Drew Wilson

Spatial Regression Analysis

An Evaluation of Heatwaves and Mental Health Related Emergency Department Visits in North Carolina

Introduction

Implications of Mental Health:

  • 1 Billion People have some form of mental health disorder 1
  • $5 trillion dollars lost in economic activity just due to complications within this people group .

Implications of Extreme Heat:

  • 1℃ increase in global temperature estimates reduction in economic growth by 8.5%
  • Over $1 billion dollars is lost annually in the United States alone due to heat related complications

Hot temperatures lead to negative mental health effects .

(not heatwaves)

Few studies have evaluated the relationship between heatwaves and mental health. First of its kind study for NC can showcase vulnerabilities.

Study Area and Methods

Data:

  • Sheps NC Emergency Department Visits
  • NC Climate Data
    • Spatial Resolution: ZCTA

Mental Health Definition:

  • ICD10 Codes: F00 – F99

Study Period:

  • Summer (May-September) 2016 – 2019

Heatwave Definition:

  • Daily Mean Temp >= 95th percentile of the Mean 3 Day Temperature

Statistical Method:

  • Linear Regression Models

Study Question:

  • Do Number of Heatwave/ Non-Heatwave Days predict Mental Health Related Emergency Department visits controlled for by population?

Preliminary Data

Linear Regression Models

Original Data Model

lm(Mental Health Patients ~ Number of Days + Population, data)

Log Transformed Data Model

lm(log1p(Mental Health Patients) ~ Number of Days + log1p(Population), data)

Model with Predictors

lm(log1p(Mental Health Patients) ~ Number of Days + log1p(Population) + log1p(Income) + log1p(Medicare/ Medicaid) + log1p(Racial Demographics), data)

Results

Discussion

Study Question:

  • Do Number of Heatwave/ Non-Heatwave Days predict Mental Health Related Emergency Department visits controlled for by population and other socioeconomic factors?

Limitations:

  • Limited data (just 4 years)
  • Heatwave calculation definition

Further Research:

  • More robust methods for the analysis – time series analysis, multiple regression
  • More predictor variables such as age, mental health condition, rural/ urban
  • Greater spatial area, entire United States