This short project was originally done as a Kaggle Task for the dataset UNCOVER COVID-19 Challenge, Task 5, sponsored by the Roche Data Science Coalition. The date for the task had passed in June, but I liked the idea of the task so I did it anyway :)
Abstract
The link between economics and mental health is well established. In 2011, the World Health Organization discussed the impact of the 2007 economic crisis on Europe in the document “Impact of Economic Crises on Mental Health”, showing the link between suicide and economics.
Now the United States is experiencing its own economic crisis. During the current COVID-19 pandemic, many states within the U.S. are experiencing economic downturn. Per a recent article by Mark Muro, this downturn may hit certain industries especially hard (mining/oil and gas, transportation, employment services, travel arrangements, and leisure and hospitality). Based on the WHO article mentioned above, it may be assumed that individuals in these industries may soon be in need of mental health services. For the purposes of this notebook, these industries will be termed “high-risk”.
Additionally, many first responders and health care workers are asked to work longer hours in more stressful situations than ever before. As more and more COVID patients are brought to hospitals, first-responders and medical staff face the risk of burnout, as well as the risk of contracting the virus, as described by the U.S. Fire Administration in a May 2020 article. This can take a high toll on mental health, even leading to depression (“What the Coronavirus..”, 2020).
Furthermore, teachers are now being asked to transition from mainly in-person classroom teaching with established protocols to online teaching, in which new environments, rules, and curriculums may need to be developed. For some educators, this experience has been extremely stressful (Gerwertz, 2020).
For the purpose of this notebook, these combined industries of healthcare and education will be termed “high-stress”.
This project provides a unique perspective by dividing U.S. Bureau of Labor Statistics (BLS) industry sectors data into “high-risk/high-stress” jobs. This was done by grouping BLS industry sectors similar to those mentioned in Muro’s article with BLS’s healthcare and education industry sector. March employment numbers across from these industry sectors were then aggregated, yielding a total of high-risk/high stress jobs per state. These jobs are then plotted against a map of the United States, providing potentially a quick-and-dirty view of upcoming state needs.
This method is fast-to-market as it is based only on data already collected, published, and verified by the U.S. Bureau of Labor Statistics, with very little data manipulation. It resolves an upcoming need by providing a possible predictor of U.S. states that may experience a high need for mental health treatment.
This solution could easily be used in other countries with similar economic data. This project has been made public for use among the Kaggle community as a Kaggle notebook.
Visualization
The map below shows which states have the most employment in high-risk/high-stress industries. As mentioned above, high-risk jobs are those that may decline due to the declining U.S. economy, and high-stress jobs are those that are likely to increase in stress due to the COVID-19 pandemic. Darker colors indicate more jobs, while lighter colors indicate fewer jobs.
From the map, it appears that Texas and California have a lot of jobs in high-risk/high-stress industries, as well as smaller states on the East Coast. This may mean that mental health services in these states could soon be needed, or may be already needed.
Conclusion
In conclusion, this code provides a quick way to assess potential health care needs throughout the U.S. by examining the number of high-risk/high-stress jobs in each state.
This is based on the research that shows a higher need for mental health services for countries experiencing economic downturn, as well as the need for mental health services for individuals experiencing stress, which may lead to conditions such as depression.
It is recommended that this analysis be used as a rough estimate going forward, and that each state be examined in more detail before policy decisions are made. However, it may act as a good jumping-off point for those attempting to determine how to allocate mental health resources currently or in the near future.
Code
The following packages were used in to make this visualization:
industry <- read_excel("~/Google Drive/R/Projects/health/industry.xlsx", col_names = FALSE)#renaming cols
colnames(industry)[1:13]<-c("STATE","TOT_MAY_2019","TOT_MAR_2020","TOT_APR_2020","TOT_MAY_2020_p","CONSTRUCTION_MAY_2019","CONSTRUCTION_MAR_2020","CONSTRUCTION_APR_2020","CONSTRUCTION_MAY_2020_p","MANUFACTURING_MAY_2019","MANUFACTURING_MAR_2020","MANUFACTURING_APR_2020","MANUFACTURING_MAY_2020_p")
industry<-industry[7:215,]#remove rows with footnotes
industry$STATE <- stringr::str_replace_all(industry$STATE, "[:digit:]", "")
industry$STATE <- stringr::str_replace_all(industry$STATE, "[:punct:]", "")
industry<-na.omit(industry)
industry$STATE<-stringr::str_trim(industry$STATE, side=c("both"))
industry<-industry %>% mutate_at(vars(TOT_MAY_2019,TOT_MAR_2020,TOT_APR_2020,TOT_MAY_2020_p,
CONSTRUCTION_MAY_2019,CONSTRUCTION_MAR_2020, CONSTRUCTION_APR_2020, CONSTRUCTION_MAY_2020_p,
MANUFACTURING_MAY_2019, MANUFACTURING_MAR_2020, MANUFACTURING_APR_2020, MANUFACTURING_MAY_2020_p),
as.numeric)#subsetting into vulnerable job sectors(trade,transportation & utils, education&health, leisure&hospitality)
vuln<-ind_clean[c(1:53),c(1:5,14:17,26:33)]
#Assumption: using March 2020 numbers will be most accurate b/c ppl not laid off yet
tots<-vuln[c(1:53),c(1,3,7,11,15)]
tots$VULN_TOT<-rowSums(tots[,3:5])#col sum to get tot vuln. jobs per state
per<-function(x,y) {
new_data<-(x/y)*100
return(new_data)
}
p<-per(tots$VULN_TOT,tots$TOT_MAR_2020)#calculating vuln. percent of state
tots$VULN_PERCENT<-p #adding percentage to df
#function to change job numbers to thousands so don't have to use label "in thousands" for map legend
scale<-function(x){
x*1000
}
tots[2:6]<-lapply(tots[2:6], scale)#changing scale#########################################
#Maps Prep
geo<-read.csv("~/Google Drive/R/Projects/state_lats.csv", stringsAsFactors = TRUE)#loading state lats and longs
tots<-merge(tots,geo,by.x = c("STATE"), by.y = c("location"))
names(tots)[7]<-"ABBREV" # renaming new col to "abbrev" to remove duplicate col names#usmap
library(usmap)
names(tots)[1]<-"state" #renaming state col b/c plot_usmap() needs a "state" column or fips col.
tots2<-tots
plot_usmap(regions = "states", data=tots2, values = "VULN_TOT")+
theme(legend.position = "right")+
labs(title = "US Potential Mental Health Needs", subtitle = "Share of employment in high-risk/high-stress industries, Mar.2020",caption = "Analysis by Morgana Kinlan-Mickel")+
scale_fill_continuous(low="white", high="blue", name="Jobs")
Data Sources: Data sets used are attached to this notebook. Also available in original form here: https://www.bls.gov/sae/tables/state-news-release/home.htm
For information about the data source, see: https://www.bls.gov/iag/tgs/iag40.htm
Abstract Sources:
“Burnout, Self-Care and COVID-19 Exposure for First Responders.” U.S. Fire Administration, 7 May 2020, www.usfa.fema.gov/coronavirus/behavioral_health/first_responder_self_care.html.
Gewertz, Catherine. “Exhausted and Grieving: Teaching During the Coronavirus Crisis.” Education Week, 28 Apr. 2020, www.edweek.org/ew/articles/2020/04/16/exhausted-and-grieving-teaching-during-the-coronavirus.html.
“Impact of Economic Crises on Mental Health.” World Health Organization Regional Office for Europe, WHO/Europe, 2011, www.euro.who.int/__data/assets/pdf_file/0008/134999/e94837.pdf.
Muro, Mark, et al. “The Places a COVID-19 Recession Will Likely Hit Hardest.” Brookings, Brookings, 17 June 2020, www.brookings.edu/blog/the-avenue/2020/03/17/the-places-a-covid-19-recession-will-likely-hit-hardest/
“What Coronavirus Is Doing to Stressed US Health Workers – and Why It Will Be Felt for Years.” The Guardian, Guardian News and Media, 15 May 2020, www.theguardian.com/us-news/2020/may/15/us-nurses-doctors-mental-health-coronavirus.
Written by Morgana Kinlan-Mickel