Interpretable machine learning for high-dimensional trajectories of aging health.

We have built a computational model for individual aging trajectories of health and survival, which contains physical, functional, and biological variables, and is conditioned on demographic, lifestyle, and medical background information. We combine techniques of modern machine learning with an interpretable interaction network, where health variables are coupled by explicit pair-wise interactions within a stochastic dynamical system. Our dynamic joint interpretable network (DJIN) model is scalable to large longitudinal data sets, is predictive of individual high-dimensional health trajectories and survival from baseline health states, and infers an interpretable network of directed interactions between the health variables. The network identifies plausible physiological connections between health variables as well as clusters of strongly connected health variables. We use English Longitudinal Study of Aging (ELSA) data to train our model and show that it performs better than multiple dedicated linear models for health outcomes and survival. We compare our model with flexible lower-dimensional latent-space models to explore the dimensionality required to accurately model aging health outcomes. Our DJIN model can be used to generate synthetic individuals that age realistically, to impute missing data, and to simulate future aging outcomes given arbitrary initial health states.

A quantile frailty index without dichotomization.

measures of health quantify the aging process of individuals. They should be interpretable, associated with future adverse outcomes, and straightforward to assemble. We use the rank-ordering of risk within a population to construct a quantile frailty index (QFI) that avoids dichotomization, is convenient and interpretable, and is associated with adverse outcomes. We show that the QFI outperforms previous frailty index (FI) measures on cross-sectional laboratory data (NHANES, CSHA, and ELSA). We construct the QFI by ranking the risk of individuals with respect to a reference population. Sex-specific reference populations narrow male-female FI differences as a function of age, and improve predictive performance. With a fixed reference population of 80-85 year olds, our QFI appears similar to earlier FI measures. With an age-matched reference population for each individual, we obtain a QFI that contains very little age information and that has similar predictive performance as other age-controlled FI measures. Adding age as an auxiliary variable leads to significantly better performance. We conclude that age should be controlled for when evaluating the predictive performance of summary measures of health. This is straight-forward to do with the QFI.

The potential for complex computational models of aging.

The gradual accumulation of damage and dysregulation during the aging of living organisms can be quantified. Even so, the aging process is complex and has multiple interacting physiological scales - from the molecular to cellular to whole tissues. In the face of this complexity, we can significantly advance our understanding of aging with the use of computational models that simulate realistic individual trajectories of health as well as mortality. To do so, they must be systems-level models that incorporate interactions between measurable aspects of age-associated changes. To incorporate individual variability in the aging process, models must be stochastic. To be useful they should also be predictive, and so must be fit or parameterized by data from large populations of aging individuals. In this perspective, we outline where we have been, where we are, and where we hope to go with such computational models of aging. Our focus is on data-driven systems-level models, and on their great potential in aging research.

Frailty among middle-aged and older Canadians: population norms for the frailty index using the Canadian Longitudinal Study on Aging.

BACKGROUND: frailty is a public health priority now that the global population is ageing at a rapid rate. A scientifically sound tool to measure frailty and generate population-based reference values is a starting point. OBJECTIVE: in this report, our objectives were to operationalize frailty as deficit accumulation using a standard frailty index (FI), describe levels of frailty in Canadians ≥45 years old and provide national normative data. DESIGN: this is a secondary analysis of the Canadian Longitudinal Study on Aging (CLSA) baseline data. SETTING/PARTICIPANTS: about 51,338 individuals (weighted to represent 13,232,651 Canadians), aged 45-85 years, from the tracking and comprehensive cohorts of CLSA. METHODS: after screening all available variables in the pooled dataset, 52 items were selected to construct an FI. Descriptive statistics for the FI and normative data derived from quantile regressions were developed. RESULTS: the average age of the participants was 60.3 years (95% confidence interval [CI]: 60.2-60.5), and 51.5% were female (95% CI: 50.8-52.2). The mean FI score was 0.07 (95% CI: 0.07-0.08) with a standard deviation of 0.06. Frailty was higher among females and with increasing age, and scores >0.2 were present in 4.2% of the sample. National normative data were identified for each year of age for males and females. CONCLUSIONS: the standardized frailty tool and the population-based normative frailty values can help inform discussions about frailty, setting a new bar in the field. Such information can be used by clinicians, researchers, stakeholders and the general public to understand frailty, especially its relationship with age and sex.

Generating synthetic aging trajectories with a weighted network model using cross-sectional data.

We develop a computational model of human aging that generates individual health trajectories with a set of observed health attributes. Our model consists of a network of interacting health attributes that stochastically damage with age to form health deficits, leading to eventual mortality. We train and test the model for two different cross-sectional observational aging studies that include simple binarized clinical indicators of health. In both studies, we find that cohorts of simulated individuals generated from the model resemble the observed cross-sectional data in both health characteristics and mortality. We can generate large numbers of synthetic individual aging trajectories with our weighted network model. Predicted average health trajectories and survival probabilities agree well with the observed data.

Blood-brain barrier leakage in systemic lupus erythematosus is associated with gray matter loss and cognitive impairment.

OBJECTIVES: To examine the association between blood-brain barrier (BBB) integrity, brain volume and cognitive dysfunction in adult patients with systemic lupus erythematosus (SLE). METHODS: A total of 65 ambulatory patients with SLE and 9 healthy controls underwent dynamic contrast-enhanced MRI scanning, for quantitative assessment of BBB permeability. Volumetric data were extracted using the VolBrain pipeline. Global cognitive function was evaluated using a screening battery consisting of tasks falling into five broad cognitive domains, and was compared between patients with normal versus extensive BBB leakage. RESULTS: Patients with SLE had significantly higher levels of BBB leakage compared with controls (p=0.04). Extensive BBB leakage (affecting over >9% of brain volume) was identified only in patients with SLE (16/65; 24.6%), who also had smaller right and left cerebral grey matter volumes compared with controls (p=0.04). Extensive BBB leakage was associated with lower global cognitive scores (p=0.02), and with the presence of impairment on one or more cognitive tasks (p=0.01). CONCLUSION: Our findings provide evidence for a link between extensive BBB leakage and changes in both brain structure and cognitive function in patients with SLE. Future studies should investigate the mechanisms underlying BBB-mediated cognitive impairment, validate the diagnostic utility of BBB imaging, and determine the potential of targeting the BBB as a therapeutic strategy in patients with SLE.

Informative frailty indices from binarized biomarkers.

Frailty indices (FIs) based on continuous valued health data, such as obtained from blood and urine tests, have been shown to be predictive of adverse health outcomes. However, creating FIs from such biomarker data requires a binarization treatment that is difficult to standardize across studies. In this work, we explore a "quantile" methodology for the generic treatment of biomarker data that allows us to construct an FI without preexisting medical knowledge (i.e. risk thresholds) of the included biomarkers. We show that our quantile approach performs as well as, or even slightly better than, established methods for the National Health and Nutrition Examination Survey and the Canadian Study of Health and Aging data sets. Furthermore, we show that our approach is robust to cohort effects within studies as compared to other data-based methods. The success of our binarization approaches provides insight into the robustness of the FI as a health measure, and the upper limits of the FI observed in various data sets, and also highlights general difficulties in obtaining absolute scales for comparing FIs between studies.

Associations between a laboratory frailty index and adverse health outcomes across age and sex.

OBJECTIVE: Early frailty may be captured by a frailty index (FI) based entirely on vital signs and laboratory tests. Our aim was to examine associations between a laboratory-based FI (FI-Lab) and adverse health outcomes, and investigate how this changed with age. METHODS: Up to 8988 individuals aged 20+ years from the 2003-2004 and 2005-2006 National Health and Nutrition Examination Survey cohorts were included. Characteristics of the FI-Lab were compared to those of a self-reported clinical FI. Associations between each FI and health care use, self-reported health, and disability were examined in the full sample and across age groups. RESULTS: Laboratory-based FI scores increased with age but did not demonstrate expected sex differences. Women aged 20-39 years had higher FI scores than men; this pattern reversed after age 60 years. FI-Lab scores were associated with poor self-reported health (odds ratio[95% confidence interval]: 1.46[1.39-1.54]), high health care use (1.35[1.29-1.42]), and high disability (1.41[1.32-1.50]), even among those aged 20-39 years. CONCLUSION: Higher FI-Lab scores were associated with poor health outcomes at all ages. Associations in the youngest group support the notion that deficit accumulation occurs across the lifespan. FI-Lab scores could be utilized as an early screening tool to identify deficit accumulation at the cellular and molecular level before they become clinically visible.

Is frailty a stable predictor of mortality across time? Evidence from the Cognitive Function and Ageing Studies.

Background: age-specific mortality reduction has been accompanied by a decrease in the prevalence of some diseases and an increase in others. Whether populations are becoming 'healthier' depends on which aspect of health is being considered. Frailty has been proposed as an integrative measure to quantify health status. Objective: to investigate changes in the near-term lethality of frailty before and after a 20-year interval using the frailty index (FI), a summary of age-related health deficit accumulation. Design: baseline data from the Cognitive Function and Ageing Studies (CFAS) in 1991 (n = 7,635) and 2011 (n = 7,762). Setting: three geographically distinct UK centres (Newcastle, Cambridgeshire and Nottingham). Subjects: individuals aged 65 and over (both institutionalised and community-living). Methods: a 30-item frailty score was used, which includes morbidities, risk factors and subjective measures of disability. Missing items were imputed using multiple imputations by chained equations. Binomial regression was used to investigate the relationship between frailty, age, sex and cohort. Two-year mortality was modelled using logistic regression. Results: mean frailty was slightly higher in CFAS II (0.19, 95% confidence interval (CI): 0.19-0.20) than CFAS I (0.18, 95% CI: 0.17-0.18). Two-year mortality in CFAS I was higher than in CFAS II (odds ratio (OR) = 1.16, 95% CI: 1.03-1.30). The association between frailty and 2-year mortality was non-linear with an OR of ~1.6 for each 0.10 increment in the FI. Conclusions: the relationship between frailty and mortality did not significantly differ across the studies. Severe frailty as an indicator of mortality is shown to be a stable construct.

Unifying aging and frailty through complex dynamical networks.

To explore the mechanistic relationships between aging, frailty and mortality, we developed a computational model in which possible health attributes are represented by the nodes of a complex network, with the connections showing a scale-free distribution. Each node can be either damaged (i.e. a deficit) or undamaged. Damage of connected nodes facilitates local damage and makes local recovery more difficult. Our model demonstrates the known patterns of frailty and mortality without any assumption of programmed aging. It helps us to understand how the observed maximum of the frailty index (FI) might arise. The model facilitates an initial understanding of how local damage caused by random perturbations propagates through a dynamic network of interconnected nodes. Very large model populations (here, 10 million individuals followed continuously) allow us to exploit new analytic tools, including information theory, showing, for example that highly connected nodes are more informative than less connected nodes. This model permits a better understanding of factors that influence the health trajectories of individuals.

Frailty status at admission to hospital predicts multiple adverse outcomes.

Aims: frailty is proposed as a summative measure of health status and marker of individual vulnerability. We aimed to investigate the discriminative capacity of a frailty index (FI) derived from interRAI Comprehensive Geriatric Assessment for Acute Care (AC) in relation to multiple adverse inpatient outcomes. Methods: in this prospective cohort study, an FI was derived for 1,418 patients ≥70 years across 11 hospitals in Australia. The interRAI-AC was administered at admission and discharge by trained nurses, who also screened patients daily for geriatric syndromes. Results: in adjusted logistic regression models an increase of 0.1 in FI was significantly associated with increased likelihood of length of stay >28 days (odds ratio [OR]: 1.29 [1.10-1.52]), new discharge to residential aged care (OR: 1.31 [1.10-1.57]), in-hospital falls (OR: 1.29 [1.10-1.50]), delirium (OR: 2.34 [2.08-2.63]), pressure ulcer incidence (OR: 1.51 [1.23-1.87]) and inpatient mortality (OR: 2.01 [1.66-2.42]). For each of these adverse outcomes, the cut-point at which optimal sensitivity and specificity occurred was for an FI > 0.40. Specificity was higher than sensitivity with positive predictive values of 7-52% and negative predictive values of 88-98%. FI-AC was not significantly associated with readmissions to hospital. Conclusions: the interRAI-AC can be used to derive a single score that predicts multiple adverse outcomes in older inpatients. A score of ≤0.40 can well discriminate patients who are unlikely to die or experience a geriatric syndrome. Whether the FI-AC can result in management decisions that improve outcomes requires further study.

A Scoping Review of Frailty and Acute Care in Middle-Aged and Older Individuals with Recommendations for Future Research.

There is general agreement that frailty is a state of heightened vulnerability to stressors arising from impairments in multiple systems leading to declines in homeostatic reserve and resiliency, but unresolved issues persist about its detection, underlying pathophysiology, and relationship with aging, disability, and multimorbidity. A particularly challenging area is the relationship between frailty and hospitalization. Based on the deliberations of a 2014 Canadian expert consultation meeting and a scoping review of the relevant literature between 2005 and 2015, this discussion paper presents a review of the current state of knowledge on frailty in the acute care setting, including its prevalence and ability to both predict the occurrence and outcomes of hospitalization. The examination of the available evidence highlighted a number of specific clinical and research topics requiring additional study. We conclude with a series of consensus recommendations regarding future research priorities in this important area.

Clinical meaningfulness of Alzheimer's Disease Assessment Scale-Cognitive subscale change in relation to goal attainment in patients on cholinesterase inhibitors.

INTRODUCTION: The clinical meaningfulness of Alzheimer's Disease Assessment Scale-Cognitive subscale (ADAS-Cog) subscale change is disputed. We compared 2- to 4-point ADAS-Cog changes with changes in Goal Attainment Scaling (GAS) and everyday function across initial ADAS-Cog scores and treatment responses. METHODS: This exploratory analysis evaluated mild-moderate Alzheimer's disease patients treated with donepezil (12 months) or galantamine (8 months). Clinical meaningfulness was defined as concomitant ADAS-Cog and GAS changes of ±3 points and/or functional improvement. RESULTS: Patients with ≥3-point ADAS-Cog improvement significantly improved on GAS but not on standard tests of everyday function. ADAS-Cog "no change" (≤±3 points) was seen with mean GAS improvement. Initial ADAS-Cog improvement made endpoint improvement (ADAS-Cog 3 points and GAS 1 point) more likely (odds ratio = 6.9; 95% confidence interval = 2.5-19.5). In contrast, initial deterioration made endpoint improvement unlikely (0.33; 0.14-0.64). DISCUSSION: ADAS-Cog improvement and no change were each associated with GAS improvement. Initial ADAS-Cog worsening was unlikely to result in later improvement. CLINICAL TRIAL REGISTRATION NUMBER: ISRCTN26167328.

Frailty affects the initial treatment response and time to recovery of mobility in acutely ill older adults admitted to hospital.

Objectives: to investigate how frailty and mobility impairment affect recovery of balance and mobility in acutely ill older patients. Design: secondary analysis of cohort study. Setting: general and geriatric medicine inpatient units, QEII Health Sciences Centre, Dalhousie University, Canada. Subjects: four hundred and nine older adults (mean age = 81 ± 7 standard deviation, 64% women). Methods: we constructed a frailty index based on a comprehensive geriatric assessment (FI-CGA), at baseline (2 weeks before admission; mean 0.31 ± 0.10), and on admission (mean 0.40 ± 0.10), and recorded Hierarchical Assessment of Balance and Mobility (HABAM) scores daily. Recovery was measured as the difference in HABAM scores between discharge and admission. Results: the odds of no or incomplete recovery increased by 1.06 (95% confidence interval: 1.01-1.11) for each 0.1 increment in the baseline FI-CGA. Recovery odds were similarly dependent on age, but independent of baseline HABAM scores. Recovery time was related to Day 1 HABAM scores, initial treatment response and change in the FI-CGA from baseline to admission (r = 0.35, P < 0.001). Recovery time was independent of age. Patients whose mobility improved within 48 h (n = 113; 28%) showed greater improvement and quicker recovery. Conclusions: frailer patients are at a greater risk of incomplete or lengthier recovery from impaired mobility and balance. Tracking mobility and balance might help providers, patients and families understand the course of acute illness in older adults.

Heterogeneity of Human Aging and Its Assessment.

Understanding the heterogeneity in health of older adults is a compelling question in the biology of aging. We analyzed the performance of five measures of health heterogeneity, judging them by their ability to predict mortality. Using clinical and biomarker data on 1,013 participants of the Canadian Study of Health and Aging who were followed for up to 6 years, we calculated two indices of biological age using the Klemera and Doubal method, which controversially includes using chronological age as a "biomarker," and three frailty indices (FIs) that do not include chronological age: a standard clinical FI, an FI from standard laboratory blood tests and blood pressure, and their combination (FI-combined). Predictive validity was tested using Cox proportional hazards analysis and discriminative ability by the area under the receiver-operating characteristic curves. All five measures showed moderate performance that was improved by combining measures to evaluate larger numbers of items. The greatest addition in explanatory power came from the FI-combined that showed the best mortality prediction in an age-adjusted model. More extensive comparisons across different databases are required, but these results do not support including chronological age as a biomarker.

Changes in the Lethality of Frailty Over 30 Years: Evidence From Two Cohorts of 70-Year-Olds in Gothenburg Sweden.

BACKGROUND: With aging, health deficits accumulate: people with few deficits for their age are fit, and those with more are frail. Despite recent reports of improved health in old age, how deficit accumulation is changing is not clear. Our objectives were to evaluate changes over 30 years in the degree of deficit accumulation and in the relationship between frailty and mortality in older adults. METHODS: We analyzed data from two population based, prospective longitudinal cohorts, assembled in 1971-1972 and 2000-2001, respectively. Residents of Gothenburg Sweden, systematically drawn from the Swedish population registry. The 1901-1902 cohort (N = 973) had a response rate of 84.8%; the 1930 cohort (N = 500) had a response rate of 65.1%. A frailty index using 36 deficits was calculated using data from physical examinations, assessments of physical activity, daily, sensory and social function, and laboratory tests. We evaluated mortality over 12.5 years in relation to the frailty index. RESULTS: Mean frailty levels were the same (x¯ = 0.20, p = .37) in the 1901-1902 cohort as in the 1930 cohort. Although the frailty index was linked to the risk of death in both cohorts, the hazards ratio decreased from 1.67 per 0.1 increment in the frailty index for the first cohort to 1.32 for the second cohort (interaction term p = .005). DISCUSSION: Although frailty was as common at age 70 as before, its lethality appears to be less. Just why this is so should be explored further.

Network model of human aging: Frailty limits and information measures.

Aging is associated with the accumulation of damage throughout a persons life. Individual health can be assessed by the Frailty Index (FI). The FI is calculated simply as the proportion f of accumulated age-related deficits relative to the total, leading to a theoretical maximum of f≤1. Observational studies have generally reported a much more stringent bound, with f≤f_{max}<1. The value of f_{max} in observational studies appears to be nonuniversal, but f_{max}≈0.7 is often reported. A previously developed network model of individual aging was unable to recover f_{max}<1 while retaining the other observed phenomenology of increasing f and mortality rates with age. We have developed a computationally accelerated network model that also allows us to tune the scale-free network exponent α. The network exponent α significantly affects the growth of mortality rates with age. However, we are only able to recover f_{max} by also introducing a deficit sensitivity parameter 1-q, which is equivalent to a false-negative rate q. Our value of q=0.3 is comparable to finite sensitivities of age-related deficits with respect to mortality that are often reported in the literature. In light of nonzero q, we use mutual information I to provide a nonparametric measure of the predictive value of the FI with respect to individual mortality. We find that I is only modestly degraded by q<1, and this degradation is mitigated when increasing number of deficits are included in the FI. We also find that the information spectrum, i.e., the mutual information of individual deficits versus connectivity, has an approximately power-law dependence that depends on the network exponent α. Mutual information I is therefore a useful tool for characterizing the network topology of aging populations.

Socioeconomic gradient in health in Canada: Is the gap widening or narrowing?

BACKGROUND: Notwithstanding a general improvement in health status, the socioeconomic gradient in health remains a public health challenge worldwide. OBJECTIVE: Using longitudinal data from the National Population Health Survey (NPHS, n=17,276), we examined trends in socioeconomic gradients in two health indicators, viz. the Health Utility Index (HUI) and the Frailty Index (FI), among Canadian adults (25 years and older) between 1998/9-2010/11. METHODS: The relative and slope indices of inequality (RII and SII, respectively) were employed to summarize income- and education-based inequality in the FI and the HUI in Canada as whole, and in five regions: the Atlantic provinces, Quebec, Ontario, the Prairies and British Columbia. RESULTS: We found that education- and income-related inequalities in health were present in all five regions of Canada. The estimated RIIs and SIIs suggested that education-related inequalities in the FI and the HUI increased among women. The results also revealed that relative and absolute income-related inequalities in the HUI increased in Canada, especially among women. Both absolute and relative inequalities indicated that income-related inequalities in the HUI increased in Quebec and in the Prairies over time. CONCLUSION: Persistent and growing socioeconomic inequalities in health in Canada over the past one and half decades should warrant more attention. The mechanisms underlying socioeconomic-related inequalities in Canada are less clear. Therefore, further studies are required to identify effective polices to reduce the socioeconomic gradient in health in Canada.

Frailty and life satisfaction in Shanghai older adults: The roles of age and social vulnerability.

PURPOSE: This study aims to examine the relationship between frailty and life satisfaction and the roles of age and social vulnerability underlying the links in Chinese older adults. MATERIAL AND METHODS: Using a cross-sectional sample of 1970 adults aged 65 and older in 2013 in Shanghai, we employed regression analyses to investigate the interaction between frailty and age on life satisfaction in the whole sample and in different social vulnerability groups. Life satisfaction was measured using a sum score of satisfaction with thirteen domains. Using a cumulative deficit approach, frailty was constructed from fifty-two variables and social vulnerability was derived from thirty-five variables. RESULTS: Frailty was negatively associated with life satisfaction. The interaction between frailty and age was significant for life satisfaction, such that the negative association between frailty and life satisfaction was stronger among the young-old aged 65-79 than among the old-old aged 80+. Moreover, frailty's stronger association with life satisfaction in the young-old than in the old-old was only found among those in the 2nd and 3rd tertiles of social vulnerability, but not for those in the 1st tertile of social vulnerability. CONCLUSIONS: Relation between frailty and life satisfaction likely weakens with age. A higher level of social vulnerability enlarges the negative impact of frailty on life satisfaction with a greater extent in the young-old.