Abstract
Background
Guidelines emphasize using atherosclerotic cardiovascular disease (ASCVD) risk prediction models for treatment decisions, but risk of cognitive impairment is an equally important concern in older adults. Current ASCVD risk prediction models were derived in younger adults and do not include holistic measures of health or predict cognitive impairment.
Methods
We utilized data from the Framingham, Framingham Offspring, CHS (Cardiovascular Health Study), and ARIC (Atherosclerosis Risk in Communities) cohorts to derive and validate 2 Selective Functional Prediction models to estimate an older person's (aged ≥75 years) risk within 5 years of developing incident: (1) cognitive impairment; and (2) ASCVD, while accounting for the competing risk of death. Variable selection, including functional status, was based on the least absolute shrinkage and selection operator method. The cognitive impairment (N=3466) and ASCVD (N=4403) model populations were split into derivation and validation cohorts with external validation, then performed in MESA (Multi‐Ethnic Study of Atherosclerosis).
Results
In the derivation and external validation cohorts (median age, 79 years), 579 (16.7%) and 67 (15.3%) participants developed incident cognitive impairment, respectively; 748 (17.0%) and 80 (8.4%), respectively, experienced an ASCVD event. The cognitive impairment model (baseline Mini‐Mental State Examination (MMSE), atrial fibrillation, antidepressant use, mobility impairment, and dependence for grocery shopping) had good discrimination in the internal and external validation cohorts (C index 0.75 and 0.73, respectively). The ASCVD model (employment status, MMSE, aspirin, lipid‐lowering medications, blood pressure medications, systolic blood pressure, general health status, high‐density lipoprotein cholesterol, triglycerides, creatinine, and mobility impairment) had satisfactory discrimination (C index 0.67) on internal validation and outperformed the pooled cohort equations, but had modest discrimination (C index 0.59) on external validation. Although both models were well calibrated in the internal validation cohorts, they overpredicted risk in the external validation cohort.
Conclusions
Accurate prediction of an older person's risk of developing cognitive impairment is possible, but predicting future ASCVD events remains more challenging.
Keywords: cardiovascular disease, cognitive impairment, geriatric, older adults, prevention, risk prediction
Subject Categories: Primary Prevention, Cardiovascular Disease, Aging, Risk Factors
Nonstandard Abbreviations and Acronyms
- ARIC
Atherosclerosis Risk in Communities
- BioLINCC
Biologic Specimen and Data Repository Information Coordinating Center
- CASI
Cognitive Abilities Screening Instrument
- CHS
Cardiovascular Health Study
- LASSO
least absolute shrinkage and selection operator
- MESA
Multi‐Ethnic Study of Atherosclerosis
- PCE
pooled cohort equation
- PREVENTABLE
Pragmatic Evaluation of Events and Benefits of Lipid‐Lowering in Older Adults
- SENIOR
selective functional prediction
- SPRINT
Systolic Blood Pressure Intervention Trial
Clinical Perspective.
What Is New?
This study introduces holistic prediction models that integrate cardiovascular and cognitive risk factors to better predict adverse outcomes in older adults.
What Are the Clinical Implications?
The combined assessment of cognitive function, physical function, and traditional cardiovascular markers allows for more accurate identification of high‐risk older patients.
Tailored preventive strategies targeting both cognitive and cardiovascular domains could improve long‐term patient outcomes and reduce the incidence of cognitive impairment and atherosclerotic cardiovascular disease–related events.
As lifespan increases, preserving quality of life for older adults (aged ≥75 years) with effective preventive strategies has taken on greater individual and global importance. At the age of 75, the average US adult has a 12‐year lifespan, which can be characterized by variably good health and independence or health declines and disability. 1 Metrics are needed to guide clinicians to better manage the risk factors that contribute to this heterogeneity but prior US guidelines for blood pressure (BP) and cholesterol management in older adults, which utilize the pooled cohort equations (PCEs), 2 , 3 have poor discrimination and calibration when applied to adults 75 years or older and were not derived for use in adults older than 80 years. 4 , 5 Most traditional risk assessment approaches developed in younger cohorts do not account for geriatric factors such as functional and cognitive status or competing risks, which contribute to suboptimal performance. 6
The American Geriatrics Society Guiding Principles for the Care of Older Adults With Multimorbidity statement identifies the development of new prognostic measures for older adults as a key area for future investigation. While atherosclerotic cardiovascular disease (ASCVD) is the number one cause of death among older adults, 7 , 8 cognitive impairment is the number one cause of disability. 9 , 10 Older adults want to live longer but also prioritize remaining cognitively independent. 11 While both cognitive impairment and ASCVD are associated with their own unique risk factors, many risk factors overlap between the 2 conditions. 12 , 13 , 14 Moreover, nontraditional risk factors including measures of mobility, function, and cognition have been identified as potential predictors of future health risk in older persons. 4 , 5 , 15 , 16 Consequently, the development of separate but complementary risk estimation tools to predict cognitive impairment and ASCVD could guide prevention decisions in older persons.
Thus, we describe the derivation and validation of the selective functional prediction (SENIOR) risk models in older adults (aged ≥75 years). These models predict the risk over 5 years of incident cognitive impairment and ASCVD, respectively. We used pooled data from several National Heart Lung, and Blood Institute (NHLBI) cohorts to study a large population free of cognitive impairment and ASCVD at baseline. We included certain key measures of function as candidate predictors, such as mobility, function, and cognitive status, which had not been considered in previous risk models. We externally validated both models in the MESA (Multi‐Ethnic Study of Atherosclerosis) cohort.
METHODS
Study Population
We pooled data from NHLBI‐sponsored cohort studies obtained from the NHLBI Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC), including all participants 75 years and older free of cognitive impairment and ASCVD at the initial visit used for this analysis. The data set is available through the NHLBI BioLINCC data repository by request (https://biolincc.nhlbi.nih.gov/). ASCVD at baseline was defined as having a history of myocardial infarction, stroke or transient ischemic attack, stable angina, prior percutaneous coronary intervention, prior coronary artery bypass grafting, or peripheral vascular disease. The initial/baseline and follow‐up visit schedules used for each of the cohorts are provided in Figure S1. We included participants with available data at the Framingham Heart Study Original cohort (Framingham) Exam 24, 17 the CHS (Cardiovascular Health Study) Year 5 Exam (including functional status variables using data from the National Center for Biotechnology Information, database of Genotypes and Phenotypes (dbGaP)), 18 and the ARIC (Atherosclerosis Risk in Communities) study Exam 5 19 for the derivation cohort (training and internal validation data sets) and the MESA (Multi‐Ethnic Study of Atherosclerosis) Exam 5 (2010–2011) for the external validation cohort. 20 Framingham Offspring (Exam 7) 21 was also used for the derivation cohort for the ASCVD model, but not the cognitive impairment model because data on cognitive status were collected more than 6.5 years after the baseline examination for this analysis (Exam 7). MESA is a population‐based sample of individuals aged 45 to 84 years across 6 US field centers to measure and characterize the prevalence and progression of subclinical cardiovascular disease. 20 MESA was used for the external validation data set because it includes baseline and follow‐up assessments of both cognition and ASCVD events in an ethnically diverse patient population. The study was approved by the institutional review board of Duke University Health System. No informed consent was required.
Predictors
Operational details of the potential predictors, covariates, and outcome measures included in these cohort studies, which were harmonized across studies, have been previously published and the selected variables are summarized in Table S1. 17 , 18 , 20 , 21 Notably, a number of variables not traditionally considered in prior prediction models were evaluated in both the derivation and validation cohorts, including sociodemographic data, mobility, cognitive function, depression, physical and mental function, and social network, with individual measures specified within the cohort studies of interest. 17 , 18 , 20 , 21 Candidate predictors at baseline considered for the models included age, sex, race (Black versus other race), education, income, health insurance, occupation, marital status, body mass index, height, weight, systolic BP, diastolic BP, alcohol use, smoking status, atrial fibrillation, diabetes, depression, total cholesterol, low‐density lipoprotein cholesterol, high‐density lipoprotein cholesterol, triglycerides, creatinine, fasting glucose, anticoagulants, antidepressants, Parkinson medications, antiplatelets, aspirin, insulin, antihypertensives, statins, other cholesterol‐lowering medications, thiazide diuretics, general health status, and measures of functional status including dressing, bathing, eating, transferring, cooking, grocery shopping, general health status, mobility, and Mini‐Mental State Examination (MMSE) score. Missing values were imputed using single imputation with fully conditional specification methods. 22
Outcomes
The outcomes of interest included: (1) cognitive impairment defined as an MMSE score ≤23, or converted equivalent on the Cognitive Abilities Screening Instrument (CASI) 23 , 24 in MESA for the cognitive risk model (Table S2); and (2) a composite of cardiovascular death, myocardial infarction, or stroke at 5 years for the ASCVD risk model. In CHS, MMSE scores collected at years 6 to 10 were used to define cognitive impairment. Similarly, in the Framingham Original cohort, MMSE scores collected at examinations 25 to 28 were used to define cognitive impairment. For ARIC, given the lack of interval cognitive assessments between baseline and the final examination, assessments of cognitive impairment were augmented with neuropsychological data collected at Exam 6, including the 6‐item screener, dementia screening interview, and surveillance for dementia hospitalizations and death. Because cognitive status was assessed only at examination visits, we ascertained cognitive impairment up to 5 years +6 months. All‐cause death was considered a competing risk for the cognitive outcome and noncardiovascular death was considered a competing risk for the cardiovascular outcome.
Statistical Analysis
Data sets for cognitive impairment and ASCVD risk models were randomly split into training (70%) and validation (30%) cohorts. Candidate predictors were summarized by cohort and by the presence or absence of the outcomes of interest. Mean/median and SDs/quartiles were calculated for continuous variables and frequency and proportions were calculated for categorical variables. We used a logistic regression model to estimate the probability of cognitive impairment at 5 years (pCI). All‐cause death was considered a competing risk event and included as a nonevent in the logistic regression model. This model can be used to quantify the risk of cognitive impairment at 5 years. Because a competing risk approach was used, the model does not provide an estimate of the risk of not developing cognitive impairment at 5 years; instead, (1–pCI) estimates the risk of not developing cognitive impairment or dying at 5 years. Predictors were selected in the training sample using the group least absolute shrinkage and selection operator (LASSO) method with stopping criteria based on the Schwarz Bayesian Criteria. The tuning (base regularization) parameter and the final coefficients were estimated in the training sample. PROC LOGSELECT in SAS Viya (SAS Institute Inc.) was used to derive the model. ASCVD events were analyzed using a time‐to‐event approach censoring follow‐up for events at 5 years (or last available follow‐up). For the ASCVD model, the group LASSO method as implemented in PROC PHSELECT in SAS Viya was used to identify predictors associated with ASCVD. The selected model was refit using the Fine‐Gray subdistribution hazards model to account for competing risk and to derive predictions of the risk of ASCVD at 5 years using the estimated cumulative incidence function. For the cognitive impairment and ASCVD risk models, restricted cubic splines were used to test for linearity of continuous variables, with nonlinear terms added as necessary. Discrimination of the models was assessed with the C index and calibration was measured comparing observed and predicted risks and by decile of predicted risk, and by calculating the χ2 statistic using the Hosmer‐Lemeshow statistic (cognitive impairment outcome) and Greenwood‐Nam‐D'Agostino statistic using cumulative incidence (ASCVD outcome), respectively. The estimated model was then applied to the testing data set (MESA) to evaluate external validation. Finally, performance of the ASCVD model was compared with the PCE in the derivation and validation data sets.
RESULTS
Baseline Characteristics
Descriptive characteristics for the cognitive impairment model cohort are provided in Table 1. A total of 3466 participants aged 75 years or older free of cognitive impairment and ASCVD at the selected baseline visit (CHS: n=1407, ARIC: n=1691, and Framingham Original: n=368) were included in the cognitive impairment derivation cohort (59.0% female, 8.2% Black race). The proportion of female participants was 54.9% in the MESA external validation cohort (n=437) with 24.9% of Black race. The median age was 79 years in both the derivation and validation cohorts (interquartile range [IQR], 77–82 years). The majority of participants in both cohorts were independent at baseline and reported good health status, but functional dependence and fair or poor health status were both more common in the MESA external validation cohort. The median MMSE was 28 (IQR, 27–29) in the overall derivation cohort and the median converted MMSE score was 27 (26–29) in the external validation cohort. Similar pooled information stratified by the presence or absence of cognitive impairment is provided in Table S3.
Table 1.
Population Description for Cognitive Impairment Outcome*
| Characteristic | CHS year 5 (n=1407) | ARIC Exam 5 (n=1691) | Framingham original Exam 24 (n=368) | Overall derivation cohort (n=3466) | MESA (external validation cohort) (n=437) |
|---|---|---|---|---|---|
| Age, median (Q1–Q3), y | 78 (76–82) | 79 (76–82) | 81 (79–84) | 79 (77–82) | 79 (77–82) |
| Women | 824 (58.6%) | 949 (56.1%) | 272 (73.9%) | 2045 (59.0%) | 240 (54.9%) |
| Black race | 45 (3.2%) | 239 (14.1%) | 0 (0.0%) | 284 (8.2%) | 109 (24.9%) |
| Education | |||||
| No schooling or grade ≤8 | 158 (11.2%) | 59 (3.5%) | 60 (16.3%) | 277 (8.0%) | 27 (6.2%) |
| Some high school | 178 (12.7%) | 136 (8.0%) | 26 (7.1%) | 340 (9.8%) | 20 (4.6%) |
| High school graduate | 374 (26.6%) | 575 (34.0%) | 142 (38.6%) | 1091 (31.5%) | 71 (16.2%) |
| Some college or technical/vocational school | 339 (24.1%) | 142 (8.4%) | 30 (8.2%) | 511 (14.7%) | 123 (28.1%) |
| College (completed) | 190 (13.5%) | 550 (32.5%) | 38 (10.3%) | 778 (22.4%) | 84 (19.2%) |
| Graduate school or professional school | 168 (11.9%) | 229 (13.5%) | 72 (19.6%) | 469 (13.5%) | 112 (25.6%) |
| Annual family income category, $ | |||||
| 0–<5000 | 38 (2.7%) | 16 (0.9%) | 0 (0.0%) | 54 (1.6%) | 8 (1.8%) |
| 5–15,000 | 458 (32.6%) | 187 (11.1%) | 0 (0.0%) | 645 (18.6%) | 64 (14.6%) |
| 15–25,000 | 285 (20.3%) | 222 (13.1%) | 0 (0.0%) | 507 (14.6%) | 53 (12.1%) |
| 25–50,000 | 349 (24.8%) | 567 (33.5%) | 0 (0.0%) | 916 (26.4%) | 140 (32.0%) |
| 50–75,000 | 205 (14.6%) | 308 (18.2%) | 0 (0.0%) | 513 (14.8%) | 90 (20.6%) |
| ≥75,000 | 0 (0.0%) | 254 (15.0%) | 0 (0.0%) | 254 (7.3%) | 45 (10.3%) |
| Unknown | 72 (5.1%) | 137 (8.1%) | 368 (100.0%) | 577 (16.6%) | 37 (8.5%) |
| Insurance: Medicare/Medicaid plus any additional insurance | 1257 (89.3%) | 1499 (88.6%) | 326 (88.6%) | 3082 (88.9%) | 289 (66.1%) |
| Employed | 109 (7.7%) | 724 (42.8%)† | 31 (8.4%) | 864 (24.9%) | 13 (3.0%) |
| Marital status | |||||
| Single/never married | 63 (4.5%) | 32 (1.9%) | 31 (8.4%) | 126 (3.6%) | 32 (7.3%) |
| Married | 803 (57.1%) | 1366 (80.8%) | 147 (39.9%) | 2316 (66.8%) | 219 (50.1%) |
| Widowed | 515 (36.6%) | 191 (11.3%) | 174 (47.3%) | 880 (25.4%) | 139 (31.8%) |
| Divorced/separated | 26 (1.8%) | 102 (6.0%) | 16 (4.3%) | 144 (4.2%) | 47 (10.8%) |
| Body mass index, median (Q1–Q3) | 25.3 (22.8–28.0) | 27.2 (24.4–30.8) | 26.0 (23.0–29.1) | 26.2 (23.6–29.5) | 26.7 (23.6–29.8) |
| Weight, median (Q1–Q3), kg | 68.3 (59.4–78.0) | 75.1 (64.7–86.5) | 65.8 (56.7–77.1) | 71.4 (61.2–81.8) | 72.9 (62.6–83.0) |
| Height, median (Q1–Q3), cm | 163 (157–171) | 164 (158–173) | 157 (152–165) | 163 (157–171) | 164 (157–171) |
| Systolic BP, median (Q1–Q3), mm Hg | 136 (123–153) | 130 (119–142) | 138 (123–153) | 133 (121–147) | 125 (113–140) |
| Diastolic BP, median (Q1–Q3) mm Hg | 69 (63–77) | 65 (57–72) | 66 (58–73) | 67 (59–74) | 65 (60–72) |
| Alcohol use | |||||
| No alcohol | 832 (59.1%) | 1102 (65.2%) | 182 (49.5%) | 2116 (61.1%) | 281 (64.3%) |
| Mild/moderate | 530 (37.7%) | 562 (33.2%) | 170 (46.2%) | 1262 (36.4%) | 143 (32.7%) |
| Heavy | 45 (3.2%) | 27 (1.6%) | 16 (4.3%) | 88 (2.5%) | 13 (3.0%) |
| Atrial fibrillation | 47 (3.3%) | 196 (11.6%) | 8 (2.2%) | 251 (7.2%) | 9 (2.1%) |
| Current smoker | 80 (5.7%) | 63 (3.7%) | 28 (7.6%) | 171 (4.9%) | 16 (3.7%) |
| Diabetes | 187 (13.3%) | 524 (31.0%) | 18 (4.9%) | 729 (21.0%) | 74 (16.9%) |
| HDL‐C, median (Q1–Q3), mg/dL | 52 (43–62) | 50 (42–60) | 51 (40–62) | 51 (42–61) | 55 (47–67) |
| LDL‐C, median (Q1–Q3), mg/dL | 123 (101–145) | 98 (78–122) | 116 (94–139) | 112 (88–135) | 94 (75–120) |
| Triglycerides, median (Q1–Q3), mg/dL | 126 (90–168) | 111 (84–150) | 140 (100–205) | 119 (87–164) | 88 (67–119) |
| Total cholesterol, median (Q1–Q3), mg/dL | 205 (182–229) | 175 (150–204) | 200 (175–225) | 191 (164–220) | 175 (152–201) |
| Creatinine, median (Q1–Q3), mg/dL | 0.95 (0.85–1.15) | 0.94 (0.81–1.14) | 1.10 (1.00–1.30) | 0.98 (0.85–1.17) | 0.90 (0.78–1.07) |
| Fasting glucose, median (Q1–Q3), mg/dL | 99 (93–109) | 106 (98–120) | 95 (81–111) | 102 (94–115) | 95 (89–104) |
| Hemoglobin, median (Q1–Q3), g/dL | 14 (13–15) | 13 (12–14) | 13 (13–14) | 13 (13–14) | 13 (12–14) |
| Taking anticoagulants | 23 (1.6%) | 141 (8.3%) | 9 (2.4%) | 173 (5.0%) | 16 (3.7%) |
| Taking antidepressants | 65 (4.6%) | 241 (14.3%) | 26 (7.1%) | 332 (9.6%) | 49 (11.2%) |
| Taking anti‐Parkinson medication | 13 (0.9%) | 36 (2.1%) | 2 (0.5%) | 51 (1.5%) | 5 (1.1%) |
| Taking antiplatelets | 0 (0.0%) | 92 (5.4%) | 0 (0.0%) | 92 (2.7%) | 7 (1.6%) |
| Taking aspirin | 34 (2.4%) | 1206 (71.3%) | 86 (23.4%) | 1326 (38.3%) | 194 (44.4%) |
| Taking insulin | 27 (1.9%) | 66 (3.9%) | 1 (0.3%) | 94 (2.7%) | 8 (1.8%) |
| Taking antihypertensives | 680 (48.3%) | 1140 (67.4%) | 176 (47.8%) | 1996 (57.6%) | 277 (63.4%) |
| Taking lipid‐lowering medication | 73 (5.2%) | 956 (56.5%) | 28 (7.6%) | 1057 (30.5%) | 189 (43.2%) |
| Taking statins | 48 (3.4%) | 884 (52.3%) | 20 (5.4%) | 952 (27.5%) | 180 (41.2%) |
| Taking thiazide diuretics | 134 (9.5%) | 560 (33.1%) | 49 (13.3%) | 743 (21.4%) | 55 (12.6%) |
| No. of medications, median (Q1–Q3) | 2 (1–4) | 9 (6–12) | 3 (2–4) | 5 (2–9) | 5 (3–8) |
| Functional status: dressing‡ | |||||
| Independent | 1388 (98.6%) | 1597 (94.4%) | 359 (97.6%) | 3344 (96.5%) | 200 (45.8%) |
| Need some help | 19 (1.4%) | 94 (5.6%) | 9 (2.4%) | 122 (3.5%) | 198 (45.3%) |
| Dependent | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 39 (8.9%) |
| Functional status: bathing‡ | |||||
| Independent | 1375 (97.7%) | 1668 (98.6%) | 287 (78.0%) | 3330 (96.1%) | 200 (45.8%) |
| Need some help | 25 (1.8%) | 3 (0.2%) | 79 (21.5%) | 107 (3.1%) | 198 (45.3%) |
| Dependent | 7 (0.5%) | 20 (1.2%) | 2 (0.5%) | 29 (0.8%) | 39 (8.9%) |
| Functional status: eating§ | |||||
| Independent | 1404 (99.8%) | 1667 (98.6%) | 365 (99.2%) | 3436 (99.1%) | 407 (93.1%) |
| Need some help | 3 (0.2%) | 24 (1.4%) | 3 (0.8%) | 30 (0.9%) | 30 (6.9%) |
| Functional status: transferring‡ | |||||
| Independent | 1293 (91.9%) | 1486 (87.9%) | 283 (76.9%) | 3062 (88.3%) | 200 (45.8%) |
| Need some help | 93 (6.6%) | 199 (11.8%) | 85 (23.1%) | 377 (10.9%) | 198 (45.3%) |
| Dependent | 21 (1.5%) | 6 (0.4%) | 0 (0.0%) | 27 (0.8%) | 39 (8.9%) |
| Functional status: cooking‡ | |||||
| Independent | 1384 (98.4%) | 1628 (96.3%) | 337 (91.6%) | 3349 (96.6%) | 200 (45.8%) |
| Need some help | 14 (1.0%) | 37 (2.2%) | 2 (0.5%) | 53 (1.5%) | 198 (45.3%) |
| Dependent | 9 (0.6%) | 26 (1.5%) | 29 (7.9%) | 64 (1.8%) | 39 (8.9%) |
| Functional status: grocery shopping‡ | |||||
| Independent | 1341 (95.3%) | 1561 (92.3%) | 315 (85.6%) | 3217 (92.8%) | 200 (45.8%) |
| Need some help | 36 (2.6%) | 72 (4.3%) | 20 (5.4%) | 128 (3.7%) | 198 (45.3%) |
| Dependent | 30 (2.1%) | 58 (3.4%) | 33 (9.0%) | 121 (3.5%) | 39 (8.9%) |
| General health status | |||||
| Excellent/very good/good | 1138 (80.9%) | 1495 (88.4%) | 313 (85.1%) | 2946 (85.0%) | 302 (69.1%) |
| Fair | 223 (15.8%) | 185 (10.9%) | 49 (13.3%) | 457 (13.2%) | 115 (26.3%) |
| Poor | 46 (3.3%) | 11 (0.7%) | 6 (1.6%) | 63 (1.8%) | 20 (4.6%) |
| Mobility | |||||
| Independently mobile | 1061 (75.4%) | 1592 (94.1%) | 311 (84.5%) | 2964 (85.5%) | 264 (60.4%) |
| Assistance needed | 325 (23.1%) | 88 (5.2%) | 50 (13.6%) | 463 (13.4%) | 120 (27.5%) |
| Immobile or largely immobile | 21 (1.5%) | 11 (0.7%) | 7 (1.9%) | 39 (1.1%) | 53 (12.1%) |
| Depression | |||||
| None/rarely | 1008 (71.6%) | 1244 (73.6%) | 265 (72.0%) | 2517 (72.6%) | 338 (77.3%) |
| Some of the time | 291 (20.7%) | 414 (24.5%) | 66 (17.9%) | 771 (22.2%) | 82 (18.8%) |
| Much or most of the time | 108 (7.7%) | 33 (2.0%) | 37 (10.1%) | 178 (5.1%) | 17 (3.9%) |
| MMSE score at baseline, median (Q1–Q3) | 28 (27–29) | 28 (27–29) | 29 (27–30) | 28 (27–29) | 27 (26–29) |
Values are expressed as frequencies (percentages) except where indicated otherwise.
ARIC indicates Atherosclerosis Risk in Communities Study; BP, blood pressure; CHS, Cardiovascular Health Study; HDL‐C, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; MESA, Multi‐Ethnic Study of Atherosclerosis; MMSE, Mini‐Mental State Examination; and Q1–Q3, quartile 1–quartile 3.
Descriptive statistics after imputation.
Employment status was imputed from a prior examination for the ARIC cohort (Exam 4) and, thus, was likely significantly overestimated.
All functional status variables in MESA aside from “eating” were considered the same and derived and imputed from a combination of questions around household chores and moderate walking.
Functional status‐eating was fully imputed in MESA.
Descriptive characteristics for the ASCVD model cohort are provided in Table 2. A total of 4403 participants aged 75 years or older free of cognitive impairment and ASCVD at baseline visit were included in the ASCVD derivation cohort (59.7% female, 7.8% Black race): CHS (n=1485), ARIC (n=2316), Framingham Original (n=381), and Framingham Offspring (n=221). The MESA external validation cohort (n=957) was 56.5% female and 24.3% Black race. The median age was 79 years in both the derivation and validation cohorts. Most participants reported functional independence and good health status, although rates were again lower in the MESA external validation cohort than in the derivation cohort. Similar pooled information stratified by the presence or absence of ASCVD and noncardiovascular death, respectively, is provided in Table S4.
Table 2.
Population Description for the ASCVD Outcome*
| Characteristic | CHS year 5 (n=1485) | ARIC Exam 5 (n=2316) | Framingham Original Exam 24 (n=381) | Framingham Offspring Exam 7 (n=221) | Overall derivation cohort (n=4403) | MESA (External validation cohort) (n=957) |
|---|---|---|---|---|---|---|
| Age, median (Q1–Q3), y | 78 (76–82) | 79 (77–82) | 81 (79–84) | 77 (76–79) | 79 (77–82) | 79 (77–83) |
| Women | 868 (58.5%) | 1346 (58.1%) | 279 (73.2%) | 137 (62.0%) | 2630 (59.7%) | 541 (56.5%) |
| Black race | 46 (3.1%) | 299 (12.9%) | 0 (0.0%) | 0 (0.0%) | 345 (7.8%) | 233 (24.3%) |
| Education | ||||||
| No schooling or grade ≤8 | 167 (11.2%) | 84 (3.6%) | 61 (16.0%) | 3 (1.4%) | 315 (7.2%) | 70 (7.3%) |
| Some high school | 185 (12.5%) | 186 (8.0%) | 26 (6.8%) | 20 (9.0%) | 417 (9.5%) | 45 (4.7%) |
| High school graduate | 396 (26.7%) | 802 (34.6%) | 144 (37.8%) | 75 (33.9%) | 1417 (32.2%) | 172 (18.0%) |
| Some college or technical/vocational school | 359 (24.2%) | 197 (8.5%) | 32 (8.4%) | 72 (32.6%) | 660 (15.0%) | 272 (28.4%) |
| College (completed) | 199 (13.4%) | 739 (31.9%) | 39 (10.2%) | 25 (11.3%) | 1002 (22.8%) | 178 (18.6%) |
| Graduate school or professional school | 179 (12.1%) | 308 (13.3%) | 79 (20.7%) | 26 (11.8%) | 592 (13.4%) | 220 (23.0%) |
| Annual family income category, $ | ||||||
| 0–<5000 | 42 (2.8%) | 22 (0.9%) | 0 (0.0%) | 0 (0.0%) | 64 (1.5%) | 15 (1.6%) |
| 5–15,000 | 487 (32.8%) | 252 (10.9%) | 0 (0.0%) | 20 (9.0%) | 759 (17.2%) | 152 (15.9%) |
| 15–25,000 | 297 (20.0%) | 299 (12.9%) | 0 (0.0%) | 40 (18.1%) | 636 (14.4%) | 154 (16.1%) |
| 25–50,000 | 368 (24.8%) | 786 (33.9%) | 0 (0.0%) | 57 (25.8%) | 1211 (27.5%) | 284 (29.7%) |
| 50–75,000 | 214 (14.4%) | 411 (17.7%) | 0 (0.0%) | 30 (13.6%) | 655 (14.9%) | 176 (18.4%) |
| ≥75,000 | 0 (0.0%) | 341 (14.7%) | 0 (0.0%) | 0 (0.0%) | 341 (7.7%) | 103 (10.8%) |
| Unknown | 77 (5.2%) | 205 (8.9%) | 381 (100.0%) | 74 (33.5%) | 737 (16.7%) | 73 (7.6%) |
| Insurance: Medicare/Medicaid plus any additional insurance | 1327 (89.4%) | 2063 (89.1%) | 334 (87.7%) | 207 (93.7%) | 3931 (89.3%) | 609 (63.6%) |
| Employed | 113 (7.6%) | 980 (42.3%)† | 32 (8.4%) | 31 (14.0%) | 1156 (26.3%) | 26 (2.7%) |
| Marital status | ||||||
| Single/never married | 64 (4.3%) | 36 (1.6%) | 33 (8.7%) | 8 (3.6%) | 141 (3.2%) | 65 (6.8%) |
| Married | 853 (57.4%) | 1862 (80.4%) | 151 (39.6%) | 131 (59.3%) | 2997 (68.1%) | 478 (49.9%) |
| Widowed | 541 (36.4%) | 256 (11.1%) | 180 (47.2%) | 75 (33.9%) | 1052 (23.9%) | 304 (31.8%) |
| Divorced/separated | 27 (1.8%) | 162 (7.0%) | 17 (4.5%) | 7 (3.2%) | 213 (4.8%) | 110 (11.5%) |
| Body mass index, median (Q1–Q3) | 25.3 (22.8–28.1) | 27.3 (24.5–30.6) | 26.1 (23.1–29.2) | 26.9 (24.2–29.8) | 26.4 (23.7–29.6) | 26.4 (23.9–29.5) |
| Weight, median (Q1–Q3), kg | 68.5 (59.4–78.0) | 74.9 (64.6–85.9) | 65.8 (56.7–77.1) | 72.6 (61.2–81.6) | 71.7 (61.7–82.0) | 71.4 (61.7–82.1) |
| Height, median (Q1–Q3), cm | 163 (157–171) | 164 (157–172) | 157 (152–165) | 162 (155–169) | 163 (157–171) | 163 (157–171) |
| Systolic BP, median (Q1–Q3), mm Hg | 137 (123–153) | 131 (119–142) | 138 (123–153) | 136 (121–151) | 133 (121–147) | 127 (114–142) |
| Diastolic BP, median (Q1–Q3), mm Hg | 69 (63–77) | 65 (58–72) | 66 (59–73) | 69 (63–75) | 67 (60–74) | 66 (60–72) |
| Alcohol use | ||||||
| No alcohol | 886 (59.7%) | 1467 (63.3%) | 190 (49.9%) | 102 (46.2%) | 2645 (60.1%) | 634 (66.2%) |
| Mild/moderate | 548 (36.9%) | 811 (35.0%) | 172 (45.1%) | 109 (49.3%) | 1640 (37.2%) | 293 (30.6%) |
| Heavy | 51 (3.4%) | 38 (1.6%) | 19 (5.0%) | 10 (4.5%) | 118 (2.7%) | 30 (3.1%) |
| Atrial fibrillation | 51 (3.4%) | 260 (11.2%) | 9 (2.4%) | 3 (1.4%) | 323 (7.3%) | 20 (2.1%) |
| Current smoker | 85 (5.7%) | 99 (4.3%) | 28 (7.3%) | 14 (6.3%) | 226 (5.1%) | 34 (3.6%) |
| Diabetes | 195 (13.1%) | 709 (30.6%) | 20 (5.2%) | 29 (13.1%) | 953 (21.6%) | 177 (18.5%) |
| HDL‐C, median (Q1–Q3), mg/dL | 52 (43–62) | 50 (43–60) | 50 (40–62) | 55 (44–66) | 51 (43–61) | 55 (47–69) |
| LDL‐C, median (Q1–Q3), mg/dL | 123 (101–145) | 100 (78–123) | 114 (94–136) | 115 (94–140) | 110 (87–134) | 96 (78–122) |
| Triglycerides, median (Q1–Q3), mg/dL | 126 (90–169) | 112 (84–151) | 138 (98–204) | 113 (84–151) | 117 (87–161) | 90 (68–122) |
| Total cholesterol, median (Q1–Q3), mg/dL | 205 (181–229) | 177 (152–205) | 198 (173–225) | 197 (173–220) | 190 (163–219) | 178 (153–205) |
| Creatinine, median (Q1–Q3), mg/dL | 0.95 (0.85–1.15) | 0.93 (0.79–1.12) | 1.10 (1.00–1.30) | 1.10 (0.95–1.28) | 0.96 (0.83–1.15) | 0.90 (0.77–1.07) |
| Fasting glucose, median (Q1–Q3), mg/dL | 99 (93–109) | 106 (98–120) | 103 (90–119) | 102 (93–112) | 103 (95–116) | 95 (88–105) |
| Hemoglobin, median (Q1–Q3), g/dL | 14 (13–14) | 13 (12–14) | 13 (13–14) | 14 (13–15) | 13 (13–14) | 13 (12–14) |
| Taking anticoagulants | 25 (1.7%) | 177 (7.6%) | 9 (2.4%) | 7 (3.2%) | 218 (5.0%) | 39 (4.1%) |
| Taking antidepressants | 69 (4.6%) | 318 (13.7%) | 26 (6.8%) | 15 (6.8%) | 428 (9.7%) | 93 (9.7%) |
| Taking anti‐Parkinson medication | 14 (0.9%) | 46 (2.0%) | 4 (1.0%) | 5 (2.3%) | 69 (1.6%) | 7 (0.7%) |
| Taking antiplatelets | 0 (0.0%) | 123 (5.3%) | 0 (0.0%) | 1 (0.5%) | 124 (2.8%) | 17 (1.8%) |
| Taking aspirin | 35 (2.4%) | 1648 (71.2%) | 89 (23.4%) | 65 (29.4%) | 1837 (41.7%) | 442 (46.2%) |
| Taking insulin | 27 (1.8%) | 81 (3.5%) | 1 (0.3%) | 3 (1.4%) | 112 (2.5%) | 21 (2.2%) |
| Taking antihypertensives | 721 (48.6%) | 1560 (67.4%) | 184 (48.3%) | 108 (48.9%) | 2573 (58.4%) | 611 (63.8%) |
| Taking anticholesterol medication | 76 (5.1%) | 1293 (55.8%) | 28 (7.3%) | 45 (20.4%) | 1442 (32.8%) | 414 (43.3%) |
| Taking statins | 49 (3.3%) | 1201 (51.9%) | 20 (5.2%) | 43 (19.5%) | 1313 (29.8%) | 396 (41.4%) |
| Taking thiazide diuretics | 144 (9.7%) | 778 (33.6%) | 49 (12.9%) | 22 (10.0%) | 993 (22.6%) | 121 (12.6%) |
| No. of medications, median (Q1–Q3) | 2 (1–4) | 9 (6–12) | 3 (2–4) | 2 (1–4) | 5 (2–9) | 5 (3–8) |
| Functional status: dressing‡ | ||||||
| Independent | 1463 (98.5%) | 2182 (94.2%) | 369 (96.9%) | 219 (99.1%) | 4233 (96.1%) | 420 (43.9%) |
| Need some help | 20 (1.3%) | 119 (5.1%) | 11 (2.9%) | 0 (0.0%) | 150 (3.4%) | 459 (48.0%) |
| Dependent | 2 (0.1%) | 15 (0.6%) | 1 (0.3%) | 2 (0.9%) | 20 (0.5%) | 78 (8.2%) |
| Functional status: bathing‡ | ||||||
| Independent | 1449 (97.6%) | 2268 (97.9%) | 293 (76.9%) | 192 (86.9%) | 4202 (95.4%) | 420 (43.9%) |
| Need some help | 24 (1.6%) | 3 (0.1%) | 84 (22.0%) | 24 (10.9%) | 135 (3.1%) | 459 (48.0%) |
| Dependent | 12 (0.8%) | 45 (1.9%) | 4 (1.0%) | 5 (2.3%) | 66 (1.5%) | 78 (8.2%) |
| Functional status: eating§ | ||||||
| Independent | 1481 (99.7%) | 2275 (98.2%) | 376 (98.7%) | 216 (97.7%) | 4348 (98.8%) | 957 (100.0%) |
| Need some help | 4 (0.3%) | 30 (1.3%) | 4 (1.0%) | 4 (1.8%) | 42 (1.0%) | 0 (0.0%) |
| Dependent | 0 (0.0%) | 11 (0.5%) | 1 (0.3%) | 1 (0.5%) | 13 (0.3%) | 0 (0.0%) |
| Functional status: transferring‡ | ||||||
| Independent | 1361 (91.6%) | 2021 (87.3%) | 285 (74.8%) | 206 (93.2%) | 3873 (88.0%) | 420 (43.9%) |
| Need some help | 97 (6.5%) | 286 (12.3%) | 94 (24.7%) | 15 (6.8%) | 492 (11.2%) | 459 (48.0%) |
| Dependent | 27 (1.8%) | 9 (0.4%) | 2 (0.5%) | 0 (0.0%) | 38 (0.9%) | 78 (8.2%) |
| Functional status: cooking‡ | ||||||
| Independent | 1465 (98.7%) | 2212 (95.5%) | 344 (90.3%) | 206 (93.2%) | 4227 (96.0%) | 420 (43.9%) |
| Need some help | 12 (0.8%) | 49 (2.1%) | 4 (1.0%) | 15 (6.8%) | 80 (1.8%) | 459 (48.0%) |
| Dependent | 8 (0.5%) | 55 (2.4%) | 33 (8.7%) | 0 (0.0%) | 96 (2.2%) | 78 (8.2%) |
| Functional status: grocery shopping‡ | ||||||
| Independent | 1405 (94.6%) | 2139 (92.4%) | 325 (85.3%) | 206 (93.2%) | 4075 (92.6%) | 420 (43.9%) |
| Need some help | 45 (3.0%) | 104 (4.5%) | 19 (5.0%) | 15 (6.8%) | 183 (4.2%) | 459 (48.0%) |
| Dependent | 35 (2.4%) | 73 (3.2%) | 37 (9.7%) | 0 (0.0%) | 145 (3.3%) | 78 (8.2%) |
| General health status | ||||||
| Excellent/very good/good | 1195 (80.5%) | 2039 (88.0%) | 322 (84.5%) | 199 (90.0%) | 3755 (85.3%) | 650 (67.9%) |
| Fair | 237 (16.0%) | 264 (11.4%) | 52 (13.6%) | 20 (9.0%) | 573 (13.0%) | 273 (28.5%) |
| Poor | 53 (3.6%) | 13 (0.6%) | 7 (1.8%) | 2 (0.9%) | 75 (1.7%) | 34 (3.6%) |
| Mobility | ||||||
| Independently mobile | 1115 (75.1%) | 2184 (94.3%) | 320 (84.0%) | 196 (88.7%) | 3815 (86.6%) | 576 (60.2%) |
| Assistance needed | 345 (23.2%) | 117 (5.1%) | 53 (13.9%) | 18 (8.1%) | 533 (12.1%) | 260 (27.2%) |
| Immobile or largely immobile | 25 (1.7%) | 15 (0.6%) | 8 (2.1%) | 7 (3.2%) | 55 (1.2%) | 121 (12.6%) |
| Depression | ||||||
| None/rarely | 1063 (71.6%) | 1686 (72.8%) | 274 (71.9%) | 162 (73.3%) | 3185 (72.3%) | 740 (77.3%) |
| Some of the time | 305 (20.5%) | 580 (25.0%) | 70 (18.4%) | 39 (17.6%) | 994 (22.6%) | 175 (18.3%) |
| Much or most of the time | 117 (7.9%) | 50 (2.2%) | 37 (9.7%) | 20 (9.0%) | 224 (5.1%) | 42 (4.4%) |
| MMSE score at baseline, median (Q1–Q3) | 28 (27–29) | 28 (27–29) | 29 (27–30) | 29 (27–29) | 28 (27–29) | 27 (25–28) |
Values are expressed as frequencies (percentages) except where indicated otherwise.
ARIC indicates Atherosclerosis Risk in Communities Study; BP, blood pressure; CHS, Cardiovascular Health Study; HDL‐C, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; MESA, Multi‐Ethnic Study of Atherosclerosis; MMSE, Mini‐Mental State Examination; and Q1–Q3, quartile 1–quartile 3.
Descriptive statistics after imputation.
Employment status was imputed from a prior examination for the ARIC cohort (Exam 4) and, thus, was likely significantly overestimated.
All functional status variables in MESA aside from “eating” were considered the same and derived and imputed from a combination of questions around household chores and moderate walking.
Functional status‐eating was fully imputed in MESA.
Incident Cognitive and ASCVD Events
Outcome event rates for cognitive impairment, ASCVD, and the competing risks of all‐cause death and noncardiovascular death, respectively, are provided in Table 3. While the rates of cognitive impairment were 16.7% in the derivation cohort and 15.3% in the external validation cohort, the competing risk of all‐cause death was 14.8% in the cognitive impairment derivation cohort and 23.1% in the external validation cohort. The risk of an ASCVD event at follow‐up was 17.0% in the derivation cohort and 8.4% in the external validation cohort, while the competing risk of noncardiovascular death was 9.7% in the derivation cohort and 10.4% in the external validation cohort.
Table 3.
Outcome Events for Cognitive Impairment, ASCVD, and Competing Risk of Death
| Characteristic | CHS y 5 | ARIC Exam 5 | Framingham Original Exam 24 | Framingham Offspring Exam 7 | Overall derivation cohort | MESA (external validation cohort) |
|---|---|---|---|---|---|---|
| Cognitive impairment at 5 y | ||||||
| No cognitive impairment | 951 (67.6%) | 1155 (68.3%) | 268 (72.8%) | N/A | 2374 (68.5%) | 269 (61.6%) |
| Cognitive impairment | 229 (16.3%) | 307 (18.2%) | 43 (11.7%) | N/A | 579 (16.7%) | 67 (15.3%) |
| All‐cause death (competing risk) | 227 (16.1%) | 229 (13.5%) | 57 (15.5%) | N/A | 513 (14.8%) | 101 (23.1%) |
| ASCVD at 5 y | ||||||
| No ASCVD | 923 (62.2%) | 1865 (80.5%) | 269 (70.6%) | 173 (78.3%) | 3230 (73.4%) | 777 (81.2%) |
| ASCVD | 409 (27.5%) | 260 (11.2%) | 58 (15.2%) | 21 (9.5%) | 748 (17.0%) | 80 (8.4%) |
| Noncardiovascular death (competing risk) | 153 (10.3%) | 191 (8.2%) | 54 (14.2%) | 27 (12.2%) | 425 (9.7%) | 100 (10.4%) |
Values are expressed as frequencies (percentages). ASCVD indicates atherosclerotic cardiovascular disease; and N/A, not available.
Cognitive Impairment Model
A summary of the final selected predictors and parameter estimates for the final cognitive impairment model is provided in Table 4. The final cognitive impairment model included baseline MMSE score, prior atrial fibrillation, antidepressant use, mobility impairment, and dependence with grocery shopping. Baseline cognitive status as measured by the MMSE was the strongest predictor of future incident cognitive impairment based on the χ2 results. Discrimination was consistent in the internal validation (C index=0.75) and testing data set (C index=0.73) for the cognitive impairment model. The cognitive impairment model was well calibrated in the (A) training and internal validation samples (χ2=10.8), but overpredicted risk across risk deciles in the (B) external validation (testing) sample (χ2=62.3) (Figure 1).
Table 4.
Summary of Cognitive Impairment Model
| Parameter | Parameter estimate | Standard error | χ2 | P value | Odds ratio represents… | Odds ratio (95% CI) |
|---|---|---|---|---|---|---|
| Intercept | 5.0223 | 1.5954 | 9.91 | 0.002 | ||
| Baseline MMSE | −0.5144 | 0.0352 | 214.10 | <0.001 | 1‐unit increase in MMSE | 0.60 (0.56–0.64) |
| Age | 0.0907 | 0.0152 | 35.61 | <0.001 | 5‐y increase in age | 2.48 (1.84–3.34) |
| Taking antidepressants | 0.6920 | 0.1704 | 16.50 | <0.001 | Yes vs no | 2.00 (1.43–2.79) |
| Grocery shopping | 0.6257 | 0.1952 | 10.27 | 0.001 | Need some help/dependent vs independent | 1.87 (1.28–2.74) |
| Prior atrial fibrillation | 0.4638 | 0.1995 | 5.41 | 0.02 | Yes vs no | 1.59 (1.08–2.35) |
| Mobility | 3.86 (2 df) | 0.145 | ||||
| Assistance needed | 0.00609 | 0.1655 | 0.001 | 0.971 | Assistance needed vs independent | 1.01 (0.73–1.39) |
| Immobile | 0.8656 | 0.4417 | 3.84 | 0.05 | Immobile vs independent | 2.38 (1.00–5.65) |
MMSE indicates Mini‐Mental State Examination.
Figure 1. Calibration of the cognitive impairment model in the (A) training and internal validation, and (B) external validation (testing) samples.
Figure 1 demonstrates the calibration of the cognitive impairment model in the training and internal validation cohorts (A) and in the external validation cohort (B).
ASCVD Model
A summary of the final selected predictors and parameter estimates for the final ASCVD model is provided in Table 5, performance characteristics for both models in Table 6, and final equations for both models in Table S5. The final ASCVD model included employment status, MMSE score, aspirin, lipid‐lowering medications, BP medications, systolic BP, general health status, high‐density lipoprotein cholesterol, triglycerides, creatinine, and mobility impairment. For the ASCVD model, the C index was 0.67 in the internal validation data set and 0.59 in the testing data set. The ASCVD model was also well calibrated in the (A) training and internal validation (χ2=7.2), but overpredicted risk across risk deciles in the (B) external validation samples (χ2=217.7) (Figure 2). Relative to that of the PCE, the ASCVD model improved discrimination in the derivation data set (Framingham/CHS/ARIC) (C index 0.61 with PCE versus 0.67 with current model) and modestly improved discrimination in MESA (C index 0.57 versus 0.59). Internal calibration was modest for the PCE (χ2=56.8), but external calibration was poor (χ2=137.4) for the PCE (Figures S2A and S2B). A survival plot for the ASCVD outcome stratified by study is presented in Figure S3.
Table 5.
Summary of ASCVD Model
| Parameter | Parameter estimate | Standard error | χ2 | P value | Hazard ratio represents… | Hazard ratio (95% CI) |
|---|---|---|---|---|---|---|
| General health | 27.39 (2 df) | <0.001 | ||||
| Fair | 0.56579 | 0.11067 | 26.14 | <0.001 | Fair vs good/very good/excellent | 1.76 (1.42–2.19) |
| Poor | 0.45559 | 0.26685 | 2.91 | 0.087 | Poor vs good/very good/excellent | 1.58 (0.94–2.66) |
| SBP | 0.00943 | 0.00212 | 19.69 | <0.001 | 10‐unit increase in SBP | 1.10 (1.05–1.15) |
| HDL | −0.01500 | 0.00357 | 17.66 | <0.001 | 10‐unit increase in HDL | 0.86 (0.80–0.92) |
| Employed | −0.45471 | 0.12142 | 14.03 | 0.0002 | Employed vs not employed | 0.64 (0.50–0.81) |
| BP medications | 0.31319 | 0.09788 | 10.24 | 0.001 | Taking BP medication vs not | 1.37 (1.13–1.7) |
| Baseline MMSE | −0.07936 | 0.02625 | 9.14 | 0.003 | 1‐unit increase in MMSE | 0.92 (0.88–0.97) |
| Aspirin | −0.27845 | 0.10651 | 6.83 | 0.009 | Taking aspirin vs not | 0.76 (0.61–0.93) |
| Lipid‐lowering medications | −0.29658 | 0.11348 | 6.83 | 0.009 | Taking lipid‐lowering medication vs not | 0.74 (0.60–0.93) |
| Triglycerides (spline*) | 11.29 (3 df.) | 0.01 | ||||
| Spline 1 | −0.01098 | 0.00435 | 6.38 | 0.01 | 10‐unit increase at 70 | 0.91 (0.84–0.98) |
| Spline 2 | 0.0002882 | 0.0001336 | 4.65 | 0.03 | 10‐unit increase at 120 | 1.11 (0.98–1.06) |
| Spline 3 | −0.0004767 | 0.0002365 | 4.06 | 0.04 | 10‐unit increase at 200 | 1.02 (1.00–1.03) |
| Mobility | 6.38 (2 df) | 0.04 | ||||
| Assistance needed | 0.28921 | 0.11964 | 5.84 | 0.02 | Assistance needed vs independent | 1.34 (1.06–1.69) |
| Immobile | 0.34255 | 0.34091 | 1.01 | 0.32 | Immobile vs independent | 1.41 (0.72–2.75) |
| Creatinine | 0.10101 | 0.08565 | 1.39 | 0.24 | 0.5‐unit increase | 1.05 (0.97–1.14) |
ASCVD indicates atherosclerotic cardiovascular disease; BP, blood pressure; HDL, high‐density lipoprotein; MMSE, Mini‐Mental State Examination; SBP, systolic blood pressure.
Restricted cubic spline with 4 knots at 58, 99, 140, and 268 mg/dL.
Table 6.
Performance Characteristics of Cognitive Impairment and ASCVD Risk Models
| Discrimination | Calibration | χ2 statistic | ||
|---|---|---|---|---|
| C index (95% CI) | Observed event rate | Average predicted event rate | ||
| Cognitive impairment model | ||||
| Training | 0.78 (0.76–0.81) | 17.3% | 16.9% (16.5%–17.4%) | 12.14 |
| Internal validation | 0.75 (0.71–0.79) | 15.3% | 17.5% (16.8%–18.3%) | 10.84 |
| External validation | 0.73 (0.66–0.79) | 15.3% | 29.3% (27.3%–31.3%) | 62.31 |
| ASCVD model | ||||
| Training | 0.69 (0.67–0.72) | 17.2% | 17.2% (16.8%–17.5%) | 12.05 |
| Internal validation | 0.67 (0.62–0.71) | 17.0% | 17.1% (16.5%–17.6%) | 7.21 |
| External validation | 0.59 (0.53–0.66) | 8.5% | 20.2% (19.6%–20.8%) | 217.68 |
For cognitive impairment, the value presented is the percentage of patients classified as having cognitive impairment at 5 years. For ASCVD, it is the cumulative incidence accounting for competing risk and incomplete follow‐up.
ASCVD indicates atherosclerotic cardiovascular disease.
Figure 2. Calibration of the ASCVD model in the (A) training and internal validation, and (B) external validation (testing) samples.
Figure 2 demonstrates the calibration of the ASCVD model in the training and internal validation cohorts (A) and in the external validation cohort (B). ASCVD indicates atherosclerotic cardiovascular disease.
DISCUSSION
We set out to derive and externally validate 2 separate models to predict 2 of the leading causes of mortality and disability among older adults, cognitive impairment and ASCVD, using measures of function as candidate predictors, which had not been considered in previous risk models. We demonstrate that among patients without cognitive impairment or ASCVD at baseline, functional measures improve risk prediction for cognitive impairment and ASCVD with varying accuracy. The SENIOR‐cognitive impairment risk model had strong performance with excellent discrimination and calibration in the derivation and internal validation cohorts. The SENIOR‐ASCVD model performed less well, with only fair discrimination and good calibration in the internal validation cohort, but it improved on the performance of the PCE. In the external validation testing cohort, the SENIOR‐cognitive impairment model overpredicted risk and the SENIOR‐ASCVD model had poor performance. In summary, measures of cognitive and physical function predicted the future risk of both cognitive impairment and ASCVD in older adults, providing a potential opportunity to lower risk for both conditions. Our findings underscore the challenge of disease‐specific risk prediction in older populations and points to the importance of a person‐centered approach to preventive care in older adults (Figure 3). 25
Figure 3. A holistic approach to prevention in older adults.
Figure 3 summarizes the factors for future cognitive impairment and ASCVD in persons 75 years or older and the accompanying interventions to reduce future risk and optimize healthy aging as part of a holistic approach. ASCVD indicates atherosclerotic cardiovascular disease; BP, blood pressure; HDL, high‐density lipoprotein; and LLT, lipid‐lowering therapies.
In support of our finding that ASCVD prediction in older adults was more challenging than predicting cognitive impairment, the PCE does not predict ASCVD risk well in older adults. 4 The poor performance of previous models may be driven by changes in the relationships between traditional cardiovascular risk factors and subsequent outcomes in older adults. Notably, the relationship between low‐density lipoprotein cholesterol and future ASCVD risk becomes attenuated among those 75 years and older, female sex becomes less protective among older adults, 4 and the association between other established risk factors and future events such as Black race, diabetes, and total cholesterol shifts with age. This was evident in the current study, which found that traditional predictors such as sex and diabetes did not remain statistically significant in the final models. Conversely, prior studies have suggested that certain risk factors not included in prior risk models such as depression, low educational attainment, and physical inactivity/immobility 15 , 16 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 are associated with future risk of both ASCVD and dementia. Models to predict the risk of dementia have focused on the apolipoprotein E gene and genetic risk scores 34 rather than clinical variables available in routine practice. In fact, the SENIOR‐cognitive impairment model had comparable performance to models with apolipoprotein E and polygenic risk scores. The strongest predictors of future risk for both cognitive outcomes and ASCVD in the current study were measures of global functioning. These measures encompass geriatric syndromes and biological aging and include cognitive function (MMSE), patient‐reported health status, functional status, and mobility impairment.
Modifiable risk factors are known to contribute to the risk of cognitive impairment and ASCVD, such as diabetes, obesity, smoking, hypertension, hyperlipidemia, and social determinants of health. 6 Up to a third of incident dementia cases are due to these modifiable risk factors. 25 , 35 Moreover, intensifying therapies that target modifiable risk factors may reduce the risk of developing both conditions. The SPRINT (Systolic Blood Pressure Intervention Trial) and SPRINT MIND substudy tested the effect of more intensive BP control on cardiovascular, renal, and cognitive outcomes, finding that more aggressive BP control reduced cardiovascular events in older adults and the risk of a composite of mild cognitive impairment or probable dementia. 36 , 37 The ongoing PREVENTABLE (Pragmatic Evaluation of Events and Benefits of Lipid‐Lowering in Older Adults; NCT04262206) trial is randomizing community‐dwelling US older adults (≥75 years) without cardiovascular disease or dementia to atorvastatin 40 mg/daily versus placebo to determine whether lipid lowering can reduce the primary composite of death, dementia, and persistent disability, with mild cognitive impairment and cardiovascular events as secondary end points. Accurate prediction models can identify the highest‐risk patients who are most likely to derive the greatest benefit.
The modest discrimination of the PCE and the SENIOR‐ASCVD risk models suggests that we must look beyond the currently available predictors to consider alternative markers of risk. While cognitive function and mobility stood out as predictors for both SENIOR models, other geriatric syndromes have strong associations with risk of cardiovascular events. 38 , 39 , 40 , 41 Frailty is strongly associated with future cardiovascular events in older adults. 42 Coronary artery calcium scoring has strong predictive value across domains of aging. 43 , 44 , 45 Both cognitive and ASCVD risk models performed less well and overpredicted risk in the MESA external validation cohort. Limited calibration in the testing cohort was possibly driven by differences in study measures and clinical phenotypes across cohorts. For example, there were differences between the derivation cohort and the MESA cohort, with a much higher prevalence of Black race and a lower prevalence of functional independence in MESA. For the cognitive impairment model, the strongest predictor of future cognitive impairment from the derivation cohort (baseline MMSE) was not captured in MESA and had to be derived from a separate measure (CASI). 23 , 24 Unsurprisingly, the association between the converted baseline MMSE and future cognitive impairment in MESA was weaker than that seen in the derivation data set. Other predictors in the cognitive impairment model, mobility and independence with grocery shopping, were not available in MESA and were imputed based on participant response about moderate walking and effort with household chores, respectively. For the ASCVD model, the derivation cohort had approximately twice the burden of ASCVD at follow‐up (17.0% versus 8.4%) compared with the external validation cohort (MESA), which likely led to poorer performance in the testing sample. Contemporary practice patterns have likely shifted in the use of preventive therapies, which may limit the external validity of our results. We would expect this to lead to overprediction of risk with better utilization of preventive therapies, compared with the higher‐risk population used for model development. The poor external validation performance of the ASCVD model in particular may limit its value in different patient populations. Ultimately, the current equations may benefit from external validation in a separate cohort with harmonized assessments of cognition, mobility, and functional status.
Our study has some limitations. First, the models were developed and validated from existing pooled cohort study data, which is inherently limited. We were also limited by the granularity of the derivation and validation data sets and ability to harmonize across data sets, which were missing certain promising candidate predictors, such as a frailty, the presence of subclinical ASCVD, inflammatory biomarkers, and duration of exposure to specific risk factors. In addition, employment status was imputed from a prior examination for the ARIC cohort (Exam 4) and, thus, was likely overestimated. Some of the final predictors included in the models may be considered as “markers of risk” rather than “drivers of risk.” This is similar to the inclusion of BP medication treatment in the PCE. A number of factors may influence the baseline cognitive assessment in older adults, including the current use of certain medications such as sleeping pills, sedating antihistamines, and neuroleptics, as well as depressive symptoms, and model performance may be limited in those contexts. Finally, clinicians must consider baseline risk within a complex array of other considerations around potential treatments in older adults, balancing potential benefits against possible harms and tradeoffs.
CONCLUSION
Global functioning variables are the strongest predictors of risk for cognitive and ASCVD outcomes in older adults without these conditions at baseline. The SENIOR models improve upon other available models, but challenges remain in the prediction of cognitive and cardiovascular events in older adults. Future models developed from more contemporary data sets and incorporating additional novel predictors will be crucial to improving both performance and generalizability of risk prediction tools in the geriatric population.
Disclosures
Dr Nanna reports consulting from Merck, Novo Nordisk, and HeartFlow, Inc. Dr Peterson receives research support from Amgen, Bristol Myer Squibb, Janssen, and Esperion; serves on Scientific Advisory Boards for Novartis, Pfizer, Novo Nordisk, and Bayer; and is a consultant for Cerner, Inc. Dr Navar has received funding for research to her institution from BMS, Esperion, Amgen, and Janssen, and honoraria and consulting fees from Cerner, Pfizer, AstraZeneca, Bayer, Novartis, BI, Lilly, Novo Nordisk, Janssen, and New Amsterdam. Dr Colantonio receives research support from Amgen. Dr Goyal receives personal fees from Agepha Pharma, Akros Pharma, Axon therapies, Bayer HealthCare Pharmaceuticals, and Sensorum Health. The remaining authors have no disclosures to report.
Source of Funding
National Institute on Aging/National Institutes of Health from R03AG074067 (GEMSSTAR award). Dr Nanna and Dr Gill are also supported by the Yale Claude D. Pepper Older Americans Independence Center (P30AG021342). The study sponsor was not involved in any aspect of this research project, including formulating the research question, analysis, drafting, reviewing, or critically appraising the article.
Supporting information
Tables S1–S5
Figures S1–S3
Acknowledgments
The authors acknowledge Devon Stuart for assistance with medical illustration. In addition to the National Institute on Aging/National Institutes of Health from R03AG074067 (GEMSSTAR award) that directly funded this work, Dr Nanna reports current research support from the American College of Cardiology Foundation supported by the George F. and Ann Harris Bellows Foundation, the Patient‐Centered Outcomes Research Institute (PCORI), the Yale Claude D. Pepper Older Americans Independence Center (P30AG021342), and the National Institute on Aging (K76AG088428). Dr Chaudhry is supported by the National Institutes of Health from R01HL160822. Dr Damluji receives research funding from the Pepper Scholars Program of the Johns Hopkins University Claude D. Pepper Older Americans Independence Center funded by the National Institute on Aging P30‐AG021334; mentored patient‐oriented research career development award from the National Heart, Lung, and Blood Institute K23‐HL153771; the National Institutes of Health National Institute of Aging R01‐AG078153; and the Patient‐Centered Outcomes Research Institute (PCORI). Dr Gill is supported by the Yale Claude D. Pepper Older Americans Independence Center (P30AG021342). CHS data were supplemented via dbGaP: this research was supported by contracts HHSN268201200036C, HHSN268200800007C, N01‐HC85079, N01‐HC‐85080, N01‐HC‐85081, N01‐HC‐85082, N01‐HC‐85083, N01‐HC‐85084, N01‐HC‐85085, N01‐HC‐85086, N01‐HC‐35129, N01‐HC‐15103, N01‐HC‐55222, N01‐HC‐75150, N01‐HC‐45133, and N01‐HC‐85239; grant numbers U01 HL080295 and U01 HL130014 from the NHLBI, and R01 AG023629 from the National Institute on Aging, with additional contribution from the National Institute of Neurological Disorders and Stroke. A full list of principal CHS investigators and institutions can be found at https://chs‐nhlbi.org/pi. This article was not prepared in collaboration with CHS investigators and does not necessarily reflect the opinions or views of CHS, or the NHLBI. MESA and the MESA SHARe project are conducted and supported by the NHLBI in collaboration with MESA investigators. Support for MESA is provided by contracts N01‐HC‐95159, N01‐HC‐95160, N01‐HC‐95161, N01‐HC‐95162, N01‐HC‐95163, N01‐HC‐95164, N01‐HC‐95165, N01‐HC‐95166, N01‐HC‐95167, N01‐HC‐95168, N01‐HC‐95169, and CTSA UL1‐RR‐024156. The authors thank the other investigators, the staff, and the participants of MESA for their valuable contributions. A full list of participating MESA investigators and institutions can be found at http://www.mesa‐nhlbi.org. This paper has been reviewed and approved by the MESA Publications and Presentations Committee. Dr. Nanna is also supported from the National Institute on Aging (K76AG088428).
This article was sent to Jose R. Romero, MD, Associate Editor, for review by expert referees, editorial decision, and final disposition.
Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/JAHA.124.038949
For Sources of Funding and Disclosures, see page 15.
REFERENCES
- 1. Centers for Disease Control and Prevention NCfHS . Table 4. Life expectancy at birth, age 65, and age 75, by sex, race, and Hispanic origin: United States, selected years 1900–2018. Accessed December 22, 2024. https://www.cdc.gov/nchs/data/hus/2019/004‐508.pdf.
- 2. Arnett DK, Blumenthal RS, Albert MA, Buroker AB, Goldberger ZD, Hahn EJ, Himmelfarb CD, Khera A, Lloyd‐Jones D, McEvoy J, et al. 2019 ACC/AHA guideline on the primary prevention of cardiovascular disease. Circulation. 2019;140:e596–e646. doi: 10.1161/cir.0000000000000678 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Grundy SM, Stone NJ, Bailey AL, Beam C, Birtcher KK, Blumenthal RS, Braun LT, de Ferranti S, Faiella‐Tommasino J, Forman DE, et al. 2018 AHA/ACC guideline on the Management of Blood Cholesterol: a report of the American College of Cardiology/American Heart Association task force on clinical practice guidelines. Circulation. 2018;139:e1082–e1143. doi: 10.1161/cir.0000000000000625 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Nanna MG, Peterson ED, Wojdyla D, Navar AM. The accuracy of cardiovascular pooled cohort risk estimates in US older adults. J Gen Intern Med. 2019;35:1701–1708. doi: 10.1007/s11606-019-05361-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Dalton JE, Rothberg MB, Dawson NV, Krieger NI, Zidar DA, Perzynski AT. Failure of traditional risk factors to predict cardiovascular events in older populations. J Am Geriatr Soc. 2020;68:754–761. doi: 10.1111/jgs.16329 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Damluji AA, Ijaz N, Chung SE, Xue QL, Hasan RK, Batchelor WB, Orkaby AR, Kochar A, Nanna MG, Roth DL, et al. Frailty, cognitive impairment, and cardiovascular disease. JACC Adv. 2023;2:100318. doi: 10.1016/j.jacadv.2023.100318 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Heron M. Deaths: leading causes for 2016. Natl Vital Stat Rep. 2018;67:1–77. [PubMed] [Google Scholar]
- 8. Roberts AW, Blakeslee L, Rabe MA. The Population 65 Years and Older in the United States: 2016. American Community Survey Reports; 2018. [Google Scholar]
- 9. Kramarow EA, Tejada‐Vera B. Dementia mortality in the United States, 2000‐2017. Natl Vital Stat Rep. 2019;68:1–29. [PubMed] [Google Scholar]
- 10. Feigin VL, Vos T, Nichols E, Owolabi MO, Carroll WM, Dichgans M, Deuschl G, Parmar P, Brainin M, Murray C. The global burden of neurological disorders: translating evidence into policy. Lancet Neurol. 2020;19:255–265. doi: 10.1016/s1474-4422(19)30411-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Nanna MG, Peterson ED, Wu A, Harding T, Galanos AN, Wruck L, Alexander KP. Age, knowledge, and risk tolerance for cardiac care. Am Heart J. 2020;219:99–108. doi: 10.1016/j.ahj.2019.09.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Norton S, Matthews FE, Barnes DE, Yaffe K, Brayne C. Potential for primary prevention of Alzheimer's disease. Lancet Neurol. 2014;13:788–794. doi: 10.1016/s1474-4422(14)70136-x [DOI] [PubMed] [Google Scholar]
- 13. Resende EPF, Llibre Guerra JJ, Miller BL. Inequities as contributors to brain health. JAMA Neurol. 2019;76:633–634. doi: 10.1001/jamaneurol.2019.0362 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Hohman TJ, Kaczorowski CC. Modifiable lifestyle factors in Alzheimer disease. JAMA Neurol. 2020;77:1207. doi: 10.1001/jamaneurol.2020.1114 [DOI] [PubMed] [Google Scholar]
- 15. Koutsavlis AT, Wolfson C. Mobility predictors of survival in dementia patients. Chronic Dis Can. 2000;21:93–103. [PubMed] [Google Scholar]
- 16. Dodson JA, Hajduk AM, Geda M, Krumholz HM, Murphy TE, Tsang S, Tinetti ME, Nanna MG, McNamara R, Gill TM, et al. Predicting mortality after myocardial infarction. Ann Intern Med. 2020;172:12–21. doi: 10.7326/m19-0974 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Dawber TR, Kannel WB, Lyell LP. Longitudinal studies in a community: the Framingham study. Ann N Y Acad Sci. 1963;107:539–556. doi: 10.1111/j.1749-6632.1963.tb13299.x [DOI] [PubMed] [Google Scholar]
- 18. Fried LP, Borhani NO, Enright P, Furberg CD, Gardin JM, Kronmal RA, Kuller LH, Manolio TA, Mittelmark MB, Newman A, et al. Cardiovascular health study: design and rationale. Ann Epidemiol. 1991;1:263–276. doi: 10.1016/1047-2797(91)90005-w [DOI] [PubMed] [Google Scholar]
- 19. The Atherosclerosis Risk in Communities (ARIC) study. Am J Epidemiol. 1989;129:687–702. doi: 10.1093/oxfordjournals.aje.a115184 [DOI] [PubMed] [Google Scholar]
- 20. Bild DE, Bluemke DA, Burke GL, et al. Multi‐ethnic study of atherosclerosis: objectives and design. Am J Epidemiol. 2002;156:871–881. doi: 10.1093/aje/kwf113 [DOI] [PubMed] [Google Scholar]
- 21. Feinleib M, Kannel WB, Garrison RJ, McNamara PM, Castelli WP. The Framingham offspring study. Prev Med. 1975;4:518–525. doi: 10.1016/0091-7435(75)90037-7 [DOI] [PubMed] [Google Scholar]
- 22. Liu Y, De A. Imputation for missing data in large studies. Int J Stat Med Res. 2015;4:287–295. doi: 10.6000/1929-6029.2015.04.03.7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Teng EL, Hasegawa K, Homma A, Imai Y, Larson E, Graves A, Sugimoto K, Yamaguchi T, Sasaki H, Chiu D, et al. The cognitive abilities screening instrument. Int Psychogeriatr. 1994;6:45–58. doi: 10.1017/s1041610294001602 [DOI] [PubMed] [Google Scholar]
- 24. Lin KN, Wang PN, Liu CY, Chen WT, Lee YC, Liu HC. CASI cutoff scores for dementia screening. Dement Geriatr Cogn Disord. 2002;14:176–182. doi: 10.1159/000066024 [DOI] [PubMed] [Google Scholar]
- 25. Ijaz N, Jamil Y, Brown CH, et al. Cognitive frailty in cardiovascular disease. J Am Heart Assoc. 2024;13:e033594. doi: 10.1161/jaha.123.033594 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Barnes DE, Yaffe K. Effect of risk factor reduction on Alzheimer's disease prevalence. Lancet Neurol. 2011;10:819–828. doi: 10.1016/s1474-4422(11)70072-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Daviglus ML, Bell CC, Berrettini W, et al. NIH conference statement on Alzheimer disease and cognitive decline. Ann Intern Med. 2010;153:176–181. doi: 10.7326/0003-4819-153-3-201008030-00260 [DOI] [PubMed] [Google Scholar]
- 28. Welmer AK, Angleman S, Rydwik E, Fratiglioni L, Qiu C. Cardiovascular burden and mobility limitation in older adults. PLoS One. 2013;8:e65815. doi: 10.1371/journal.pone.0065815 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Fitzgerald JD, Johnson L, Hire DG, Ambrosius WT, Anton SD, Dodson JA, Marsh AP, McDermott MM, Nocera JR, Tudor‐Locke C, et al. Physical activity and cardiovascular risk in older adults. J Am Heart Assoc. 2015;4:e001288. doi: 10.1161/JAHA.114.001288 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Newman AB, Simonsick EM, Naydeck BL, Boudreau RM, Kritchevsky SB, Nevitt MC, Pahor M, Satterfield S, Brach JS, Studenski SA, et al. Long‐distance walk performance and cardiovascular outcomes. JAMA. 2006;295:2018–2026. doi: 10.1001/jama.295.17.2018 [DOI] [PubMed] [Google Scholar]
- 31. Hardy SE, Kang Y, Studenski SA, Degenholtz HB. Walking ability predicts disability and mortality. J Gen Intern Med. 2011;26:130–135. doi: 10.1007/s11606-010-1543-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Montero‐Odasso M, Speechley M, Muir‐Hunter SW, Sarquis‐Adamson Y, Sposato LA, Hachinski V, Borrie M, Wells J, Black A, Sejdić E, et al. Motor and cognitive trajectories before dementia. J Am Geriatr Soc. 2018;66:1676–1683. doi: 10.1111/jgs.15341 [DOI] [PubMed] [Google Scholar]
- 33. Dodson JA, Hajduk AM, Murphy TE, Geda M, Krumholz HM, Tsang S, Nanna MG, Tinetti ME, Goldstein D, Forman DE, et al. Readmission risk in older adults with myocardial infarction. Circ Cardiovasc Qual Outcomes. 2019;12:e005320. doi: 10.1161/CIRCOUTCOMES.118.005320 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Stocker H, Perna L, Weigl K, Möllers T, Schöttker B, Thomsen H, Holleczek B, Rujescu D, Brenner H. Polygenic risk score and APOE in dementia diagnosis. Mol Psychiatry. 2021;26:5812–5822. doi: 10.1038/s41380-020-0764-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Livingston G, Huntley J, Sommerlad A, Ames D, Ballard C, Banerjee S, Brayne C, Burns A, Cohen‐Mansfield J, Cooper C, et al. Dementia prevention, intervention, and care: 2020 report of the lancet commission. Lancet. 2020;396:413–446. doi: 10.1016/S0140-6736(20)30367-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Williamson JD, Supiano MA, Applegate WB, Berlowitz DR, Campbell RC, Chertow GM, Fine LJ, Haley WE, Hawfield AT, Ix JH, et al. Intensive blood pressure control and cardiovascular outcomes in older adults. JAMA. 2016;315:2673–2682. doi: 10.1001/jama.2016.7050 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Williamson JD, Pajewski NM, Auchus AP, et al. Blood pressure control and probable dementia. JAMA. 2019;321:553–561. doi: 10.1001/jama.2018.21442 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Kumar M, Orkaby A, Tighe C, Villareal DT, Billingsley H, Nanna MG, Kwak MJ, Rohant N, Patel S, Goyal P, et al. Life's essential 8 and cardiovascular risk in older adults. JACC Adv. 2023;2:100560. doi: 10.1016/j.jacadv.2023.100560 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Nanna MG, Sutton NR, Kochar A, Rymer JA, Lowenstern AM, Gackenbach G, Hummel SL, Goyal P, Rich MW, Kirkpatrick JN, et al. Percutaneous coronary interventions in older adults. JACC Adv. 2023;2:100421. doi: 10.1016/j.jacadv.2023.100421 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Nanna MG, Sutton NR, Kochar A, Rymer JA, Lowenstern AM, Gackenbach G, Hummel SL, Goyal P, Rich MW, Kirkpatrick JN, et al. Assessment of older adults undergoing PCI: a JACC expert panel. JACC Adv. 2023;2:100389. doi: 10.1016/j.jacadv.2023.100389 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Goyal P, Kwak MJ, Al Malouf C, et al. Geriatric cardiology: current perspectives. JACC Adv. 2022;1:100070. doi: 10.1016/j.jacadv.2022.100070 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Yin J, John A, Cadar D. Depressive symptoms and cognitive function over time. JAMA Netw Open. 2024;7:e2416305. doi: 10.1001/jamanetworkopen.2024.16305 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Tota‐Maharaj R, Blaha MJ, McEvoy JW, et al. Coronary artery calcium for mortality prediction in young and older adults. Eur Heart J. 2012;33:2955–2962. doi: 10.1093/eurheartj/ehs230 [DOI] [PubMed] [Google Scholar]
- 44. Yano Y, O'Donnell CJ, Kuller L, et al. Coronary artery calcium score and cardiovascular risk in older adults. JAMA Cardiol. 2017;2:986–994. doi: 10.1001/jamacardio.2017.2498 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Obisesan OH, Boakye E, Wang FM, Dardari Z, Dzaye O, Cainzos‐Achirica M, Meyer ML, Gottesman R, Palta P, Coresh J, et al. Coronary artery calcium as a marker of aging in adults aged 75 and older. Atherosclerosis. 2024;392:117475. doi: 10.1016/j.atherosclerosis.2024.117475 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Tables S1–S5
Figures S1–S3