Harnessing smartphone and environmental data to explore the impact of temperature on real-time mental wellbeing

Harnessing smartphone and environmental data to explore the impact of temperature on real-time mental wellbeing | Research Square Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Harnessing smartphone and environmental data to explore the impact of temperature on real-time mental wellbeing Giulia Vivaldi, Emeka Chukwusa, Arturo de la Cruz Libardi, Antonio Gasparrini, and 8 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8907997/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract Background Smartphone-based ecological momentary assessments (EMAs) offer a unique opportunity to actively engage and involve the public in scientific research and study the impact of environment on mental wellbeing. Objective tracking of environmental exposures is not easily achieved with commonly used portable devices, however linkage with external datasets can enhance geotagged EMA data with detailed environmental information. Methods We illustrate a novel approach matching geotagged EMA assessments from the Urban Mind study with ambient temperature, air pollution, and green space exposure data. We applied a case-time series design, self-matched analytic methods (distributed lag non-linear models) and a multistep approach to missing data to analyse associations between ambient temperature and momentary wellbeing, calculated using self-reported contemporaneous levels of positive (5 items) and negative affect (five items). Results 7088 assessments from 266 participants over 14 days were successfully matched to ambient environmental data. We found little evidence that daily maximum temperatures are associated with momentary mental wellbeing, nor for effect modification by air pollution or green space exposure. Conclusions This study illustrates the value of this methodology for allowing researchers to answer a broader range of questions in climate change, environment and human health. Ecological Momentary Assessment Mental Health Well-being Environmental Exposure Temperature Self report Urban Health Figures Figure 1 Figure 2 Figure 3 Figure 4 Background Human beings are not separate from their environments; indeed, the environments in which we live, work, and move through can have important effects on our mental health and wellbeing. Mental health outcomes have been shown to be affected by ambient temperature ( 1 ), urban density ( 2 ), air pollution ( 3 ), green space exposure, and biodiversity ( 4 ), among other environmental factors. Many existing studies on the impact of our environments have focused on severe mental health outcomes, such as hospitalisations ( 5 ), suicide risk ( 6 ), and prevalence of mental health disorders ( 3 ). However, our environments can also affect transient mental health experiences, such as emotions, mood, and mental wellbeing ( 7 ). Smartphone-based ecological momentary assessments (EMAs) are a useful tool for understanding and tracking real-time changes in people’s mood and mental wellbeing as they navigate different environments. By capturing repeated snapshots of people’s psychological experiences over time and across different settings, EMA-based studies provide rich, intensive longitudinal data that can improve the ecological validity of psychological research ( 8 ). However, a careful balance needs to be struck between the frequency and length of assessments to maximise assessment completion rates and participant retention ( 8 ), as substantial bias can be introduced by missingness in such data ( 9 ). Achieving this balance means that EMA questionnaires alone may not be able to capture the level of detail on outcomes, exposures, and confounders required to characterise complex associations between one’s environment and outcomes such as mental wellbeing. There are several ways to enhance EMA data without increasing the burden on study participants. Occasional detailed questionnaires—for instance at study start or finish—can be used to obtain information on time-invariant factors such as sociodemographic characteristics. Wearable and portable technologies, such as smartphones, smartwatches, or portable monitors, can offer passive tracking of time-varying factors such as location or periods of activity. However, these technologies come with their share of limitations when it comes to tracking environmental exposures such as temperature, air pollution, or green space exposure: as such tracking is not routinely enabled in smartphones and other commonly used devices, studies considering environmental exposures will likely have to supply their own portable monitors, leading to high upfront cost and inconvenience to the user when required to carry or wear additional devices ( 10 ). Linkage with external environmental datasets therefore provides a unique opportunity to enhance EMA data with detailed information on objective environmental exposures. Smartphone-based EMA studies can include the option for participants to share their locations, which can then be linked via real-time geotagging of longitude and latitude to a wide range of high-resolution environmental datasets, allowing researchers to create a detailed, time-varying picture of the environments participants are exposed to. In this study, we illustrate the potential of combining subjective EMA data with objective environmental datasets to enhance current approaches in the field of climate change, environment and health. We link smartphone-based mental wellbeing assessments with data on ambient air temperature, air pollution levels, and green space exposure to (A) explore the association of heat exposure with real-time mental wellbeing and (B) investigate the potential effect modification of air pollution and green space exposure, hypothesising that high levels of pollution would strengthen the association and green space would weaken the association (Fig. 1 ). We highlight the strengths of this approach, as well as the challenges, and offer suggestions for how these challenges may be overcome. Methods Urban Mind Urban Mind is a smartphone-based tool that examines the impact of the urban environment on mental health ( http://www.urbanmind.info ). Launched in 2018, the app uses EMAs to capture participants’ perceptions of their surrounding built and social environments alongside their momentary mental wellbeing and activities. The study has been described in detail elsewhere ( 11 ). In brief, participants are initially invited to complete a baseline assessment. After the baseline assessment, participants are asked to complete three EMAs for the next 14 days, comprising a maximum of 42 EMAs. Each assessment is time-stamped, and its location is geotagged with corresponding longitude and latitude. For this study, we used data from study launch in 2018 to Dec 31, 2023, collected from assessments recorded in the UK. Assessments As part of their baseline assessment, participants report on demographics (eg, age, gender, ethnicity), socioeconomic status (eg, educational attainment and employment status), their living environment (eg, building type, living companions, feelings about the neighbourhood), and their self-reported mental health history. For each EMA, they are asked about their location, company, and current activities; their mental wellbeing; and their experiences across six key themes (sensory, nature, neighbourhood deprivation, safety, and inclusivity from a planning and social perspective). Further details and the questionnaires used can be found in previous publications ( 11 ). Mental wellbeing Mental wellbeing is measured in each assessment with an overall score, created from ten items of momentary mental wellbeing: five positive affect items (feeling confident, relaxed, happy, connected with other people, or energetic) and five negative affect items (feeling anxious, stressed, down, lonely, or tired). Participants are asked to indicate their level of agreement with statements of the form “Right now I feel…”, using a 5-point Likert scale ranging from “Strongly disagree” to “Strongly agree”. An overall momentary mental wellbeing score can then be calculated, giving a total score ranging from 10 (low mental wellbeing) to 50 (high mental wellbeing). Linkage with environmental data Geotagged smartphone-based assessments, such as those collected by Urban Mind, enable linkage with external environmental datasets. There is a wealth of publicly and privately available environmental data, including meteorological, air quality, and green space data, among many others. We used longitude and latitude to link each geotagged Urban Mind assessment to external data on air temperature, pollution levels, and green space exposure, as described in detail elsewhere ( 12 ). Temperature and extreme heat Daily maximum temperatures on a 1 km² grid were extracted from HadUK-Grid, which provides gridded climate observations on a for the UK ( 13 ). We took two approaches to modelling temperature: as a continuous variable, across the full range of temperatures, and as a binary variable identifying extreme heat events. We defined an extreme heat event as any day with a maximum daily temperature of at least 28°C, based on the higher of the UK Health Security Agency (UKHSA) decision-making aid temperature thresholds ( 14 ). Air pollution Air pollution was measured using a satellite-based spatiotemporal machine learning model, which estimates daily levels of PM 2.5 , PM 10 , and NO 2 at a spatial resolution of 1 km² ; further details on this model have been published elsewhere ( 15 ). For our analysis, we converted each pollutant measure into a binary variable, using the World Health Organization’s air quality guidelines interim target 4 as the threshold (10 µg/m³ for PM 2.5 and 20 µg/m³ for PM 10 and NO 2 ) ( 16 ). Green space We used the Normalised Difference Vegetation Index (NDVI) as a measure of green space exposure. 10 m resolution multispectral satellite images were obtained from the Sentinel 2 satellite programmes to extract the NDVI within a 300 m search radius of each participant’s location, according to the season of exposure, from which mean NDVI was calculated. We converted mean NDVI into a binary variable, where an NDVI value of 0.27 (the 75th percentile in our dataset) was used to indicate high levels of greenness. Missing data Intensive longitudinal studies, such as EMA studies, are particularly prone to missing data owing to the increased burden on participants ( 9 ). The format of EMA studies—comprised of short, frequent assessments—means that missingness is likely to present as a missed assessment rather than an incomplete one. Simultaneously, these studies offer a data structure that is ideal for exploring lagged effects of exposures, with multiple repeated measurements taken over a short period. Therefore, to fully harness the power of EMA data, the problem of missing data needs to be addressed. To make the most of available data while exploring lagged effects, we need to impute missing exposure data. For each study, it is worth considering whether available data on participants’ locations provides insights into their likely exposures during missing observations, particularly when exposure data can be collected independently from the EMA study. Within the Urban Mind dataset, nearly 90% of participants with high levels of completion completed the majority of their assessments within one local authority district (LAD)—50% of whom completed 80% of assessments within that same LAD. This suggests that historical location data can be a good tool for imputing missing data on environmental exposures. We therefore imputed environmental exposures for missing assessments by considering participants’ recorded location history. For each participant, we identified the LADs they had completed at least 10% of their assessments in. We then calculated average daily measures of environmental exposures for each LAD, and used these to create a weighted daily average of likely exposure for each participant based on the proportion of assessments they had completed in each LAD. We refer to this as our complete-case dataset. Missingness in EMA data can be further addressed by imputing not just exposures but outcomes. Multiple imputation has rapidly become a go-to tool for this task, as it can be used to impute outcome as well as covariate data ( 17 ). It can be applied to clustered data ( 18 ), such as repeated measures in longitudinal datasets, and reduces bias when compared with listwise deletion ( 9 ). We therefore imputed missing values of wellbeing score using multiple imputation with chained equations, accounting for clustering by participant ( 18 ). We imputed 50 datasets, with a maximum of ten iterations. We used these imputed datasets for all analyses, pooling estimates using Rubin’s rules, and repeated the main analysis with the complete-case dataset as a sensitivity analysis. Finally, as the proportion of missing data may influence the success of imputation methods, we restricted our main analyses to participants who had completed at least 21 (50%) assessments and conducted sensitivity analyses with participants who had completed at least 11 (25%) of assessments. Outcomes, exposures, and covariates Our outcome was mental wellbeing score and our main exposure was maximum daily temperature. Our effect modifiers were daily levels of PM 2.5 , PM 10 , and NO 2 and seasonal NDVI. We included a mix of static person-level and exogenous covariates as potential confounders. From the baseline assessment, we obtained age (continuous), gender (female, male, or other), ethnicity (White, Asian, or other), highest educational achievement (high school or less, apprenticeship, or university), occupational status (student, employed, self-employed, unemployed, or retired), baseline mental health status (very poor, poor, fair, good, or very good), and any history of mental health conditions. We extracted the following temporal covariates from each follow-up assessment: assessment time (morning, afternoon, or evening), day (Monday–Friday vs Saturday–Sunday), season, and year. Statistical analysis We used the case time series design for our main analysis: this self-matched method enables the analysis of transient health associations with time-varying exposures ( 19 ). A strength of this self-matched approach is that it eliminates confounding arising from constant between-person characteristics by using each participant as their own control. This design allows for exploration of non-linear exposure–lag–response relationships through the use of distributed lag non-linear models (DLNMs) ( 20 ). DLNMs combine the prediction and lag effect into a cross-basis, allowing us to estimate the change in wellbeing score across a range of temperatures and lag periods ( 20 ). Studies examining the impact of temperature on mental health service use have often considered lags of up to 7 days in their models ( 21 , 22 ); however, studies examining mental health symptoms and mood have rarely considered lagged effects ( 23 , 24 ). We chose a 3-day lag for our models (considering the effect on the day of exposure and the two following days), based on our hypothesis that the effect of temperature on mental wellbeing was likely to be acute. We tested this assumption in sensitivity analyses by increasing the lag period and comparing model fit with the Akaike information criterion. We fit a fixed-effect model with a Gaussian distribution, matching observations by participant. We modelled the exposure–response relationship with a natural spline with knots at the 10th, 50th, and 90th percentiles of maximum temperature, and the lag–response relationship with a natural spline with a knot at lag = 1. We fit this cross-basis within a generalised non-linear model and used median maximum daily temperature (16.7°C) as the reference value to predict changes to wellbeing score. To account for common baseline trends, we included a natural spline of time with 7 degrees of freedom per year, as used in previous studies ( 25 ). We further adjusted for assessment time, day, season and year. To explore the effect of extreme heat, we repeated this analysis, replacing the exposure–response function with a step function with a break at 28°C. We tested whether there was evidence of effect modification by levels of PM 2.5 , PM 10 , NO 2 , and mean seasonal NDVI. We refit the main model for each environmental exposure, first including the main effect of the exposure and then adding an interaction term between the temperature cross-basis and environmental exposure, checking whether the interaction term improved model fit with analysis of variance testing ( 26 ). Similarly, we explored whether there was evidence of effect modification by age (above or below median age), gender, and history of mental health conditions. We were interested in comparing the case time series approach with non-matched methods, which are widely used for EMA data ( 27 ). Therefore, as a sensitivity analysis, we repeated the main analyses using a generalised additive mixed model (GAMM) for maximum daily temperature and mixed-effects linear regression for extreme heat events. For the GAMM, we included three cubic spline smooths for temperature at lags 0–2, with knots at the 10th, 50th, and 90th percentiles. For the mixed-effects regression, we included an indicator variable for extreme heat at lags 0–2. In both models, we adjusted for age, gender, ethnicity, educational attainment, occupation, baseline mental health, history of mental health conditions, assessment time, day, season, and year, and included a random intercept for participant. All analyses were done with Stata/MP 18.0 and R (version 4.4.1), with R packages mice , dlnm, gnm , and mgcv . Results 266 participants with 7088 assessments were included in the analyses (Figure S1 ). The majority of participants were female and White, with a median age of 30 years (IQR 24–47; Table 1 ). A third of participants had a previous or current diagnosis of a mental health disorder (Table 1 ). Across their assessments, participants recorded a mean wellbeing score of 35 (SD 5) out of 50. 81% of assessments were recorded while participants were indoors. Participants were exposed to an average daily maximum temperature of 16.5°C, with 139 (2.0%) assessments recorded in days with maximum temperatures of 28°C or higher (Table S1 ). Analysis with the case time series model showed no significant association between maximum daily temperature exposure and changes in wellbeing score (Fig. 2 ). When modelling temperature as continuous, the overall cumulative exposure–response curve showed a dip in wellbeing score for temperatures between 5°C and 10°C, compared with the median temperature of 16.7°C, and a slight increase for higher temperatures; however, these changes were not significant (Fig. 2 B). Across the three lags, the greatest changes in wellbeing score were observed at very low and very high maximum temperatures: high maximum temperatures showed a negative effect on wellbeing score at lags 0 and 2, with a positive effect at lag 1 (Fig. 2 A). Similar results were observed when considering extreme heat as a binary variable (Fig. 2 C). The complete-case analysis presented with a similar cumulative exposure–response relationship, where narrower confidence intervals enabled a significant decrease in wellbeing score for temperatures down to 6°C and an increase in wellbeing score for temperatures up to 31°C, compared with the median temperature (Figure S2). Analysis on the 25% response dataset had little impact on our results (Figure S3). Modelling with longer lags did not improve model fit (data not shown). Table 1 Participant characteristics Participants (N = 266) Age, years 30 (24–47) <25 79 (29.7%) 25 to < 35 75 (28.2%) 35 to < 45 32 (12.0%) 45 to < 55 40 (15.0%) ≥55 40 (15.0%) Gender Male 71 (26.7%) Female 193 (72.6%) Other 2 (0.8%) Ethnicity White 198 (74.4%) Other 26 (9.8%) Asian 42 (15.8%) Highest educational attainment High school or less 30 (11.3%) Apprenticeship 20 (7.5%) University 216 (81.2%) Occupation Employed 145 (54.5%) Student 82 (30.8%) Self-employed 17 (6.4%) Retired 10 (3.8%) Unemployed 12 (4.5%) Previous or current mental health diagnosis 93 (35.0%) Baseline mental health Very poor 4 (1.5%) Poor 33 (12.4%) Fair 80 (30.1%) Good 111 (41.7%) Very good 38 (14.3%) Season most EMAs were recorded Spring 139 (52.3%) Summer 59 (22.2%) Autumn 35 (13.2%) Winter 33 (12.4%) Wellbeing score* 35 ( 5 ) Data are n (%), mean (SD), or median (IQR). EMA=ecological momentary assessment. *Wellbeing score ranges from 10 (low wellbeing) to 50 (high wellbeing). Table shows distribution of mean wellbeing score, adjusted for clustering by participant. GAMMs and mixed models produced similar results to the case time series models, although effect sizes at the ends of the temperature distribution were smaller for the GAMM (Fig. 3 ; Figure S4). By contrast with the case time series model, the GAMM predicted a small reduction in wellbeing score at lag 2 for low maximum temperatures; this reduction, however, was not significant and the shape of the lagged associations otherwise remained similar (Figure S4). We observed little evidence of effect modification by different levels of environmental exposures (Fig. 4 ; Figure S5). Similarly, we found little evidence of effect modification by age, gender, or pre-existing mental health diagnosis, with the largest difference seen by age group at high temperatures (Figure S6; p = 0.62). Discussion In this study, we demonstrate how smartphone-based EMAs can be successfully linked to external environmental datasets to better understand the effect of objective environmental factors on subjective participant-reported measures. Our example analysis explores the association of extreme heat and momentary mental wellbeing, finding little evidence of an effect: only small changes were recorded across a range of maximum daily temperatures, few of which were statistically significant. We additionally found no evidence that the effect of maximum daily temperature on mental wellbeing was modified by other environmental factors such as levels of air pollution and green space exposure, nor by age, gender, or history of mental health conditions. Previous research has shown negative effects of extreme heat on mental health ( 1 , 28 ); however, most studies to date have focused on hospital attendance for mental health problems, rather than mental wellbeing in the general population. Research on the impact of ambient temperature on mental wellbeing is sparse and often conflicting. While there is evidence that the thermal strain experienced during a 37°C heatwave can negatively impact mood ( 29 ), existing observational studies have largely been conducted at much lower ambient temperatures ( 23 , 24 , 30 ). Two cohort studies have found a decrease in low mood or depressive symptoms with increasing ambient temperature, among both the general population ( 23 ) and people with existing mental health disorders ( 23 , 24 ). However, a large cross-sectional study found that exposure to elevated ambient temperatures the previous day was associated with a reduction in wellbeing ( 30 ). Importantly, these studies have been conducted in countries with different temperature distributions (Switzerland and the USA), using a wide range of outcome measures and variable definitions of heat exposure (e.g. ranging from daily maximum to average temperature over the previous 2 weeks), making the findings difficult to compare directly. While we found no effect of extreme heat on mental wellbeing, results from sensitivity analyses suggested that maximum daily temperatures up to approximately 31°C were associated with small improvements in mental wellbeing when compared with a reference temperature of 17°C, supporting findings from cohort studies using EMA data ( 23 , 24 ); however, the functional form of our curve suggests that the relationship between maximum daily temperature and mental wellbeing may not be linear, as modelled in these studies. By using location-tagged EMAs, we were able to link granular environmental datasets to the mental wellbeing data, enabling exploration of the potential effect modification of air pollution levels and green space exposure. Chronic exposure to high levels of air pollution has been found to increase odds of mental health problems ( 3 ), with some evidence that short-term peaks in pollutants may also have an acute impact ( 31 ). Air pollution has also been found to intensify the negative effects of extreme heat on mental health-related emergency department visits ( 32 ), but little is known about its impact on heat with regards to mood and mental wellbeing. Our findings suggest that daily levels of air pollution do not modify the relationship between maximum daily temperature and mental wellbeing; however, future studies should consider whether chronic exposure to high levels of pollution affects the relationship between extreme heat and mental wellbeing. We also found no evidence of effect modification by green space exposure, in contrast with a previous study suggesting that tree canopy mitigated the negative effects of extreme heat on mental health outcomes ( 32 ). There are several mechanisms by which green space exposure may support mental health during extreme heat events, including cooling of ambient air temperatures and provision of shade ( 33 ). Our measure of green space exposure was mean NDVI over a 300 m radius, which will also have captured private parks and gardens not accessible to study participants, it is therefore possible that this measure was not sufficiently granular to reflect the acute impacts of green space on ambient temperatures. We must also note that the majority of the EMAs in this study were completed indoors, where the cooling effect of green spaces are unlikely to have been acutely felt. The evidence around how age, gender, and mental health diagnoses may affect the relationship between extreme heat and mental wellbeing is unclear. Our finding of no effect modification by these factors supports a previous study in adults with depression and bipolar disorder ( 24 ). However, a population-based study with a wider range of mental health diagnoses found that increasing ambient temperatures affected participants with bipolar disorder differently than those with anxiety disorders or schizophrenia ( 23 ). While Urban Mind collects information on specific mental health diagnoses, our sample size precluded subgroup analyses by diagnosis. Further research is therefore needed to assess whether the relationship between extreme heat and mood—as well as other mental health outcomes—is modified by different mental health disorders, to better inform public health efforts. This study has successfully shown that subjective smartphone based EMAs, can be enhanced with geotagging and environmental data linkage, allowing for a rich and granular analysis of the person-environment interplay. This approach captures the benefits of real-time real-life participant engagement while preserving objectivity for variables that are unsuitable for self-report ( 34 ). This study has several strengths. We used EMAs, which enable longitudinal, real-time assessment of mood, thus avoiding recall bias. Sophisticated, self-matched analytic methods allowed us to consider lagged effects of heat exposure while robustly adjusting for time-invariant confounders. Linkage of multiple, detailed environmental datasets to EMA data enabled exploration of effect modification of the heat–wellbeing relationship by air pollution and green space exposure, both of which have been found to affect mental health outcomes ( 32 , 35 ). Use of the UKHSA definition of extreme heat exposure and the WHO air quality thresholds allows us to interpret our results within the context of existing policy. Among the limitations of this study is one shared with many EMA studies: namely, a large proportion of missing data ( 9 ). However, we demonstrated the use of various imputation methods to reduce the potential bias of missing data ( 9 ) and to make the most of the data available. As Urban Mind enrolment is not specific to the summer, our study was likely underpowered to assess extreme heat events, with only 2% of assessments recorded during days with temperatures of at least 28°C. Additionally, more than 80% of assessments were indoors, making it difficult to understand the level of exposure participants had to the environmental factors under consideration. Finally, the self-selected Urban Mind cohort is predominantly female, White, young, and with high levels of educational attainment, limiting the generalisability of our findings to the general population. Conclusions In conclusion, this study found little evidence that extreme heat affects momentary mental wellbeing in a moderate UK climate. Additionally, this relationship was not modified by environmental (air pollution and green space exposure) or sociodemographic (age, gender, and history of mental health diagnosis) factors. The study demonstrates how subjective, smartphone-based measures can be successfully integrated with objective, environmental datasets to enable detailed exploration of how various elements of our environment interplay to affect our mental wellbeing. The approach outlined here can serve as a blueprint to make the most of rich EMA data to understand the mental health impacts of our environment. Abbreviations EMA Ecological Momentary Assessment DLNM distributed lag non-linear models UK United Kingdom UKHSA UK Health Security Agency PM 2.5 Particulate Matter < 2.5 micrometers PM 10 Particulate Matter < 10 micrometers NO 2 Nitrogen dioxide NDVI Normalised Difference Vegetation Index LAD Local Authority District GAMM Generalised Additive Mixed Model IQR Interquartile Range SD Standard Deviation USA United States of America WHO World Health Organization Declarations Clinical trial numbe r: not applicable Ethics approval and consent to participate Participants self-selected into the study by downloading and installing the Urban Mind app where they were presented with a description of the project’s aim and methodology and provided informed consent. The research project had received full ethical approval from the Psychiatry, Nursing, and Midwifery Research Ethics Sub-Committee at King's College London (no. PNM/14/15–59). Consent for publication Not applicable Funding This study was supported by a Climate Impacts Award from Wellcome (grant ref 228033/Z/23/Z). Author Contribution GV (methodology, data curation, data analysis, writing – original draft, writing – review and editing). EC (methodology, writing – review and editing). ADLC (methodology, data curation, writing – review and editing). AG (methodology, writing – review and editing). YL (methodology, data curation, writing – review and editing). AG (methodology, writing – review and editing). RH (methodology, data curation, writing – review and editing). Lisa Mandle (methodology, writing – review and editing). RT (writing- review and editing). MS (conceptualisation, writing – review and editing). AM (conceptualisation, funding acquisition, methodology, writing – review and editing). IB (conceptualisation, methodology, writing – review and editing). Acknowledgement We are grateful to all Urban Mind study participants for their contributions to the study. Data Availability The datasets analysed during the current study are available from the corresponding author on reasonable request References Thompson R, Lawrance EL, Roberts LF, Grailey K, Ashrafian H, Maheswaran H, et al. Ambient temperature and mental health: a systematic review and meta-analysis. Lancet Planet Health. 2023;7(7):e580–9. Zijlema W, Cerin E, Cirach M, Bartoll X, Borrell C, Dadvand P, et al. Cities and mental health: The role of the built environment, and environmental and lifestyle factors in Barcelona. Environ Pollut. 2024;346:123559. Bakolis I, Hammoud R, Stewart R, Beevers S, Dajnak D, MacCrimmon S, et al. Mental health consequences of urban air pollution: prospective population-based longitudinal survey. Soc Psychiatry Psychiatr Epidemiol. 2021;56(9):1587–99. Methorst J, Bonn A, Marselle M, Böhning-Gaese K, Rehdanz K. Species richness is positively related to mental health – A study for Germany. Landsc Urban Plann. 2021;211:104084. Hu T, Xu ZY, Wang J, Su Y, Guo BB. Meteorological factors, ambient air pollution, and daily hospital admissions for depressive disorder in Harbin: A time-series study. World J Psychiatry. 2023;13(12):1061–78. Bär S, Bundo M, de Schrijver E, Müller TJ, Vicedo-Cabrera AM. Suicides and ambient temperature in Switzerland: A nationwide time-series analysis. Swiss Med Wkly. 2022;152:w30115. Shin J-h, Dennis S, Mohammed H. Mental Health Outcome Measures in Environmental Design Research: A Critical Review. HERD: Health Environ Res Des J. 2021;14(3):331–57. Fritz J, Piccirillo ML, Cohen ZD, Frumkin M, Kirtley O, Moeller J, et al. So You Want to Do ESM? 10 Essential Topics for Implementing the Experience-Sampling Method. Adv Methods Practices Psychol Sci. 2024;7(3):25152459241267912. Ji L, Chow SM, Schermerhorn AC, Jacobson NC, Cummings EM. Handling Missing Data in the Modeling of Intensive Longitudinal Data. Struct Equ Model. 2018;25(5):715–36. Bernasconi S, Angelucci A, Aliverti A. A Scoping Review on Wearable Devices for Environmental Monitoring and Their Application for Health and Wellness. Sens (Basel). 2022;22(16). Bakolis I, Hammoud R, Smythe M, Gibbons J, Davidson N, Tognin S, et al. Urban Mind: Using Smartphone Technologies to Investigate the Impact of Nature on Mental Well-Being in Real Time. Bioscience. 2018;68(2):134–45. Chukwusa E, Vivaldi G, Baecker L, Bowerman E, Bridge N, Davidson N, et al. A longitudinal, population-based, record-linked natural experiment on the effects of extreme heat events on mental health in urban communities: a study protocol. BMJ Open. 2025;15(10):e099771. Hollis D, McCarthy M, Kendon M, Legg T, Simpson I. HadUK-Grid—A new UK dataset of gridded climate observations. Geosci Data J. 2019;6(2):151–9. UK Health Security Agency. User guide: Weather-health alerting system. London: UK Health Security Agency; 2024. Report No. de la Cruz Libardi A, Masselot P, Schneider R, Nightingale E, Milojevic A, Vanoli J, et al. High resolution mapping of nitrogen dioxide and particulate matter in Great Britain (2003–2021) with multi-stage data reconstruction and ensemble machine learning methods. Atmospheric Pollution Res. 2024;15(11):102284. World Health Organization. WHO global air quality guidelines. 2021. van Ginkel JR, Linting M, Rippe RCA, van der Voort A. Rebutting Existing Misconceptions About Multiple Imputation as a Method for Handling Missing Data. J Pers Assess. 2020;102(3):297–308. Ji L, Li Y, Potter LN, Lam CY, Nahum-Shani I, Wetter DW, et al. Multiple imputation of missing data in multilevel ecological momentary assessments: an example using smoking cessation study data. Front Digit Health. 2023;5:1099517. Gasparrini A. The Case Time Series Design. Epidemiology. 2021;32(6):829–37. Gasparrini A. Modeling exposure-lag-response associations with distributed lag non-linear models. Stat Med. 2014;33(5):881–99. Bundo M, de Schrijver E, Federspiel A, Toreti A, Xoplaki E, Luterbacher J, et al. Ambient temperature and mental health hospitalizations in Bern, Switzerland: A 45-year time-series study. PLoS ONE. 2021;16(10):e0258302. Cohen G, Rowland ST, Benavides J, Lindert J, Kioumourtzoglou M-A, Parks RM. Daily temperature variability and mental health-related hospital visits in New York State. Environ Res. 2024;257:119238. Bundo M, Preisig M, Merikangas K, Glaus J, Vaucher J, Waeber G, et al. How ambient temperature affects mood: an ecological momentary assessment study in Switzerland. Environ Health. 2023;22(1):52. Clery P, Hayes JF, Launders N, Thompson R, Kandola A, Osborn DPJ, et al. The association between outdoor ambient temperature and depression and mania: An ecological momentary assessment study. J Affect Disord. 2025;379:457–66. Peng Z, Wang Q, Kan H, Chen R, Wang W. Effects of ambient temperature on daily hospital admissions for mental disorders in Shanghai, China: A time-series analysis. Sci Total Environ. 2017;590–591:281–6. Faraway JJ. Practical regression and ANOVA using R. University of Bath Bath; 2002. Oleson JJ, Jones MA, Jorgensen EJ, Wu YH. Statistical Considerations for Analyzing Ecological Momentary Assessment Data. J Speech Lang Hear Res. 2022;65(1):344–60. Liu J, Varghese BM, Hansen A, Xiang J, Zhang Y, Dear K, et al. Is there an association between hot weather and poor mental health outcomes? A systematic review and meta-analysis. Environ Int. 2021;153:106533. McGarr GW, Meade RD, Kenny GP. Indoor overheating influences self-reported symptoms and mood-state in older adults during a simulated heatwave: Effects of mid-day cooling centre use. Physiol Behav. 2023;271:114335. Noelke C, McGovern M, Corsi DJ, Jimenez MP, Stern A, Wing IS, et al. Increasing ambient temperature reduces emotional well-being. Environ Res. 2016;151:124–9. Pignon B, Borel C, Lajnef M, Richard JR, Szöke A, Hemery F, et al. PM(2.5) and PM(10) air pollution peaks are associated with emergency department visits for psychotic and mood disorders. Environ Sci Pollut Res Int. 2022;29(59):88577–86. Lavigne E, Maltby A, Côté J-N, Weinberger KR, Hebbern C, Vicedo-Cabrera AM, et al. The effect modification of extreme temperatures on mental and behavior disorders by environmental factors and individual-level characteristics in Canada. Environ Res. 2023;219:114999. Yang Y, Zhang Y, Sheng S. Residential greenness for mitigating impacts of extreme heat events on depression and supporting mental health. Front Public Health. 2023;11:1310410. Kosmala M, Wiggins A, Swanson A, Simmons B. Assessing data quality in citizen science. Front Ecol Environ. 2016;14(10):551–60. Xu J, Liu N, Polemiti E, Garcia-Mondragon L, Tang J, Liu X, et al. Effects of urban living environments on mental health in adults. Nat Med. 2023;29(6):1456–67. Additional Declarations No competing interests reported. Supplementary Files UrbanMindTemperaturepaperBMCPHFEB2026Supplementarymaterial.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 15 Apr, 2026 Reviewers agreed at journal 02 Apr, 2026 Reviewers agreed at journal 19 Mar, 2026 Reviewers invited by journal 18 Mar, 2026 Editor invited by journal 22 Feb, 2026 Editor assigned by journal 19 Feb, 2026 Submission checks completed at journal 19 Feb, 2026 First submitted to journal 18 Feb, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information

Related documents

• The effect of temperature on physical activity: an aggregated timeseries analysis of smartphone users in five major Chinese cities#178276
• Physiological, Cognitive, and Behavioral Effects Related to Thermal Exposure#678128
• Temperature exposure and time adolescents spend in physical activity across intensity levels#273398
• Perceived heat threshold and transit use in a humid subtropical climate#818532
• Physical Activity Dependence on Relative Temperature and Humidity Characteristics in a Young, Insufficiently Active Population: A Weather Typing Analysis.#823637