Wearables for Measuring Health Effects of Climate Change–Induced Weather Extremes: Scoping Review


IntroductionBackground

Climate change is one of the biggest global health threats of the century [], and the field of climate and health research has been rapidly growing []. Many environmental conditions such as rising temperatures, floods, wildfires, heat waves, droughts, and other extreme weather events can be linked to climate change according to the 2021 Intergovernmental Panel on Climate Change report [] and may, directly and indirectly, impact human health []. The wide-ranging health effects of these weather extremes include malnutrition from food insecurity; infectious disease; respiratory, cardiovascular, neurological, and mental health disorders; and mortality [,].

Epidemiological studies often focus on the relationship between heat and mortality or morbidity in terms of the number of hospital admissions or long-term effects but do not consider individual exposure and direct health effects []. Furthermore, most studies use weather and climate data from satellites or the nearest weather station, which is often located at the airport. These approaches do not consider granular spatial and temporal differences in weather exposure or individual factors that influence the exposure such as time spent indoors [,]. To this end, consumer-grade wearable devices (hereafter wearables) could generate high-resolution data at the individual level, measuring exposure and health parameters in the real-life environment, the ecological momentary assessment []. Wearables can cover a variety of variables and physiological data, including, among others, activity levels, sleep, sweat rate, and heart rate (HR) [], presenting a potential solution to the shortage of short-term and individual-level data in climate change and health research.

In recent years, some reviews have been conducted on the assessment of heat strain and individual heat exposure using wearable devices. However, these studies have mainly focused on urban and occupational heat exposure [,], although populations living in low- and middle-income countries and rural settings have a high vulnerability to climate change []. Although the urban heat island effect describes higher heat exposure in cities owing to human activities and dense concentrations of surfaces that absorb and retain heat, rural populations are often more exposed because of their reliance on climate-sensitive livelihoods [,]. Some reviews have examined the validity of various wearables but only in moderate climate settings [,]. Furthermore, many studies [,] used prototypes and not off-the-shelf devices, which make them difficult to reproduce in the field.

Research Objectives

Therefore, the overarching objectives of this review were (1) to map the available research on the use of off-the-shelf wearables for measuring direct health effects of and individual exposure to climate change–induced weather extremes such as heat, (2) to examine current approaches to wearable use in this field, and (3) to identify gaps in the research. We particularly focused on (1) demographic characteristics, (2) selected wearable devices and their measures, (3) extreme weather condition exposure and data collection methods, (4) analytical approaches, (5) validity of wearables in extreme weather conditions, and (6) observed effects of extreme weather exposure on health (especially of heat on sleep, physical activity, and HR, as well as occupational heat stress).


MethodsOverview

The methodology for this scoping review was based on the framework outlined by Arksey and O'Malley [] and Peters et al [] and in accordance with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) [] (). A review protocol can be obtained from the principal author (MK) upon request. A scoping review seemed most appropriate to approach the research objective, as initial research into this topic revealed a broad scope of heterogeneous studies, however, limited in their numbers.

Eligibility Criteria

We included articles that were available in English and published after January 1, 2010, because wearables have become widely available on the consumer market and were also increasingly adopted in research since then [,]. Types of studies included were case studies, observational studies, non–randomized controlled trials, and randomized controlled trials. We included any consumer- or research-grade wearables that were available off-the-shelf, could be worn on the body, and were neither invasive nor obtrusive (excluding, eg, ingestible, handheld, or wired devices). All types of sensors or measurements that measured at least one physiological parameter or individual exposure were included. For a complete list of eligibility criteria, see .

Inclusion and exclusion criteria.

Inclusion criteria

Publications:Full text availablePublished in the English languagePublished between January 2010 and September 2021Randomized controlled trials (RCTs), non-RCTs, observational studies, or case studiesWearable device:Off-the-shelf wearable electronic devicesNoninvasive and nonobtrusiveMeasuring at least one physiological parameter (eg, heart rate or sleep duration) or individual exposure (eg, ambient temperature)Climate change:Climate change–related weather extremes: heat, flood, drought, wildfire, tropical cyclone, or heavy precipitationExposure: outdoors, indoors, or in a climatic chamber or laboratoryOutcomes:Individual effect of climate change–related environmental condition measured with wearablesValidity and method comparison of wearables in extreme weather conditions

Exclusion criteria

Publications:Nonhuman study populationReviews, editorials, or commentariesWearable device:Not commercially available (eg, prototype or design study)Wearable with interventional function only (eg, cooling vest)Smartphone used as a wearableWearable not implementedClimate change:Other environmental conditionsExposure to heat in the context of mining or firefightingOutcomes:Wearable (data) not specifically included in outcomesEnvironmental exposure or condition not included in outcomesWearable only used to assess the effect of another intervention (eg, cooling)Textbox 1. Inclusion and exclusion criteria.

Individual effects of climate change were limited to those resulting from exposure to weather extremes, as the topic would have been too broad otherwise []. As per the 2021-published Intergovernmental Panel on Climate Change report [], we included exposure to heat and heat waves, heavy precipitation, floods, tropical cyclones, droughts, and wildfires. As heat and heat waves are often defined as extremes relative to the local climate (ie, daily minimum and maximum temperatures above the 95th or 99th percentile of the climatological record or a baseline period) [,], we relied on the definitions provided in the included studies. If the authors did not provide a definition, we used one of the following classifications, based on the available data in the screened articles:

If data were available on wet bulb globe temperature (WBGT) [,] or the universal thermal climate index [], we used >26 °C as a threshold.If data were available on ambient temperature and relative humidity, we calculated the heat stress index (HSI) [] and used a threshold of >26 °C HSI.If data were available on ambient temperature, we used the average relative humidity at the study location (city or country) during the study period to calculate the HSI.

Studies on the effect of temperature on sleep were included even for lower ambient temperatures, as previous research has shown that small temperature changes already have adverse effects on sleep quality and duration [] because humans only have a minimal ability to thermoregulate in rapid eye movement sleep phases []. We also included studies that reported on indoor heat exposure in climatic chambers or laboratories. We excluded studies on heat exposure during firefighting and mining, as we considered them job-specific and they predominantly assessed the microclimate inside the protective gear [].

In case no full text was available or information on the wearables was missing, the authors were contacted 3 times before exclusion.

Search Strategy and Information Sources

The full search was conducted on September 1, 2021, by 1 reviewer (MK) in 6 electronic databases: PubMed (MEDLINE), Scopus, CINAHL (EBSCOhost), IEEE Xplore, Ovid (MEDLINE[R]), and Web of Science. Gray literature was searched with Google Scholar, and the first 1000 search results were included []. We manually searched references of relevant included and excluded articles for further sources of evidence.

We followed the Population/Patients, Intervention, Comparison, and Outcome (PICO) framework to compile the search strategy. Population (P) included study participants wearing a wearable. Intervention (I) included exposure to climate change–induced weather extremes. No comparison (C) was required. Outcomes (O) included psychological and physiological health parameters or exposure measurable with wearables. Accordingly, the databases were searched using a search string including synonyms and medical subject headings terms for these concepts. Search strings were adapted to the specific requirements of each database (see for the full search strings). We applied a search filter for publications after January 1, 2010.

Study Selection

The search results were imported into the literature reference management system EndNote 20 (Clarivate Analytics) and then imported into the systematic review management software Covidence (Veritas Health Innovation) where duplicates were removed automatically as well as manually. We screened titles and abstracts, as well as full texts, with application of the inclusion and exclusion criteria (see for a full list of criteria). Subsequently, we extracted data from the included literature. The screening process was piloted prior with a sample of 20 articles. The literature was screened by 2 independent reviewers (MK and IM). Any disagreements were resolved by consensus between the 2 reviewers (MK and IM) and an independent researcher (SB).

Data Extraction

A data-charting form was developed using the Covidence software template and piloted on 3 articles; data were charted by the 2 reviewers independently, and any disagreements were mutually resolved. The following data categories were extracted and synthesized [,]: title, author, year, country of study, objectives of study, demographics of the study population, sample size, methods, intervention type, outcomes, and key findings related to the scoping review question. In addition, the following items were extracted: wearable models, measured parameters with wearable, study setting, climate change–related environmental conditions including the measurement method, and methods used for data analysis or correlation.

Synthesis of Results

The characteristics of the included studies and the study populations were summarized using Microsoft Excel (version 2206; Microsoft Corporation), and the study outcomes were narratively synthesized. The purpose of the use of the wearables was identified according to three categories: (1) validity and comparison in extreme conditions, (2) measuring individual exposure, or (3) measuring direct health effects.


ResultsOverview

The initial search yielded 1831 results, and 40 references were added after a manual search. We removed 419 duplicates and screened the titles and abstracts of the remaining 1452 nonduplicates. From a total of 190 screened full-text articles (186 studies), we included 53 studies (56 articles; see for the PRISMA [Preferred Reporting Items for Systematic Reviews and Meta-Analyses] flow diagram) including 1 preprint article []. For the conducting of this scoping review the preprint article was used and is therefore cited throughout the manuscript instead of the accepted article [] that was published after our last search and data extraction process.

Figure 1. PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flowchart. View this figureStudy Characteristics

In total, we included a study population of 55,284 participants in this review (the characteristics of the included studies are summarized in and ). Overall, there have been an increasing number of publications using wearables in the context of climate change and health research since 2010 (). The included studies were mostly observational (35/53, 66%) and crossover studies (21/53, 40%). Most studies were conducted in countries classified by the World Bank in 2022 [] as upper–middle-income (5/53, 9%) and high-income countries (47/53, 87%), especially with more than half of the total studies conducted in North America (31/53, 58%). A few studies (4/53, 8%) included lower–middle-income countries. Most studies were conducted in urban settings (25/53, 47%) or in a climatic chamber (19/53, 36%), with a short study duration of up to 1 week (16/53, 30%) or up to 5 cross-sectional data collection points (17/53, 32%).

The median number of participants per study was 39 (range 6-47,628), comprising an average of 67% of male participants ( shows the demographics of the study population). In total, of the 53 studies, 15 (28%) studies focused solely on male participants versus 3 (6%) studies that only included female participants. A few studies (3/53, 6%) specifically included nonhealthy participants. Most study populations consisted of outdoor workers (14/53, 26%), including farm workers, construction workers, traffic police officers, or other workers, as well as the general population (11/53, 21%) or university members (students and staff; 7/53, 13%). Of the 53 studies, 2 (4%) studies included older adults and 4 (8%) studies included children. In addition, the study populations of individuals in the military, athletes, and homeless individuals were each represented in 2% (1/53) of studies.

Figure 2. Map of study locations (countries). Minor et al [] mentioned 68 countries across all continents (except Antarctica) but did not further specify, so they were not included in the map. View this figureTable 1. Study characteristics.Study characteristicsStudies (N=53), n (%)Participants (N=55,284), n (%)Regions and countriesa
North America30 (56.6)3524 (6.4)

United States24 (45.3)2807 (5.1)

Canada5 (9.4)697 (1.3)

Mexico1 (1.9)20 (0)
Asia11 (20.8)1226 (2.2)

Hong Kong1 (1.9)740 (1.3)

China3 (5.7)161 (0.3)

India3 (5.7)141 (0.3)

Japan2 (3.8)97 (0.2)

Singapore2 (3.8)87 (0.2)
Europe5 (9.4)94 (0.2)

Belgium1 (1.9)39 (0.1)

United Kingdom2 (3.8)33 (0.1)

Germany1 (1.9)15 (0)

Cyprus1 (1.9)7 (0)
Oceania4 (7.5)597 (1.1)

Australia4 (7.5)597 (1.1)
Middle East3 (5.7)2192 (4)

Qatar1 (1.9)2088 (3. 8)

Israel1 (1.9)104 (0.2)

Saudi Arabia1 (1.9)23 (0)
South America0 (0.0)0 (0)
Africa0 (0.0)0 (0)
Countries not specified (68 countries: 42 high-income countries; 17 upper–middle-income countries; 9 lower–middle-income countries)1 (1.9)47,628 (86.2)Study settinga,b
Urban31 (58.5)—c

Outdoor9 (17.0)—

Indoor5 (9.4)—

Indoor and outdoor17 (32.1)—
Rural11 (20.8)—

Outdoor7 (13.2)—

Indoor1 (1.9)—

Indoor and outdoor3 (5.7)—
Climatic chamber or laboratory19 (35.8)—Study duration
Cross-sectional data collection points (up to 4 hours each)

≤5 data collection points17 (32.1)496 (0.9)

≥6 and ≤10 data collection points3 (5.7)174 (0.3)

≥11 and ≤50 data collection points3 (5.7)116 (0.2)
Continuous monitoring (at least 1 [work] day)

≤7 days16 (30.2)3614 (6.5)

≤1 months6 (11.3)171 (0.3)

≤6 months5 (9.4)542 (1)

≤2 years3 (5.7)50,171 (90.8)Study design
Experimental crossover studyd7 (13.2)210 (0.4)
Prospective cohort study20 (37.7)49,690 (89.9)
Retrospective cohort study1 (1.9)104 (0.2)
Prospective observational crossover studyd14 (26.4)5017 (9.1)
Method comparison or evaluation study11 (20.8)263 (0.5)

aMultiple characteristics may apply per study.

bInformation for study settings is not available for all study participants and therefore not summarized here as the number of participants per study setting.

cNot available.

dEach participant serves as their own control or comparison.

Table 2. Years of publication.Year of publicationIncluded publications (N=56), n (%)20133 (5)20146 (11)20156 (11)20165 (9)20175 (9)20186 (11)20198 (14)202011 (20)2021 (until September 1)6 (11)Table 3. Demographics of included studies.StudyParticipants monitored with wearables, nStudy populationSex (male), %Age (years)Ethnicity, %Al-Bouwarthan et al [], 202023Construction worker100Mean 42.7 (SD 8.8)—aAl-Mohannadi et al [], 20162088General population67Range 18-65—Al Sayed et al [], 201712Male100Mean 24.8 (SD 3.8)—Bailey et al [], 201938University member50Group 1: mean 32.6 (SD 13); group 2: mean 21.5 (SD 3)92% WhiteBenita et al [], 2020; Benita and Tuncer [], 201910University student; female0Mean 22.8 (SD 1.5)—Benjamin et al [], 202019Athlete; female0Mean 20.6 (SD 1.4)—Bernhard et al [], 201581Outdoor worker or general population35Mean 52 (rural), 50.5 (urban), and 44.5 (outdoor worker)93% Black or African AmericanCedeño Laurent et al [], 201844University student; healthy51Mean 20.2 (SD 1.8)40% WhiteCheong et al [], 20209Older adult22Range 65-8767% White, 11% Black, 11% Hispanic or Latino, and 11% otherCuddy et al [], 201356Male100Mean 22 (SD 3)—Culp and Tonelli [], 201920Farm worker; male100Range 18-65100% HispanicEdwards et al [], 2015372Children (age 3 years at recruitment); healthy52Mean 3.4 (SD 0.3)22% Black or African AmericanHamatani et al [], 201713General population92——Hass and Ellis [], 201945General population37Range 18-≥6564% White and 11% Black or African AmericanHondula et al [], 202084General population———Ioannou et al [], 20177Farm worker; healthy71Male: mean 39 (SD 10.8); female: mean 39.5 (SD 13.4)—Jehn et al [], 201415Clinically stable NYHA II-IVb patients with PAHc60Mean 66.7 (SD 5.2)—Kakamu et al [], 202184Construction worker100Mean 48.4 (SD 14)—Ketko et al [], 2014104Military; male100Range 18-21—Kim et al [], 201312Male100Mean 25.5 (SD 4.1)—Kuras et al [], 201523General population39Range 25-7974% White and 26% Black or African AmericanLam et al [], 2021145University student (first-year student)34Mean 18.1 (range 17-21)—Larose et al [], 201460Male; healthy100Mean 45.4 (range 20-70)—Lewis et al [], 20161095Children aged 9-11 years43Mean 10.6 (SD 0.4)—Li et al [], 202010Construction worker; healthy; male100Mean 39.4 (SD 3.6)—Lisman et al [], 201446Military or university community member; healthy or previous exertional heat stroke74Mean 29.7 (SD 5.9)—Longo et al [], 201720Homeless individual or university student75Range 18-60—Lundgren et al [], 201477Outdoor worker86——MacLean et al [], 202012Male; healthy100Mean 24.2 (SD 3.7)—Minor et al [], 202047,628General population69Age distribution: 19-25, 6%; 25-65, 91%; ≥65, 3%—Mitchell et al [], 2018587Farm worker66Mean 38.698% LatinoNazarian et al [], 202177General population52Range 18-48100% AsianNotley et al [], 202150Young (18-30) and healthy or older (50-70) and healthy; older and T2Dd or HTNe100Mean 50 (SD 17); mean per group: 22 (young), 58 (older), 60 (T2D), and 61 (HTN)—Ojha et al [], 202010University student70——Pancardo et al [], 201520Outdoor worker; healthy55Mean 28.6 (range 22-51)—Quante et al [], 2017669Adolescents aged 12-14 years49Mean 12.9 (SD 0.6)68% White, 14% Black, 3% Hispanic, 3% Asian, and 13% OtherRaval et al [], 201816Traffic police worker100Range 19-57—Ravanelli et al [], 2016; Ravanelli et al [], 20158Male; healthy100Mean 24 (SD 3)—Relf et al [], 201814Female; healthy0Mean 26 (SD 7)—Relf et al [], 202019General population; healthy79Mean 41 (SD 23)—Rosenthal et al [], 2020455General population42——Runkle et al [], 2019; Sugg et al [], 201835Outdoor worker100Mean 39.274% White, 14% Black or African American, 9% Hispanic, and 2% American Indian or Alaska NativeSahu et al [], 201348Farm worker100Range 25-34—Seo et al [], 201612Male; healthy100Group 1: mean 23 (SD 1); group 2: mean 23 (SD 2); group 3: mean 24 (SD 2)—Shakerian et al [], 202118University student78Female: mean 24 (SD 3.2); male: mean 24 (SD 2.8)—Shin et al [], 20159Young; healthy67Mean 23.3 (SD 4.1)—Suwei et al [], 201951Outdoor worker35Mean 42.9 (range 21-60)96% African AmericanUejio et al [], 201850Outdoor worker92Mean 44 (SD 11.1)59% Black, 39% White, and 2% HispanicVan Hoye et al [], 201439University student; healthy54Mean 21.4 (SD 1.41)—Williams et al [], 201951Older adult43Mean 65.467% WhiteXiong et al [], 202048General population46Mean 36 (SD 12)—Zheng et al [], 2019740Adolescent or secondary school student52Mean 14.7 (SD 1.6)100% AsianZhu et al [], 20166General population50Males: mean 27.3 (SD 2.5); female: mean 22.3 (SD 1.2)—

aThe respective information was missing in the article.

bNYHA II-IV: New York Heart Association Functional Classification for heart failure stage II-IV.

cPAH: pulmonary arterial hypertension.

dT2D: type 2 diabetes.

eHTN: hypertension.

Wearable Devices

Most of the included studies used 1 (39/53, 74%) or 2 (12/53, 23%) wearables; a few studies (2/53, 4%) used ≥3 devices (study methods and objectives detailed in ). The 70 wearables in the included studies were from 23 different companies overall with Polar Electro (16/53, 30%), Maxim Integrated (13/53, 25%), and Fitbit (5/53, 9%) providing the most frequently used wearables. The most commonly reported use for wearables was the measurement of HR (30/53, 57%), physical activity (15/53, 28%), or individually experienced temperature (IET; the air temperature surrounding the individuals; 14/53, 26%). Other parameters included sleep (duration, onset, wake time, etc), energy expenditure, skin temperature, electrodermal activity, local sweat rate, respiratory rate, or geoposition. Some wearables measured multiple parameters. The devices were mostly wristbands (25/70, 36%), chest straps (18/70, 25%), clipped to clothing or accessories (15/70, 21%), or directly taped to the skin (5/70, 7%). All included studies additionally used questionnaires and further health parameters (eg, blood pressure, weight, height, and urine samples).

Table 4. Study methods and objectives.Methods and objectivesStudies (N=53), n (%)Number of wearables per study
137 (74)
212 (23)
≥32 (4)Wearable company (models)a
Polar Electro (RCX3, H7, RS800XC, FT1, FT7, Team 2 [Pro], RS800, RS400, WearLink, Accurex Plus, A300, and M400)16 (30)
Maxim Integrated (iButton Hygrochron and Thermochron)13 (25)
Fitbit (Ionic, Charge 2, and Flex)5 (9)
Medtronic (Zephyr BioHarness)4 (8)
Philips Respironics (Actical and Actiwatch 2), Onset Corp (HOBO Pendant), and Empatica (E4)3 (6; each)
Crossbridge Scientific (KuduSmart), Actigraph (GT3X and GT3X+), Intel (Basis Peak Watch), BodyMedia (SenseWear Pro 3), Sony (SmartBand Talk SWR30 and SWR12)2 (4; each)
Omron Healthcare (HJ-720 ITC pedometer), STATSports (Viper Pod), Microsoft (Band), Garmin (Vivoactive HR), Aipermon (APM), Stayhealthy (RT3), GISupply (LW-360HR), Lifensense (Mambo 2), LASCAR (EL-USB-2-LCD+), Easylog (Easylog), PAL Technologies (activPAL and activPAL3C)1 (2; each)Measured parameter with wearablea
Heart rate30 (57)
Physical activity15 (28)
Energy expenditure8 (15)
Skin temperature12 (23)
Electrodermal activity5 (9)
Sleep (onset, offset duration, and efficiency)7 (13)
Individually experienced temperature14 (26)
Others (local sweat rate, respiratory rate, and GPS location)7 (13)Wear location of wearablea
Wristband25 (47)
Chest strap18 (34)
Attached to clothing or accessories15 (28)
Taped to the skin5 (9)
Other: shirt, back strap, around upper arm, or not specified8 (15)Climate change–related extreme weather
Heat52 (98)
Wildfire1 (2)Measured environmental conditiona
Temperature50 (94)
Relative humidity40 (75)
Precipitation7 (13)
Other (wind speed, wet bulb temperature, dry bulb temperature, dew point, mean radiant temperature, barometric pressure, visibility, CO2 concentration, and air quality)22 (42)Measurement location or data source for environmental conditiona
Nearest weather station20 (38)
Sensors placed on study site18 (34)
Climatic chamber or laboratory18 (34)
Locally installed weather station4 (8)
Smartphone sensor2 (4)
Satellite data2 (4)Heat stress measurea
Wet bulb globe temperature14 (26)
Heat stress index5 (9)
Humidex2 (4)
Others (universal thermal climate index, heating or cooling degrees, heat stroke index, heat stress days, heat stress level estimation, heat balance equation, extreme heat degree minutes, and physiological equivalent temperature)1 (2; each)
None27 (51)Method of analysis (statistical test)a
Regression (linear, logistic, and Cox)16 (30)
Linear mixed effect model16 (30)
Time-series analysis1 (2)
t test (2-tailed or 1-tailed)21 (30)
Correlation (Pearson, Spearman, etc)13 (25)
ANOVA (one-way, repeated measures, and mixed design)14 (26)
MANOVA1 (2)
Nonparametric test (Wilcoxon U test and Kruskal-Wallis test)7 (13)
Chi-square and Fisher Exact Test4 (8)
Bland Altman plot5 (9)
Spatial correlation1 (2)
Cohen kappa1 (2)
Descriptive analysis only5 (9)Study objectives and use of wearablesa
Studies measuring the correlation of wearables’ data and environmental conditions

Effect of heat on sleep7 (13)

Effect of heat on physical activity7 (13)

Effect of heat on heart rate10 (19)

Other physical responses to heat6 (11)

Occupational heat stress8 (15)

Effect of wildfires on physical activity1 (2)
Studies measuring the individual experienced temperature and comparing it to local or area measurements10 (19)
Studies assessing the validity and applicability of wearables for their use in extreme weather14 (26)

aMultiple characteristics may apply per study.

Weather or Climate Data

The primary focus was on the use of wearables to measure physiological responses to heat exposure (52/53, 98%). Of the 53 studies, 1 (2%) study assessed the impact of forest fires on individual activity, and 5 (9%) measured the effect of precipitation on activity in addition to heat. The weather or climate conditions were predominantly assessed using data from the nearest weather station (20/53, 38%), sensors placed on the study site (18/53, 34%), or measured in a climatic chamber or laboratory (18/53, 34%). Others accessed weather data from locally installed weather stations, built-in sensors of participants’ smartphones, or satellite data. Besides the primarily focused measurements of temperature, precipitation, and relative humidity, 49% (26/53) of the included studies calculated different heat stress indices (eg, WBGT, HSI, or universal thermal climate index).

Statistical Analysis

The methods of statistical analysis of wearables’ data and correlation to climate or weather data were primarily regression, linear mixed effect models, correlation, ANOVA, and 1- or 2-tailed t tests (). Linear regression models or linear mixed effect models, for example, were often used to correlate IETs and area-level temperature data [,,,,], but t tests were also used for the comparison between both methods [,]. Data sources differed between group-level data and participant-level data [,]. The associations of heat exposure and wearables-measured parameters were mostly examined with linear mixed effect models or different regression models (linear, logistic, or Cox), adjusted for age, sex, and education [,,,,,,,,,-]. For the comparison of the effect of heat between groups with different characteristics such as sex or age and for the comparison of heat-stress and non–heat-stress days, t tests, Chi-square tests, and ANOVAs were used [,,,,-,,,,,]. Studies that compared wearables measurements with standard devices applied; in addition to t tests and ANOVAs, different correlation coefficients and Bland Altman plots for the appraisal of disagreement [,,,,]. Other analysis methods were also used. One study [,] spatially correlated different urban environmental exposures and bod

留言 (0)

沒有登入
gif