Allostatic Load, Morbidity, and Mortality Among Older Adults: A Multi-Wave Analysis From the National Health and Aging Trends Study (2024)

Associated Data

Supplementary Materials

Abstract

Although allostatic load (AL) is a key concept to reflect physiologic wear and tear from stress, older adults are underrepresented in AL-related research, especially the oldest old (≥80). Further, attenuative factors are often unaccounted for. This longitudinal analysis using data from National Health and Aging Trends Study investigated relationships of AL in 2017 and multi-wave (1) comorbidity accumulation using multilevel Poisson modeling and (2) mortality risk using survival analysis. By year five (2022), each incremental AL increase that older adults (n = 3614) experienced was associated with a 47% increase in comorbidity (p < .001), and a 33% increased mortality risk (p < .001). This research supports a shift to a more proactive, health promotion/risk mitigation paradigm through informing intervention research targeting AL, which is currently scarce. Identifying potentially modifiable and key driving factors influencing the relationship between AL and health among older adults is an important next step to inform intervention design.

Keywords: chronic conditions, morbidity, mortality, Allostatic Load, Older Adults

Background

The impact of stress and adversity across the lifespan is a growing area of research with foci spanning single disease factors to population level health outcomes (Grummitt et al., 2021; Polick et al., 2023). Lifetime exposure to psychosocial stressors is believed to increase the risk of diseases in later life by disrupting physiological regulatory systems that govern the initiation, inhibition, and maintenance of the stress response (e.g., the sympathetic nervous system and the hypothalamic-pituitary axis) which in turn increases vulnerability to disease (McCrory et al., 2023; Seeman et al., 1997). While some health outcomes are driven, in part, from health risk behaviors (e.g., smoking and alcohol misuse), many negative health outcomes have been attributed to physiological sequelae of immune and inflammatory cascades elicited during stress responses (Guidi et al., 2020; McEwen, 2017). For example, many of the top ten causes of mortality including obesity, diabetes, heart disease, dementia, and lung disease have been linked to inflammatory response (Furman et al., 2019; Kochanek et al., 2019; Slavich, 2020).

Allostatic load (AL) is a multi-system composite index for quantifying physiological dysregulation caused by life stressors (McCrory et al., 2023). While there is no gold standard measurement approach (), AL typically reflects function across the cardiovascular, neuroendocrine, and immune systems and is measured with a combination of biomarkers and anthropometric measures (Li et al., 2019). Elevated AL is commonly associated with increased risk of cardiovascular disease and mortality across many populations (Parker et al., 2022), and additionally with frailty, disability, and number of chronic conditions in older adults (Freire et al., 2020; Gruenewald et al., 2009; Karlamangla et al., 2006; Seeman et al., 2001, 2004).

The older adult population is rapidly expanding in the United States. Promoting quality of life among older populations is an increasing priority, yet many older adults, especially those in the oldest-old category (i.e., 80 or more) are generally underrepresented in AL research (Guidi et al., 2020). One reason may be that older studies suggest that AL plateaus around age 70 (Crimmins et al., 2003). More recent studies, however, which account for more covariates highlight a persistent linear trajectory of AL with age (). Yet, more focus is needed on whether the effects of AL continue to rise with age in later life. Further, when participants within the oldest-old group are included in AL research, they are often reported in wide age categories (e.g., 65+ and 70+), or only included based on their high functionality (Gruenewald et al., 2009), which lacks nuance that would help inform health promotion strategies for adults as they reach advanced ages.

Previous studies indicate significant associations between stressors (e.g., discrimination stress and neighborhood disadvantage) and AL, highlighting AL as an intermediary to worse health outcomes (Gustafsson et al., 2014; Obaoye et al., 2023; Rosemberg et al., 2022). Much of the literature is limited, however, by failing to examine whether positive influences (e.g., social networks) attenuate the effects of AL on health. Further, health risk behavior covariates (e.g., smoking) are not consistently included (Rosemberg et al., 2020). A further limitation of the literature examining AL is that many studies have relied on older data, not yet evaluating more recent data since the COVID-19 pandemic. The utility of AL as an emerging prognostic tool to detect morbidity and mortality risk remains understudied and may be critical to improving health and longevity among diverse groups of older adults, especially the oldest-old population (>=80years). The aim of the present study was to investigate relationships and trajectories between AL and morbidity and mortality in a large representative national cohort of older adults (ages 65+) in the United States.

Methods

This study used a retrospective, observational design using longitudinal data to examine the relationship between AL and morbidity, defined as number of chronic conditions or health events, and mortality over five years spanning before the COVID-19 pandemic to 2022. This secondary data analysis was deemed exempt by Duke University IRB.

Statistical Analysis

Given multiple observations per person corresponding to rounds 7–12 (i.e. multiple years nested within a single participant), we used multilevel modeling (MLM) to assess the association of allostatic load with number of poor health outcomes. Unlike repeated-measures ANOVA, MLM can handle data missing at random via maximum likelihood estimation, thus using all available data. Hence, incomplete data through round 12 did not pose a substantial risk of bias. Given the skewed nature of count data, Poisson and negative binomial parameterization with log link were used to model the morbidity outcome (Elhai et al., 2008). The Poisson model would serve as the default model unless the Bayesian information criterion (BIC) for the negative binomial model was 10 less than that of the Poisson model. The selected model examined count of health outcomes as a function of allostatic load, time (dummy-coded for waves 8–12, with 7 representing the reference value), the interaction of allostatic load with time, and the aforementioned sociodemographic variables (age, sex, race, education, smoking status, and social support). To improve interpretability, social support was centered.

In a second analysis, we additionally assessed risk of mortality as a function of allostatic load. Given an indeterminable date of death, discrete-time survival analysis was used to model risk of mortality. This approach uses latent variable analysis to model proportional hazards, akin to Cox survival models, but accounts for uncertainty in the situations where events are assessed at regular intervals rather than at the time of those events. Specifically, we modeled time until mortality, with time defined as the number of rounds since round 7 in which death was reported. For instance, death at wave 8 was marked as time 1; at wave 9, time 2; and so on. Respondents who were still alive at round 12 were censored. Allostatic load was the primary predictor, with the aforementioned sociodemographic variables entered as covariates.

Results

The analytic sample consisted of 3614 participants. Weighted sample statistics for this group are reported in Table 1. High-risk threshold cut points for HS-CRP (>2.789), IL-6 (>0.775), Hgb A1c (>5.9), BMI (>30.78), and WC (>43.5). The mean AL score was 1.92, although the distribution was positively skewed, with a mode of 1 (n = 1,113, 30.8%). Approximately 13.7% (n = 496) had a score of 0, and just 0.1% (n = 36) had a score of 6 (Supplemental Table 1).

Discussion

In this nationally representative study among older Americans, increased AL was associated with significantly increased morbidity and mortality each year over five years. The number of comorbid health issues increased over time (nearly 50% by year 5), even when controlling for sociodemographic, health behaviors (i.e., smoking), and positive factors (i.e., social support). Mortality increased by 33% with each one-point increase in AL. Collectively these findings demonstrate a compounding detrimental effect of cumulative wear and tear on the body, which is generalizable to US-based adults aged 64years and older. This work advocates for more inclusion to help inform interventions to promote healthier trajectories as people age.

Findings from seminal work support that AL is higher among older adults compared to younger adults and that it plateaus around the age of 70, with increases in mortality risk observed with increased AL among older adults (Parker et al., 2022; Seeman et al., 2001). These studies were limited in that they focused on high functioning older adults (not inclusive of those over 80) who were more likely to have fewer health issues at baseline, which could underestimate the compounding effect of AL on mortality overtime. These early findings have been used to justify the exclusion of the oldest-old and the prioritization of younger adults in more recent AL research. In contrast, by including older adults across all ages, using a larger sample, and a longitudinal design this study highlights a critical period in time in which health outcomes are significantly affected by lifetime stress (regardless of whether a plateau in AL exists). Future AL research should include older adults across all ages in order to provide a more comprehensive understanding of this critical period and to inform potential strategies to mitigate the effects of AL in older adulthood.

Implications

This study further demonstrates the utility of examining AL as a predictor of important health outcomes such as morbidity and mortality especially among the oldest of old who have been excluded from prior work. While trending a personalized AL risk assessment aligns with the clinical shift towards precision medicine and a more holistic view of patients, there are significant barriers to implementing AL as a standard clinical practice. Under a more reactive, disease management paradigm, current practice typically involves trending pertinent biomarker lab tests for patients who have already been diagnosed with a disease. Implementing AL would involve additional lab tests and associated costs, which insurance companies may not reimburse without a clear indication (e.g., existing diagnosis) for being ordered. To the best of our knowledge, there are not any studies testing the feasibility, acceptability, and cost of increased workload due to implementing AL as a proactive risk indicator in the clinical setting.

AL can still support a shift to a more proactive, health promotion/risk mitigation, paradigm through informing intervention research. Evidence-based health interventions are more easily adopted by health systems and other health entities to help avoid or delay the accumulation of morbidity. A recent review highlights the paucity of intervention research in this area, with only six studies that specifically target al, most of which have average participant ages of 65 or younger (Rosemberg et al., 2020). Cognitive behavioral therapy and tai chi show promise in lowering AL among older adults and warrant further investigation (Carroll et al., 2015). Although our indicator of social support, the number of people in the participant’s social network, was not significant in this study, a more sensitive measure that captures the quality of social support may yield different results. Interventions that enhance social support or mitigate loneliness are another potential mechanism to explore. Previous studies have shown that social support across the lifespan partially mediates the relationship from childhood adversity to AL in middle adulthood (). More recently, the negative effects of social isolation on the health of older adults have been highlighted since COVID-19. Comprehensive multi-modal interventions to address issues spanning the stress response, supporting resilience/coping and decreasing health risk behaviors like smoking may be particularly helpful to older adults.

Limitations of this research should be noted. Since participants needed to have the ability to self-report to consent to blood sampling, the AL scores may be skewed toward a healthier sample. There is potential for a survivor bias, where older adults that survive are typically healthier than those who died, which may underestimate the relationships between AL and morbidity accumulation. However, using five years of data helps capture a wider range of experiences compared to just one outcome time point. Further, exact dates were unavailable so years were used, and thus, lacks some granularity. Strengths of this study include a large sample that is nationally representative of adults aged 65 years and older. Our study is also novel for using updated outcome data spanning from before COVID-19 onset to 2022, which helps inform older adult research in the context of ongoing respiratory threats.

Conclusion

This study highlights the importance of improving the inclusion of older adults in AL research, especially intervention research. As the US population continues to age, supporting holistic, healthy longevity will continue to be at the forefront of aging research and policy (National Academy of Medicine, 2022). The rising emphasis on wellness necessitates interventions to mitigate the impact of AL on morbidity and mortality among older adults.

Supplemental Material

Ethical Statement

ORCID iD

Carri S. Polick https://orcid.org/0000-0002-0555-5190

References