Introduction
Mental health problems among young people are a serious public health concern in the United States. One in six US children and adolescents between the ages of 6 and 17 years experiences a mental health disorder such as depression or anxiety problems (Whitney and Peterson, Reference Whitney and Peterson2019). An estimated 30% of young adults aged 18 to 25 years meet the diagnostic criteria for a mental health disorder (Substance Abuse and Mental Health Services Administration, 2021). The mental health crisis in young people continues to grow in recent years. Adolescents and young adults experiencing serious mental health problems have increased by over 50% from the mid-2000s to 2017, and mental health–related emergency department (ED) visits by adolescents and young adults have increased by 56% from 2009 to 2015 (Santillanes et al., Reference Santillanes, Axeen, Lam and Menchine2020; Twenge et al., Reference Twenge, Cooper, Joiner, Duffy and Binau2019). The burden of mental health disorders on the healthcare system is significant, with 9%–12% of ED visits and nearly 10% of inpatient stays by young people were for a primary mental health diagnosis (Bardach et al., Reference Bardach, Coker, Zima, Murphy, Knapp, Richardson, Edwall and Mangione-Smith2014; Santillanes et al., Reference Santillanes, Axeen, Lam and Menchine2020).
A growing body of literature demonstrates links between ambient heat and acute mental health problems. Research shows that hospital and ED encounters for mental and behavioural disorders tend to increase during hot weather (Basu et al., Reference Basu, Gavin, Pearson, Ebisu and Malig2018; Hansen et al., Reference Hansen, Bi, Nitschke, Ryan, Pisaniello and Tucker2008; Nori-Sarma et al., Reference Nori-Sarma, Sun, Sun, Spangler, Oblath, Galea, Gradus and Wellenius2022; Sherbakov et al., Reference Sherbakov, Malig, Guirguis, Gershunov and Basu2018). For example, Mullins and White (Reference Mullins and White2019) reported a positive association of ambient temperature with multiple mental health outcomes among US individuals of all ages, including mental health–related ED visits, suicides and self-reported days of poor mental health. In a California study, Basu et al. (Reference Basu, Gavin, Pearson, Ebisu and Malig2018) found an increase in ED visits for mental health disorders associated with high temperatures during the warm season. Nori-Sarma et al. Reference Nori-Sarma, Sun, Sun, Spangler, Oblath, Galea, Gradus and Wellenius(2022) showed an association between heat and ED visits for mental health diagnoses for US adults in a national study. Positive associations between temperature and mental health–related healthcare encounters were reported in countries from different temperature zones, including Sweden, Australia, Canada, Vietnam, China and Korea (Carlsen et al., Reference Carlsen, Oudin, Steingrimsson and Oudin Åström2019; Chan et al., Reference Chan, Lam, So, Goggins, Ho, Liu and Chung2018; Hansen et al., Reference Hansen, Bi, Nitschke, Ryan, Pisaniello and Tucker2008; Lee et al., Reference Lee, Lee, Myung, Kim and Kim2018; Min et al., Reference Min, Shi, Ye, Wang, Yao, Tian, Zhang, Liang, Qu and Bi2019; Peng et al., Reference Peng, Wang, Kan, Chen and Wang2017; Trang et al., Reference Trang, Rocklöv, Giang, Kullgren and Nilsson2016; Wang et al., Reference Wang, Lavigne, Ouellette-kuntz and Chen2014). Previous research has also shown an association between higher ambient temperature and an increased risk of suicidal behaviour in multiple countries (Burke et al., Reference Burke, González, Baylis, Heft-Neal, Baysan, Basu and Hsiang2018; Kim et al., Reference Kim, Kim, Gasparrini, Armstrong, Honda, Chung, Ng, Tobias, Íñiguez and Lavigne2019; Mullins and White, Reference Mullins and White2019; Sugg et al., Reference Sugg, Dixon and Runkle2019). There are several potential pathways by which heat can impact mental health illness. Exposure to ambient heat can exacerbate existing mental health conditions through altering thermosensitive physiological processes and causing negative mental states including distress and exhaustion (Noelke et al., Reference Noelke, McGovern, Corsi, Jimenez, Stern, Wing and Berkman2016). Heat can also impact mental health through disrupted sleep, including difficulty falling asleep, poor sleep quality and fewer sleep hours (Mullins and White, Reference Mullins and White2019; Obradovich et al., Reference Obradovich, Migliorini, Mednick and Fowler2017). Heat exposure has also been shown to increase irritability, aggression and violence, which are associated with adverse mental health outcomes (Anderson, Reference Anderson2001).
The existing literature on links between ambient temperature and mental health has focused on the adult population; however, little is known about heat impacts on mental health across childhood, adolescence and young adulthood, which are key developmental periods with special risk for mental health disorders (Qiu et al., Reference Qiu, Danesh-Yazdi, Wei, Di, Just, Zanobetti, Weisskopf, Dominici and Schwartz2022). Children and adolescents are less able to regulate their emotions and behaviours, are more prone to impulsive behaviours and are less likely to take adaptive measures such as avoidance of high-intensity physical activities compared to adults, which can increase the risk for mental health illness during periods of high temperatures (Santillanes et al., Reference Santillanes, Axeen, Lam and Menchine2020). Young adults experience high emotional burden from social, school and work responsibilities, and many face economic and social barriers that further contribute to mental health challenges (Rahal et al., Reference Rahal, Chiang, Fales, Fuligni, Haselton, Slavich and Robles2020). Only a few studies have examined associations between ambient temperature and acute mental health outcomes in young people, and those few have reported mixed results (Basu et al., Reference Basu, Gavin, Pearson, Ebisu and Malig2018; Chan et al., Reference Chan, Lam, So, Goggins, Ho, Liu and Chung2018; Trang et al., Reference Trang, Rocklöv, Giang, Kullgren and Nilsson2016; Wang et al., Reference Wang, Lavigne, Ouellette-kuntz and Chen2014).
This study examines associations between high ambient temperature and risk of mental health–related healthcare encounters (i.e., ED visits and hospitalization) among young people aged 6 to 25 years living in New York City (NYC). We hypothesized that all three age groups (children, 6–11 years; adolescents, 12–17 years young adults, 18–25 years) would be at increased risk for heat impact on mental health. We further identify subgroups of young people who are particularly vulnerable to heat by stratifying by sex, race/ethnicity, payment source and mental health diagnostic groups. Based on previous work that heat increases emotional distress and induces impulsiveness and aggression (Anderson, Reference Anderson1989; Kim et al., Reference Kim, Kim, Honda, Guo, Chen, Woo and Ebi2016), we expected that emotional (e.g., depression and anxiety) and behavioural disorders (e.g., externalizing) would be more strongly impacted by heat than other mental health categories. Given that children and adolescents from racially marginalized groups and low socio-economic positions are disproportionately impacted by heat and often lack resources (e.g., air conditioning) to cope with heat (Gronlund, Reference Gronlund2014; Renteria et al., Reference Renteria, Grineski, Collins, Flores and Trego2022), we expected that racial/ethnic minority groups and those paying with Medicaid (indicating low income) would be more vulnerable to high temperature in terms of mental health risk. Due to mixed findings related to sex (Basu et al., Reference Basu, Gavin, Pearson, Ebisu and Malig2018; Kim et al., Reference Kim, Kim, Gasparrini, Armstrong, Honda, Chung, Ng, Tobias, Íñiguez and Lavigne2019; Nori-Sarma et al., Reference Nori-Sarma, Sun, Sun, Spangler, Oblath, Galea, Gradus and Wellenius2022), no specific hypothesis was made for how sex would moderate the association between temperature and mental health risk.
Methods
Health data and study population
Data on all daily ED visits and hospitalizations in NYC for the period from 2005 to 2011 were extracted from the New York Statewide Planning and Research Cooperative System (SPARCS). SPARCS is a comprehensive all-payer administrative claims database for inpatient stays and outpatient (ambulatory surgery, ED and outpatient services) visits in New York State (New York State Department of Health, 2015). It includes patient-level information on demographic characteristics, diagnoses and treatments, payment sources and charges for each hospital discharge record. We restricted our analysis to ED visits and hospitalizations that occurred during the warm-season months of June, July and August between 2005 and 2011, as we sought to study the heat effects on mental health–related encounters (Sheffield et al., Reference Sheffield, Herrera, Kinnee and Clougherty2018). Because of the relative acute nature of the hypothesized heat effect on mental health, we excluded scheduled inpatient visits (e.g., elective surgery) as well as inpatient and outpatient visits that resulted from circumstances other than a disease or injury (e.g., routine child health check). We focused on ED visits and hospitalizations for young people between the ages of 6 and 25 years to capture developmental stages when mental health disorders emerge, increase and peak (Bitsko, Reference Bitsko2022).
Health outcomes
Mental health–related ED and hospital encounters were identified using each patient’s primary diagnosis coded with the International Classification of Diseases, Ninth Revision (ICD-9). The primary outcome of interest is the presence of any primary diagnosis indicative of mental health disorders (ICD-9 codes 290–299), following several prior studies studying acute weather impacts on mental health encounters (Basu et al., Reference Basu, Gavin, Pearson, Ebisu and Malig2018; Chan et al., Reference Chan, Lam, So, Goggins, Ho, Liu and Chung2018; Sherbakov et al., Reference Sherbakov, Malig, Guirguis, Gershunov and Basu2018). We defined our outcome as those with mental health disorders as the primary diagnosis and excluded cases with another cause as the primary diagnosis and a mental health diagnosis as the secondary diagnosis (e.g., patients with existing mental health disorders who become ill with other causes during hot weather). We grouped the diagnoses into 17 subcategories of mental health diagnosis using the definitions developed for the paediatric patient population by Bardach et al. Reference Bardach, Coker, Zima, Murphy, Knapp, Richardson, Edwall and Mangione-Smith(2014): anxiety disorders, bipolar disorder, depression, externalizing disorders (e.g., conduct and oppositional defiant disorders), psychosis, reaction disorders (i.e., an unhealthy or excessive emotional or behavioural response to a stressful event or change in a person’s life), substance use disorders, attention–deficit/hyperactivity disorder (ADHD), attachment, autism, developmental, eating, elimination, motor, personality, sexuality and other mental health disorders. We also coded a separate category for suicide and self-inflicted injury if an encounter had an associated external cause of injury code identified as self-inflicted injury/suicide (ICD-9 codes E95). The ICD-9 codes used for the identification of mental health diagnostic categories were included as Supplementary Table S1.
We also extracted from SPARCS information about patients’ race/ethnicity (mutually exclusive categories of Hispanic, non-Hispanic Black, non-Hispanic White or non-Hispanic other) and payment sources based on the primary source of payment (coded as mutually exclusive categories of commercial, Medicaid, self-pay or other payment sources including worker’s compensation, Medicare, other federal programme, blue cross, CHAMPUS (a Department of Defense health care program now called TRICARE) and other non-federal programme).
Climate exposure and meteorological data
We obtained data on daily temperatures from the four NYC-area meteorological stations (Central Park, JFK International Airport, LaGuardia Airport and Newark International Airport) from the National Oceanic and Atmospheric Administration (NOAA) National Climate Data Center (NCDC) (NOAA National Centers for Environmental Information, 2020). The data from the four stations were averaged to create a citywide daily trend. Because disruption of sleep is a critical mechanism through which temperature may affect mental health episodes (Obradovich et al., Reference Obradovich, Migliorini, Mednick and Fowler2017), we used the daily minimum temperature (T min) – a strong indicator of nighttime temperature – as our main exposure variable (Barnett et al., Reference Barnett, Tong and Clements2010; Sheffield et al., Reference Sheffield, Herrera, Kinnee and Clougherty2018; Winquist et al., Reference Winquist, Grundstein, Chang, Hess and Sarnat2016). We found T min to be highly correlated with the daily maximum temperature (r = 0.96) and mean temperature (r = 0.99). We calculated relative humidity (RH) from mean temperature and dew point temperature following the standard NOAA equation using the R package weathermetrics (Anderson et al., Reference Anderson, Peng and Ferreri2016).
Statistical analysis
We used a case-crossover approach to study associations between daily T min and mental health–related ED and hospital visits (Maclure, Reference Maclure1991). For each subject who had an ED or hospital visit, we used the date of the visit as the case day and then matched control days to each case day based on day of the week, month and year, resulting in 3 or 4 control days per case day (Janes et al., Reference Janes, Sheppard and Lumley2005). Because cases serve as their own controls, this approach adjusts for individual time-independent factors (Maclure, Reference Maclure1991). By matching the case day to all of the same days of week within the same month, it also controls for long-term time trends, seasonality and day of week (Schinasi et al., Reference Schinasi, Bloch, Melly, Zhao, Moore and De Roos2020).
We then performed conditional logistic regression to estimate the effects of temperature on the risk of mental health–related ED visits and hospitalizations (Buteau et al., Reference Buteau, Goldberg, Burnett, Gasparrini, Valois, Brophy, Crouse and Hatzopoulou2018). We modelled the temperature variable using the distributed lag non-linear model (DLNM) function, a modelling framework that allows, simultaneously, for non-linear and lagged effects in exposure–outcome associations. We modelled the effect of T min over lag days 0 to 5. We selected a lag period of 6 days based on prior research showing that the heat effects on mental health are relatively acute, with the most adverse effects of heat apparent within several days (Lee et al., Reference Lee, Lee, Myung, Kim and Kim2018; Min et al., Reference Min, Shi, Ye, Wang, Yao, Tian, Zhang, Liang, Qu and Bi2019; Wang et al., Reference Wang, Lavigne, Ouellette-kuntz and Chen2014). In keeping with previous practice, we modelled T min as a natural cubic spline of 3 degrees of freedom (df) with integer lag functions. Exploratory analyses using an alternative 5 df produced a similar pattern in exposure–response associations (Supplementary Appendix A). As such, we chose 3 df because it was parsimonious and provided a reliable fit to the data.
For all models, we estimated the cumulative odds ratios (ORs) over lags 0 to 5 for a temperature increment from the minimum risk temperature (MRT) to an elevated temperature, defined as the 95th percentile of distribution of the warm-season (June–August) temperature from 2005 to 2011. The MRT were defined as the T min at which the least mental health–related encounters were observed. We used it as the reference point for calculating the OR to allow flexibility in interpreting heat impacts for different age groups (e.g., the optimal temperature for children may be different in adolescents or young adults).
We examined the association between T min and mental health–related disorders separately for three age groups: 6–11 years (children), 12–17 years (adolescence) and 18–25 years (young adults). These stratified analyses were conducted because childhood, adolescence and young adulthood represent development stages with distinct healthcare utilization patterns, contributing factors and underlying mechanisms associated with mental health and heat vulnerability. We combined mental health–related ED visits and hospital encounters in the main analysis because an investigation of ED and hospitalizations as separate outcomes yielded similar shapes of temperature–outcome associations in each age group (Supplementary Appendix B).
To identify subgroups of individuals at heightened risk for mental health episodes associated with heat, we examined temperature–outcome associations by sex, race/ethnicity, payment source and mental health diagnostic categories separately for children, adolescents and young adults. Subgroups with small sample sizes (i.e., less than 500) were not examined due to insufficient power.
As sensitivity analyses, we tested whether including RH as a covariate would change the temperature–outcome associations. We also fitted parallel models limiting to cases during July and August to examine the possible influence of the school year (e.g., the public school is closed from June to late August).
All statistical analyses were conducted using the R statistical software version 4.0.2 (R Core Team, 2015) with the ‘dlnm’ and ‘coxph’ packages (Gasparrini, Reference Gasparrini2011). Institutional review boards at the authors’ institutions approved this study.
Results
As shown in Table 1, during the period from June to August between 2005 and 2011 in NYC, there were 82,982 mental health–related healthcare encounters (59,731 ED visits and 23,251 hospitalizations) among children (n = 6,268), adolescents (n = 22,206) and young adults (n = 54,508). Females comprised 44% of the healthcare encounters (32% for 6–11 years, 53% for 12–17 years and 41% for 18–25 years). Non-Hispanic Black (35%) individuals represented the largest group, followed by Hispanic (24%), non-Hispanic other (23%) and non-Hispanic White individuals (18%). Nearly a third of the encounters were paid by commercial insurance (31%); Medicaid accounted for 25% and self-pay accounted for 19% of the encounters. The most common mental health diagnoses were externalizing disorders for children aged 6 to 11 years (27%), depression for adolescents aged 12 to 17 years (21%) and substance use disorders for young adults aged 18 to 25 years (27%). Descriptive statistics of mental health–related encounters, separate for hospitals and EDs, are provided in Supplementary Table S2. The range of T min over the study period was 50–83°F, with a mean of 67.9°F and a standard deviation of 5.4°F (see details of exposure data in Supplementary Table S3).
Table 1. Descriptive statistics of mental health–related ED and hospital admissions in NYC during the warm season (June–August), 2005–2011
Notes: Externalizing disorders include conduct and oppositional defiant disorders. Other mental health diagnosis category includes ADHD, attachment, autism, developmental, eating disorders, elimination, others, motor, personality, sexuality and other mental health disorders. Other payment sources include worker’s compensation, Medicare, blue cross, CHAMPUS, other federal programme and other non-federal programme.
Associations between temperature and mental health–related healthcare encounters
The shape of the exposure–outcome associations differed by age groups (Figure 1). Among children, we observed a ‘U’-shaped relationship between daily minimum temperature and risk of mental health–related encounters, with the lowest risk at 69.2°F (i.e., MRT). Compared to the MRT, an elevated temperature (76°F, 95th percentile) was associated with higher odds of mental health events (OR: 1.28, 95% confidence interval [CI]: 1.13–1.46) accumulative over lag 0 to 5 days. Among adolescents, we observed a similar, yet attenuated, ‘U’ shape, with a slightly cooler MRT (65.2°F); the OR comparing an elevated temperature (76°F, 95th percentile) to MRT was 1.17 (95% CI: 1.09–1.25). For young adults, the exposure–outcome curve was almost flat, with a small increase in risk of mental health event from around 65°F to 72°F, which then levelled off with increasing temperatures. The OR comparing an elevated temperature (76°F, 95th percentile) with the MRT (58.7°F) was 1.09 (95% CI: 1.04–1.15).
Figure 1. Cumulative odds ratio and 95% confidence intervals of mental health–related ED and hospital admissions associated with daily minimum temperature over lag 0–5 days, June to August.
Stratification by sex, race/ethnicity, payment sources and mental health diagnostic categories
In children, we identified several subgroups at significant risk for mental health encounters associated with an elevated temperature (95th percentile compared to the MRT). They included children from non-Hispanic Black (OR: 1.43, 95% CI: 1.12–1.83) or other racial/ethnic groups (OR: 1.60, 95% CI: 1.12–2.27) and those paid with Medicaid (OR: 1.48, 95% CI: 1.09–2.02). Both males (OR: 1.28, 95% CI: 1.09–1.52) and females (OR: 1.34, 95% CI: 1.06–1.71) were at risk for mental health encounters at an elevated temperature. Among all mental health diagnostic groups examined, the strongest associations were found among children diagnosed with reaction disorders (OR: 2.21, 95% CI: 1.22–3.99) (Table 2).
Table 2. Cumulative ORs and 95% CIs of mental health–related ED and hospital admissions associated with elevated minimum temperature (95th percentile) relative to the MRT, overall and by sex, race, payment source and mental health diagnosis subgroups
We did not present results for small subgroups (less than 500) due to large margins of error. Italic estimates indicate significant associations with a P-value greater than 0.05
In adolescents, we identified subgroups at significant risk for mental health encounters associated with an elevated temperature (95th percentile) compared to the MRT, including adolescents from non-Hispanic Black (OR: 1.21, 95% CI: 1.06–1.39) or other racial/ethnic groups (OR: 1.26, 95% CI: 1.07–1.49) and those who paid with commercial insurance (OR: 1.25, 95% CI: 1.08–1.44). Both male (OR: 1.11, 95% CI: 1.00–1.22) and female adolescents (OR: 1.26, 95% CI: 1.14–1.39) were at risk for mental health encounters at an elevated temperature. Adolescents diagnosed with anxiety (OR: 1.42, 95% CI: 1.08–1.87) or bipolar disorders (OR: 1.39, 95% CI: 1.06–1.83) showed significant vulnerability to elevated temperature, among all diagnostic categories (Table 2).
For young adults, elevated temperatures (95th percentile compared to the MRT) conferred significant risk for mental health encounters among several subgroups, including young adults who were male (OR: 1.07, 95% CI: 1.00–1.14), those from other racial/ethnic groups (OR: 1.12, 95% CI: 1.01–1.25), those who paid with Medicaid (OR: 1.20, 95% CI: 1.08–1.33) or self-paid (OR: 1.33, 95% CI: 1.06–1.67), as well as those diagnosed with psychosis (OR: 1.14, 95% CI: 1.03–1.27) or reaction disorders (OR: 1.32, 95% CI: 1.04–1.67) (Table 2).
The shape of exposure–outcome associations for all subgroup analyses is shown in Supplementary Figures S3–S6. Within each of the three age groups, we performed heterogeneity test to assess significant differences in ORs by sex, race, payment source and mental health diagnosis group (Kaufman and MacLehose, Reference Kaufman and MacLehose2013). We did not find statistical evidence for heterogeneous effect by these factors, as indicated by the Cochran’s Q test statistic (all P > 0.05, Supplementary Table S8).
Sensitivity analysis
We found that temperature–outcome associations were robust to adjustment for RH (Supplementary Appendix C), with very similar results for all age groups (children, OR: 1.28, 95% CI: 1.13–1.46; adolescents, OR: 1.17, 95% CI: 1.09–1.26; young adults, OR: 1.09, 95% CI: 1.04–1.15). Restricting analyses to only July and August produced stronger effects of elevated temperatures on mental health–related encounters (children, OR: 1.44, 95% CI: 1.20–1.74; adolescents, OR: 1.62, 95% CI: 1.27–2.08; young adults, OR: 1.29, 95% CI: 1.11–1.50) (Supplementary Appendix D).
Discussion
We found associations between daily minimum temperature and mental health–related healthcare encounters among children, adolescents and young adults in a large urban setting. Associations were strongest among children and adolescents, though still elevated in young adults. We also identified subgroups of young people at heightened risk for mental health encounters associated with heat. Our findings contribute to the understanding of the complex links between heat exposure and mental health among young people. As extreme temperatures become more frequent and severe with global climate change, findings from this study can help inform targeted prevention strategies to improve mental health during periods of extreme heat in different youth populations.
Prior studies have found mixed results on whether children and adolescents have higher risk for mental health encounters in association with high ambient heat. For example, stronger heat–outcome associations were observed among individuals aged 6 to 18 years in Basu et al. (Reference Basu, Gavin, Pearson, Ebisu and Malig2018) and individuals aged 0 to 14 years in Wang et al. Reference Wang, Lavigne, Ouellette-kuntz and Chen(2014) compared to older age groups (Basu et al., Reference Basu, Gavin, Pearson, Ebisu and Malig2018; Wang et al., Reference Wang, Lavigne, Ouellette-kuntz and Chen2014), whereas two other studies have found that young people are less impacted by heat than are adults or elderly persons (Chan et al., Reference Chan, Lam, So, Goggins, Ho, Liu and Chung2018; Trang et al., Reference Trang, Rocklöv, Giang, Kullgren and Nilsson2016). Our study provided additional evidence that the impact of heat on acute mental health encounters were strongest among children and adolescents, though still elevated in young adults. The younger age groups were more affected by heat than young adults likely because of age-related physiological vulnerability (e.g., younger children are less able to regulate core body temperature). In addition, young adults may have better ability to protect themselves in terms of avoiding heat exposure (e.g., staying indoor and turning on air conditioning) and taking early-stage interventions to prevent severe mental healthcare outcomes (e.g., seeking social support and taking treatments and consulting therapists) and also possibly more hours per day exposure to air conditioning in workplace settings than children and adolescents (Sugg et al., Reference Sugg, Dixon and Runkle2019). Several common mechanisms could underlie the vulnerability to high temperature in terms of mental health risk among the three age groups. These include sleep disturbance, heat stress and/or dehydration contributing to altered neuroendocrine chemistry and also social aggression mechanisms that contribute to conflict and violence (Manchia et al., Reference Manchia, Carpiniello, Valtorta and Comai2017; Tiihonen et al., Reference Tiihonen, Halonen, Tiihonen, Kautiainen, Storvik and Callaway2017). Also, individuals with mental health conditions can have impaired cognitive awareness that make them less able to cope with heat, resulting in an exacerbation of mental health episodes.
The types of mental health diagnoses most strongly associated with temperature differed by age group: reaction disorder for children, anxiety and bipolar disorders among adolescents and psychosis and reaction disorders for young adults. Our findings for young adults were consistent with a recent national study of adults that also found associations between heat and schizotypal and anxiety disorders, which had overlapping diagnostic codes with our psychosis and reaction disorder categories (Nori-Sarma et al., Reference Nori-Sarma, Sun, Sun, Spangler, Oblath, Galea, Gradus and Wellenius2022). They also found significant positive associations between ambient heat and hospitalizations related to substance use disorders, where we observed positive but non-significant associations possibly due to our smaller sample size (Nori-Sarma et al., Reference Nori-Sarma, Sun, Sun, Spangler, Oblath, Galea, Gradus and Wellenius2022). Other studies have also shown increases in alcohol consumption and drug abuse during heatwaves (Cusack et al., Reference Cusack, de Crespigny and Athanasos2011). In contrast to previous studies that have reported a positive association between temperature and suicide (Burke et al., Reference Burke, González, Baylis, Heft-Neal, Baysan, Basu and Hsiang2018; Kim et al., Reference Kim, Kim, Gasparrini, Armstrong, Honda, Chung, Ng, Tobias, Íñiguez and Lavigne2019; Mullins and White, Reference Mullins and White2019; Sugg et al., Reference Sugg, Dixon and Runkle2019), we find essentially no evidence of an association between heat exposure and suicidal behaviour. It is likely that our small number of hospital and ED encounters identified with suicide or self-inflicted injury diagnostic codes led to imprecise and unstable effect estimates. Future research with large data sets and in different locations could help to clarify developmental changes in heat vulnerability related to mental health problems and suicide and shed light on risk factors and mechanisms.
We also found significant positive associations between heat and mental health–related encounters among children and adolescents who were non-Hispanic Black or from other racial/ethnic groups. Our results suggest that racially marginalized youth, a population disproportionately impacted by micro-urban heat islands effects, may need targeted prevention efforts during heatwaves. We found that payment sources with the highest heat vulnerability differed by age group. Although heat conferred risk for mental health encounters paid with Medicaid among children, heat risk was found for those who self-paid among adolescents, perhaps indicating the complicated and often inconsistent health insurance coverage particularly experienced by young people from low-income backgrounds.
Similar to a few previous studies (Kim et al., Reference Kim, Kim, Gasparrini, Armstrong, Honda, Chung, Ng, Tobias, Íñiguez and Lavigne2019; Sugg et al., Reference Sugg, Dixon and Runkle2019), our study has shown a ‘U’-shaped relationship between temperature and mental health ED and hospital encounters among children and adolescents, which suggests that the risk of mental health encounters may increase at both high and low temperature extremes. We found that the MRT differed across age groups and across racial/ethnic, payment source and mental health diagnostic subgroups. MRT indicates the temperature at which the lowest risk of mental health–related encounters is observed. While not well-studied, the MRT variation we observed complements our findings of variations in the magnitude of association with high temperature and mental health outcomes. It also suggests that behavioural and potentially physiologic susceptibility to heat may vary by age and social factors and depend on the type of mental health outcome.
This study has a number of limitations. First, our study examined associations, not causal effects, of heat and mental health. Second, the study is not generalizable beyond large US cities, as it includes data only from NYC health facilities, though there are some consistencies with results from other states or nationwide data (Mullins and White, Reference Mullins and White2019; Nori-Sarma et al., Reference Nori-Sarma, Sun, Sun, Spangler, Oblath, Galea, Gradus and Wellenius2022; Palinkas and Wong, Reference Palinkas and Wong2020). Third, we were unable to control for potential biases in mental health diagnostic patterns (Bailey et al., Reference Bailey, Mokonogho and Kumar2019; Fadus et al., Reference Fadus, Ginsburg, Sobowale, Halliday-Boykins, Bryant, Gray and Squeglia2020; Snowden, Reference Snowden2003), particularly those across racial or socio-economic subgroups. Fourth, although the study design of a case-crossover analysis helps us remove many of the long-term psychosocial, climate or other factors that may confound the associations between temperature and mental health, there are weather co-exposures (e.g., daily precipitation level) and air pollution exposures that have a potential to correlate with both the temperature level and mental health risk. Future research needs to consider how temperatures, in conjunction with weather and air pollution co-exposures, contribute to increased mental health issues in young people. Fifth, while many of the participants presenting at ED or being hospitalized for mental disorders have pre-existing mental health problems, we do not have data on participants’ pre-existing mental health conditions. Previous research has shown that the use of psychotropic medications can disrupt normal thermoregulation, and children and adolescents may experience more severe reactions to psychotropic medications (Cooper et al., Reference Cooper, Callahan, Shintani, Fuchs, Shelton, Dudley, Graves and Ray2014; Stöllberger et al., Reference Stöllberger, Lutz and Finsterer2009). Future research needs to elucidate the role of pre-existing mental health conditions and use of medications on increasing mental health risk associated with heat. Further, this study does not have data on sleep patterns, which suggests caution in making interpretations about the impact of high temperatures on sleep disturbance as a potential pathway to mental health disorders. Finally, our study did not examine the effect of heat during the cold season and our data are from 2005 to 2011 and thus do not capture the present mental health crises precipitated by the COVID-19 pandemic of 2020 and beyond. To address these limitations and other research gaps, future studies will hopefully include larger sample sizes and investigate broader exposures, including indoor temperatures, to better understand the sleep disruption mechanism and the impact of multiple extreme weather events on mental health.
In summary, we observed increased healthcare utilization for multiple mental health diagnoses associated with high temperatures across childhood, adolescence and young adulthood. As ED and hospital encounters represent the more severe cases of mental health crises, these findings may underestimate the true toll of heat on mental health. As sleep disturbance is a hypothesized mechanism, future studies could examine indoor temperature and the potential protective role well-functioning air conditioning may play, particularly for those communities disproportionately burdened by micro-urban heat islands.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S2045796023000161.
Data availability statement
Data are available from New York State Department of Health by application and signed data user agreement.
Acknowledgements
This publication was produced from raw data purchased from or provided by the New York State Department of Health (NYSDOH). However, the conclusions derived and views expressed herein are those of the authors and do not reflect the conclusions or views of NYSDOH. NYSDOH, its employees, officers and agents make no representation, warranty or guarantee as to the accuracy, completeness, currency or suitability of the information provided here.
Financial support
Research reported in this publication was supported by the National Institute of Environmental Health Sciences (R01ES030717). The funder/sponsor did not participate in the work.
Competing interest
The authors have no conflicts to disclose.
Open access
