Abstract
Objectives
Our first objective was to estimate empirically-derived subgroups (latent profiles) of observed carbohydrate, protein, and fat intake density in a nationally representative sample of older U.S. adults. Our second objective was to determine whether membership in these groups was associated with levels of, and short term change in, physical mobility limitations.
Design and Setting
Measures of macronutrient density were taken from the 2013 Health Care and Nutrition Study, an off-year supplement to the Health and Retirement Study, which provided indicators of physical mobility limitations and sociodemographic and health-related covariates.
Participants
3,914 community-dwelling adults age 65 years and older.
Measurements
Percent of daily calories from carbohydrate, protein, and fat were calculated based on responses to a modified Harvard food frequency questionnaire. Latent profile analysis was used to describe unobserved heterogeneity in measures of carbohydrate, protein, and fat density. Mobility limitation counts were based on responses to 11 items indicating physical limitations. Poisson regression models with autoregressive controls were used to identify associations between macronutrient density profile membership and mobility limitations. Sociodemographic and health-related covariates were included in all Poisson regression models.
Results
Four latent subgroups of macronutrient density were identified: “High Carbohydrate Moderate with Fat” “Moderate” and “Low Carbohydrate/High Fat”. Older adults with the lowest percentage of daily calories coming from carbohydrate and the greatest percentage coming from fat (“Low Carbohydrate/High Fat”) were found to have greater reported mobility limitations in 2014 than those identified as having moderate macronutrient density, and more rapid two-year increases in mobility limitations than those identified as “Moderate with Fat” or “Moderate”.
Conclusion
Older adults identified as having the lowest carbohydrate and highest fat energy density were more likely to report a greater number of mobility limitations and experience greater increases in these limitations than those identified as having moderate macronutrient density. These results suggest that the interrelation of macronutrients must be considered by those seeking to reduce functional limitations among older adults through dietary interventions.
Key words: Macronutrient density, mobility limitations, latent profile analysis, Health Care and Nutrition Study (HCNS), Health and Retirement Study (HRS)
Introduction
The older adult population is challenged by age-related declines in muscle mass and function, reducing their ability to perform tasks of daily living and remain independent (1). Functional limitations are significantly increasing in older adults (2), resulting in rising health care costs and need for long-term care services (3). As the disablement process proceeds from disease pathology to impairment in body systems that limit basic abilities and result in restricted capacity to function (4), identifying potentially modifiable risk factors to reduce the development and impact of functional limitations can lessen the burden of disability on our growing population of older adults.
Nutritional status is associated with the risk of physical limitations, disability, and frailty in older adults (5, 6). There has been significant research on the importance of adequate protein intakes in older adults for maintenance of muscle mass, function, and mobility (7, 8, 9, 10), yet there is limited research on the role of the other macronutrients and the interrelation between macronutrients (11). The contribution of macronutrients to overall energy intake can vary considerably among individuals (12), but the impact of these varying dietary patterns on risk of mobility limitations is largely unknown. Furthermore, there are limited prospective, population-based studies of older adults that examine the relationship between dietary intakes and progression of functional limitations.
To identify whether variation in the percentage of daily calories coming from carbohydrate, protein, and fat, hereafter referred to as macronutrient density, was associated with change in physical mobility limitations among older adults, our first goal was to estimate empirically-derived latent subgroups of macronutrient density. Second, we investigated whether the latent profiles of macronutrient density identified in step one were associated with reported levels of, and short-term change in, physical mobility limitations.
Methods
Data and Participants
Observations were drawn from the Health and Retirement Study (HRS), a biennial national panel survey of older Americans funded by the National Institute on Aging and the Social Security Administration (grant number: NIA U01AG009740) (13). The 2013 Health Care and Nutrition Study (HCNS), an off-year mail-out HRS supplement, collected information about food consumption using a modified Harvard food frequency questionnaire originally proposed by Willett and colleagues (14). The University of Michigan's institutional review board approved the HRS protocol and participants were read a confidentiality statement and provided oral or implied consent when first contacted and were provided a written informed consent form at each interview (15).
The HCNS contains 8,073 observations with complete information on consumption of 170 food items and calculated macronutrient and micronutrient content. Participants with implausible daily energy intakes of less than 500 kilocalories or greater than 5,000 kilocalories (n = 97) were excluded to reduce outlier bias in estimation of the latent profiles. Individuals were removed from the analytic sample if they were under the age of 65 (n = 3,792), were missing population weighting values from the 2014 HRS (n = 250), or had missing physical mobility limitation measures in 2014 (n = 20), resulting in an analytic sample of 3,914. Supplementary online tables provide comparison of the analytic sample to those with complete HCNS data but who were excluded from analyses. Briefly, measures of macronutrient density and measures related to fat intake were similar among those included and excluded from the analytic sample and the analytic sample generally reflected characteristics of the older adults being analyzed, including a greater number of mobility limitations, lower mean daily calories, a greater number of chronic conditions, and being more likely to report no vigorous physical activity. (Supplementary online tables).
Indicators of physical mobility limitation were taken from the 2012 and 2014 core HRS files, and the RAND HRS data file (Version P) (16), produced by the RAND Center for the Study of Aging with funding from the U.S. National Institute on Aging and the U.S. Social Security Administration, provided all participant characteristics measured in 2014. All measures were collected through participant self-report and proxy response was allowed when the respondent was unable to complete the interview.
Measures
Macronutrient Density
Estimated daily calories from separate macronutrient sources were taken from the HCNS constructed variables dataset. Estimated daily calories from carbohydrate, protein, and fat were used to calculate the estimated percent of daily calories coming from each macronutrient. The proportion of each macronutrient's contribution to estimated daily calories (% kcal) has also been described as nutrient density or energy percentage (En %).
Physical Mobility Limitations
Physical mobility limitations, hereafter referred to as mobility limitations, were operationalized as a count of 11 indicators of limitation in physical mobility (17, 18). Respondents were asked whether they had difficulty in each of the following activities: stooping or crouching, climbing one flight of stairs without resting, climbing several flights of stairs without resting, moving large objects, sitting in a chair for two hours, getting up from a chair after sitting for long periods, lifting weights more than 10 pounds, raising arms above shoulder level, walking one block, walking several blocks, and picking up a dime from a table (1 = yes; 0 = no). These items have been shown to load together with a Kuder-Richardson Formula 20 of .85 (19). Respondents with seven or more missing values on these 11 items were coded as missing. Due to non-normality in the distribution of mobility limitations in both 2012 and 2014, mobility limitations in 2014 were modeled as a count outcome using a Poisson distribution and quartiles identified for mobility limitations reported in 2012 were used as controls in the autoregressive Poisson models.
Mobility limitations represent one possible operationalization of physical functioning in older adulthood, with other popular measures including limitations in activities of daily living (ADLs) and limitations in instrumental activities of daily living (IADLs). Mobility limitations represent an indicator of underlying physical capacity that are less dependent on built environment, social roles, and assistive technology than ADLs and IADLs (20, 21). Mobility limitations represent less severe physical limitations than ADLs and IADLs, providing a leading indicator of physical limitation that may develop into more severe disability.
Participant Characteristics
A number of measures collected during the 2014 interview were included as covariates in models predicting mobility limitations. Analyses controlled for respondents' gender (1 = female, 0 = male), age, race/ethnicity (White (reference), Black, Hispanic, Other), and marital status (1 = partnered or married, 0 = single, divorced, or widowed). Education (< 12 years of education, 12 years of education (reference), > 12 years of education), longest occupational tenure (white-collar (reference), blue-collar, female homemaker, other occupational tenure), retirement status (1 = retired, 0 = not retired), and log-transformed household income and assets accounted for potential associations between mobility limitations, socioeconomic status, and employment history. Covariates indicating health behaviors included body mass index (BMI; underweight (BMI < 18.5 kg/m2), normal weight (18.5 kg/m2 ≤ BMI < 25 kg/m2; reference), overweight (25 kg/m2 ≤ BMI < 30 kg/m2), obese (BMI ≥ 30 kg/m2)), vigorous physical activity (participation in activities such as sports, heavy housework, or a job that involves physical labor; no vigorous physical activity (reference), vigorous physical activity less than 1 time per week, vigorous physical activity more than one time per week), current smoking status (1 = current smoker, 0 = not current smoker), and alcohol consumption (non-drinkers (reference), moderate drinkers (men drinking between 1 and 14 drinks per week and females drinking between 1 and 7 drinks per week), and heavy drinkers (men drinking more than 14 drinks per week and females drinking more than 7 drinks per week)). Chronic conditions were measured as the count of doctor-diagnosed high blood pressure, cancer, diabetes, lung disease, heart problems, stroke, psychiatric problems, and arthritis. A measure of estimated daily calories was included to capture variation in absolute nutrient intake unaccounted for by macronutrient density. Finally, an indicator of proxy response in 2014 was included to account for completion of the mobility limitation questions by representative on behalf of the primary respondent.
Statistical Methods
Latent Profile Analysis of Macronutrient Density
Latent profile analysis (LPA) was used to identify underlying subgroups of macronutrient density. LPA is a form of finite mixture modeling allowing classification of unobserved heterogeneity in responses to multiple observed continuous variables (22). To identify the number of profiles that best represented the underlying response distributions, model fit tests were conducted using the Vuong-Lo-Mendell-Ruben (VLMR) likelihood ratio test of whether the estimated model significantly improved model fit compared to a model with one less profile. Model fit indices including Akaike information criterion (AIC), Bayesian information criterion (BIC), and sample-size adjusted AIC (SSA-AIC) were also used to determine if identification of additional profiles improved model fit. Model entropy, representing the accuracy of assigning individuals to profiles, was also considered.
Autoregressive Poisson Modeling of Mobility Limitations
To examine if membership in the latent macronutrient density profiles was associated with reported mobility limitations, Poisson regression was used to account for the nonnormal count distribution of mobility limitations (23). Variation in levels of mobility limitations reported in 2014 was analyzed using standard Poisson regression and short-term change in mobility limitations from 2012 to 2014 was assessed by regressing 2014 mobility limitation count on mobility limitation quartiles identified in 2012, described as a generalization of the autoregressive model or as the regressor variable method of analyzing change scores (23, 24).
The LPA and Poisson regression models were conducted using maximum likelihood estimation with robust standard errors, providing treatment of missing data with maximum likelihood, estimation of standard errors robust to nonnormality, and adjustment for correlations between independent variables. Both LPA and Poisson regression models adjusted for person-level weighting, household clustering, and stratum used to calculate sampling error, resulting in estimates representing the U.S. population of community-dwelling older adults age 65 and over in 2014. Mplus version 7.3 was used to conduct all LPA and Poisson regression models (25). SAS version 9.4 (26) was used to conduct supplementary analyses. Bivariate ANOVA with Tukey post-hoc tests were used to test mean differences across profiles for continuous variables. Chi-square with pairwise statistical tests using the Bonferroni correction for multiple testing were used to identify significant differences in categorical variables across profiles with standardized residuals used to estimate the direction of observed versus expected cell frequencies. Variance inflation factor (VIF) and Durbin-Watson D are reported as diagnostics for multicollinearity and serial correlation in the predictive models.
Results
Identification of Latent Macronutrient Density Profiles
Table 1 presents model fit statistics for LPA model solutions estimating between one and five distinct profiles of carbohydrate, protein, and fat macronutrient density. As the number of profiles identified increased from one through five, AIC, BIC, aBIC, and entropy measures reflected better model fit. The VLMR test indicated that the four-class model was a significant improvement on the three-class solution (p < .001), but the five-class solution did not provide significantly better fit than the four-class model (p = .206). Based on model fit tests and the necessity of parsimony, the four-class solution was identified as the best description of the latent macronutrient density profiles.
Table 1.
Model fit statistics for latent profile analysis of macronutrient density
| Number of profiles | VLMR | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| df | LL | AIC | aBIC | Entropy | Mean | SD | p | ||
| 1 | 6 | -36900.10 | 73812.20 | 73849.83 | 73830.76 | - | - | - | - |
| 2 | 10 | -36028.24 | 72076.48 | 72139.20 | 72107.43 | 0.64 | 24.55 | 25.17 | 0.000 |
| 3 | 14 | -35593.05 | 71214.10 | 71301.91 | 71257.43 | 0.73 | 105.48 | 151.50 | 0.004 |
| 4 | 18 | -35339.86 | 70715.73 | 70828.63 | 70771.43 | 0.76 | -62.47 | 152.50 | 0.000 |
| 5 | 22 | -35144.98 | 70333.95 | 70471.94 | 70402.04 | 0.80 | 249.21 | 327.61 | 0.206 |
Notes: AIC = Akaike information criterion, BIC = Bayesian information criterion, SSA-BIC = sample size adjusted Bayesian information criterion; VLMR = Vuong-Lo-Mendell-Rubin Likelihood ratio test.
Description of Latent Profiles and Participant Characteristics
Table 2 presents descriptive statistics for macronutrient density, mobility limitations in 2014 and 2012, and the average number of mobility limitations in each quartile identified in 2012 for the overall sample and by latent profile. Table 3 presents participant characteristics measured in 2014.
Table 2.
Descriptive statistics for macronutrient density and mobility limitations
| Identified Latent Profiles | |||||||
|---|---|---|---|---|---|---|---|
| Macronutrient Density Mean ± SD | Complete Sample (n = 3,914) | High Carbohydrate (n = 279) | Moderate with Fat (n = 1,786) | Moderate (n = 1,504) | Low Carbohydrate/High Fat (n = 345) | Overall Test | |
| F | p | ||||||
| % daily kcal carbohydrate | 53.48±8.24 | 69.38±4.10 | 49.35±3.87 | 58.65±3.45 | 39.43±4.44 | 2637.42 | >.001 |
| % daily kcal protein | 16.45±3.32 | 13.92±3.04 | 17.03±3.17 | 15.78±2.95 | 18.39±3.91 | 147.00 | >.001 |
| % daily kcal fat | 34.55±6.44 | 23.72±3.41 | 37.30±3.82 | 30.80±3.31 | 45.46±4.77 | 4857.95 | >.001 |
| Mobility Limitations | |||||||
| Count 2014 | 3.39±3.06 | 3.77±3.15 | 3.36±3.02 | 3.38±3.07 | 3.35±3.08 | 1.52 | 0.207 |
| Count 2012 | 3.01±2.92 | 3.30±2.99 | 3.00±2.88 | 3.03±2.98 | 2.83±2.80 | 1.38 | 0.248 |
| 2012 MobilityLimitation Quartiles | |||||||
| 1st quartile | 0.00±0.00 | 0.00±0.00 | 0.00±0.00 | 0.00±0.00 | 0.00±0.00 | 15.26 | 0.084 |
| 2nd quartile | 1.44±0.50 | 1.45±0.50 | 1.46±0.50 | 1.42±0.49 | 1.39±0.49 | ||
| 3rd quartile | 3.92±0.81 | 3.93±0.81 | 3.89±0.80 | 3.97±0.80 | 3.89±0.85 | ||
| 4th quartile | 7.63±1.46 | 7.81±1.48 | 7.57±1.45 | 7.69±1.48 | 7.48±1.42 | ||
Notes: a-c When overall test identified as significant, means in a row without a common superscript letter differ (p < .05), as analyzed by one-way ANOVA and Tukey post-hoc tests.
Table 3.
Descriptive statistics for covariates
| Identified Latent Profiles | |||||||
|---|---|---|---|---|---|---|---|
| Continuous Covariates Mean ± SD | Complete Sample (n = 3,914) | High Carbohydrate (n = 279) | Moderate with Fat (n = 1,786) | Moderate (n = 1,504) | Low Carbohydrate /High Fat (n = 345) | Overall test | |
| F | p | ||||||
| Age | 76.91±6.79 | 78.54±6.66a | 76.42±6.68 | 77.63±6.95a | 75.01±6.09 | 23.53 | >.001 |
| Estimated daily calories | 1,779±750 | 1,591±710a | 1,821 ±765b,c | 1,773±749b,d | 1,737±681a,c,d | 8.16 | >.001 |
| Household income ($) | 58,904±79,412 | 37,210±33,282 | 63,441±88,814a | 55,510±73,496 | 67,757±75,439a | 28.22 | >.001 |
| Household assets ($) | 533,773±1,111,252 | 397,298±1,277,31a | 597,178±1,318,785b,c | 469,099±789,607a,b | 597,843 ±959,782c | 5.04 | 0.002 |
| # chronic conditions | 2.61±1.38 | 2.65±1.38 | 2.66±1.38 | 2.54±1.38 | 2.54±1.42 | 2.31 | 0.07 |
| Categorical Covariates n (%) | Identified Latent Profiles | ||||||
| Complete Sample(n = 3,914) | High Carbohydrate(n = 279) | Moderate with Fat(n = 1,786) | Moderate(n = 1,504) | Low Carbohydrate/High Fat (n = 345) | Overall test | ||
| χ2 | p | ||||||
| Female | 2,296(58.66) | 186(66.67)1 | 1,002(56.10)2 | 900(59.84) | 208(60.29) | 13.43 | 0.004 |
| White | 3,086(78.85) | 182(65.23)2 | 1,477(82.70)1 | 1,125(74.80)2 | 302(87.54)1 | 85.35 | >.001 |
| Black | 457(11.68) | 54(19.35)1 | 174(09.74)2 | 206(13.70)1 | 23(06.67)2 | ||
| Hispanic | 288(07.36) | 30(10.75) | 106(05.94)2 | 141(09.38)1 | 11(03.19)2 | ||
| Other race/ethnicity | 82(02.10) | 13(04.66)1 | 29(01.62) | 32(02.13) | 8(02.32) | ||
| Married/partnered | 2,286(58.41) | 136(48.75)2 | 1,103(61.76)1 | 832(55.32)2 | 215(62.32) | 27.05 | >.001 |
| Retired | 2,981(77.35) | 222(80.73) | 1,338(75.85) | 1,175(79.72)1 | 246(72.14) | 14.04 | 0.003 |
| < HS Degree | 777(19.85) | 71(25.45) | 320(17.92) | 337(22.41)1 | 49(14.20) | 33.41 | >.001 |
| HS Degree | 1,416(36.18) | 107(38.35) | 648(36.28) | 547(36.37) | 114(33.04) | ||
| > HS Degree | 1,721(43.97) | 101(36.20) | 818(45.80) | 620(41.22) | 182(52.75)1 | ||
| White-collar occ. tenure | 2,117(54.09) | 136(48.75) | 988(55.32) | 767(51.00)2 | 226(65.51)1 | 35.52 | >.001 |
| Blue-collar occ. tenure | 1,274(32.55) | 103(36.92) | 572(32.03) | 510(33.91) | 89(25.80) | ||
| Homemaker occ. tenure | 94(02.40) | 12(04.30) | 37(02.07) | 41(02.73) | 4(01.16) | ||
| Other occ. tenure | 430(10.99) | 28(10.04) | 189(10.58) | 187(12.43) | 26(07.54) | ||
| BMI underweight | 82(02.12) | 10(03.61) | 36(02.04) | 35(02.35) | 1(00.29) | 41.28 | >.001 |
| BMI normal | 1,171(30.22) | 104(37.55) | 499(28.21) | 474(31.83) | 94(27.65) | ||
| BMI overweight | 1,482(38.25) | 100(36.10) | 650(36.74) | 598(40.16) | 134(39.41) | ||
| BMI obese | 1,140(29.42) | 63(22.74) | 584 (33.01)1 | 382(25.65)2 | 111(32.65) | ||
| No vigorous activity | 2,338(59.92) | 180(64.98) | 1,057(59.28) | 896(59.81) | 205(59.59) | 4.40 | 0.622 |
| Some vigorous activity | 671 (17.20) | 37 (13.36) | 312 (17.50) | 259 (17.29) | 63 (18.31) | ||
| Regular vigorous activity | 893 (22.89) | 60 (21.66) | 414 (23.22) | 343 (22.90) | 76 (22.09) | ||
| Current smoker | 263 (06.77) | 14 (05.07) | 144 (08.13)1 | 73 (04.89)2 | 32 (09.33) | >.001 | |
| No alcohol | 2,531 (65.76) | 228 (82.91)1 | 1,077 (61.19)2 | 1,018 (69.11)1 | 208 (61.00) | >.001 | |
| Moderate alcohol | 1,136 (29.51) | 45 (16.36)2 | 573 (32.56)2 | 413 (28.04) | 105 (30.79) | ||
| Heavy alcohol | 182 (04.73) | 2 (00.73)2 | 110 (06.25)2 | 42 (02.85)2 | 28 (08.21)1 | ||
| Proxy interview | 149 (03.81) | 11 (03.94) | 64 (03.58) | 63 (04.19) | 11 (03.19) | 1.22 | 0.749 |
Notes: a-d When overall test identified as significant, means in a row without a common superscript letter differ (p < .05), as analyzed by one-way ANOVA and Tukey post-hoc tests; Superscript number represents identification of significant pairwise test with Bonferroni correction (p < .05)
significantly greater observed versus expected count
significantly fewer observed versus expected count in cell.
The four latent profiles that best explained unobserved heterogeneity in observed macronutrient density are displayed in Figure 1. Profile names were based on where each profile fell in relation to other profiles and the Acceptable Macronutrient Distribution Ranges (AMDRs) described in the Dietary Reference Intakes (27) as shown on Figure 1. Respondents with high carbohydrate density relative to other profiles and above the upper carbohydrate AMDR were labeled “High Carbohydrate”(n = 279). Older adults with a more balanced macronutrient density profile but who still had average fat intake higher than the recommended fat AMDR were labeled “Moderate with Fat”(n = 1,786), and those with carbohydrate, protein, and fat intake densities all within the AMDRs for these macronutrients were identified as “Moderate” (n = 1,504). Respondents with the lowest percentage of daily calories coming from carbohydrate and highest percentage of daily calories coming from fat, both of which fell outside the relevant AMDR threshold, were labeled “Low Carbohydrate/High Fat” (n = 345). Carbohydrate, protein, and fat macronutrient density were significantly different across all latent profiles.
Figure 1.
Latent Profiles of Macronutrient Density, 2013 HCNS
For all respondents, an average of 3.39 mobility limitations (SD ± 3.06) were reported in 2014 and bivariate analyses indicated that mobility limitations did not vary across macronutrient profile. In 2012, an average of 3.01 mobility limitations (SD ± 2.92) were reported, which again did not vary across latent profile. In regards to the 2012 mobility limitation quartiles used as controls in the autoregressive Poisson model, the average number of mobility limitations in the first, second, third, and fourth quartile were 0.00 (SD ± 0.00), 1.44 (SD ± 0.50), 3.92 (SD ± 0.81), and 7.63 (SD ± 1.46), respectively. The number of individuals identified in each of the 2012 mobility limitation quartiles was not found to significantly vary by macronutrient profile.
Regarding select participant characteristics measured in 2014 as displayed in Table 3, the mean age of the overall sample was 76.91 (SD ± 6.79), with those in the “Low Carbohydrate/High Fat” group on average being significantly younger than members of all other profiles. Members of the “High Carbohydrate” group had significantly lower daily caloric intake than those in the “Moderate with Fat” and “Moderate” profiles. Those in the “High Carbohydrate” and “Moderate” profiles tended to have lower household income and assets than those in the “Moderate with Fat” and “Low Carbohydrate/High Fat” profiles. Older adults in the “High Carbohydrate” profile were more likely to be female, Black, and report consuming no alcohol, than expected. Individuals identified in the “Low Carbohydrate/High Fat” profile were more likely to be White, have greater than a high school degree, report white-collar occupational tenure, and report heavy alcohol use, than expected.
Poisson Regression Results
Exponentiated Poisson regression estimates, 95% confidence intervals, VIFs, and Durbin-Watson D are presented for the mobility limitation models in Table 4. Both VIFs and Durbin-Watson D fell within acceptable ranges [28,29]. Model 1 regressed mobility limitations in 2014 on macronutrient density profile membership using the “Low Carbohydrate/High Fat” profile as the reference category, controlling for participant characteristics measured in 2014. Individuals identified as members of the “Moderate” macronutrient density profile had an expected mobility limitation count 12.7% less than those in the “Low Carbohydrate/High Fat” reference category [100 * (exp (-0.14) − 1) = -12.72]. Age and number of chronic diseases were positively associated with mobility limitations and household income and assets were both negatively associated with mobility limitations. Females, Blacks, Hispanics, those who were retired, those reporting blue-collar or other occupational tenure, those who were obese, and current smokers had significantly more predicted mobility limitations than those in their respective reference categories. Individuals with greater than a high school degree, reported any vigorous physical activity, and reported either moderate or heavy alcohol consumption had fewer predicted mobility limitations than their reference counterparts.
Table 4.
Poisson regression estimates from mobility limitation count and mobility limitation autoregressive models
| Model 1 Mobility Limitation Count | Model 2Mobility Limitation Change | |||||||
|---|---|---|---|---|---|---|---|---|
| Exp(B) | 95% CI | p | VIF | Exp(B) | 95% CI | p | VIF | |
| Intercept | 2.63 | [2.23;3.10] | >.001 | - | 0.94 | [0.79;1.13] | 0.511 | - |
| Latent Profile | ||||||||
| (“Low Carbohydrate /High Fat reference) | ||||||||
| High Carbohydrate | 0.91 | [0.79;1.04] | 0.165 | 1.76 | 0.94 | [0.84;1.04] | 0.205 | 1.76 |
| Moderate with Fat | 0.92 | [0.82;1.02] | 0.108 | 3.39 | 0.92 | [0.85;1.00] | 0.041 | 3.38 |
| Moderate | 0.87 | [0.78;0.98] | 0.014 | 3.41 | 0.90 | [0.83; 0.98] | 0.008 | 3.40 |
| 2012 Mobility Limitations | ||||||||
| (1st quartile reference) | ||||||||
| Second quartile | 2.45 | [2.11;2.85] | >.001 | 1.59 | ||||
| Third quartile | 4.26 | [3.66;4.96] | >.001 | 1.76 | ||||
| Fourth quartile | 6.30 | [5.39;7.36] | >.001 | 2.02 | ||||
| Covariates | ||||||||
| Age | 1.02 | [1.02;1.03] | >.001 | 1.35 | 1.01 | [1.01;1.01] | >.001 | 1.38 |
| Female | 1.22 | [1.14;1.31] | >.001 | 1.29 | 1.08 | [1.02;1.14] | 0.009 | 1.32 |
| Black | 1.10 | [1.00;1.20] | 0.039 | 1.13 | 1.07 | [1.00;1.14] | 0.042 | 1.13 |
| Hispanic | 1.12 | [1.00;1.25] | 0.045 | 1.17 | 1.12 | [1.01;1.24] | 0.029 | 1.18 |
| Other race/ethnicity | 1.03 | [0.86;1.23] | 0.780 | 1.03 | 0.98 | [0.86;1.13] | 0.821 | 1.03 |
| Married/partnered | 1.01 | [0.94;1.09] | 0.833 | 1.47 | 0.99 | [0.94;1.05] | 0.840 | 1.47 |
| Retired | 1.21 | [1.11;1.32] | >.001 | 1.13 | 1.08 | [1.01;1.16] | 0.016 | 1.14 |
| < HS degree | 1.06 | [0.98;1.14] | 0.171 | 1.44 | 1.02 | [0.96;1.09] | 0.450 | 1.44 |
| > HS degree | 0.93 | [0.87;1.00] | 0.042 | 1.44 | 0.98 | [0.93;1.03] | 0.455 | 1.45 |
| Blue-collar occ. tenure | 1.09 | [1.01;1.17] | 0.022 | 1.46 | 1.06 | [1.00;1.12] | 0.038 | 1.47 |
| White-collar occ. tenure | 1.10 | [0.92;1.33] | 0.286 | 1.10 | 1.08 | [0.95;1.24] | 0.243 | 1.10 |
| Other occ. tenure | 1.11 | [1.02;1.21] | 0.016 | 1.13 | 1.02 | [0.94;1.10] | 0.613 | 1.13 |
| Household income (log) | 0.95 | [0.91;0.99] | 0.006 | 1.62 | 0.99 | [0.96;1.01] | 0.361 | 1.66 |
| Household assets (log) | 0.99 | [0.99;1.00] | 0.036 | 1.25 | 1.00 | [0.99;1.01] | 0.703 | 1.25 |
| BMI underweight | 1.11 | [0.92;1.34] | 0.276 | 1.07 | 1.00 | [0.87;1.16] | 0.958 | 1.07 |
| BMI overweight | 1.08 | [1.00;1.16] | 0.053 | 1.48 | 0.99 | [0.93;1.05] | 0.729 | 1.49 |
| BMI obese | 1.31 | [1.21;1.42] | >.001 | 1.62 | 1.08 | [1.01;1.15] | 0.015 | 1.66 |
| Some vigorous activity | 0.69 | [0.63;0.76] | >.001 | 1.14 | 0.86 | [0.79;0.92] | >.001 | 1.16 |
| Regular vigorous activity | 0.58 | [0.53;0.64] | >.001 | 1.18 | 0.73 | [0.68;0.79] | >.001 | 1.21 |
| Current smoker | 1.13 | [1.01;1.27] | 0.031 | 1.08 | 1.06 | [0.97;1.16] | 0.165 | 1.08 |
| Moderate alcohol | 0.83 | [0.77;0.90] | >.001 | 1.17 | 0.93 | [0.88;0.98] | 0.010 | 1.17 |
| Heavy alcohol | 0.85 | [0.74;0.99] | 0.029 | 1.07 | 0.88 | [0.79;0.98] | 0.021 | 1.08 |
| # chronic conditions | 1.20 | [1.17;1.23] | >.001 | 1.14 | 1.07 | [1.05;1.09] | >.001 | 1.29 |
| Estimated daily calories | 1.02 | [0.98; 1.06] | 0.248 | 1.03 | 1.01 | [0.98;1.04] | 0.451 | 1.04 |
| Proxy Interview | 1.23 | [1.06;1.41] | 0.004 | 1.04 | 1.12 | [0.98;1.27] | 0.087 | 1.05 |
| Durbin-Watson D | 1.97 | 2.00 | ||||||
Notes: Reference categories: race/ethnicity (White), education (HS degree), occupational tenure (white-collar), BMI (normal), vigorous physical activity (no vigorous physical activity), alcohol use (no alcohol use); †VIF – variance inflation factor calculated using complete-case analysis.
To investigate short-term change in mobility limitations, Model 2 included all independent variables that were included in Model 1 with the addition of 2012 mobility limitation quartiles, using those in the first quartile as reference. Mobility limitation quartiles identified in 2012 displayed a positive monotonic association with 2014 mobility limitations where membership in upper quartiles was associated with more predicted mobility limitations in 2014 than was membership in lower quartiles. Members in either the “Moderate with Fat” or “Moderate” macronutrient profiles had significantly less rapid increase in mobility limitations than those in the “Low Carbohydrate/High Fat” reference category. Specifically, respondents identified in the “Moderate with Fat” profile had an expected increase in mobility limitations 7.60% less than [100 * (exp (-0.08) − 1) = -07.60], and those is the “Moderate” profile had an expected increase in mobility limitations 10.15% less than [100 * (exp (-0.11) − 1) = -10.15], respondents in the “Low Carbohydrate/High Fat” profile, respectively.
Follow-Up Analysis
To investigate whether the composition of fat intake differed across latent profile, overall fat intake was separated into saturated, monounsaturated, and polyunsaturated sources, then both percent of overall daily calories attributable to each fat type and percent of daily fat intake represented by each type of fat was calculated. Bivariate ANOVA with Tukey post-hoc tests were used to test whether overall daily calories and daily fat calories attributable to each type of fat differed across the identified latent profiles (Table 5).
Table 5.
Fat type as percentage of daily calories and fat type as percentage of daily fat intake by macronutrient density profile
| Identified Latent Profiles | |||||||
|---|---|---|---|---|---|---|---|
| Fat Type as % of Daily Calories Mean ± SD | Complete Sample (n = 3,914) | High Carbohydrate (n = 279) | Moderate with Fat (n = 1,786) | Moderate (n = 1,504) | Low Carbohydrate /High Fat(n = 345) | Overall test | |
| F | p | ||||||
| Saturated Fat | 11.74±2.69 | 7.93±1.73 | 12.73±2.18 | 10.54±1.85 | 14.95±2.53 | 902.01 | >.001 |
| Monounsaturated Fat | 12.24±2.78 | 8.09±1.47 | 13.32±1.86 | 10.70±1.51 | 16.75±2.68 | 1794.92 | >.001 |
| Polyunsaturated Fat | 7.08±2.16 | 5.20±1.36 | 7.49±1.97 | 6.44±1.63 | 9.26±3.05 | 315.46 | >.001 |
| Fat Type as % of Daily Fat Intake | |||||||
| Saturated Fat | 34.02±4.72 | 33.32±4.86a | 34.15±4.67b | 34.21±4.62b | 33.07±5.18a | 8.28 | >.001 |
| Monounsaturated Fat | 35.32±3.46 | 34.01±3.39 | 35.73±3.38 | 34.74±3.23 | 36.81±4.03 | 58.80 | >.001 |
| Polyunsaturated Fat | 20.52±4.71 | 21.99±4.86 | 20.04±4.55a | 20.89±4.64b | 20.19±5.30a,b | 18.51 | >.001 |
Notes: a-d When overall test identified as significant, means in a row without a common superscript letter differ (p < .05), as analyzed by one-way ANOVA and Tukey post-hoc tests.
Members of the “Low Carbohydrate/High Fat” profile had the greatest percentage of daily calories coming from saturated, monounsaturated, and polyunsaturated fat, with those in the “High Carbohydrate” profile having the lowest percentage of daily calories attributable to the separate fats. In the complete sample, around 34% of overall fat intake came from sources of saturated fat, 35% came from monounsaturated sources, and close to 21% came from polyunsaturated sources. The percent of overall fat intake attributable to each fat source significantly differed across latent profiles, though the absolute differences in fat source as percentage of overall fat intake were generally small.
Conclusion
Using LPA we identified four unique subgroups of macronutrient consumption in the sample of U.S. adults age 65 and over. When the identified macronutrient profiles were included as predictors of reported number of mobility limitations and short-term change in mobility limitations, members of the “Low Carbohydrate/High Fat” profile were found to have both greater reported mobility limitations in 2014 than members of the “Moderate” profile, and more rapid increase in mobility limitations from 2012 to 2014 than members of either the “Moderate” or “Moderate with Fat” profiles. These results suggest that dietary patterns characterized by low carbohydrate density and high fat density may have a detrimental impact on mobility limitations among older adults.
As recently reviewed by Borg et al., the majority of older adults do not meet recommended intake of macronutrients, as intakes of energy and carbohydrates fall below reference values, while saturated fat exceeds recommendations (30). Furthermore, mean fat intakes in the general population of older adults are at the upper range of the AMDR, and over 40% of studies included in the Borg et al. review reported fat intakes that exceeded the AMDR (30). Thus, the “Low Carbohydrate/High Fat” dietary profile we found to be positively associated with mobility limitations exhibits characteristics similar to those prevalent in the general population of older adults.
Although it is well-established that inadequate protein intakes in older adults may increase the risk of age-related muscle loss and subsequent risk of frailty, falls, and mobility limitations, there is little evidence on the impact of the other macronutrients on these health outcomes (31). All four of the dietary profiles identified had protein intakes that fell within the recommended AMDR of 10-35% of kilocalories. Additionally, the “Low Carbohydrate/High Fat” profile had significantly higher protein intakes as a percentage of calories compared to all of the other dietary patterns, yet this dietary pattern was associated with greater mobility limitations. Thus, our findings support the importance of considering the balance and interrelation between macronutrients.
A negative relationship between a high fat dietary pattern and physical function has been identified in other samples of mid-life and older adults. Higher fat intakes in mid-aged women predicted greater subsequent functional limitations (32), and in a sample of very old adults, a dietary pattern characterized by higher consumption of red meats, potatoes, gravy, and butter was associated with greater declines in lower muscle strength and physical performance (33). Though members of the “Low Carbohydrate/High Fat” profile had a greater percentage of daily calories coming from saturated fat, monounsaturated fat, and polyunsaturated fat than those in other profiles, there were relatively small differences across profiles in the composition of daily fat intake. As such, the greater overall proportion these fats represent in daily calorie intake, rather than variation in fat type, may be linked to the increased mobility limitations observed in the “Low Carbohydrate/High Fat” group.
By favoring foods high in fat, individuals may be displacing nutrient dense, carbohydrate-rich foods. Inadequate intake of micronutrients and fiber found in carbohydrate-rich foods including whole grains, fruits, and vegetables may contribute to an increased risk of functional limitations as observed in the “Low Carbohydrate/High Fat” profile. Lower fruit, vegetable, and fiber intake at mid-life has been related to reporting greater functional limitations (32). The inverse association between fruit and vegetable intake and subsequent functional limitations and disability was also identified in a biracial cohort of middle aged and older adults (34). Lower intakes and serum levels of carotenoids, which are predominantly consumed through fruits and vegetables, have been shown to be predictive of declines in walking speed (35), severe walking disability (36), and frailty (37). In a Taiwanese sample of older adults, individuals who consumed higher amounts of dietary fiber performed significantly better on a battery of physical performance measures (38). Considered together, our findings contribute to the limited but growing body of evidence supporting a relationship between dietary patterns and functional limitations in older adults. More research is needed to identify the causal mechanisms though which suboptimal macronutrient distributions influence health outcomes such as disability in older adults.
Limitations and Future Directions
Though the HRS and HCNS provide valuable nationally-representative data on the health, well-being, and nutritional status of older U.S. adults, the observational nature of survey data does not allow the internal validity necessary to identify causal associations between relative macronutrient profiles and disability. This area of research could be strengthened by the application of experimental designs able to eliminate potential confounding variables and ensure proper temporal ordering of exposures and outcomes. Similarly, the results of our investigation come from self-reported food frequency questionnaires and reports of mobility limitations, introducing potential measurement error into the data collected. Further studies using careful measurement of dietary intake and clinical measures of mobility limitations including handgrip strength (39) are necessary to increase accuracy in estimating the association between dietary patterns and mobility limitations among older adults.
Cross-sectional dietary assessment does not allow for continued measurement of dietary patterns that may be associated with levels and change in mobility limitations. Longitudinal measurement of both dietary intake and physical health and mobility are required to disentangle the association between co-occurring changes in dietary intake and mobility limitations, especially when using observational sources of data. As reported, a cross-sectional snapshot of dietary intake provides important information on the dietary choices of older adults, but cannot inform considerations of how dietary choices made across the life course and related health outcomes mutually influence one another's development over time. Further research is warranted that measures both dietary intake and mobility limitations longitudinally.
Our findings provide evidence that macronutrient balance may influence the development of mobility limitations, an early stage of the disablement process. Registered dietitians, clinicians, and others tasked with monitoring and influencing dietary choices made by older adults should consider the interrelation of macronutrients in the diets of their patients and seek to promote moderate dietary choices, in turn potentially reducing the impact of physical mobility limitations on our older adult population.
Conflict of Interest
The authors declare no conflicts of interest.
Declaration of Ethical Standards
All research contained herein complies with the current laws of the country in which they were performed.
Electronic supplementary material
Supplementary material is available for this article at https://doi.org/10.1007/s12603-017-0986-0 and is accessible for authorized users.
Supplementary Table 1 Descriptive statistics for macronutrient density and mobility limitations by analytic sample inclusion
Supplementary Table 2 (continued) Descriptive statistics for covariates by analytic sample inclusion
Supplementary Table 3 Descriptive statistics for fat type as percentage of daily calories and fat type as percentage of daily fat intake by analytic sample inclusion
References
- 1.den Ouden ME, Schuurmans MJ, Arts IE, van der Schouw YT. Association between physical performance characteristics and independence in activities of daily living in middle-aged and elderly men. Geriatr Gerontol Int. 2013;13:274–280. doi: 10.1111/j.1447-0594.2012.00890.x. 10.1111/j.1447-0594.2012.00890.x [DOI] [PubMed] [Google Scholar]
- 2.Seeman TE, Merkin SS, Crimmins EM, Karlamangla AS. Disability trends among older Americans: National Health and Nutrition Examination surveys, 1988–1994 and 1999–2004. Am J Public Health. 2010;100:100–107. doi: 10.2105/AJPH.2008.157388. 10.2105/AJPH.2008.157388 PubMed PMID: 19910350, PMCID 2791257. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Luppa M, Luck T, Weyerer S, Konig H-H, Brahler E, Riedel-Heller SG. Prediction of institutionalization in the elderly. A systematic review. Age Ageing. 2010;39:31–38. doi: 10.1093/ageing/afp202. 10.1093/ageing/afp202 PubMed PMID: 19934075. [DOI] [PubMed] [Google Scholar]
- 4.Verbrugge LM, Jette AM. The disablement process. Soc Sci Med. 1994;38(1):1–14. doi: 10.1016/0277-9536(94)90294-1. 10.1016/0277-9536(94)90294-1 PubMed PMID: 8146699. [DOI] [PubMed] [Google Scholar]
- 5.Inzitari M, Doets E, Bartali B, et al. Nutrition in the age-related disablement process. J Nutr Health Aging. 2011;15:599–604. doi: 10.1007/s12603-011-0053-1. 10.1007/s12603-011-0053-1 PubMed PMID: 21968852. [DOI] [PubMed] [Google Scholar]
- 6.Leon-Munoz LM, Garcia-Esquinas E, Lopez-Garcia E, Banegas JR, Rodriguez-Artalejo F. Major dietary patterns and risk of frailty in older adults: a prospective cohort study. BMC Med. 2015;13:11. doi: 10.1186/s12916-014-0255-6. 10.1186/s12916-014-0255-6 PubMed PMID: 25601152, PMCID 4298966. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Beasley JM, Lacroix AZ, Neuhouser ML, et al. Protein intake and incident frailty in the Women’s Health Initiative Observational Study. J Am Geriatr Soc. 2010;58:1063–1071. doi: 10.1111/j.1532-5415.2010.02866.x. 10.1111/j.1532-5415.2010.02866.x PubMed PMID: 20487071, PMCID 2924946. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Gregorio L, Brindisi J, Kleppinger A, et al. Adequate dietary protein is associated with better physical performance among post-menopausal women 60–90 years. J Nutr Heal Aging. 2014;18:155–160. doi: 10.1007/s12603-013-0391-2. 10.1007/s12603-013-0391-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Houston DK, Nicklas BJ, Ding J, et al. Dietary protein intake is associated with lean mass change in older, community-dwelling adults: the Health, Aging, and Body Composition (Health ABC) Study. Am J Clin Nutr. 2008;87:150–155. doi: 10.1093/ajcn/87.1.150. 10.1093/ajcn/87.1.150 PubMed PMID: 18175749. [DOI] [PubMed] [Google Scholar]
- 10.Bartali B, Frongillo EA, Stipanuk MH, et al. Protein intake and muscle strength in older persons: does inflammation matter. J Am Geriatr Soc. 2012;60:480–484. doi: 10.1111/j.1532-5415.2011.03833.x. 10.1111/j.1532-5415.2011.03833.x PubMed PMID: 22283208, PMCID 3390191. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Mulla UZ, Cooper R, Mishra GD, Kuh D, Stephen AM. Adult macronutrient intake and physical capability in the MRC National Survey of Health and Development. Am J Clin Nutr. 2013;42:81–87. doi: 10.1093/ageing/afs101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.De Groot C, Van Staveren W, De Graaf C. Determinants of macronutrient intake in elderly people. Eur J Clin Nutr. 2000;54:s70–s76. doi: 10.1038/sj.ejcn.1601028. 10.1038/sj.ejcn.1601028 PubMed PMID: 11041078. [DOI] [PubMed] [Google Scholar]
- 13.Hauser RM, Weir D. Recent developments in longitudinal studies of aging in the United States. Demography. 2010;47:s111–s131. doi: 10.1353/dem.2010.0012. 10.1353/dem.2010.0012 PubMed PMID: 21302430, PMCID 4677668. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Willett WC, Sampson L, Stampfer MJ, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol. 1985;122:51–65. doi: 10.1093/oxfordjournals.aje.a114086. 10.1093/oxfordjournals.aje.a114086 PubMed PMID: 4014201. [DOI] [PubMed] [Google Scholar]
- 15.Weir DR, Langa KM, Ofstedal MB, Hurd MD, Sharon R, Kardia L nd, et al, Health and Retirement Study Institutional Review Board Information, 2017
- 16.Chien S, Campbell N, Chan C, et al. RAND HRS Data Documentation, Version P. RAND Center for the Study of Aging; Santa Monica, CA: 2015. [Google Scholar]
- 17.Nagi SZ. Disability and rehabilitation: legal, clinical, and self-concepts and measurement. Ohio State U. Press; Oxford: 1969. [Google Scholar]
- 18.Rosow I, Breslau N. A Guttman health scale for the aged. J Gerontol. 1966;21:556–559. doi: 10.1093/geronj/21.4.556. 10.1093/geronj/21.4.556 PubMed PMID: 5918309. [DOI] [PubMed] [Google Scholar]
- 19.Fonda S, Herzog AR, Documentation of physical functioning measured in the Health and Retirement Study and the Asset and Health Dynamics among the Oldest Old Study, 2004
- 20.Martin LG, Freedman VA, Schoeni RF, Andreski PM. Health and functioning among baby boomers approaching 60. Journals Gerontol Ser B Psychol Sci Soc Sci. 2009;64B:369–377. doi: 10.1093/geronb/gbn040. 10.1093/geronb/gbn040 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Haas S. Trajectories of functional health: the “long arm of childhood health and socioeconomic factors. Soc Sci Med. 2008;66:849–861. doi: 10.1016/j.socscimed.2007.11.004. 10.1016/j.socscimed.2007.11.004 PubMed PMID: 18158208. [DOI] [PubMed] [Google Scholar]
- 22.McLachlan G, Peel D, Finite mixture models, 2004
- 23.Cameron AC, Trivedi PK. Regression analysis of count data. second. Cambridge University Press; Cambridge: 2013. 10.1017/CBO9781139013567 [Google Scholar]
- 24.Allison P. Change scores as dependent variables in regression analysis. Sociol Methodol. 1992;20:93–114. 10.2307/271083 [Google Scholar]
- 25.Muthén LK, Muthén B 1-, Mplus user’s guide, 2017, eighth
- 26.SAS Institute Inc. SAS/STAT® 14.1 User’s Guide. SAS Institute Inc; Cary, NC: 2015. [Google Scholar]
- 27.Institute of Medicine. Dietary reference intakes for energy, carbohydrate, fiber, fat, fatty acids, cholesterol, protein, and amino acids. The National Academies Press; Washington, DC: 2005. [Google Scholar]
- 30.ter Borg S, Verlaan S, Mijnarends DM, Schols JM, de Groot LC, Luiking YC. Macronutrient intake and inadequacies of community-dwelling older adults, a systematic review. Ann Nutr Metab. 2015;66:242–255. doi: 10.1159/000435862. 10.1159/000435862 PubMed PMID: 26183836. [DOI] [PubMed] [Google Scholar]
- 31.Bauer J, Biolo G, Cederholm T, et al. Evidence-based recommendations for optimal dietary protein intake in older people: a position paper From the PROT-AGE Study Group. J Am Med Dir Assoc. 2013;14:542–559. doi: 10.1016/j.jamda.2013.05.021. 10.1016/j.jamda.2013.05.021 PubMed PMID: 23867520. [DOI] [PubMed] [Google Scholar]
- 32.Tomey KM, Sowers MFR, Crandall C, Johnston J, Jannausch M, Yosef M. Dietary intake related to prevalent functional limitations in midlife women. Am J Epidemiol. 2008;167:935–943. doi: 10.1093/aje/kwm397. 10.1093/aje/kwm397 PubMed PMID: 18250080, PMCID 2730458. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Granic A, Jagger C, Davies K, et al. Effect of dietary patterns on muscle strength and physical performance in the very old: findings from the Newcastle 85+ Study. PLoS One. 2016;11:1–18. doi: 10.1371/journal.pone.0149699. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Houston DK, Stevens J, Cai J, Haines PS. Dairy, fruit, and vegetable intakes and functional limitations and disability in a biracial cohort: the Atherosclerosis Risk in Communities Study. Am J Clin Nutr. 2005;81:515–522. doi: 10.1093/ajcn.81.2.515. 10.1093/ajcn.81.2.515 PubMed PMID: 15699243. [DOI] [PubMed] [Google Scholar]
- 35.Alipanah N, Varadhan R, Sun K, Ferrucci L, Fried LP, Semba RD. Low serum carotenoids are associated with a decline in walking speed in older women. J Nutr Health Aging. 2009;13:170–175. doi: 10.1007/s12603-009-0053-6. 10.1007/s12603-009-0053-6 PubMed PMID: 19262947, PMCID 2748676. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Semba RD, Varadhan R, Bartali B, Ferrucci L, Ricks MO, Blaum C, Fried LP. Low serum carotenoids and development of severe walking disability among older women living in the community: the Women’s Health and Aging Study I. Age Ageing. 2007;36:62–67. doi: 10.1093/ageing/afl122. 10.1093/ageing/afl122 PubMed PMID: 17114201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Michelon E, Blaum C, Semba RD, Xue Q-L, Ricks MO, Fried LP. Vitamin and carotenoid status in older women: associations with the frailty syndrome. Journals Gerontol Ser A Biol Sci Med Sci. 2006;61:600–607. doi: 10.1093/gerona/61.6.600. 10.1093/gerona/61.6.600 [DOI] [PubMed] [Google Scholar]
- 38.Wu I-C, Chang H-Y, Hsu C-C, et al. Association between dietary fiber intake and physical performance in older adults: a nationwide study in Taiwan. PLoS One. 2013;8:80209. doi: 10.1371/journal.pone.0080209. 10.1371/journal.pone.0080209 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.de Souza Vasconcelos KS, Domingues Dias JM, de Carvalho Bastone A, et al. Handgrip strength cutoff points to identify mobility limitation in communitydwelling older people and associated factors. J Nutr Health Aging. 2016;20:306–315. doi: 10.1007/s12603-015-0584-y. 10.1007/s12603-015-0584-y [DOI] [PubMed] [Google Scholar]
Uncited references
- 28.O’brien RM. A caution regarding rules of thumb for variance inflation factors. Qual Quant. 2007;41:673–690. 10.1007/s11135-006-9018-6 [Google Scholar]
- 29.Cohen J, Cohen P, West SG, Aiken LS, Applied multiple correlation/regression analysis for the behavioral sciences, 2003, third
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Table 1 Descriptive statistics for macronutrient density and mobility limitations by analytic sample inclusion
Supplementary Table 2 (continued) Descriptive statistics for covariates by analytic sample inclusion
Supplementary Table 3 Descriptive statistics for fat type as percentage of daily calories and fat type as percentage of daily fat intake by analytic sample inclusion