Home Ambient Temperature and Self-Reported Attention in Community-Dwelling Older Adults

Abstract

Background

Climate change is expected to disrupt weather patterns across the world, exposing older adults to more intense and frequent periods of hot weather. Meanwhile, lab-based studies have established a causal relationship between ambient temperature and cognitive abilities, suggesting the expected rise in temperature may influence older adults’ cognitive functioning. Nevertheless, it is not clear whether, and to what extent, the temperature variations in older adults’ own homes—which unlike lab settings are under their control—influence their cognitive functioning. Our objective was to provide proof of concept that home ambient temperature influences self-reported ability to maintain attention in older adults.

Methods

We conducted a longitudinal observational study, continuously monitoring the home ambient temperature and self-reported difficulty keeping attention for 12 months in 47 of community-dwelling older adults living in Boston, Massachusetts.

Results

We observed a U-shaped relationship between home ambient temperature at the time of assessment and the odds ratio (OR) of reporting difficulty keeping attention such that the OR was lowest between 20°C and 24°C and doubled when moving away from this range by 4°C in either direction.

Discussion

Our results suggest that even under the current climate, a considerable portion of older adults encounter indoor temperatures detrimental to their cognitive abilities. Climate change may exacerbate this problem, particularly among low-income and underserved older adults. Addressing this issue in public health and housing policy is essential to building climate resiliency in this vulnerable population.

Aging diminishes the physiological response to changes in ambient temperature because it impairs thermoregulation, fluid regulation, and cardiovascular system functions (1). The ability for heat (or cold) adaptation in older adults is also reduced by chronic conditions that are prevalent in this population (2), such as type 2 diabetes, which reduces density of sweat glands (3), or Parkinsonism, which alters the neuronal substrate of thermoregulation (4). Thermal perception and adaptation are also diminished by medications commonly used by older adults (5), such as beta blockers, diuretics, and antihistamines. As a result of these physiological vulnerabilities, older adults are more likely to experience the adverse effects of exposure to hot or cold environments (1,6,7), and old age is a well-established determinant of vulnerability to extreme weather (8–11).

Among the many aspects of health that are negatively affected by suboptimal ambient temperatures, reduced cognitive abilities (12–14) is particularly detrimental to the overall health and well-being of older adults, especially those who already suffer from existing cognitive impairments. The link between ambient temperature and cognitive function has been established both in controlled experiments with (7,14–19), and population-level studies that use aggregated exposure estimates and “snapshot” measurement of cognitive function (20–24). Although both types of studies offer valuable insights, they are not directly applicable to the daily lives of older adults within their own homes, where many behavioral and environmental effect modifiers are present, and the individuals have at least some levels of control over the ambient temperature. Despite the well-established link between ambient temperature and cognitive abilities, very few studies (25) have examined it within older adults’ own home. As a result, there is a lack of knowledge on whether, and to what extent, older adults experience temperatures that are detrimental to their cognitive abilities during their daily life and within their own environment.

Establishing a link between the temperature the individual is habitually exposed to in their daily life and their cognitive abilities requires home-based monitoring over extended periods of time, which poses many practical and technological challenges. This article presents one of the first efforts to use smart sensors and smartphone-based surveys for long-term home-based monitoring of environment and attention in older adults, using a recently developed data collection platform (26). The objective was to demonstrate, as a proof of concept, the relationship between home ambient temperature and the self-reported ability to maintain attention on daily tasks in a cohort of community-dwelling older adults.

Method

We conducted a longitudinal observational study to monitor the home thermal environment and subjective attention in community-dwelling older adults. The project was approved by the Advarra IRB (protocol ID: Pro00047567). This study was conducted in Boston, Massachusetts, USA.

Participant Recruitment

The rolling enrollment spanned from October 2021 to September 2022. Data collection started on October 28, 2021, with the first participant’s enrollment, and ended on March 29, 2023. We recruited older adults who were at least 65 years old, were able to complete daily smartphone-based surveys, resided within the greater Boston metropolitan area, lived independently in the community, and had access to a reliable internet connection at home. Exclusion criteria: (1) intending to move out of their current place of residence in the 12 months following the enrollment; (2) spending most of their time away from home; (3) having any acute or unstable medical condition; (4) self-reported physician-diagnosed dementia or any condition that could hinder their ability to understand the study protocol and safely follow study procedures.

Recruitment was completed via distribution of flyers in senior living facilities and neighborhood community centers, as well as both subsidized and private senior housing sites. Those who were interested and met the eligibility criteria were contacted by phone and scheduled for an in-person screening and baseline health assessment, which took place in their own home. Participants who met the above criteria were provided with a written informed consent form, which they read and signed. During the same visit, we collected medical history information, completed the baseline Montreal Cognitive Assessment (MoCA) (27–29), installed environmental sensors in participants’ living rooms, and instructed/trained them on how to complete daily smartphone-based surveys. We then monitored each participant for a year, promptly addressing gaps in incoming data to minimize data loss. Some participants were part of the study for less than a year due to the rolling enrollment.

Outcome, Exposure, and Covariates

The outcome of interest was self-report of having difficulty keeping attention. The exposure variables were objective and subjective measures of ambient temperature at the time of assessment (modeled separately). The objective measure was the sensor-measured dry-bulb ambient temperature of the room. The subjective measure was self-reported thermal sensation on a 7-point scale ([very cold] [cold] [slightly cold] [neutral] [slightly hot] [hot] [very hot]). The corresponding scores in our analysis ranged from −3 (very cold) to +3 (very hot) with 0 as neutral. Continuously measured covariates were time-synced home ambient humidity, day of the year, and time of the day. Patient-centered covariates were baseline measurements of health (assessed using the Charlson Comorbidity Index; CCI (30,31)) and cognitive function (assessed by MOCA), biological sex, housing type (subsidized vs private market-rate housing), and living status (alone vs with a partner).

Measurements

Ambient temperature

We continuously measured indoor air temperature and relative humidity in an area of the home identified by the participant as the area they spend most of their time in using Netatmo Indoor Sensors (Netatmo, France). We used Netatmo’s Application Programming Interface to transmit measured ambient temperature and humidity to our cloud storage every 15 minutes. Self-reported thermal sensation and attention were captured twice per day via smartphone surveys. Participants received a text message with a link to the survey that asked them to report their thermal sensation and answer the question “Right now, is it difficult to keep your attention on what you are doing?” Surveys were sent at different times of the day to be able to capture diurnal variation in both temperature and attention. However, to minimize disruption to participants’ daily life, we limited this time to 9 am–6 pm. All surveys began by asking participants whether they were at home, allowing us to exclude responses that were completed outside. The surveys were hosted on a Redcap database that instantaneously provided the study staff with time-stamped responses. Throughout the study, we monitored the incoming data from all participants on a weekly basis and promptly addressed gaps to ensure high compliance. Patient-centered covariates (CCI, MOCA, housing type, sex, and living status) were measured/recorded during the baseline visit.

Analysis

All surveys and sensor-measured environmental data (indoor temperature and humidity) were time-stamped, allowing us to time-sync each survey with the environmental data measured at the time of survey completion. We excluded the surveys that were completed outside of the home. We removed outliers in sensor-measured temperature data by applying the interquartile range method (32) of outlier detection to each participant’s ambient temperature data.

Our hypothesis was that both objective and subjective measures of home ambient temperature were associated with difficulty keeping attention. We therefore created 2 logistic regression models with the same dependent variable (self-reported difficulty keeping attention) but different exposure variables (sensor-measured ambient temperature and subjective thermal sensation). We allowed each participant to have their own intercept, acknowledging the serial correlation in assessments arising from the repeated measures design. Further, we used generalized additive models (33) (GAMs) to fit a curve, instead of a line, to the data because we expected the relationship between the exposure and the predictors to be nonlinear. GAMs are a subset of generalized linear models that can incorporate nonlinear dependent variables based on unknown smooth functions of the predictors, replacing the traditional beta coefficient with the smooth function S. Equations 1 and 2 show the fully adjusted models:

$${\displaystyle \begin{array}{l}{\displaystyle \begin{array}{llllll}& {\displaystyle Logit\left(Difficultykeepingattention\right)\sim \beta 0_i}& +\beta 1\times baselineCC{I}_i+\beta 2\times se{x}_i+\beta 3\times MOC{A}_i& {\displaystyle +\beta 4\times livingstatu{s}_i+\beta 5\times housingtyp{e}_i}& +S_1\left(ambienttemperatur{e}_{j,i}\right){\displaystyle +S_2\left(relativehumidit{y}_{j,i}\right)}& +S_3\left(dayoftheyea{r}_j\right)+S_4\left(hourofda{y}_j\right)\end{array}}\end{array}}$$ (1)

$${\displaystyle \begin{array}{l}{\displaystyle \begin{array}{llllll}& {\displaystyle Logit\left(Difficultykeepingattention\right)\sim \beta 0_i}& +\beta 1\times baselineCC{I}_i+\beta 2\times se{x}_i+\beta 3MOC{A}_i& {\displaystyle +\beta 4livingstatu{s}_i+\beta 5housingtyp{e}_i}& +S_1\left(thermalsensatio{n}_{j,i}\right)+S_2\left(relativehumidit{y}_{j,i}\right)& {\displaystyle +S_3\left(dayoftheyea{r}_j\right)+S_4\left(hourofda{y}_j\right)}\end{array}}\end{array}}$$ (2)

Where i and j represent the ith participant and the jth observation, and S_1–4 represent the smooth functions in the GAM. Because GAMs do not result in a traditional beta coefficient for the nonlinear terms, we plotted the relationships between outcome and exposure variables, holding other variables at their median.

Results

Recruited Participants and Descriptive Data

Figure 1 and Table 1 show the recruitment diagram and characteristics of our cohort. On average (SD), participants completed 373 (±158) surveys while they were at home. In total, our analysis included 17 201 surveys synced with environmental data. Figure 2 (top panel) shows the overall distribution of both sensor-measured home ambient temperature (left) and individuals’ subjective thermal sensation. In the bottom panel of Figure 2, we plotted the sensor-measured air temperature at the time of survey completion against subjective thermal sensation. These data indicate that ambient temperature tracks with individuals’ subjective thermal comfort, which is most likely to be “neutral” between 22°C and 23°C. This is an expected trend previously observed in numerous studies of human comfort. Further, our data show that subjective thermal sensation of “cold” is associated with a noticeably wider range of ambient temperatures compared to other thermal sensations.

Figure 1.

Enrollment and retention of participants.

Table 1.

Participants’ Characteristics and Overview of Survey Results

Characteristic	Mean	SD
Age (y)	79	7.2
MoCA score (out of 30)	25.6	3.12
Charlson Comorbidity Index (CCI)	5.5	2.3
Number of surveys completed at home	373	158.2
% time feeling hotter than neutral	6.5	7.38
% time feeling colder than neutral	16.8	17.60
% of time experiencing poor attention	5.7	10.08
Sensor-measured home temperature at the time of survey completion (C)	22.3	1.47
Biological sex	Female: 38, male: 9
Race	White: 43, Black: 3, non-White Latino: 1
Living status	Alone: 29, with a partner: 18
Housing type	Subsidized: 13, private: 34

Note: MoCA = Montreal Cognitive Assessment.

Figure 2.

Top panel: Histograms of sensor-measured home ambient temperature and subjective thermal sensation at the time of survey completion. Bottom panel: Range of sensor-measured air temperature at each level of subjective thermal sensation (all measurements from all participants).

Main Results

Regression results are presented in Table 2. We did not report β coefficients for temperature, humidity, day of year, and time of the day because we modeled them as nonlinear predictors. The relationship between both objective and subjective temperature and model-predicted odds ratio (OR) of reporting difficulty keeping attention is presented graphically in Figure 3.

Table 2.

Fully Adjusted Regression Results for Factors Associated With Difficulty Keeping Attention

Variable	Exposure = Sensor-Measured Ambient Temperature		Exposure = Thermal Sensation
	p Value	β	p Value	β
Temperature (exposure)	<.001	NA	NA	NA
Humidity	<.001		<.001
Day of the year	<.001		.07
Hour of the day	.01		.01
Baseline cognitive function (MoCA score)	.09	−0.03	.015	−0.04
Baseline health status (CCI index)	<.001	−0.09	<.001	−0.13
Biological sex (f = 0, m = 1)	<.001	1.99	<.001	1.85
Living alone (N = 0, Y = 1)	<.001	0.81	<.001	0.90
Affordable housing (N = 0, Y = 1)	<.001	1.62	<.001	1.44
Model AIC	6 239		6 164

Note: AIC = Akaike Information Criterion; CCI = Charlson Comorbidity Index; MoCA = Montreal Cognitive Assessment.

Figure 3.

Relationship between both home ambient air temperature (left panel) and subjective thermal sensation (right panel) and odds ratio of experiencing difficulty keeping attention.

These results reveal a relationship between home ambient temperature at the time of assessment and the OR of experiencing difficulty keeping attention. There was a U-shaped relationship between the sensor-measured ambient temperature and difficulty keeping attention such that the OR of reporting difficulty keeping attention was lowest between temperatures of 20°C and 24°C, and essentially doubled when moving away from this range by 4°C in either direction. Similarly, we observed a relationship between the subjective report of temperature (thermal sensation) and the OR of experiencing difficulty keeping attention. This relationship was also U-shaped, with the OR increasing in both “hot” and “cold” directions. Considering the descriptive data above, these results suggest that self-reported ability to pay attention is optimal when the thermal sensation is “neutral,” and the objective temperature is around 22°C.

Discussion

Interpretation of Findings

In a cohort of community-dwelling older adults living in Boston, Massachusetts, home ambient temperature was linked to the likelihood of experiencing difficulty keeping attention. We observed U-shaped relationships between both objective and subjective measures of home thermal environment and time-synced attention, whereby both hot and cold temperatures are associated with difficulty keeping attention. It is important to note that in this purely observational study, we neither influenced the thermal environment nor limited participants’ behavioral adaptations to changes in ambient temperature. Further, this cohort was relatively healthy, lived in a wealthy (in global terms) society, and had a relatively high standard of housing and socioeconomic status even compared to the general older adult population in the same city. These observations therefore build upon past lab-based research and reveal that cognitive abilities could be affected by the thermal environment even within individuals’ own homes where they are at least partially in control of the ambient temperature.

In the direction of heat, the increase in the OR of self-reported difficulty keeping attention was similar using both subjective and objective measures of ambient temperature as the exposure variable. In other words, as the objective temperature rises, individuals are more likely to report hotter than neutral thermal sensations and complain about self-reported difficulty keeping attention. However, in the direction of cold, a change in thermal sensation from “neutral” to “cold” was associated with a substantially larger increase in the OR of self-reported difficulty keeping attention than the increase predicted by the corresponding cooling of objective ambient temperature. In addition, the “cold” thermal sensation was associated with a larger range of objective air temperature than any other thermal sensation. Several environmental, physiological, and behavioral processes could contribute to this interesting observation.

One possible environmental explanation is the difference in radiant heat exchange between the body and its surroundings during hot and cold seasons. The surface temperature of walls and windows is typically close to indoor air temperature during mild or even hot weather, meaning that indoor air temperature and mean radiant temperature (MRT; representing the average temperature that accounts for the temperature of all objects surrounding a person with which the person exchanges radiant heat) are very close. On the other hand, due to the large temperature difference between indoors and outdoors during the winter, the wintertime MRT is generally lower than air temperature. As a result, during the winter, the thermal interaction between the body and the surrounding environment cannot be completely captured by ambient temperature alone. Future research should therefore include radiant temperature (or wall surface temperature) measurements. Another potential explanation for greater sensitivity to cold may be that physiologic mechanisms that enable adaptation to cold (eg, brown fat, vasoconstriction, mitochondrial energy production) may be less effective than those that enable adaptation to heat (eg, sweating, vasodilatation, thirst) in the older participants in our study. Unfortunately, we were unable to assess these mechanisms.

Although important limitations (described below) exist that prevent us from generalizing these results to the entire older adult population in Boston, the potential clinical implications are substantial because they provide strong proof of concept that the naturally occurring variations in home ambient temperature contribute to cognitive complaints in older adults. Notably, home ambient temperature, compared to other pharmaceutical or medical interventions, presents a substantially cheaper and safer point of intervention to improve habitual cognitive function and performance in older adults.

Effect of Climate Change and Need for Interventions

Climate change is expected to disrupt weather patterns across the world, thus threatening public health and well-being, especially among older adults who are more vulnerable to heat and cold (34–42). Within this context, the home ambient environment plays a critical role (43) because many older adults, even in wealthier societies, either cannot afford heating and cooling systems, or they do not have sufficient motor and cognitive abilities to properly use them (44–48). This study reveals yet another potential impact of suboptimal home ambient temperature on health and well-being of older adults, making the case for much-needed interventions in public health policy, housing regulations, and healthcare practices to help protect vulnerable populations from the near-future impacts of climate change. Notably, our past work, along with the relevant building science literature (49–56), demonstrates that the housing stock in general, including houses built specifically for older adults, is not resilient to the expected increase in frequency and intensity of hot and cold weather events. We therefore recommend exploring technological, financial, and policy interventions that enable older adults to maintain a comfortable and health-promoting home thermal environment. These include development of automated temperature control systems responsive to the specific needs of older adults with cognitive or physical impairments, financial support to protect them from energy poverty, and investment in housing retrofit and policies that improve the passive resilience of the housing stock to future weather.

Limitations

Sample

Our sample had a relatively high socioeconomic status and housing quality. Low-income and underserved populations are less likely to be able to maintain optimal home ambient temperature due to energy poverty and substandard housing. Possibly, this results in an underestimation of the effects observed in this study.

Measurement of attention

To ensure compliance with survey completion, we had to design short and simple surveys that can be quickly completed on a smartphone. Unfortunately, to date, no well-established survey of attention or other cognitive outcomes exists that is suitable for this type of repeated measurements at home. The specific question used in this study has not been validated against objective measures of attention or cognitive performance. Future studies can use technology (eg, smartphone-based assessments) to build upon this work and study more objective measures of attention and cognitive performance.

Medication

We did not monitor changes in medication throughout the study and therefore could not account for this potentially important variable.

Generalizability to other climates

This study was conducted in one city. It is noteworthy that climate, demographics, and housing characteristics can vary substantially between cities.

Conclusion

In a cohort of community-dwelling older adults living in Boston, Massachusetts, we observed U-shaped relationships between both objective and subjective measures of home thermal environment and time-synced attention. This finding is in line with past lab-based work on ambient temperature and cognitive performance/function and reveals that the daily variations in home ambient temperature, which are supposedly controlled by the individual, can also influence the ability to pay attention in older adults. Given that climate change is expected to increase both frequency and intensity of extreme weather events, it is important to develop technological, financial, and policy interventions that enable older adults to maintain a comfortable and health-promoting home thermal environment.

Funding

A.B. was supported by a T32 fellowship through U.S. National Institute on Aging (T32AG023480). Authors also acknowledge funding from the TMCITY foundation.

Conflict of Interest

None.

Author Contributions

A.B.: conceptualization, methodology, formal analysis, investigation, data curation, writing—original draft, visualization; W.Y.: project administration, investigation, supervision, writing—review and editing; T.T.: conceptualization, formal analysis, methodology, writing—review and editing; B.M.: conceptualization, supervision, funding acquisition, methodology, project administration, investigation, resources, writing—review and editing; L.L.: conceptualization, supervision, funding acquisition, methodology, project administration, investigation, resources, writing—review and editing; R.D.: investigation, writing-review and editing.