https://orcid.org/0000-0003-3487-0795
Abstract
The association between ambient temperature and morbidity has been explored previously. However, the association between temperature and mental health-related outcomes, including violence and self-harm, remains relatively unexamined. For the period 2005–2013, we obtained daily counts of mental health-related emergency room visits involving injuries with an external cause for 16 California climate zones from the California Office of Statewide Health Planning and Development and combined them with data on mean apparent temperature, a combination of temperature and humidity. Using Poisson regression models, we estimated climate zone-level associations and then used random-effects meta-analyses to produce overall estimates. Analyses were stratified by season (warm: May–October; cold: November–April), race/ethnicity, and age. During the warm season, a 10°F (5.6°C) increase in same-day mean apparent temperature was associated with 4.8% (95% confidence interval (CI): 3.6, 6.0), 5.8% (95% CI: 4.5, 7.1), and 7.9% (95% CI: 7.3, 8.4) increases in the risk of emergency room visits for mental health disorders, self-injury/suicide, and intentional injury/homicide, respectively. High temperatures during the cold season were also positively associated with these outcomes. Variations were observed by race/ethnicity, age group, and sex, with Hispanics, whites, persons aged 6–18 years, and females being at greatest risk for most outcomes. Increasing mean apparent temperature was found to have acute associations with mental health outcomes and intentional injuries, and these findings warrant further study in other locations.
The association between elevated ambient temperatures and morbidity from heat stroke and dehydration has been well established (1). The mentally ill have been identified as being at risk from extreme heat, especially those taking medications that interfere with physiological homeostasis; extreme heat has been associated with higher mortality (2, 3), heat-related illnesses (4), and physiologically impaired heat tolerance (5). Very few studies have examined the direct impact of temperature on mental health-related outcomes (6–8), which would identify specific populations at enhanced risk and help providers anticipate health-care utilization for mental disorders. Violence and self-harm, which are more prominent during hotter weather (9–13), may also be related to mental states of aggression. There is limited evidence that the relationship between temperature and mental disorders, violence, and self-harm is mediated by individual factors such as age, sex, and education (6, 7, 14), and the issue warrants further investigation.
In this study, we examined the association between apparent temperature and emergency room (ER) visits for mental health-related outcomes in California from 2005 to 2013. We had data on a large number of ER visits, so we were able to examine effect modification by race/ethnicity, age, sex, and region. Additionally, we explored specific mental disorder categories, specifically psychosis and neurotic disorders, and violent outcomes possibly related to mental health, including self-inflicted injury/suicide and inflicted injury/homicide. Though the latter two outcomes are not directly documented as mental health outcomes, exploring their association with temperature could be indicative of overall trends in mental health as they relate to outdoor temperature.
METHODS
Study population
We obtained data on daily numbers of ER visits in 16 California climate zones from the California Office of Statewide Health Planning and Development for the period 2005–2013. Climate zone boundaries were classified on the basis of weather and energy-use factors (15). We focused on primary diagnoses and limited our analyses to visits involving injuries with external causes, to focus on high-risk groups. The outcomes examined included all mental disorders (International Classification of Diseases, Ninth Revision (ICD-9) codes 290–319) and subcategories of psychosis (ICD-9 codes 290–299) and neurotic disorders (ICD-9 codes 300–316), including all neurotic, personality, and other nonpsychotic mental disorders. We also examined ER visits with external causes of injury identified as self-inflicted injury/suicide (ICD-9 codes E950–E959) or being a victim of intentionally inflicted injury/homicide (ICD-9 codes E960–E969), regardless of primary diagnosis. Data on race/ethnicity (white non-Hispanic (white), black non-Hispanic (black), Asian non-Hispanic (Asian), or Hispanic), age (6–18, 19–64, or ≥65 years), and sex (male, female) were also abstracted. To examine potential seasonal effect modification, we stratified ER visits by season (warm (May 1–October 31) or cold (November 1–April 30)) for all analyses. We were unable to differentiate between incident and prevalent visits, and multiple visits may have occurred for the same person.
Exposure data
Because previous studies have shown that temperature, rather than humidity, sunshine, or other meteorological factors, is most associated with mental health-related outcomes such as self-injury and suicide (14, 16), we considered mean, minimum, and maximum apparent temperature to take into account the combined exposure of temperature and humidity. Apparent temperature data were provided by the Environmental Protection Agency through the Air Quality System Data Mart (17), the California Irrigation Management Information System (18), and the National Oceanic and Atmospheric Administration (19). These data were restricted to December 2004 (1 month prior to the study period) to account for lagged temperature for ER visits occurring from the beginning of January 2005 to December 31, 2013. In total, 401 meteorological monitors had daily temperature and humidity data for the study period, and we restricted the data to monitors with at least 98% completeness.
For each population-weighted zip-code tabulation area centroid in a climate zone, we estimated daily apparent temperature using monitored data in the same climate zone, weighting by distance from each monitor. These values were combined to create an estimate of daily apparent temperature for each climate zone, while also weighting by the US Census population count associated with the zip-code tabulation area centroid. Distance calculations were completed with Geospatial Modelling Environment software (Spatial Ecology LLC, Marshfield, Wisconsin) (20), with additional data manipulation done using SAS, version 9.4 (SAS Institute, Inc., Cary, North Carolina) (21). The final data set consisted of variables for apparent temperature lags of 0–30 days for mean, maximum, and minimum apparent temperature.
Data analysis
We analyzed the data using a 2-stage time-series method with Poisson regression in R, version 3.2.2 (R Foundation for Statistical Computing, Vienna, Austria) (22) to estimate the association of mean, maximum, or minimum daily apparent temperature with daily numbers of ER visits, adjusting for holidays, day of the week, and seasonal/long-term trends using a natural spline smoothing function of time, measured in days with 2 degrees of freedom per season, for long-term temporal adjustment. Effect estimates were first calculated for each climate zone and then combined using random-effects models for meta-analyses to obtain an overall estimate (23). Daily temperature data were merged with ER daily count data and binomial (0/1) indicators for holidays (New Year’s Day, Memorial Day, Independence Day, Labor Day, Thanksgiving, and Christmas). Day of the week was included because homicides and suicides have been shown to vary by day of the week (11, 24). We excluded the last month, to avoid the issue of counts declining from the inclusion of persons admitted to the ER in 2013 who were not discharged until 2014.
The final model was chosen on the basis of Akaike’s Information Criterion, comparing models with mean, maximum, and minimum apparent temperatures. Several lag times were considered: the same day (lag 0), the previous day (lag 1), the average of the same day and the previous 6 days (lag 0–6), and the average of the same day and the previous 29 days (lag 0–29), and the metric of apparent temperature with the best model fit was used. We tested an additional squared term for apparent temperature in the final model with the best lag fit. Results were further stratified by race/ethnicity, age group, sex, and region (coastal/noncoastal). The race/ethnicity recorded for inflicted injury/homicide is typically that of the victim(s), since this was the person(s) admitted to the ER; but the race/ethnicity of the victim(s) is usually that of the perpetrator(s) as well (27). All main results are reported as the percent change in the risk of ER visits per 10°F (5.6°C) increase in apparent temperature.
We also considered extreme apparent temperatures, defined as the bottom 10th percentile and the upper 90th percentile of the distribution for each climate zone, using indicator variables (0/1). Additionally, we evaluated heat waves, which were defined as 2 or more consecutive days of apparent temperature above the 95th percentile for each climate zone.
Prior to beginning this study, we obtained institutional review board approval from California’s Committee for the Protection of Human Subjects.
RESULTS
Our study population consisted of 219,942 ER visits for mental health-related external-cause injuries, 62,812 of which were categorized as psychosis visits and 157,050 of which were neurosis visits (Table 1). Additionally, there were 322,478 self-injury/suicide visits and 1,041,433 inflicted injury/homicide visits recorded (Table 1). White patients made 56% of mental disorder visits, while 26% of such visits were made by Hispanic patients, 8% by black patients, and 3% by Asian patients. Most patients were 19–64 years of age (74%), with smaller percentages of cases being aged 6–18 years and ≥65 years (13% each). Fewer than 1% of mental disorder visits were made by children under age 6 years (an insufficient number for analysis). The majority of visits (56%) were made by males.
Characteristics of the Study Population (%) in a Study of Apparent Temperature and Mental Health-Related Emergency Room Visits, California, 2005–2013
| Characteristic | Mental Health Outcome | ||||
|---|---|---|---|---|---|
| All Mental Disorders (n = 219,942) | Psychosis (n = 62,812) | Neurotic Disorders (n = 157,050) | Self-Injury/Suicide (n = 322,478) | Inflicted Injury/Homicide (n = 1,041,433) | |
| Racial/ethnic group | |||||
| White | 56 | 63 | 53 | 55 | 34 |
| Black | 8 | 8 | 8 | 8 | 18 |
| Asian | 3 | 3 | 3 | 4 | 3 |
| Hispanic | 26 | 20 | 28 | 25 | 36 |
| Age group, years | |||||
| 0–5 | <1 | <1 | 1 | <1 | 1 |
| 6–18 | 13 | 7 | 16 | 21 | 18 |
| 19–64 | 74 | 67 | 76 | 75 | 79 |
| ≥65 | 13 | 26 | 7 | 3 | 2 |
| Region | |||||
| Coastal | 28 | 28 | 29 | 23 | 25 |
| Noncoastal | 72 | 72 | 71 | 77 | 75 |
| Sex | |||||
| Male | 56 | 54 | 56 | 41 | 68 |
| Female | 44 | 46 | 44 | 59 | 32 |
| Characteristic | Mental Health Outcome | ||||
|---|---|---|---|---|---|
| All Mental Disorders (n = 219,942) | Psychosis (n = 62,812) | Neurotic Disorders (n = 157,050) | Self-Injury/Suicide (n = 322,478) | Inflicted Injury/Homicide (n = 1,041,433) | |
| Racial/ethnic group | |||||
| White | 56 | 63 | 53 | 55 | 34 |
| Black | 8 | 8 | 8 | 8 | 18 |
| Asian | 3 | 3 | 3 | 4 | 3 |
| Hispanic | 26 | 20 | 28 | 25 | 36 |
| Age group, years | |||||
| 0–5 | <1 | <1 | 1 | <1 | 1 |
| 6–18 | 13 | 7 | 16 | 21 | 18 |
| 19–64 | 74 | 67 | 76 | 75 | 79 |
| ≥65 | 13 | 26 | 7 | 3 | 2 |
| Region | |||||
| Coastal | 28 | 28 | 29 | 23 | 25 |
| Noncoastal | 72 | 72 | 71 | 77 | 75 |
| Sex | |||||
| Male | 56 | 54 | 56 | 41 | 68 |
| Female | 44 | 46 | 44 | 59 | 32 |
Characteristics of the Study Population (%) in a Study of Apparent Temperature and Mental Health-Related Emergency Room Visits, California, 2005–2013
| Characteristic | Mental Health Outcome | ||||
|---|---|---|---|---|---|
| All Mental Disorders (n = 219,942) | Psychosis (n = 62,812) | Neurotic Disorders (n = 157,050) | Self-Injury/Suicide (n = 322,478) | Inflicted Injury/Homicide (n = 1,041,433) | |
| Racial/ethnic group | |||||
| White | 56 | 63 | 53 | 55 | 34 |
| Black | 8 | 8 | 8 | 8 | 18 |
| Asian | 3 | 3 | 3 | 4 | 3 |
| Hispanic | 26 | 20 | 28 | 25 | 36 |
| Age group, years | |||||
| 0–5 | <1 | <1 | 1 | <1 | 1 |
| 6–18 | 13 | 7 | 16 | 21 | 18 |
| 19–64 | 74 | 67 | 76 | 75 | 79 |
| ≥65 | 13 | 26 | 7 | 3 | 2 |
| Region | |||||
| Coastal | 28 | 28 | 29 | 23 | 25 |
| Noncoastal | 72 | 72 | 71 | 77 | 75 |
| Sex | |||||
| Male | 56 | 54 | 56 | 41 | 68 |
| Female | 44 | 46 | 44 | 59 | 32 |
| Characteristic | Mental Health Outcome | ||||
|---|---|---|---|---|---|
| All Mental Disorders (n = 219,942) | Psychosis (n = 62,812) | Neurotic Disorders (n = 157,050) | Self-Injury/Suicide (n = 322,478) | Inflicted Injury/Homicide (n = 1,041,433) | |
| Racial/ethnic group | |||||
| White | 56 | 63 | 53 | 55 | 34 |
| Black | 8 | 8 | 8 | 8 | 18 |
| Asian | 3 | 3 | 3 | 4 | 3 |
| Hispanic | 26 | 20 | 28 | 25 | 36 |
| Age group, years | |||||
| 0–5 | <1 | <1 | 1 | <1 | 1 |
| 6–18 | 13 | 7 | 16 | 21 | 18 |
| 19–64 | 74 | 67 | 76 | 75 | 79 |
| ≥65 | 13 | 26 | 7 | 3 | 2 |
| Region | |||||
| Coastal | 28 | 28 | 29 | 23 | 25 |
| Noncoastal | 72 | 72 | 71 | 77 | 75 |
| Sex | |||||
| Male | 56 | 54 | 56 | 41 | 68 |
| Female | 44 | 46 | 44 | 59 | 32 |
Climate-zone-specific mean apparent temperature distributions are listed in Table 2. Noncoastal climate zones had the majority of ER visits, with 71%–77% of visits, while coastal climate zones had the remaining 23%–29% of visits. The overall mean apparent temperature was 65.5°F (18.6°C) during the warm season and 48.9°F (9.4°C) during the cold season. During the warm season, individual zone mean apparent temperatures ranged from 51.9°F (11.1°C) to 83.5°F (28.6°C), while mean apparent temperatures ranged from 37.3°F (2.9°C) to 56.1°F (13.4°C) during the cold season.
Mean Apparent Temperature Exposure (°F) by Region, Climate Zone, and Season, California, 2005–2013
| Region and Representative Location | Climate Zone | Temperature Measure and Seasona | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean (SD) | Lower 5th Percentile | Lower 10th Percentile | Upper 90th Percentile | Upper 95th Percentile | |||||||
| Warm | Cold | Warm | Cold | Warm | Cold | Warm | Cold | Warm | Cold | ||
| Coastal | 59.9 (7.2) | 49.6 (6.4) | 48.7 | 38.8 | 50.9 | 41.1 | 69.6 | 57.6 | 72.7 | 59.7 | |
| Eureka | 1 | 51.9 (4.3) | 43.5 (5.5) | 45.0 | 34.8 | 46.4 | 36.6 | 57.5 | 50.6 | 59.4 | 52.5 |
| San Francisco | 3 | 57.5 (4.1) | 48.0 (5.0) | 50.7 | 39.7 | 52.1 | 41.4 | 62.7 | 54.3 | 64.3 | 56.3 |
| Santa Maria | 5 | 58.7 (4.7) | 48.9 (5.4) | 51.1 | 40.4 | 52.5 | 42.0 | 64.0 | 55.3 | 66.7 | 57.7 |
| LAX Airport | 6 | 64.0 (4.8) | 53.3 (4.9) | 56.6 | 45.3 | 58.2 | 46.8 | 70.3 | 59.2 | 71.9 | 61.2 |
| San Diego | 7 | 67.3 (5.9) | 54.0 (4.9) | 58.7 | 45.9 | 60.3 | 47.4 | 75.6 | 60.0 | 77.8 | 61.7 |
| Noncoastal | 68.0 (10.4) | 48.6 (8.7) | 52.2 | 34.5 | 55.4 | 37.4 | 81.3 | 59.5 | 86.5 | 62.8 | |
| Napa | 2 | 59.9 (6.3) | 44.2 (6.3) | 49.6 | 34.0 | 51.4 | 36.1 | 67.4 | 52.3 | 70.2 | 54.9 |
| San Jose | 4 | 61.5 (5.6) | 47.9 (5.9) | 52.4 | 38.1 | 54.2 | 40.3 | 68.8 | 55.3 | 71.0 | 57.7 |
| Long Beach | 8 | 68.8 (5.9) | 55.2 (5.3) | 59.9 | 47.0 | 61.7 | 48.4 | 76.9 | 61.6 | 78.7 | 63.5 |
| LA Civic Center | 9 | 67.8 (7.2) | 53.8 (6.8) | 56.5 | 43.3 | 58.6 | 45.1 | 77.4 | 62.7 | 80.1 | 65.4 |
| Riverside | 10 | 69.5 (8.3) | 52.1 (6.5) | 56.3 | 41.8 | 58.8 | 43.5 | 80.5 | 60.4 | 83.2 | 62.8 |
| Red Bluff | 11 | 69.9 (9.4) | 47.2 (7.5) | 54.0 | 35.7 | 57.1 | 37.9 | 81.4 | 57.0 | 84.5 | 60.0 |
| Stockton | 12 | 66.7 (7.7) | 47.0 (7.1) | 54.0 | 36.2 | 56.3 | 38.0 | 76.2 | 56.0 | 79.4 | 58.8 |
| Fresno | 13 | 71.5 (9.3) | 48.0 (7.8) | 55.8 | 35.9 | 58.3 | 37.9 | 82.4 | 58.0 | 85.6 | 61.3 |
| Barstow | 14 | 69.7 (9.7) | 45.7 (7.6) | 52.4 | 34.6 | 55.5 | 36.5 | 81.3 | 55.5 | 83.2 | 59.1 |
| Brawley | 15 | 83.5 (10.9) | 56.1 (7.9) | 64.8 | 43.8 | 67.3 | 46.3 | 97.1 | 66.7 | 98.9 | 69.5 |
| Bishop | 16 | 59.5 (9.0) | 37.3 (6.8) | 43.9 | 26.9 | 46.8 | 28.8 | 70.0 | 46.3 | 72.1 | 49.3 |
| Region and Representative Location | Climate Zone | Temperature Measure and Seasona | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean (SD) | Lower 5th Percentile | Lower 10th Percentile | Upper 90th Percentile | Upper 95th Percentile | |||||||
| Warm | Cold | Warm | Cold | Warm | Cold | Warm | Cold | Warm | Cold | ||
| Coastal | 59.9 (7.2) | 49.6 (6.4) | 48.7 | 38.8 | 50.9 | 41.1 | 69.6 | 57.6 | 72.7 | 59.7 | |
| Eureka | 1 | 51.9 (4.3) | 43.5 (5.5) | 45.0 | 34.8 | 46.4 | 36.6 | 57.5 | 50.6 | 59.4 | 52.5 |
| San Francisco | 3 | 57.5 (4.1) | 48.0 (5.0) | 50.7 | 39.7 | 52.1 | 41.4 | 62.7 | 54.3 | 64.3 | 56.3 |
| Santa Maria | 5 | 58.7 (4.7) | 48.9 (5.4) | 51.1 | 40.4 | 52.5 | 42.0 | 64.0 | 55.3 | 66.7 | 57.7 |
| LAX Airport | 6 | 64.0 (4.8) | 53.3 (4.9) | 56.6 | 45.3 | 58.2 | 46.8 | 70.3 | 59.2 | 71.9 | 61.2 |
| San Diego | 7 | 67.3 (5.9) | 54.0 (4.9) | 58.7 | 45.9 | 60.3 | 47.4 | 75.6 | 60.0 | 77.8 | 61.7 |
| Noncoastal | 68.0 (10.4) | 48.6 (8.7) | 52.2 | 34.5 | 55.4 | 37.4 | 81.3 | 59.5 | 86.5 | 62.8 | |
| Napa | 2 | 59.9 (6.3) | 44.2 (6.3) | 49.6 | 34.0 | 51.4 | 36.1 | 67.4 | 52.3 | 70.2 | 54.9 |
| San Jose | 4 | 61.5 (5.6) | 47.9 (5.9) | 52.4 | 38.1 | 54.2 | 40.3 | 68.8 | 55.3 | 71.0 | 57.7 |
| Long Beach | 8 | 68.8 (5.9) | 55.2 (5.3) | 59.9 | 47.0 | 61.7 | 48.4 | 76.9 | 61.6 | 78.7 | 63.5 |
| LA Civic Center | 9 | 67.8 (7.2) | 53.8 (6.8) | 56.5 | 43.3 | 58.6 | 45.1 | 77.4 | 62.7 | 80.1 | 65.4 |
| Riverside | 10 | 69.5 (8.3) | 52.1 (6.5) | 56.3 | 41.8 | 58.8 | 43.5 | 80.5 | 60.4 | 83.2 | 62.8 |
| Red Bluff | 11 | 69.9 (9.4) | 47.2 (7.5) | 54.0 | 35.7 | 57.1 | 37.9 | 81.4 | 57.0 | 84.5 | 60.0 |
| Stockton | 12 | 66.7 (7.7) | 47.0 (7.1) | 54.0 | 36.2 | 56.3 | 38.0 | 76.2 | 56.0 | 79.4 | 58.8 |
| Fresno | 13 | 71.5 (9.3) | 48.0 (7.8) | 55.8 | 35.9 | 58.3 | 37.9 | 82.4 | 58.0 | 85.6 | 61.3 |
| Barstow | 14 | 69.7 (9.7) | 45.7 (7.6) | 52.4 | 34.6 | 55.5 | 36.5 | 81.3 | 55.5 | 83.2 | 59.1 |
| Brawley | 15 | 83.5 (10.9) | 56.1 (7.9) | 64.8 | 43.8 | 67.3 | 46.3 | 97.1 | 66.7 | 98.9 | 69.5 |
| Bishop | 16 | 59.5 (9.0) | 37.3 (6.8) | 43.9 | 26.9 | 46.8 | 28.8 | 70.0 | 46.3 | 72.1 | 49.3 |
Abbreviations: LA, Los Angeles; LAX, Los Angeles International; SD, standard deviation.
a Warm season: May–October; cold season: November–April.
Mean Apparent Temperature Exposure (°F) by Region, Climate Zone, and Season, California, 2005–2013
| Region and Representative Location | Climate Zone | Temperature Measure and Seasona | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean (SD) | Lower 5th Percentile | Lower 10th Percentile | Upper 90th Percentile | Upper 95th Percentile | |||||||
| Warm | Cold | Warm | Cold | Warm | Cold | Warm | Cold | Warm | Cold | ||
| Coastal | 59.9 (7.2) | 49.6 (6.4) | 48.7 | 38.8 | 50.9 | 41.1 | 69.6 | 57.6 | 72.7 | 59.7 | |
| Eureka | 1 | 51.9 (4.3) | 43.5 (5.5) | 45.0 | 34.8 | 46.4 | 36.6 | 57.5 | 50.6 | 59.4 | 52.5 |
| San Francisco | 3 | 57.5 (4.1) | 48.0 (5.0) | 50.7 | 39.7 | 52.1 | 41.4 | 62.7 | 54.3 | 64.3 | 56.3 |
| Santa Maria | 5 | 58.7 (4.7) | 48.9 (5.4) | 51.1 | 40.4 | 52.5 | 42.0 | 64.0 | 55.3 | 66.7 | 57.7 |
| LAX Airport | 6 | 64.0 (4.8) | 53.3 (4.9) | 56.6 | 45.3 | 58.2 | 46.8 | 70.3 | 59.2 | 71.9 | 61.2 |
| San Diego | 7 | 67.3 (5.9) | 54.0 (4.9) | 58.7 | 45.9 | 60.3 | 47.4 | 75.6 | 60.0 | 77.8 | 61.7 |
| Noncoastal | 68.0 (10.4) | 48.6 (8.7) | 52.2 | 34.5 | 55.4 | 37.4 | 81.3 | 59.5 | 86.5 | 62.8 | |
| Napa | 2 | 59.9 (6.3) | 44.2 (6.3) | 49.6 | 34.0 | 51.4 | 36.1 | 67.4 | 52.3 | 70.2 | 54.9 |
| San Jose | 4 | 61.5 (5.6) | 47.9 (5.9) | 52.4 | 38.1 | 54.2 | 40.3 | 68.8 | 55.3 | 71.0 | 57.7 |
| Long Beach | 8 | 68.8 (5.9) | 55.2 (5.3) | 59.9 | 47.0 | 61.7 | 48.4 | 76.9 | 61.6 | 78.7 | 63.5 |
| LA Civic Center | 9 | 67.8 (7.2) | 53.8 (6.8) | 56.5 | 43.3 | 58.6 | 45.1 | 77.4 | 62.7 | 80.1 | 65.4 |
| Riverside | 10 | 69.5 (8.3) | 52.1 (6.5) | 56.3 | 41.8 | 58.8 | 43.5 | 80.5 | 60.4 | 83.2 | 62.8 |
| Red Bluff | 11 | 69.9 (9.4) | 47.2 (7.5) | 54.0 | 35.7 | 57.1 | 37.9 | 81.4 | 57.0 | 84.5 | 60.0 |
| Stockton | 12 | 66.7 (7.7) | 47.0 (7.1) | 54.0 | 36.2 | 56.3 | 38.0 | 76.2 | 56.0 | 79.4 | 58.8 |
| Fresno | 13 | 71.5 (9.3) | 48.0 (7.8) | 55.8 | 35.9 | 58.3 | 37.9 | 82.4 | 58.0 | 85.6 | 61.3 |
| Barstow | 14 | 69.7 (9.7) | 45.7 (7.6) | 52.4 | 34.6 | 55.5 | 36.5 | 81.3 | 55.5 | 83.2 | 59.1 |
| Brawley | 15 | 83.5 (10.9) | 56.1 (7.9) | 64.8 | 43.8 | 67.3 | 46.3 | 97.1 | 66.7 | 98.9 | 69.5 |
| Bishop | 16 | 59.5 (9.0) | 37.3 (6.8) | 43.9 | 26.9 | 46.8 | 28.8 | 70.0 | 46.3 | 72.1 | 49.3 |
| Region and Representative Location | Climate Zone | Temperature Measure and Seasona | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean (SD) | Lower 5th Percentile | Lower 10th Percentile | Upper 90th Percentile | Upper 95th Percentile | |||||||
| Warm | Cold | Warm | Cold | Warm | Cold | Warm | Cold | Warm | Cold | ||
| Coastal | 59.9 (7.2) | 49.6 (6.4) | 48.7 | 38.8 | 50.9 | 41.1 | 69.6 | 57.6 | 72.7 | 59.7 | |
| Eureka | 1 | 51.9 (4.3) | 43.5 (5.5) | 45.0 | 34.8 | 46.4 | 36.6 | 57.5 | 50.6 | 59.4 | 52.5 |
| San Francisco | 3 | 57.5 (4.1) | 48.0 (5.0) | 50.7 | 39.7 | 52.1 | 41.4 | 62.7 | 54.3 | 64.3 | 56.3 |
| Santa Maria | 5 | 58.7 (4.7) | 48.9 (5.4) | 51.1 | 40.4 | 52.5 | 42.0 | 64.0 | 55.3 | 66.7 | 57.7 |
| LAX Airport | 6 | 64.0 (4.8) | 53.3 (4.9) | 56.6 | 45.3 | 58.2 | 46.8 | 70.3 | 59.2 | 71.9 | 61.2 |
| San Diego | 7 | 67.3 (5.9) | 54.0 (4.9) | 58.7 | 45.9 | 60.3 | 47.4 | 75.6 | 60.0 | 77.8 | 61.7 |
| Noncoastal | 68.0 (10.4) | 48.6 (8.7) | 52.2 | 34.5 | 55.4 | 37.4 | 81.3 | 59.5 | 86.5 | 62.8 | |
| Napa | 2 | 59.9 (6.3) | 44.2 (6.3) | 49.6 | 34.0 | 51.4 | 36.1 | 67.4 | 52.3 | 70.2 | 54.9 |
| San Jose | 4 | 61.5 (5.6) | 47.9 (5.9) | 52.4 | 38.1 | 54.2 | 40.3 | 68.8 | 55.3 | 71.0 | 57.7 |
| Long Beach | 8 | 68.8 (5.9) | 55.2 (5.3) | 59.9 | 47.0 | 61.7 | 48.4 | 76.9 | 61.6 | 78.7 | 63.5 |
| LA Civic Center | 9 | 67.8 (7.2) | 53.8 (6.8) | 56.5 | 43.3 | 58.6 | 45.1 | 77.4 | 62.7 | 80.1 | 65.4 |
| Riverside | 10 | 69.5 (8.3) | 52.1 (6.5) | 56.3 | 41.8 | 58.8 | 43.5 | 80.5 | 60.4 | 83.2 | 62.8 |
| Red Bluff | 11 | 69.9 (9.4) | 47.2 (7.5) | 54.0 | 35.7 | 57.1 | 37.9 | 81.4 | 57.0 | 84.5 | 60.0 |
| Stockton | 12 | 66.7 (7.7) | 47.0 (7.1) | 54.0 | 36.2 | 56.3 | 38.0 | 76.2 | 56.0 | 79.4 | 58.8 |
| Fresno | 13 | 71.5 (9.3) | 48.0 (7.8) | 55.8 | 35.9 | 58.3 | 37.9 | 82.4 | 58.0 | 85.6 | 61.3 |
| Barstow | 14 | 69.7 (9.7) | 45.7 (7.6) | 52.4 | 34.6 | 55.5 | 36.5 | 81.3 | 55.5 | 83.2 | 59.1 |
| Brawley | 15 | 83.5 (10.9) | 56.1 (7.9) | 64.8 | 43.8 | 67.3 | 46.3 | 97.1 | 66.7 | 98.9 | 69.5 |
| Bishop | 16 | 59.5 (9.0) | 37.3 (6.8) | 43.9 | 26.9 | 46.8 | 28.8 | 70.0 | 46.3 | 72.1 | 49.3 |
Abbreviations: LA, Los Angeles; LAX, Los Angeles International; SD, standard deviation.
a Warm season: May–October; cold season: November–April.
After analyzing several lag times, we found that same-day lags most frequently had the best model fit according to Akaike’s Information Criterion (Figure 1). While some elevated risks were also observed for lag 1 and lag 0–6, lag 0–30 had all significant associations for the cold season (see Web Figure 1, available at https://academic.oup.com/aje). After comparing associations between mean, minimum, and maximum apparent temperatures (Figure 2, Web Figure 2), we found mean apparent temperature to have the best model fit. Squared terms were generally not statistically significant when data were stratified by season (warm or cold), implying that the linear term was sufficient. Thus, our final model included mean apparent temperature for lag 0, and all results were stratified by season.
Percent change in the risk of mental health-related emergency room visits per 10°F (5.6°C) increase in mean apparent temperature, by lag period and season, California, 2005–2013. Bars, 95% confidence intervals.
Percent change in the risk of mental health-related emergency room visits per 10°F (5.6°C) increase in mean apparent temperature, by lag period and season, California, 2005–2013. Bars, 95% confidence intervals.
Percent change in the risk of mental health-related emergency room visits per 10°F (5.6°C) increase in minimum, mean, or maximum apparent temperature (lag 0), by season, California, 2005–2013. Bars, 95% confidence intervals.
Percent change in the risk of mental health-related emergency room visits per 10°F (5.6°C) increase in minimum, mean, or maximum apparent temperature (lag 0), by season, California, 2005–2013. Bars, 95% confidence intervals.
We observed significant associations between same-day mean apparent temperature and overall and specific categories of mental health-related ER visits (Table 3). During the warm season, a 10°F (5.6°C) increase in mean apparent temperature was associated with a 4.8% (95% confidence interval (CI): 3.6, 6.0) increase in risk of overall mental health events, a 2.9% (95% CI: 0.7, 5.2) increase in risk of psychotic events, and a 5.7% (95% CI: 3.8, 7.6) increase in risk of neurotic events. Increased temperatures during the cold season were associated with 5.3% (95% CI: 4.0, 6.7), 3.9% (95% CI: 1.6, 6.3), and 6.0% (95% CI: 4.5, 7.5) increases in the risk of all mental health events, psychotic events, and neurotic events, respectively. During the warm season, a 10°F (5.6°C) increase in mean apparent temperature was associated with a 5.8% (95% CI: 4.5, 7.1) increase in the risk of self-injury/suicide ER visits and a 7.9% (95% CI: 7.3, 8.4) increase in risk of inflicted injury/homicide. During the cold season, a mean apparent temperature increase of 10°F (5.6°C) was associated with a 7.2% (95% CI: 5.7, 8.8) increase in risk of self-injury/suicide and a 10.6% (95% CI: 9.8, 11.4) increase in risk of inflicted injury/homicide visits.
Percent Change in the Risk of Mental Health-Related Emergency Room Visits per 10°F (5.6°C) Increase in Mean Apparent Temperature, by Seasona, California, 2005–2013
| Outcome | Warm Season | Cold Season | ||||
|---|---|---|---|---|---|---|
| No. of ER Visits | % Changeb | 95% CI | No. of ER Visits | % Changeb | 95% CI | |
| All mental disorders | 115,962 | 4.8 | 3.6, 6.0 | 103,980 | 5.3 | 4.0, 6.7 |
| Psychosis | 32,959 | 2.9 | 0.7, 5.2 | 29,853 | 3.9 | 1.6, 6.3 |
| Neurotic disorders | 82,962 | 5.7 | 3.8, 7.6 | 74,088 | 6.0 | 4.5, 7.5 |
| Suicide | 168,596 | 5.8 | 4.5, 7.1 | 153,882 | 7.2 | 5.7, 8.8 |
| Homicide | 561,891 | 7.9 | 7.3, 8.4 | 479,542 | 10.6 | 9.8, 11.4 |
| Outcome | Warm Season | Cold Season | ||||
|---|---|---|---|---|---|---|
| No. of ER Visits | % Changeb | 95% CI | No. of ER Visits | % Changeb | 95% CI | |
| All mental disorders | 115,962 | 4.8 | 3.6, 6.0 | 103,980 | 5.3 | 4.0, 6.7 |
| Psychosis | 32,959 | 2.9 | 0.7, 5.2 | 29,853 | 3.9 | 1.6, 6.3 |
| Neurotic disorders | 82,962 | 5.7 | 3.8, 7.6 | 74,088 | 6.0 | 4.5, 7.5 |
| Suicide | 168,596 | 5.8 | 4.5, 7.1 | 153,882 | 7.2 | 5.7, 8.8 |
| Homicide | 561,891 | 7.9 | 7.3, 8.4 | 479,542 | 10.6 | 9.8, 11.4 |
Abbreviations: CI, confidence interval; ER, emergency room.
a Warm season: May–October; cold season: November–April.
b Percent change in the risk of ER visits.
Percent Change in the Risk of Mental Health-Related Emergency Room Visits per 10°F (5.6°C) Increase in Mean Apparent Temperature, by Seasona, California, 2005–2013
| Outcome | Warm Season | Cold Season | ||||
|---|---|---|---|---|---|---|
| No. of ER Visits | % Changeb | 95% CI | No. of ER Visits | % Changeb | 95% CI | |
| All mental disorders | 115,962 | 4.8 | 3.6, 6.0 | 103,980 | 5.3 | 4.0, 6.7 |
| Psychosis | 32,959 | 2.9 | 0.7, 5.2 | 29,853 | 3.9 | 1.6, 6.3 |
| Neurotic disorders | 82,962 | 5.7 | 3.8, 7.6 | 74,088 | 6.0 | 4.5, 7.5 |
| Suicide | 168,596 | 5.8 | 4.5, 7.1 | 153,882 | 7.2 | 5.7, 8.8 |
| Homicide | 561,891 | 7.9 | 7.3, 8.4 | 479,542 | 10.6 | 9.8, 11.4 |
| Outcome | Warm Season | Cold Season | ||||
|---|---|---|---|---|---|---|
| No. of ER Visits | % Changeb | 95% CI | No. of ER Visits | % Changeb | 95% CI | |
| All mental disorders | 115,962 | 4.8 | 3.6, 6.0 | 103,980 | 5.3 | 4.0, 6.7 |
| Psychosis | 32,959 | 2.9 | 0.7, 5.2 | 29,853 | 3.9 | 1.6, 6.3 |
| Neurotic disorders | 82,962 | 5.7 | 3.8, 7.6 | 74,088 | 6.0 | 4.5, 7.5 |
| Suicide | 168,596 | 5.8 | 4.5, 7.1 | 153,882 | 7.2 | 5.7, 8.8 |
| Homicide | 561,891 | 7.9 | 7.3, 8.4 | 479,542 | 10.6 | 9.8, 11.4 |
Abbreviations: CI, confidence interval; ER, emergency room.
a Warm season: May–October; cold season: November–April.
b Percent change in the risk of ER visits.
Figure 3 and Web Figure 3 show results stratified by racial/ethnic group, age, sex, and region for all mental health events (Figure 3) and for the subcategories of psychosis (Web Figure 3a), neurotic disorders (Web Figure 3b), self-injury/suicide (Web Figure 3c), and inflicted injury/homicide (Web Figure 3d), respectively. Since the results for warm and cold seasons were similar, only warm-season results are shown in these figures. Per 10°F (5.6°C) increase in mean apparent temperature, the greatest associations were observed for whites and Hispanics for all mental health outcomes during the warm season (whites: 4.7% (95% CI: 3.1, 6.4); Hispanics: 6.3% (95% CI: 3.7, 9.1)) (Figure 3). Similar results were observed during the cold season. For all mental health outcomes, as well as psychosis and neurotic disorder visits, estimates for black and Asian patients showed no significant associations with temperature increases.
Percent change in the risk of mental health-related emergency room visits per 10°F (5.6°C) increase in mean apparent temperature during the warm season, by race/ethnicity (A), age group (B), sex (C), and region (D), California, 2005–2013. Bars, 95% confidence intervals.
Percent change in the risk of mental health-related emergency room visits per 10°F (5.6°C) increase in mean apparent temperature during the warm season, by race/ethnicity (A), age group (B), sex (C), and region (D), California, 2005–2013. Bars, 95% confidence intervals.
When results were stratified, risk of self-injury/suicide (cold season) and inflicted injury/homicide (warm season) appeared to be highest for Hispanics. For self-injury/suicide, a 10°F (5.6°C) increase in mean apparent temperature was associated with a 6.9% (95% CI: 4.3, 9.6) increase during the warm season and a 10.5% (95% CI: 7.3, 13.7) increase during the cold season. These risks were more increased in comparison with whites during both the warm season (5.1%, 95% CI: 3.4, 6.8) and the cold season (6.1%, 95% CI: 4.7, 7.5). Blacks also had significantly elevated risk during both seasons, while Asians had greater risk for the cold season only. Risk of being a victim of inflicted injury/homicide was also highest for Hispanics in both seasons. During the warm season, a 9.9% (95% CI: 9.0, 10.9) increase in inflicted injury/homicide was found for Hispanics, as compared with a 7.0% (95% CI: 6.0, 7.9) increase among whites and a 6.0% (95% CI: 4.6, 7.4) increase among blacks.
Risks were relatively the same across age categories during both the warm and cold seasons. During the warm season, mental health disorders had the highest association with apparent temperature for 6- to 18-year-olds, with a 7.3% (95% CI: 4.0, 10.8) increase in the risk of ER visits per 10°F (5.6°C) increase in mean apparent temperature. The association between temperature and self-inflicted injury/suicide was also highest in 6- to 18-year-olds (6.7%, 95% CI: 4.5, 8.9). The increase in inflicted injury/homicide was highest among persons aged 19–64 years (8.3%, 95% CI: 7.7, 9.0). Cold-season estimates followed a similar pattern, with all mental health and self-inflicted injury/suicide associations being highest for the age group 6–18 years. During the cold season, however, inflicted injury/homicide had the highest association with temperature for 6- to 18-year-olds, at 15.4% (95% CI: 13.7, 17.2) greater risk.
Effect modification by region was not found, except for psychosis, which had a greater association in noncoastal areas during the warm season (Web Figure 3a). Inflicted injury/homicide showed fairly similar associations with apparent temperature by region. During the cold season, significant associations were found among coastal and noncoastal populations for neurotic disorders, self-inflicted injury/suicide, and inflicted injury/homicide, but no significant differences between regions were observed.
In analyses that considered extreme temperatures, results for the upper 90th percentile and heat-wave exposure were similar to our estimates for temperature increments of 10°F (5.6°C) (Figure 4). Most estimates had slightly less magnitude for the 90th percentile but a slightly higher association with heat waves for inflicted injury/homicide, overall mental health, and neurotic outcomes. The lowest decile (10th percentile) had all negative associations.
Percent change in the risk of mental health-related emergency room visits in the bottom 10th percentile of apparent temperature (A), in the upper 90th percentile of apparent temperature (B), and during heat waves (≥2 days in the 95th percentile for climate zone) (C), California, 2005–2013. Bars, 95% confidence intervals.
Percent change in the risk of mental health-related emergency room visits in the bottom 10th percentile of apparent temperature (A), in the upper 90th percentile of apparent temperature (B), and during heat waves (≥2 days in the 95th percentile for climate zone) (C), California, 2005–2013. Bars, 95% confidence intervals.
DISCUSSION
We found associations between increasing apparent temperature and risk of all mental health disorders in California, as well as for the subcategories of psychotic and neurotic events, and inflicted injury/homicide and self-injury/suicide. While very few prior studies showed increased associations with mental health-related outcomes primarily during the warm season or following heat waves (6, 7), we observed greater risk of the mental health outcomes examined with increasing temperature during both the warm and cold seasons. Because temperatures in California are relatively mild, particularly in the more densely populated coastal areas, there was not a drastic difference between mean cold-season and warm-season temperatures. Extremes in hotter temperatures or heat waves were similar to the magnitude of the effect estimates for temperature per 10°F (5.6°C). Higher associations were observed in noncoastal areas for psychosis only, while other outcomes had similar risks by region. Poorer mental health-related outcomes in noncoastal areas could be related to lower socioeconomic status (SES) in comparison with coastal areas. For example, use of air conditioning is typically correlated with higher SES; however, in California, coastal regions have higher SES but lower air conditioner ownership due to relatively milder temperatures. Furthermore, low SES is also a risk factor for mental disorder itself, particularly for children and adolescents (25), as is being a victim of a violent crime, chronic stress, crowding, noise pollution, and discrimination (26).
ER visits in our study were often made for the victims of violence, not necessarily the perpetrators, unless the perpetrators were also admitted to the ER. For homicides, the victim’s race/ethnicity may be a good predictor of the perpetrator’s (27). Temperature may have also affected the mental state and aggressiveness of both the perpetrator and the victim, resulting in a higher likelihood of physical altercation and injury. Our results support the hypothesis that higher temperatures may lead to more violence, including suicides and homicides, particularly for Hispanics and whites during both seasons. During the warm season, Hispanics displayed the greatest risks for all mental health-related outcomes except psychosis, for which whites had a higher association. However, sample sizes were small for some outcomes for Asians and blacks, so we may not have captured accurate risk estimates for those racial/ethnic groups. There may be a cultural stigma regarding mental illness in some communities, so it may be underreported from lack of diagnosis. Some populations may be more likely to use the ER for primary care (28), and they would have received better coverage in this study. Prior studies did not examine disparities among racial/ethnic groups specifically for temperature and mental health-related outcomes.
Although risks were similar by age, the greatest risk was observed for 6- to 18-year-olds, except for inflicted injury/homicide, where the greatest risk was for adults aged 19–64 years. Previous studies found elevated risk of psychological emergency calls for persons aged 15–44 years (29) and elevated risk of ER admissions due to mental and behavioral disorders (8). Unlike the majority of prior studies, which focused on medication use in the elderly, we were able to evaluate both overall and specific mental health outcomes for all age groups. In a previous Australian study (6), the elderly were also found to be at increased risk for mortality during heat waves, but not for hospitalization (except those aged ≥75 years). We did not observe increased risk for the elderly, partly because we examined ER visits rather than mortality or hospital visits. Elderly persons with mental health disorders in California typically reside in medical care facilities where they spend most of their time indoors under supervision and in climate-controlled environments, with lower risk of morbidity/mortality following heat waves compared with younger age groups (30).
Apart from homicides, females had greater or similar risk as males during both the warm and cold seasons. In the Australian study (6), adult males had greater risk of dementia and schizophrenic disorders following heat waves, but females had greater risk of disorders due to psychoactive substance use. In a study in Vietnam, males aged ≥60 years were more sensitive to seasonal fluctuations for mental disorder hospital admissions (31). In a South Korean study (12), males also had slightly greater risk for suicide than females, although the finding was not statistically significant.
The biological mechanisms for temperature-related increases in mental health-related outcomes are multifactorial. As temperatures rise, stresses from everyday life in the home, social networks, and the workplace are likely to be exacerbated, particularly for persons with acute or chronic mental problems. During extended periods of heat, irritability and psychological distress (including risky behaviors such as alcohol consumption), violence, and aggression are more common (32). In one study, daytime temperature was found to be a risk factor for episodes related to schizophrenia (33), possibly because of increased stress on the nervous system (34). Medications such as beta blockers (35), benzodiazepines (36), and antidepressants (37), used to treat hypertension, schizophrenia, and depression, respectively, have been shown to increase risk of suicide, while antipsychotic medications for schizophrenia (38), as well as psychoactive substances (39), have been related to increased violence and homicides. Impaired thermoregulation and suppressed thirst, noted side effects of medications associated with mental health disorder treatment, may contribute to greater mental health admissions during periods of increased heat (40). Mental illness is also a predominant risk factor for heat-related illnesses. In animal models, specific neurotransmitters that may be involved in thermoregulation may be compromised, particularly for schizophrenic patients (41). Some antipsychotic medications, especially those commonly prescribed for dementia (42), may disrupt normal thermoregulation due to their particular pharmacological properties (4). Furthermore, persons with specific conditions such as dementia, as well as those who take particular types of medications, may lack the awareness or cognitive function to avoid going outdoors or to wear appropriate clothing during periods of heat exposure (6).
There were several limitations in this study. The most severe cases would not have been captured because they resulted in death before the person reached the ER. On the other hand, other cases may not have been severe enough for a person to visit the ER and may have been treated at the scene or in other treatment facilities. In either scenario, ER visits capture a specific subset of mental health morbidity. While we were able to evaluate some specific mental health-related outcomes, we could not look at several others in more detail because of inadequate numbers. We relied on information provided by the databases and could not consider changes in individual factors that may be on the noncausal pathway, such as medication use, SES, activity patterns, trends in seeking care for mental health issues, or comorbidity, that may have contributed to spurious false-positive associations. Since we used time-series data, these confounders should not have affected our results, unless they changed on a daily level, and other time-variant sources of confounding may exist. Race/ethnicity may be a valid measure of SES effect modification and could not account for the other factors. However, the results were generally consistent by lag period, outcome, and geographic region, providing further credibility for our findings. We also relied on ecological exposure data but tried to minimize misclassification by weighting climate zone temperature estimates on the basis of population and temperature monitor proximity. It is not feasible to conduct personal temperature exposure assessment for each individual (43) in studies with large populations, given the great amount of time and expense involved, and monitored data provided a good proxy for population-weighted exposure.
As global temperatures continue to rise and heat waves are predicted to become more intense and frequent in the future (44), it will become necessary to consider all outcomes rather than just mortality from all causes or cardiovascular and/or respiratory diseases (45) in order to target high-risk populations and prevent heat-associated deaths and illnesses. To our knowledge, this study is among the first to have quantified the association between temperature and mental health-related outcomes, and the first to do so in California. We examined both warm and cold seasons and considered extreme temperatures.
Through identification of subgroups that are most vulnerable, morbidity related to temperature exposure can be alleviated. The results of this study could be invaluable to mental health patients and their caretakers, as well as policy-makers. This study focused on short-term exposures, with the longest lag being 30 days. In future studies, longer exposure averages should also be examined for associations between temperature and mental health-related outcomes, particularly in colder climates.
ACKNOWLEDGMENTS
Author affiliations: Air and Climate Epidemiology Section, California Office of Environmental Health Hazard Assessment, Oakland, California (Rupa Basu, Dharshani Pearson, Keita Ebisu, Brian Malig); and Department of Environmental Health Sciences, School of Public Health, Yale University, New Haven, Connecticut (Lyndsay Gavin).
The opinions expressed in this article are those of the authors and do not represent those of the California Environmental Protection Agency or the Office of Environmental Health Hazard Assessment.
Conflict of interest: none declared.
Abbreviations
- CI
confidence interval
- ER
emergency room
- ICD
International Classification of Diseases, Ninth Revision
- SES
socioeconomic status
