COVID-19 and Health Care

Maddie Kentch, Harun Khan, Gavin Mullin, Sarah Ward

PSYC 500 final project

Introduction

COVID-19 has ravaged countries around the globe by shutting down economies and causing the deaths of over a million people worldwide. Because of this, our researchers wanted to know whether the quality of healthcare systems impacted the total deaths per million from COVID-19.

We asked three questions to determine whether the above was true:

Does Per Capita Spending by Domestic Governments on healthcare impact total deaths per million from COVID-19?
Does GDP Share of healthcare impact total deaths per million from COVID-19?
Do countries with Universal Healthcare have a significantly different mean total deaths per million than countries without Universal Healthcare?

Scientific Questions and Hypothesis

What is the relationship between healthcare share of GDP and total deaths per million from COVID-19?
What is the relationship between per capita spending on healthcare by domestic governments and total deaths per million from COVID-19?
Do countries with a Universal Healthcare system have a lower mean deaths per million from COVID-19 compared to countries without a Universal Healthcare system

We hypothesize countries with a Universal Healthcare system will have a lower mean deaths per million from COVID-19 than countries without Universal Healthcare system.

Data

Data Curation and Ethics

Data was collected from three sources: The World Bank, Oxford Martin School, and the New York State Department of Health.

The World Bank data contained data from countries indicating Per Capita spent on Healthcare and GDP share of healthcare by country
The Oxford Martin School contained data detailing the number of covid deaths per million by country
The New York State Department of Health contained a list of countries that are classified as having Universal Healthcare

All data files used contained only information regarding healthcare expenditure by countries, their covid deaths, and healthcare system.

None of the information in these datasets contained information that could be used to identify ‘participants’ as the participants were whole countries.

The information presented does not violate privacy and contains no legality concerns while also substantially benefiting the public. Analysis can be used to better understand how to effectively combat emerging infectious diseases

Data Preparation

Data Preparation and Exploration

Preparing Data for usability in Python:

Data from the Oxford Martin School was in a format that was ready to be converted into a workable data frame
The list of countries with Universal Healthcare from the New York State Department of Health were put into a data frame manually
Data from the World Bank was inherited in an unusable format that required rearrangement in terms of selecting specific columns, eliminating undesirable rows, and renaming columns so they could be worked with (On the Right shows the data in its original format).
- Columns were unnamed making the data difficult to work with until it was rearranged
- This was true for both datafiles from the World Bank that contained per capita spending on healthcare and healthcare share of GDP

Data Exploration

Data Frames and their Columns:

Data frame: df
- This data came from the Oxford Martin School
- Contains two columns: "Country" and "total_deaths_per_million"
- Python code was used to select rows that contained data from the most recent date (11-11-2020)
- Shape and Size: 212 x 2 & 424

2. Data frame: df1

- This data came from the World Bank
- Contains two columns: "Country" and "Per Capita Paid"
- The GDP share values are from 2017
- Shape and Size: 264 x 2 & 528

3. Data frame: df2

- This data came from the World Bank
- Contains two columns: "Country" and "GDP Share"
- The Per Capita Paid values are from 2017
- Shape and Size: 264 x 2 & 528

4. Data frame: df_med_system

- This data frame was manually made using information from the New York State Department of Health
- Contains two columns: "Country" and "total_deaths_per_million". The column of "Country" only contains countries that have implemented Universal Healthcare
- Shape and Size: 32 x 2 & 64

5. Data frame: df_med_system2

This data frame was manually made using information from the New York State Department of Health
Contains two columns: "Country" and "total_deaths_per_million". The column of "Country" only contains countries that have not implemented Universal Healthcare
Shape and Size: 119 x 2 & 238

Data Frame Merging/Rearranging

Three of the data frames were merged with data frame df that contained the column of total deaths per million:

df1: was merged on the column 'Country' with df creating a new data frame that contains the following columns:

"Country", "total_deaths_per_million", and "Per Capita Paid".
This was done in order to compare and graph deaths per million and Per Capita Paid

df2: was merged on the column 'Country' with df creating a new data frame that contains the following columns:

"Country", "total_deaths_per_million", and "GDP Share"
This was done in order to compare deaths per million and GDP share

df_med_system & df_med_system2: were both merged on the column "Country" with df creating new data frames that contained the following columns

"Country" and "total_deaths_per_million"
This was done in order to compare the mean of df_med_system to the mean of df_med_system2 to compare healthcare systems

Data Summary Statistics

Data Visualization

The mean death rate per million from COVID-19 in countries with Universal Healthcare was higher than mean death rate per million from COVID-19 in countries with Non-Universal Healthcare.

A scatterplot of per capita spent on healthcare and deaths per million reveals a slight positive relationship between the two variables, meaning that as spending on healthcare increases, so do COVID-19 deaths.

The scatterplot of GDP share and COVID-19 deaths per million shows a stronger positive relationship than the relationship between per capita and deaths per million, meaning that higher GDP share results in more COVID-19 deaths than does higher spending per capita.

Model Building and Validation

Normality check of GDP Share and Per Capita Spending

The empirical cdf of per capita spending on healthcare does not match up with the theoretical distribution of per capita spending. This suggests per capita spending is not normally distributed

The empirical cdf for Healthcare GDP share fits nicely onto theoretical cdf for Healthcare GDP. This suggests that the distribution for Healthcare GDP share is normally distributed

Linear Regression Analysis of GDP Share and Per Capita Spending

The pearson coefficient (r = .30) and the linear regression done indicates that the relationship between Per Capita Spent on healthcare by domestic governments and total deaths per million from COVID-19 is a medium strength positive relationship.

This means as per capita spending increases so do the number of total deaths per million from COVID-19

The slope value of approximately .04 means for every 1 increase in per capita spent on healthcare there is a corresponding .04 increase in the total number of total deaths per million from COVID-19

The pearson coefficient of r = .34 and linear regression model indicates a medium strength relationship between GDP share and total deaths per million from COVID-19.

This means as GDP share increases so does the total number of deaths per million from COVID-19

The slope value of approximately 32.44 means for every 1 increase in healthcare share of GDP there is a 32.44 increase in deaths per million from COVID-19

Correlation Hypothesis Test

There is a .17% chance that the observed correlation (r = .30) would occur if the true correlation were 0.

Because the p-value (.0017) is less than the alpha level (.05) we reject the null hypothesis and conclude that the correlation between per capita spending on healthcare and total deaths per million from COVID-19 is statistically significant

There is a .02% chance that the observed correlation (r = .34) would occur given that the true correlation were equal to 0.

Because the p-value (.0002) is less than the alpha level (.05) we reject the null hypothesis and conclude that the correlation between GDP share and total deaths per million from COVID-19 is statistically significant

Bootstrap Hypothesis Test for Healthcare System

There is a 0% chance of getting the observed mean difference if the true mean difference in total COVID-19 deaths per million between healthcare systems (Universal versus Non-Universal) is zero. Because the p-value is less than alpha (.05) our results are statistically significant.

Therefore we reject the null hypothesis and conclude that countries without Universal Healthcare have less deaths per million from COVID-19 on average than countries with Universal Healthcare.

Discussion

The objective of this study was to understand the relationship between countries spending per capita on healthcare as well as their GDP share of healthcare, on the total number of COVID-19 deaths per million. The type of healthcare system, universal or non-universal, was also compared. The hypothesis, as stated above, was that countries with universal healthcare systems would have lower numbers of COVID-19 deaths per million than those countries with non-universal healthcare systems.

Data was obtained from three separate sources: The World Bank, Oxford Martin School, and the New York State Department of Health. The data sets were merged to compare variables: country, total deaths per million, GDP share of healthcare, healthcare spending per capita paid, and universal versus non-universal healthcare systems.

After checking normality of the data, a linear regression analysis was performed on GPD share and deaths per million as well as per capita spending and deaths per million. The results are summarized below.

Following the linear regression analyses, a correlational hypothesis test was run on both variables, showing that the previous findings were significant.

Finally, a bootstrap hypothesis test was run to explore the two types of healthcare systems and their relationship with COVID-19 deaths per million. Again, the results are summarized below.

In short, it was found that as GDP share on healthcare and healthcare spending per capita increased, they both predicted an increase in the total number of COVID-19 deaths per million, with countries operating on a universal healthcare system having higher deaths than those with non-universal systems.

There are three main results/insights of the present research study:

Healthcare’s share of GDP predicts the number of total deaths per million from COVID-19. The higher the GDP share the higher number of total deaths per million. The correlation value of r = .34 between GDP share of healthcare and total deaths per million was statistically significant with a p-value of .0001. The relationship between the two values is a positive medium strength relationship.
Government per capita spending on healthcare predicts the number of total deaths per million from COVID-19. The higher per capita spending the higher number of total deaths per million. The correlation value of r = .30 between per capita spending and total deaths per million was statistically significant with a p-value of .0015. The relationship between the two values is a positive medium strength relationship.
Countries with Universal Healthcare systems have, on average, higher total deaths per million from COVID-19 compared to countries without Universal Healthcare systems. Countries with Universal Healthcare had a mean total death per million rate of 210.8 and countries with Non-Universal Healthcare had a mean total death per million rate of 166.07. The p-value of 0 indicates that the empirical mean difference between death rates in Universal countries and countries without Universal Healthcare is statistically significant. Countries without Non-Universal Healthcare had significantly less deaths per million on average than countries with Universal Healthcare.

There were several limitations to the present study:

- There is an inability to control for confounding variables that would affect the number of COVID-19 deaths. Examples of such variables include government restrictions implemented, lack/availability of testing and distribution of populations, such as urban versus rural locations.
- It is not guaranteed that governments are reporting all of the COVID-19 deaths, there may be many missed cases never confirmed. The cause of death is another factor that can have variation, as it's not clear whether the death counts used in the data had COVID-19 listed as primary cause of death or complications caused by COVID-19.
- The data did not include government spending in healthcare system specific to COVID-19; therefore, it may be that countries with lower death counts (per million) designated more of their overall healthcare spending on fighting COVID-19, even if their total healthcare spending per capita/GDP share was less.

Implications of the results and actions to be taken following this study:

- This study should be used as a starting point to analyze other variables that better explain the relationship between countries and their total deaths per million from COVID-19. It is clear that countries cannot solely rely on their health care systems as a means to fight emerging infectious diseases; but rather, countries should take additional measures to ensure lower total deaths per million. These additional measures could be studied by examining the relationship between COVID-19 deaths and level of governmental lockdowns (measured by length of lockdown, enforcement of lockdown, etc.), COVID-19 deaths and population distribution (rate of infection in urban versus rural communities), and how effective governments are in reporting COVID-19 cases.

Page updated

Google Sites

Report abuse