-life, and 4 measured in late-life were associated with late-life stroke (73 EIS and 124 TIS events) at the Bonferroni threshold p < 1E-5. In mid-life, N-terminal pro-B-type natriuretic peptide (NPPB) was significantly associated with EIS, but not TIS (p-difference = 9.14E-7). Nineteen mid-life proteins were strongly associated with TIS; 7 strongly associated with TIS and only nominally (p < 0.05) with EIS and the remaining 12 with TIS only. In late-life, NPPB, serine protease inhibitor Kazal-type 4, oligodendrocyte-myelin-glycoprotein, and neurocan-core protein were significantly associated with EIS, but not TIS. Ingenuity Pathway Analysis tools implicated cancer for EIS-associated proteins, whereas TIS pathways reflected cell-structure and atherogenesis. INTERPRETATION:We identified plasma proteins associated with risk of EIS versus TIS reflecting distinct stroke mechanisms: cardiac dysfunction protein in EIS (eg, NPPB) and inflammation dysregulation in TIS (eg, interleukins). ANN NEUROL 2025;98:1125-1135.

With age, hematopoietic stem cells can acquire somatic mutations in leukemogenic genes that confer a proliferative advantage in a phenomenon termed "clonal hematopoiesis of indeterminate potential" (CHIP). How these mutations confer a proliferative advantage and result in increased risk for numerous age-related diseases remains poorly understood. We conducted a multiracial meta-analysis of epigenome-wide association studies (EWAS) of CHIP and its subtypes in four cohorts (N=8196) to elucidate the molecular mechanisms underlying CHIP and illuminate how these changes influence cardiovascular disease risk. The EWAS findings were functionally validated using human hematopoietic stem cell (HSC) models of CHIP. A total of 9615 CpGs were associated with any CHIP, 5990 with DNMT3A CHIP, 5633 with TET2 CHIP, and 6078 with ASXL1 CHIP (P <1×10-7). CpGs associated with CHIP subtypes overlapped moderately, and the genome-wide DNA methylation directions of effect were opposite for TET2 and DNMT3A CHIP, consistent with their opposing effects on global DNA methylation. There was high directional concordance between the CpGs shared from the meta-EWAS and human edited CHIP HSCs. Expression quantitative trait methylation analysis further identified transcriptomic changes associated with CHIP-associated CpGs. Causal inference analyses revealed 261 CHIP-associated CpGs associated with cardiovascular traits and all-cause mortality (FDR adjusted p-value <0.05). Taken together, our study sheds light on the epigenetic changes impacted by CHIP and their associations with age-related disease outcomes. The novel genes and pathways linked to the epigenetic features of CHIP may serve as therapeutic targets for preventing or treating CHIP-mediated diseases.

Background Limited data exist regarding risk factors for aortic stenosis (AS). The plasma proteome is a promising phenotype for discovery of novel biomarkers and potentially causative mechanisms. Objectives The aim of this study was to discover novel biomarkers with potentially causal associations with AS. Methods We measured 4,877 plasma proteins (SomaScan aptamer-affinity assay) among ARIC (Atherosclerosis Risk In Communities) study participants in mid-life (visit 3 [V3]; n = 11,430; age 60 +/- 6 years) and in late-life (V5; n = 4,899; age 76 +/- 5 years). We identified proteins cross-sectionally associated with aortic valve (AV) peak velocity (AVmax) and dimensionless index by echocardiography at V5 and with incident AV-related hospitalization after V3 with the use of multivariable linear and Cox proportional hazard regression. We assessed associations of candidate proteins with changes in AVmax over 6 years and with AV calcification with the use of cardiac computed tomography, replicated analysis in an independent sample, performed Mendelian randomization, and evaluated gene expression in explanted human AV tissue. Results Fifty-two proteins cross-sectionally were associated with AVmax and dimensionless index at V5 and with risk of incident AV-related hospitalization after V3. Among 3,413 participants in the Cardiovascular Health Study, 6 of those proteins were significantly associated with adjudicated moderate or severe AS, including matrix metalloproteinase 12 (MMP12), complement C1q tumor necrosis factor-related protein 1 (C1QTNF1), and growth differentiation factor-15. MMP12 was also associated with greater increase in AVmax over 6 years, greater degree of AV calcification, and greater expression in calcific compared with normal or fibrotic AV tissue. C1QTNF1 had consistent potential causal effects on both AS and AVmax according to Mendelian randomization analysis. Conclusions These findings identify MMP12 as a potential novel circulating biomarker of AS risk and C1QTNF1 as a new putative target to prevent AS progression.

BACKGROUND AND AIMS:Despite mechanistic data implicating unresolving inflammation in stroke pathogenesis, data regarding circulating immune cell phenotypes - key determinants of inflammation propagation versus resolution - and incident stroke are lacking. Therefore, we aimed to comprehensively define associations of circulating immune phenotypes and activation profiles with incident stroke.METHODS:We investigated circulating leukocyte phenotypes and activation profiles with incident adjudicated stroke in 2104 diverse adults from the Multi-Ethnic Study of Atherosclerosis (MESA) followed over a median of 16.6 years. Cryopreserved cells from the MESA baseline examination were thawed and myeloid and lymphoid lineage cell subsets were measured using polychromatic flow cytometry and intracellular cytokine activation staining. We analyzed multivariable-adjusted associations of cell phenotypes, as a proportion of parent cell subsets, with incident stroke (overall) and ischemic stroke using Cox regression models.RESULTS:We observed associations of intermediate monocytes, early-activated CD4+ T cells, and both CD4+ and CD8+ T cells producing interleukin-4 after cytokine stimulation (Th2 and Tc2, respectively) with higher risk for incident stroke; effect sizes ranged from 35% to 62% relative increases in risk for stroke. Meanwhile, differentiated and memory T cell phenotypes were associated with lower risk for incident stroke. In sex-stratified analyses, positive and negative associations were especially strong among men but null among women.CONCLUSIONS:Circulating IL-4 producing T cells and intermediate monocytes were significantly associated with incident stroke over nearly two decades of follow-up. These associations were stronger among men and not among women. Further translational studies are warranted to define more precise targets for prognosis and intervention.

Purpose: High blood pressure (BP) is a well-known risk factor for atrial fibrillation (AF), but a single BP measurement may provide limited information about AF risk in older adults. We evaluated whether longitudinal measures of BP were associated with subclinical arrhythmias after adjustment for more conventional BP measures such as a single cross-sectional BP value. Methods: This study included 1,256 Multi-Ethnic Study of Atherosclerosis (MESA) and 1,948 Atherosclerosis Risk in Communities study (ARIC) participants who underwent extended ambulatory electrocardiographic monitoring and who were free of clinically-detected cardiovascular disease, including AF. Using BP measurements from six exams (2000-2018 in MESA, 1987-2017 in ARIC), we examined cross-sectional BP at the most recent exam, individual long-term mean BP from exams 1-5, BP trend during exams 1-5, and visit-to-visit variability during exams 1-5 in systolic BP (SBP) and pulse pressure (PP) for each participant. Logistic regression models were used to examine BP exposures in relation to subclinical AF and linear regression models were used to examine BP exposures in relation to frequency of supraventricular ectopy. Models adjusted for participant demographics, height, weight, diabetes, antihypertensive medication use and cross-sectional BP. Results from each study were combined with inverse variance-weighted meta-analysis. Results: At Exam 6, the mean age was 73 years in MESA and 79 years in ARIC, and 4% had subclinical AF. A 10 mmHg higher cross-sectional SBP and a 10 mmHg higher cross-sectional PP were associated with less AF (odds ratio [OR] SBP 0.86, 95% confidence interval [CI] 0.76, 0.96 and OR PP 0.73, 95% CI 0.63, 0.84). In contrast, a 4 mmHg higher visit-to-visit variability in SBP was associated with a greater prevalence of subclinical AF (OR 1.20, 95% CI 1.02, 1.38) and with a greater frequency of premature atrial contractions/hour (8%, 95% CI 1%, 15%). For PP as well, a 4 mmHg higher visit-to-visit variability was associated with a greater prevalence of AF (OR 1.18, 95% CI 1.00-1.37). In addition, a 10 mmHg higher long-term mean PP was associated with a greater prevalence of subclinical AF (OR 1.36, 95% CI 1.08-1.70). Conclusion: Our results indicate that information on BP assessed longitudinally for several years, especially visit-to-visit BP variability, is associated with a greater prevalence of subclinical atrial arrhythmias, while cross-sectional BP values alone were not associated with a greater prevalence of subclinical atrial arrhythmias. Prior longitudinal BP assessment, rather than current BP, may be more helpful in identifying older adults who are at higher risk of atrial arrhythmias.

Purpose: Opioids and gabapentinoids may have adverse cardiovascular effects. We evaluated whether these medications were associated with incident clinically-detected atrial fibrillation (AF) or monitor-detected supraventricular ectopy (SVE), including premature atrial contractions (PACs) and supraventricular tachycardia (SVT). Methods: We used data from the Multi-Ethnic Study of Atherosclerosis (MESA), a cohort study that enrolled 6,814 Americans without clinically-detected cardiovascular disease (CVD) in 2000-2002. At the 2016-2018 examination, a subset of participants received extended ambulatory electrocardiographic (ECG) monitoring with a device that records up to 14 days of continuous data. Longitudinal analyses investigated time-varying medication exposures at 5 exams (through 2012) and the risk of incident clinically-detected AF through 2015 using Cox proportional hazards regression models. Cross-sectional analyses investigated medication exposures at the 2016-2018 examination and the risk of monitor-detected SVE using linear and logistic regression models. Methods: The longitudinal cohort included 6,652 participants. Opioid and gabapentenoid use increased over time (Figure). During 12.4 years of mean follow-up, 961 participants (14.4%) experienced incident AF. Opioid use and gabapentinoid use were not associated with the risk of incident AF compared with no use. The cross-sectional analysis included 1,435 participants with ECG monitoring. Compared with non-use, gabapentinoid use was associated with an 84% greater count of PACs/hour (95% CI, 25%-171%) and with a 44% greater average number of runs of SVT/day (95% CI, 3%-100%) but not with a higher odds of SVT. No associations were found with use of opioids in cross-sectional analyses. Conclusions: In this study, gabapentinoid use was associated with more SVE. Given the rapid increase in gabapentinoid use, additional studies are needed to clarify whether these medications increase the risk of CVD complications.

Background: Epidemiologic studies often use diagnosis codes to identify dementia outcomes. It remains unknown to what extent cognitive screening test results add value in identifying dementia cases in big data studies leveraging electronic health record (EHR) data. We examined test scores from EHR data and compared results with dementia algorithms. Methods: This retrospective cohort study included patients 60+ years of age from Kaiser Permanente Washington (KPWA) during 2013-2018 and the Veterans Health Affairs (VHA) during 2012-2015. Results from the Mini Mental State Examination (MMSE) and the Saint Louis University Mental Status Examination (SLUMS) cognitive screening exams, were classified as showing dementia or not. Multiple dementia algorithms were created using combinations of diagnosis codes, pharmacy records, and specialty care visits. Correlations between test scores and algorithms were assessed. Results: 3,690 of 112,917 KPWA patients and 2,981 of 102,981 VHA patients had cognitive test results in the EHR. In KPWA, dementia prevalence ranged from 6.4%-8.1% depending on the algorithm used and in the VHA, 8.9%-12.1%. The algorithm which best agreed with test scores required >= 2 dementia diagnosis codes in 12 months; at KPWA, 14.8% of people meeting this algorithm had an MMSE score, of whom 65% had a score indicating dementia. Within VHA, those figures were 6.2% and 77% respectively. Conclusions: Although cognitive test results were rarely available, agreement was good with algorithms requiring >= 2 dementia diagnosis codes, supporting the accuracy of this algorithm.