Performance Under Fire: Older Adult Cognitive Risks and Protections Under Heat Strain

Abstract

Background and Objectives

Although extreme heat events disproportionately affect older adults and the importance of cognition is known, research examining older adult cognition under heat stress is limited. This study examines the relationship between risk/protective factors and heat strain on older adult cognition, employing a social-ecological model.

Research Design and Methods

Retrieved from the 1996–2016 waves of the Health and Retirement Study, our study used older adults aged 50 and older and their spouses residing in the United States. Individual-fixed effects models estimated changes in cognition as measured by fluid and crystallized intelligence scores in response to extreme heat days. This study further estimated interactions of extreme heat with protective/risk factors for cognition (i.e., education, physical activity, social engagement, and genetic risk for Alzheimer’s disease).

Results

Our results demonstrated that extreme heat days were associated with fluid but not crystallized intelligence scores. Educational attainment, mild physical activity, and social contacts with children moderated this relationship. Furthermore, Alzheimer’s disease polygenic scores moderated the correlation between extreme heat days and crystallized intelligence scores.

Discussion and Implications

An increasing frequency of extreme heat events and an aging population globally highlight the need for policies and interventions building resiliency in older adults. Actions promoting the protective modifiable behaviors to older adult cognition identified by our study can lead to healthier individuals and communities.

Cognition, Extreme heat, Education, Physical activity, Social engagement

Since 1979, over 11,000 Americans have succumbed to extreme heat events, with individuals aged 60 and older constituting up to 92% of excess mortality during heat waves (Belmin et al., 2007; U.S. EPA, 2016). Older adults are particularly susceptible to heat-related effects due to degradation of thermoregulatory systems, resulting in reduced tolerance to extreme temperatures (Blatteis, 2012; Lu et al., 2010). In the United States, regions identified as “hotspots,” where population aging and climate-driven changes intersect, are increasingly exposing older adults to unprecedented heat impacts (Carr et al., 2024). These adverse effects of heat can manifest in exacerbated cardiovascular diseases (Khatana & Groeneveld, 2022), diabetes (Song et al., 2021) and other comorbidities (Kenny et al., 2010), impaired kidney functioning (Schanz et al., 2022), sleep disturbances (Zhou et al., 2023), heightened hospital admissions for mental illness (Hansen et al., 2008), and higher suicide risk compared to other age groups (Hansen et al., 2008; Ren et al., 2020), all collectively contributing to cognitive decline later in life (Bouldin et al., 2016; Liu et al., 2019).

Extreme heat events may heighten allostatic loads, adversely affecting cognitive reserve—the brain’s adaptability in everyday functions, encompassing responses to aging, stress, or environmental influences (Crews et al., 2019; Lenart-Bugla et al., 2022; Stern et al., 2019). Allostatic load, defined as the toll from prolonged exposure to neural and neuroendocrine responses perceived as stressful (Guidi et al., 2020), may be exacerbated by prolonged exposure to heat waves, acting as a chronic stressor for older adults. This can lead to repeated physiological arousal without effective adaptation to stressors or coping mechanisms, resulting in the development of allostatic overload (Guidi et al., 2020). With allostatic overload induced by heat waves, older adults’ cognitive capacity to adapt to the stressor might be compromised, risking a loss of efficient brain function. Impaired cognition poses significant financial burdens to both households and governments. Households with cognitive impairment require 48% higher income compared to their counterparts (Morris et al., 2021). Additionally, long-term care, the costliest aspect of U.S. healthcare, reaches an annual estimate of $100 billion for Alzheimer’s disease (AD) alone (Leifer, 2003).

Given the challenge of reversing cognitive decline once it has progressed, preventive interventions are paramount. These interventions not only offer significant cost-saving effects compared to medical and long-term care treatments but also play a crucial role in mitigating or delaying cognitive decline. To develop evidence-based preventive strategies, it is imperative to identify both risk and protective factors for cognitive decline among older adults exposed to extreme heat conditions. The social-ecological model (SEM) provides a holistic approach to understanding health by considering various interconnected factors operating at different levels of the social ecology (Lee et al., 2023; Wilson & Anstey, 2023). By examining cognitive health through the lens of the SEM, interventions can be developed to address multiple levels of influence simultaneously (Peeters et al., 2023).

The SEM identifies influencing factors at various levels: (1) individual-level factors (genetics, lifestyle behaviors [e.g., diet, physical activity, and cognitive stimulation] and health conditions [e.g., chronic diseases and mental health]); (2) interpersonal-level factors (interpersonal relationships, social support networks, and social engagement); (3) community-level factors (socioeconomic status, neighborhood characteristics [e.g., walkability], availability of recreational opportunities, and opportunities for lifelong learning); (4) organizational-level factors (the quality and accessibility of healthcare systems, community centers, and senior centers, and the availability of cognitive health programs and services); and (5) societal-level factors (cultural norms, policies related to aging and healthcare, and socioeconomic disparities) (Majoka & Schimming, 2021; Peeters et al., 2023; Sisco et al., 2015; Zahodne et al., 2019).

Building on the SEM framework, our study primarily focuses on individual and interpersonal factors to delve into the necessary interventions for older adult cognition amid extreme heat conditions. Specifically, we consider education, physical activity (PA), biological factors, and social connections (Lisko et al., 2021). These individual- and interpersonal-level factors can be more readily integrated into healthcare programs, guidelines, or safety protocols designed for extreme heat events compared to macrolevel factors, which require broader community and systemic adjustments. Identifying individual- or interpersonal-level factors can also help identify older adults at higher risks from extreme heat and further guide policies addressing macrolevel factors targeting them, leading to ripple effects across other levels (Ungar & Theron, 2020).

At the individual level, protective factors such as higher educational attainment (Pettigrew & Soldan, 2019) and PA mitigate cognitive impairment by enhancing cognitive reserve. This reserve is built through education and challenged by PA, fostering neural development and problem-solving skills, and building resilience against age-related cognitive decline (Le Carret et al., 2003; Wilson et al., 2019). Physical activity, particularly aerobic exercises, boosts blood flow and oxygen delivery to the brain, promoting better cognitive function and contributing to overall brain health. Affecting cardiorespiratory fitness, PA enhances cognitive function, speed, memory, attention, and motor skills (Angevaren et al., 2008; Carvalho et al., 2014). Consequently, it contributes to a lowered risk for AD and cognitive decline (Buchman et al., 2012; Blondell et al., 2014; Stephen et al., 2017; Tan et al., 2017).

This study considered genetic factors as a risk factor. Studies identify genetics as a significant risk factor for late-onset AD and highlight apolipoprotein (APOE) ɛ4 as the strongest predictor (Holtzman et al., 2012). Inheriting at least one copy of the ɛ4 allele increases the risk threefold, and having two copies elevates it eight- and 12-fold compared to those with two copies of the ɛ3 allele (Holtzman et al., 2012). A recent genome-wide association study (GWAS) has identified 25 loci associated with AD (Kunkle et al., 2019). These findings allow the use of a polygenic score (PGS) for AD, predicting individual cognition in old age before clinical symptoms emerge. Literature indicates that PGS for AD is linked to poorer memory and a smaller hippocampus in youth, followed by progressive cognitive decline (Mormino et al., 2016) and declining crystallized intelligence with aging (Shin et al., 2021).

At the interpersonal level, active social engagement, including participation in social activities and maintaining a robust social network, is linked to a decreased risk of cognitive impairment in later life (Bourassa et al., 2017; Lövdén et al., 2005; Saczynski et al., 2006). A strong social network provides emotional support, mitigating chronic stress. Considering the established connection between climate changes and increased allostatic loads, social supports may play a crucial role in buffering against environmental stressors. Additionally, social engagement contributes to brain plasticity—the ability to adapt and form new neural connections. Interacting with others and engaging in social activities may enhance cognitive reserve, assisting the brain in coping with age-related changes (Davidson & McEwen, 2012).

Although older adult cognitive trajectories have been examined generally following heat exposure (Choi et al., 2023), this is the first study to elucidate underlying factors, contributing effects, and potential preventative measures for older adult cognitive decline from extreme heat events under the SEM framework. Some studies have delved into elements like gender moderating cognitive performance in high temperatures (Chang & Kajackaite, 2019), and the impact of high temperatures on cognition during moderate activities (Chen et al., 2020). Aligning with our study, recent research explored cumulative heat exposure and cognitive decline among older U.S. adults, finding an association only among Black Americans and those in impoverished neighborhoods (Choi et al., 2023). However, the link between heat and cognitive protective/risk factors remains understudied, particularly in a longitudinal, nationally representative sample.

This study addresses this gap by utilizing longitudinal data from the 1996–2016 waves of the Health and Retirement Study (HRS), surveying a nationally representative sample of individuals aged 50 or older and their spouses. The primary aim is to estimate the cognitive functioning trajectories of older adults over time in response to exposure to extreme heat events. Additionally, this research contributes to the existing literature by examining whether risk or protective factors, particularly individual and interpersonal ones, moderate the relationship between heat and cognition.

Method

Data and Sample

This research utilized individual-level data from a nationally representative sample of older adults aged 50 and older, along with their spouses in the United States. The data were obtained from the HRS longitudinal panel collected between 1996 and 2016, conducted by the Institute for Social Research at the University of Michigan. The HRS, administered biennially, covers a broad spectrum of information, including demographics, health, cognitive abilities, family dynamics, disabilities, wealth, employment history, healthcare access, and genetic traits for phenotypes for over 20,000 Americans (Servais, 2010).

The study utilized data from the National Environmental Public Health Tracking Network, a CDC division, offering county-level information on extreme weather events. Historical metrics on extreme heat days were extracted from this data set and then merged with individual-level HRS data based on county of residence and survey year. The unit of analysis is an individual who completed a series of cognition tests over time, aiming to examine the correlation between heat events and cognition in older adults. Refer to the right panel of Table 1 for sample characteristics.

Table 1.

Sample Characteristics

Variables	Mean (SD) or %
Explanatory variables
# of extreme heat day ≥35	15.13 (29.25)
# of extreme heat day ≥38	6.58 (18.15)
# of extreme heat day ≥40	2.27 (8.78)
Outcome variables
Fluid intelligence score	14.45 (4.66)
Crystalized intelligence score	7.26 (1.11)
Moderators
LTHS	23.64%
HS	51.10%
SC	5.21%
BA	12.48%
Graduate	7.56%
Engaging in light PA	86.61%
No. of children in contact	2.61 (2.00)
PGS for AD	−0.00 (1.02)
Covariates
# of extremely rainy day 0.1	7.63 (4.67)
Female	55.74%
Age	71.52 (10.05)
White	65.94%
Black	18.74%
Hispanic	11.85%
Other	3.47%
Coupled	60.64%
Sep./div./wid.	35.82%
Never married	3.54%
Employed	17.92%
Unemployed	11.39%
Retired	70.69%
Poor health	8.11%
Fair	21.18%
Good	32.58%
Very good	28.11%
Excellent	10.02%
ADLs	0.34 (0.88)
IADLs	0.27 (0.78)
Health insurance ownership	96.64%
Homeowners	78.02%
Household income	$64,611 (224,696)
Financial assets	$238,657 (786,481)
# people in HH	2.11 (1.13)
# living children	3.19 (2.12)

Note: Unweighted estimates. 1996–2018 waves of the HRS. N = 34,253, observations (respondent-wave) = 122,408. ADLs = activities of daily living; BA = bachelor’s; div. = divorced; HH = household; HS = high school; IADLs = instrumental activities of daily living; LTHS = less than high school; PA = physical activity; PGS for AD = polygenic score for Alzheimer’s disease; Sep. = separated; SC = some college; SD = standard deviation; wid. = widowed.

Variables

The study focused on fluid and crystallized intelligence scores as dependent variables measured each survey year. Fluid intelligence, closely tied to biological and physical mechanisms, exhibits susceptibility to age-related decline (Salthouse, 1999). It encompasses working memory, which involves handling and retaining information simultaneously (Craik, 1999). Crystallized intelligence, derived from structured learning and life experiences, generally shows a less distinct decline with age (Salthouse, 1999). The HRS cognitive functioning evaluation included various tests. Fluid intelligence scores were derived from memory assessments, involving immediate and delayed recall, a serial 7s, and backward counting tasks (Crimmins et al., 2011), resulting in scores ranging from 0 to 27. Crystallized intelligence scores, drawn from object naming and orientation assessments, ranged from 0 to 8 (Crimmins et al., 2011). The sample’s average fluid and crystallized intelligence scores were 14 (SD = 5) and 7 (SD = 1), respectively (Table 1). Note that the means and standard deviations of these scores are calculated using multiple observations of the same respondents across respondents. Therefore, they are overall estimates incorporating both between and within variations.

The study’s independent variable was extreme heat events in a respondent’s county, represented by the annual count of days with peak temperatures of 35, 38, and 40 degrees Celsius from May to September. The average annual counts for these temperature thresholds were 15, 7, and 2 days, respectively (Table 1).

Utilizing the SEM framework, the individual-level moderators identified as influential factors in cognitive decline encompass education, PA, and genetic traits for AD, whereas social engagement serves as a moderator at the interpersonal level. Educational attainment was categorized into “high school (HS),” “some college (SC),” “bachelor’s (BA),” and “graduate” degrees, with “less than high school (LTHS)” as the reference group. About two-thirds of the sample completed a high school degree or less (Table 1). Older adults’ PA was measured using engagement in mild PAs, including sports or mildly energetic household tasks like vacuuming and home repairs. Approximately 87% of the sample participated in light PAs (Table 1). Because questions about PAs have been included since 2004, the analysis using PA includes fewer waves. As a proxy for social engagement, the study used the number of children a respondent had contact with, coding zero for those without any living children for this variable. The average number of the respondent’s children in touch was 2 during the study period (Table 1).

The study included a PGS for AD as a risk factor for cognition. Respondent saliva samples collected from 2006 to 2012 were used to generate PGS for various phenotypes, including AD. The PGS for AD was established through a GWAS and included 1,406,839 SNPs overlapping with the HRS genetic data and the GWAS. The weights and identified single nucleotide polymorphisms (SNPs) associated with AD were adopted from the International Genomics of Alzheimer’s Project conducted in 2019, confirming and discovering 25 loci related to late-onset AD (Kunkle et al., 2019). The PGS sums the number of reference alleles (0, 1, or 2) at each SNP for individuals after assigning weights based on effect sizes (odds ratios or beta estimates) obtained from GWAS meta-analyses. The PGSs are normalized to have a mean of zero and a standard deviation of one (Table 1).

Most GWAS used to determine SNP weights are centered around European ancestry populations, and the analysis is recommended for those from European ancestry groups because PGSs for other ancestral groups may not have equivalent predictive power (Martin et al., 2017; Ware et al., 2017). Therefore, our data analyses include only respondents from European ancestry groups, with AD PGSs acting as a moderator for the relationship between extreme heat events and cognition.

Following the SEM, this study further controlled for time-varying individual- and household-level factors. These included age centered on 50 and its squared value, marital status, self-reported health, activities of daily living (ADLs), instrumental activities of daily living (IADLs), health insurance and home ownership, natural logarithm of household income and financial assets, and employment status. The interpersonal-level covariates encompassed the number of household members and living children. The study controlled for individual- and wave-fixed effects, incorporating community, organizational, and societal-level factors within the SEM framework. We posit that these factors are adequately captured within individual- and wave-fixed effects. Our rationale stems from the observation that our sample demonstrated minimal relocation between communities. Moreover, when such movements did occur, individuals tended to transition to communities with comparable attributes. This study also controlled for the impact of annual days with extreme precipitation. To address potential bias from population stratification, adjustments for randomly labeled principal components (PCs) 1 to 5 were incorporated, controlling for genetic elements of shared ancestry. Five PCs were absorbed into individual-fixed effects because they are time-invariant individual factors.

Empirical Models

This study employed individual-fixed effects models to assess changes in older adults’ cognition concerning variations in extreme heat days. These models are tailored to capture within-individual changes over time and were estimated as follows:

$${\displaystyle \begin{array}{l}{\displaystyle CO{G}_{it}={\mathrm{\alpha }}_0+{\mathrm{\alpha }}_{1}E{H}_{ct}+{\mathrm{\alpha }}_2{X}_{it}+i_i+w_w+{\epsilon }_{it}}\end{array}}$$

(1)

$CO{G}_{it}$ denotes respondent i’s cognition measured using fluid and crystalized intelligence scores in survey year t. $E{H}_{ct}$ is the number of extreme heat days with daily maximum temperatures of 35, 38, and 40 Celsius in respondent’s county of residence c in year t. By merging the weather data according to the county of residence of HRS respondents in the survey year, both cognition and the count of extreme heat days are measured concurrently in the same year, denoted as year t. $X_{it}$ denotes a vector of time-variant individual or household characteristics measured in survey year t. $i_i$ and $w_w$ indicate individual- and wave-fixed effects. ${\epsilon }_{it}$ is idiosyncratic error. Of note, individual time-invariant factors are absorbed in individual-fixed effects (⁠$i_i$⁠), such as genetic traits for AD, education, race/ethnicity, gender, etc.

This study investigated whether selected risk/protective factors moderated the association between extreme heat events and cognition using individual-fixed effects modeling, as outlined in the following equation:

$${\displaystyle \begin{array}{lll}{\displaystyle CO{G}_{it}=}& {\mathrm{\beta }}_0+{\mathrm{\beta }}_{1}E{H}_{ct}+{\mathrm{\beta }}_2{M}_{it}+{\mathrm{\beta }}_{3}E{H}_{ct}{M}_{it}& +{\mathrm{\beta }}_4{X}_{it}+i_i+w_w+{\epsilon }_{it}\end{array}}$$

(2)

$M_{it}$ denotes respondent i’s protective/risk factors for cognition measured in survey year t. Factors included educational attainment, engagement in light PA, AD PGSs, and number of children in contact. When moderators are time-invariant factors (i.e., education, PGSs for AD), $M_{it}$ is absorbed in individual-fixed effects, resulting in the elimination of β₂.

Results

Sample Characteristics

The right panel of Table 1 presents the sample characteristics. Slightly more than half of the sample were females (56%). A larger portion of the sample were non-Hispanic White older adults (66%). A majority of the sample observations were retired or unemployed (82%), reported at least good or better health status (71%), were homeowners (78%), and held any health insurance (97%). More than half of the observations were in coupled status (61%). Our sample reported less than one ADLs and IADLs with difficulty to perform. The mean number of household members and living children were 2 and 3, respectively.

The Effects of Extreme Heat Days on Cognition

Table 2 presents the results of the individual-fixed effects model, examining the effect of extreme heat days on fluid (Panel A) and crystallized (Panel B) intelligence scores. The study found that changes in the number of extreme heat days were associated with alterations in fluid intelligence scores. Specifically, a one-unit increase in heat days with a temperature of 38 related to a decrease of 0.0023 in older adults’ fluid intelligence scores. This significance was not observed with alternative peak temperatures, prompting the use of a threshold of 38 degrees in the subsequent analysis. The full results, including covariates, are available in Supplementary Table A1.

Table 2.

The Effect of Extreme Heat on Older Adults’ Cognition

	(A) Fluid intelligence scores			(B) Crystalized intelligence scores
	(I) Coef. (SE)	(II) Coef. (SE)	(III) Coef. (SE)	(IV) Coef. (SE)	(V) Coef. (SE)	(VI) Coef. (SE)
# Extreme heat days ≥35	−0.0014 (0.0008)			0.0002 (0.0003)
# Extreme heat days ≥38		−0.0023* (0.0011)			0.0000 (0.0003)
# Extreme heat days ≥40			−0.0030 (0.0023)			0.0003 (0.0007)
R-squared	0.2112	0.2112	0.2112	0.1810	0.1810	0.1810

Notes: Individual-fixed effects estimators. N = 34,253, Obs. = 122,408. 1998–2016 HRS. The covariates include the number of extremely rainy days, age centered on 50, centered age-squared, marital status, self-reported health, number of difficulties with performing ADLs and IADLs, health insurance ownership, logarithm of household income, homeownership, employment status, logarithm of total financial assets, number of people in the household, number of living children, and wave-fixed effects. SE = standard errors.

^*p < .05.

In our study, changes in two outcomes, namely fluid and crystallized intelligence scores, were assessed in response to variations in a single environmental factor—extreme heat events—using two separate models. Given this approach, a multiple hypothesis test might be necessary. To address this, we adjusted for the false discovery rate (FDR), defined as the expected proportion of rejections that are false positives (Anderson, 2008). Utilizing the sharpened q value to control FDR (Benjamini et al., 2006) for pairs of models (I) and (IV), (II) and (V), and (III) and (VI), we observed that our findings generally align with those presented in Table 2. However, the significance of extreme heat days, defined as maximum temperatures exceeding 38 degrees, became slightly weaker (FDR sharpened q < 0.10).

The Protective/Risk Factors for the Relationship Between Extreme Heat Days and Cognition

Table 3 displays estimates from Equation (2), showcasing the interaction of extreme heat days with educational attainment (Panel A), mild PA (Panel B), AD PGSs (Panel C), and number of children in contact (Panel D).

Table 3.

Risk/Protective Factors as a Moderator

	DV = Fluid intelligence scores	DV = Crystalized intelligence scores
	Coef. (SE)	Coef. (SE)
Panel A: Education level
# Extreme heat days	−0.0073** (0.0022)	−0.0008 (0.0007)
Heat × HS	0.0048 (0.0027)	0.0006 (0.0008)
Heat × SC	0.0225*** (0.0062)	0.0021 (0.0018)
Heat × BA	0.0065 (0.0037)	0.0018 (0.0011)
Heat × Graduate	0.0109** (0.0042)	0.0029* (0.0012)
R-squared	0.2114	0.1811
N	34,253	34,253
Obs.	122,408	122,408
Panel B: Whether or not engaging in mild physical activity
# Extreme heat days	−0.0040 (0.0022)	−0.0003 (0.0007)
Mild PA	0.3810*** (0.0468)	0.1593*** (0.0142)
Heat × PA	0.0040* (0.0020)	0.0003 (0.0006)
R-squared	0.1535	0.1632
N	27,490	27,490
Obs.	78,431	78,431
Panel C: Polygenic score for Alzheimer’s disease (AD)
# Extreme heat days	−0.0011 (0.0015)	0.0001 (0.0004)
Heat × AD PGS	−0.0000 (0.0014)	−0.0011** (0.0004)
R-squared	0.2262	0.1826
N	11,098	11,098
Obs.	59,606	59,606
Panel D: Number of children with contact
# Extreme heat days	−0.0046* (0.0019)	0.0001 (0.0005)
Children w/contact	0.0117 (0.0136)	0.0074 (0.0039)
Heat × Child	0.0008* (0.0004)	0.0003 (0.0001)
R-squared	0.2143	0.1756
N	29,451	29,451
Obs.	80,611	80,611

Notes: Individual-fixed effects estimators. 1998–2016 HRS. The covariates include the number of extremely rainy days, age centered on 50, centered age-squared, marital status, self-reported health, number of difficulties with performing ADLs and IADLs, health insurance ownership, logarithm of household income, homeownership, employment status, logarithm of total financial assets, number of people in the household, number of living children, and wave-fixed effects.

AD PGS = Alzheimer’s disease polygenic score; BA = bachelor’s; HS = high school; Obs. = observations; PA = physical activity; SC = some college; SE = standard errors.

^*p < .05.

^**p < .01.

^***p < .001.

Educational attainment primarily moderated the association between extreme heat days and fluid intelligence scores. For individuals with a LTHS degree, a one-unit increase in extreme heat days correlated with a 0.0073 decline in fluid intelligence scores. In contrast, those with SC and graduate degrees experienced increases of 0.0152 and 0.0036, respectively, even with heightened heat exposure. Graduate degree attainment also showed moderating effects on crystallized scores, revealing a 0.0021 increase despite increased exposure to intense heat.

Engaging in mild PA moderated the relationship with fluid intelligence but not crystallized intelligence. Older adults unengaged in mild PA did not exhibit changes in fluid intelligence scores with a one-unit increase in extreme heat days, whereas those engaging in light PA had a 0.3810 increase in fluid intelligence.

AD PGSs moderated the correlation between extreme heat days and crystallized, but not fluid, intelligence scores. Older adults with an average PGS for AD did not experience changes in crystallized intelligence. Conversely, those with a one standard deviation higher PGS for AD exhibited a 0.0010 decrease in crystallized intelligence scores in response to a one-unit increase in extreme heat days.

Finally, social engagement with children moderated the association of heat days with fluid intelligence but not crystallized intelligence. Older adults without a child in contact experienced a 0.0046 decline in fluid intelligence scores with a one-unit increase in extreme heat days, whereas those with more children in touch showed a 0.0079 increase.

Despite employing the FDR sharpened q value to control the FDR, our results remain generally robust. However, it’s noteworthy that the statistical significance of the findings slightly diminished following the adjustment. Notably, all estimates presented in Panels A and B retain significance, with q values of at least < 0.05, except for the interaction terms depicted in Panels B and D, which remained significant, albeit at a slightly higher threshold of q < 0.10.

Discussion

Differing from Choi et al.’s findings (2023) that extreme heat exposure showed no correlation with the baseline or rate of change in cognitive functioning (equivalent to fluid intelligence defined in our study) among U.S. older adults, this study reveals an association between extreme heat events and a decline in fluid intelligence scores. Because fluid intelligence is linked to a less age-sensitive cognitive reserve (Baltes et al., 1986), our results might support the hypothesized mechanism that extreme heat exposure may adversely influence the individual factor of cognitive reserve through allostatic overload.

Within the SEM framework, our study examined both protective and risk factors on individual and interpersonal dimensions to suggest comprehensive strategies that mitigate the adverse effects of extreme heat on cognitive functioning. Building upon previous studies that have shown a positive relationship between education and cognitive functioning in later life (e.g., Du et al., 2023; Kobayashi et al., 2017; Pettigrew & Soldan, 2019), our findings suggest that increased educational attainment acts as a protective factor for both fluid and crystallized intelligence, especially among older adults exposed to extreme heat conditions. Specifically, older adults with SC and graduate degrees demonstrated enhanced fluid intelligence compared to those without such degrees. Moreover, improvements in crystallized intelligence, despite increased exposure to extreme heat events, were observed solely among those with a graduate degree. These results underscore the importance of continuous learning and lifelong educational interventions to maintain cognitive reserve. Accessible educational initiatives or incentives promoting higher education could potentially guard against cognitive decline associated with environmental stressors like extreme heat. Tailored healthcare programs for older adults with lower educational attainment could prioritize cognitive health interventions, such as cognitive training programs or mental stimulation activities. Our findings warrant further research into the specific mechanisms through which education mitigates the risk of impaired cognitive function from extreme heat conditions, pursuing more finely tuned education interventions.

Our findings confirm that PA has protective effects on the cognitive ability of older adults (Blondell et al., 2014; Stephen et al., 2017; Tan et al., 2017). Older adults engaging in mild PA showed improvements in both fluid and crystallized intelligence compared to their peers. However, the protective effects of PA diminished for crystallized intelligence among older adults with increasing exposure to extreme heat events. Nonetheless, our findings reveal that engaging in mild PA, even in the form of mild routine household tasks such as vacuuming, is associated with an increase in fluid intelligence over time, even when older adults face more extreme heat events. Given the typical decline in fluid intelligence with age (Horn & Cattell, 1967; Salthouse, 1999), this finding of mild home activities’ protective influence against the cognitive risks posed by heat is particularly noteworthy. These findings carry profound implications for the development of preventive strategies during heat waves. Although factors like age and family history are nonmodifiable (Alzheimer’s Association, 2023), PA is a variable that can be influenced and promoted by healthcare programs and practices. Choi et al. (2023) found that extreme heat was correlated with accelerated cognitive decline in older adults, particularly in poor neighborhoods where unfavorable environmental conditions may limit opportunities for PA in public spaces. In light of this, advocating mild home activities as a form of daily PA for older adults, especially in regions susceptible to extreme heat and lacking public amenities for exercise, presents a pragmatic and accessible health guideline to support cognitive health and overall well-being in aging populations. However, it is important to caution against excessive PA during heat waves.

The correlation observed between AD PGS and crystallized intelligence underscores the importance of recognizing genetic susceptibility within disproportionately vulnerable groups. Our research elucidates the heightened cognitive risk faced by older adults with a genetic predisposition to AD when exposed to extreme heat, resulting in diminished crystallized intelligence. These findings can guide the development of targeted healthcare programs for individuals with genetic predispositions to AD. For instance, education and awareness campaigns can be implemented to highlight the elevated risks of extreme heat associated with genetic predispositions. Healthcare providers and institutions can incorporate heat-related risk assessments into routine health screenings for individuals predisposed to AD. They can also extend necessary assistance and resources, and offer tailored guidance on preventive measures to address the specific needs of this population.

At the interpersonal dimension of the SEM framework, our study underscores the importance of social engagement, especially with children, as a protective factor for cognitive functioning during extreme heat events. Although increased exposure to extreme heat events was associated with reduced fluid intelligence, this adverse effect of extreme heat was mitigated among older adults engaging more frequently with children, resulting in increased fluid intelligence. However, this protective effect did not extend to crystallized intelligence scores. Various studies demonstrate that heightened social engagement correlates with slower cognitive decline and reduced dementia risk (Bourassa et al., 2017; Lövdén et al., 2005; Saczynski et al., 2006). Older adults with frequent family interactions often exhibit higher self-esteem, reduced stress, and better cognitive health (Fratiglioni et al., 2004). Individuals with greater cognitive reserve effectively manage age-related brain changes (Stern, 2003). Increased interactions with children might foster the development of a resilient cognitive reserve among older adults amidst extreme heat exposure. Expanding upon these interpersonal factors, health programs could advocate for community events or programs that facilitate social interactions across generations, fostering cognitive health and mitigating vulnerability to environmental stressors, especially for older adults living alone.

We acknowledge certain limitations of our study. Despite efforts to control for individual- and wave-fixed effects to integrate community, organizational, and societal factors, there remains the possibility of omitted variables. Furthermore, limitations in the HRS data set hinder our ability to elucidate the precise biological or neural mechanisms underlying brain damage caused by extreme heat. Future research employing pre- and post-heat exposure brain imaging may be better suited to address this question.

Despite its limitations, our study provides empirical evidence linking extreme heat events to a decline in fluid intelligence among older U.S. adults. It emphasizes the importance of promoting modifiable lifestyles, such as exercise and social activities, to build resilience and delay extreme heat-related cognitive decline in older adults.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Conflict of Interest

None.

Data Availability

Due to the utilization of restricted data, specifically pertaining to participants’ county of residence, this study is bound by regulations that prevent the distribution of such data to the public. Approval from the Institutional Review Board (IRB) was obtained to ensure compliance with protocols governing the use of restricted data in this study. This study was not preregistered.