Introduction

Rpubs link: https://rpubs.com/SamDP222/773336

• Cardiovasvular disease (CVD) is currently the leading cause of death world wide. Accounting for 31% of all deaths world wide(World Health Organization: WHO, 2019). It is important for medical profesionals to be able to understand the disease as best as possible.
• It is for this reason that the investigation undertaken in this report will examine the survivability of heart attacks among smokers and non smokers, to help improve understanding on the mortality rates of CVD patients.

Image(Physicians Premier ER,2019)

Problem Statement

• Does smoking affect the survival rate of having a heart attack?
• It is commonly known that being a long time smoker will increase your chances of having a heart attack, but will it also effect your chance of survivng one(ROTH and SHICK, 1958)?
• A Chi-square test will be conducted on the data to see if there is a statistically significant relationship and dying from a heart attack.

Data

• The data was obtained from kaggle.com (https://www.kaggle.com/andrewmvd/heart-failure-clinical-data). It is an open data set that is free to share and adapt under Attribution 4.0 International law(Creativecommons,2018). The data was created by (Chicco and Jurman,2020) and contains 13 variables and 299 rows. The data contains information gathered on heart attack victims.
• The sampling method is not certain. However in the research paper published by (Chicco and Jurman,2020) they state that they took the data from 299 patients in 2015 which could suggest simple random sampling was used.

Data Cont.

• The variables that will be analysised are ‘DEATH_EVENT’ and ‘smoking’. ‘DEATH_EVENT’ records if the patient survived the heart attack or not. These coenside with two values ‘0’ indicating patient survival and ‘1’ indicating patient death. This variable was changed to a factor with two levels, labelled ‘Alive’ and ‘Dead’ respectivley.
  
• The variable ‘smoking’ was similar. This variable also contained two values. ‘0’ coensided with a non-smoking patient and ‘1’ for a patient who did. This variable was also set as a factor with two levels, labelled ‘Non-smoker’ and ‘Smoker’.
• Setting the variables as labelled factors will make the data easier to process as well as making the output for the statistical tests more readable and easier to understand.

HA$age <- round(HA$age, digits = 0)
HA$DEATH_EVENT <- factor(HA$DEATH_EVENT,
                         levels = c(0,1),
                         labels = c("Alive", "Dead")
                         )
HA$smoking <- factor(HA$smoking,
                         levels = c(0,1),
                         labels = c("Non-smoker","Smoker")
                         )

Checking for NA values

• Both columns were checked for NA values. They were found to be non existant.
• NA values could impact the results of the test and cause issues in the computation of the statistical values. It is for this reason that it is important to make sure that the values are free of them before any statistical tests proceed.

a<-is.na(HA$DEATH_EVENT) %>% table 
b<-is.na(HA$smoking) %>% table

knitr::kable(a,caption = 'NA values in "DEATH_EVENT"" variable')

knitr::kable(b,caption = 'NA values in "smoking"" variable')          

Descriptive Statistics

Dimensions of the total data set

dim(HA)

## [1] 299  13

Table <- table(HA$smoking,HA$DEATH_EVENT) 
TableMargin <- Table %>% addmargins()
knitr::kable(TableMargin, caption = 'Value counts of the patients within both the catagories.')

propTable <- Table %>% prop.table
knitr::kable(propTable, digits = 3, caption = 'Proportion of patients within each category.')

Visualisation

barplot(propTable, main ="Heart attack survival by smoker status",
        ylab="Proportion of smokers"
        ,ylim=c(0,1),legend=rownames(propTable),beside=TRUE,
        args.legend=c(x ="topright",horiz=TRUE,
        title="Smoker status"),xlab="Survival", col = c('BLUE','RED'))

Hypothesis Testing

• \(H_0:\) There is no association between smoker status and whether a person will survive a heart attack.
• \(H_a:\) There is an association between smoker status and whether a person will survive a heart attack.
• Assumption: No more than 25% of expected cell counts are below 5.
• Decision Rules: – Reject \(H_0:\) if the p-value <0.05 (\(\alpha\) significance level) – If not, fail reject \(H_0:\)
• Conclusion: The test is statistically significant if we reject \(H_0\).
• Otherwise, the test is not statistically significant.

chitest<- chisq.test(Table)
chitest

## 
##  Pearson's Chi-squared test with Yates' continuity correction
## 
## data:  Table
## X-squared = 0.0073315, df = 1, p-value = 0.9318

Hypthesis Testing Cont.

• Chi test mathematical equation:

\[\chi^{2}=\sum\frac{(Obs - Exp)^{2}}{Exp} \]

knitr::kable(chitest$expected, caption = 'Assumption test')

As seen from the assumption test no cell counts were below five. The Assumption of no more than 25% of expected cell counts are below 5 was met.

– \(\chi^{2}\)=0.0073315
– df = 1
– p-value = 0.931
As the p-value is not <0.05 we fail to reject the null hypothesis. This causes the conclusion of there being no significant difference between smokers and non smokers in their chance to survive a heart attack to be made.

Discussion

• The investigation found that there was no significant difference it the survival rate of having a heart attack between smokers and non smokers. This is quite suprising as previous studies have found a strong correlation between smoking status and the risk of a heart attack(Roth and Shick,1958).
• The stength of the chi tests lies in its ability to compare nominal data robustly with no assumptions about the distribution of the population having to be made. However, it is difficult to say the level of validity this investigation holds as it is uncertain what sampling method was used in the collection of the data. Chi squares assume random sampling so without knowing this it is hard to determine how acurate the investigation was. Also, as Chi squares rely heavily on the sample size being large. It may be worth replicating the test with a sample that is known to be randomly selected, as well as with a much larger sample population in order to increase the accuracy of the test results.
• Further testing might involve investigating smoking effects on known the influences of heart attack survivorship. This could include testing for the relationship between smoking and the presence of creatinine phosphokinase(CPK) which can signify injury or stress to muscle tissue, the heart, or the brain(Mount Sinai Health System, 2017). This test might give a better understanding on the relationship between smoking on the survival of a heart attack.

References

• Chicco, D. and Jurman, G. (2020). Machine learning can predict survival of patients with heart failure from serum creatinine and ejection fraction alone. BMC Medical Informatics and Decision Making, 20(1).

• Healthline. (2018). Tobacco Allergy: Can You Be Allergic to Cigarette Smoke? [online] Available at: https://www.healthline.com/health/allergic-to-cigarette-smoke.

• Mount Sinai Health System. (2017). Creatine phosphokinase test Information | Mount Sinai - New York. [online] Available at: https://www.mountsinai.org/health-library/tests/creatine-phosphokinase-test.

• Physicians Premier ER. (2019). Life-Saving Facts About Heart Attack | Calallen, TX Emergency Room. [online] Available at: https://mdpremier.com/life-saving-facts-about-heart-attack/.

• World Health Organization: WHO (2019). Cardiovascular diseases. [online] Who.int. Available at: https://www.who.int/health-topics/cardiovascular-diseases/#tab=tab_1.