Debora Oliveira
The Chi-Square test of independence, also called Pearson's chi-square test is used to determine if there is a significant relationship between two nominal (categorical) variables. It offers a great advantage once it is a nonparametric statistical procedure that does not require a normal distribution[1]. Thus, the result determines whether the association among variables is statistically significant.
The New York Heart Association (NYHA) Functional Classification[2] provides a simple way of classifying the extent of heart failure. It places patients in one of four categories based on how much they are limited during physical activity; the limitations/symptoms are in regard to normal breathing and varying degrees in shortness of breath and/or angina.
The Kansas City Cardiomyopathy Questionnaire (KCCQ) is a self-administered questionnaire developed to measure the patient’s perception of their health status, which includes heart failure symptoms, impact on physical and social function, and how their heart failure impacts their quality of life[3].
To run this project you need the following libraries:
library(redcapAPI)
library(dplyr)
library(stats)
library(corrplot)This project uses data from REDCap. The package redcapAPI contain functions such as exportfields(), exportevents() that allows to choose between different events, or moments, in a clinical trial. For this project only two functions will be important: redcapConnection(), that requires url and token, exportRecords() to bring patient's records. However, an alternative to redcapAPI would be a csv file with data easily exported from REDCap.
source("token.txt")
rcon <- redcapConnection(url=url, token=token)
rm(token)
##########
#Calling our variables
##########
vector_baseline <- c("record_id","icc_pd_adm_nyha_q01")
event_baseline <- "t0_arm_1"
vector_30days_after <- c("record_id","icc_proms_kccq_score")
event_30days_after <- "30_dias_arm_1"
data_baseline <- exportRecords(rcon, factors = FALSE,
fields = vector_baseline, events = event_baseline)
data_after_treat <- exportRecords(rcon, factors = FALSE,
fields = vector_30days_after, events = event_30days_after)For the project, the package dplyr is mainly used during the cleaning and data rearrangement. Thus, the function merge() allows to merge two dataframes combining them through the patient's record_id.
patient_outcome<- merge(data_baseline,data_after_treat, by.x = "record_id", by.y = "record_id") Another very useful function from dplyr package would be select() since exportRecords() brings few not called information.
patient_outcome <- select(patient_outcome, "record_id", "icc_pd_adm_nyha_q01", "icc_proms_kccq_score") The package still provide another two important functions: mutate_all() and filter(). The first used to turn the dataframe into numeric while, the second allows to filter missing data.
patient_outcome <- mutate_all(patient_outcome, funs(as.numeric), 1:3)
patient_outcome <- filter(patient_outcome, patient_outcome[,2] < 100,
patient_outcome[,2] != 'NA', patient_outcome[,3] != 'NA')The result of this entire cleaning process will be a two column table with both variables to compare.
## record_id icc_pd_adm_nyha_q01 icc_proms_kccq_score
## 1 1 0 92.1875
## 2 103 2 25.0000
## 3 1083 3 100.0000
## 4 1086 3 100.0000
## 5 1107 1 96.8750
## 6 1130 3 87.5000
The package stats allows to get quartiles, a contingency table and the chi-square. Firstly, the function quartile() takes a sequence of values from a vector, sorting it (ascending) and then produces sample quantiles corresponding to a given probability:
checking_quartiles <- quantile(patient_outcome$icc_proms_kccq_score,c(0.25,0.5,0.75), na.rm = TRUE)
patient_outcome$quartil <- cut(patient_outcome$icc_proms_kccq_score,
breaks=quantile(patient_outcome$icc_proms_kccq_score,probs=seq(0,1, by=0.25),
na.rm = TRUE), labels=c("Q1","Q2","Q3","Q4")) The result of this process:
## icc_pd_adm_nyha_q01 quartil
## 1 0 Q4
## 2 2 Q1
## 3 3 Q4
## 4 3 Q4
## 5 1 Q4
## 6 3 Q3
Once the data is clean and fits a xtabs() function, the following step is to create a contingency table:
outcome_table <- xtabs(~icc_pd_adm_nyha_q01+quartil, data=patient_outcome_clean) The contigency table will look like:
## quartil
## icc_pd_adm_nyha_q01 Q1 Q2 Q3 Q4
## 0 0 0 0 2
## 1 0 6 5 3
## 2 13 10 5 6
## 3 9 7 12 10
Lastly but not the least, the Pearson's chi-square test from the function chisq.test() will return de values for X-square, df and p-value.
Chi-square test examines whether rows and columns of the contingency table above are statistically significantly associated.
chi_data <- chisq.test(outcome_table)The last part of this project requires the package corroplot to visualize Pearson residuals. Hence, in the corroplot table:
- Positive residuals will be in blue. Positive values in cells specify an attraction (positive association) between the corresponding row and column variables.
- Negative residuals will be in red. This implies a repulsion (negative association) between the corresponding row and column variables.
corrplot(chi_data$residuals, is.cor = FALSE)
contrib <- 100*chi_data$residuals^2/chi_data$statistic
round(contrib, 3)## quartil
## icc_pd_adm_nyha_q01 Q1 Q2 Q3 Q4
## 0 2.745 2.870 2.745 26.671
## 1 19.215 8.222 3.529 0.191
## 2 13.079 0.766 7.912 3.023
## 3 0.144 4.751 3.612 0.526
# Visualize the contribution
corrplot(contrib, is.cor = FALSE)
http://www.sthda.com/english/wiki/chi-square-test-of-independence-in-r
- References
[1] http://www.sthda.com/english/wiki/chi-square-test-of-independence-in-r [2] NYHA https://en.wikipedia.org/wiki/New_York_Heart_Association_Functional_Classification [3] MEDICAL DEVICE DEVELOPMENT TOOL (MDDT) QUALIFICATION DECISION SUMMARY FOR KANSAS CITY CARDIOMYOPATHY QUESTIONNAIRE (KCCQ). https://www.fda.gov/media/108301/download