In this exercise, I will use synthetic data from a cancer clinic to understand how anticancer drugs are administered to patients. To do this, I will use the following three tables:
(Patients_Diagnosis.csv)(Patient_Demographics.csv)(Patient_Treatment.csv)I will assume that patients with multiple diagnoses of the same cancer type are counted once and classified based on the distinct cancer types present in their diagnosis history.
I will assume that if patient has multiple diagnoses, they will be counted in one single category, as Both.
I will assume that the cancer types are mutually exclusive and that each patient can only be classified into one category based on their diagnosis history.
I will assume that patients diagnosed with the same cancer type at different stages are treated as having a single cancer type diagnosis. Stage information is not considered in the cancer group classification, only the cancer type (breast vs colon) is used.
Before answering this question, I will go through these intermediate steps: installing and loading the packages, setting up the working directory, and then exploring the data to determine if there are any missing values, the number of columns and rows, and the variable types for each dataset.
# Install pacman only if it's not already installed
if (!requireNamespace("pacman", quietly = TRUE)) {
install.packages("pacman")
}
# install or load relevant packages
pacman::p_load(
tidyverse, # core tidy tools (includes dplyr, tidyr, readr, etc.)
#gtsummary, # for creating summary tables
here, # for constructing file paths
rmarkdown, # for rendering R Markdown documents
lubridate # for working with dates and times
)# Setup the directory containing the data files
# directory <- here::here(directory)
data_dir <- "~/Desktop/Flatiron_assessment"
# Load the data from the specified directory
patients_diagnosis <- read_csv(file.path(data_dir, "Patient_Diagnosis.csv"))
patients_demographics <- read_csv(file.path(data_dir, "Patient_Demographics.csv"))
patients_treatment <- read_csv(file.path(data_dir, "Patient_Treatment.csv"))
#Explore the data using glimpse() to understand the structure of each dataset and the types of variables they contain
#str(patients_diagnosis)
#str(patients_demographics)
#str(patients_treatment)
#glimpse(patients_diagnosis)
#glimpse(patients_demographics)
#glimpse(patients_treatment)# Create a new variable to classify patients based on their cancer types and keep only the earliest diagnosis date for each patient and cancer type
patients_diagnosis_distinct <- patients_diagnosis |>
group_by(patient_id) |>
mutate(class_diagnosis = case_when(
n_distinct(diagnosis) == 1 ~ first(diagnosis),
n_distinct(diagnosis) > 1 ~ "Both"
)) |>
arrange(diagnosis_date) |>
slice(1)|>
ungroup()
# Calculate the number and proportion of patients in each cancer group
result <- patients_diagnosis_distinct |>
group_by(class_diagnosis) |>
summarise(
n_patients = n(),
proportion = round(n() / nrow(patients_diagnosis_distinct),2)
)
print(result)# A tibble: 3 × 3
class_diagnosis n_patients proportion
<chr> <int> <dbl>
1 Both 5 0.11
2 Breast Cancer 31 0.66
3 Colon Cancer 11 0.23Table 1. Number and proportion of patients diagnosed with each type of cancer
| Type of cancer | Number of patients (N) | Proportion of patients (%) |
|---|---|---|
| Breast cancer | 31 | 66 |
| Colon cancer | 11 | 23 |
| Both | 5 | 11 |
I will assume that for patients with both cancer types, the reference diagnosis date used to calculate the time to treatment is the earliest diagnosis date across all cancer types, regardless of which cancer was diagnosed first.
I will assume that if a patient has multiple treatments, the earliest treatment date is used for the calculation of time to treatment.
# Format the date variable
patients_early_diagnosis <- patients_diagnosis_distinct |>
dplyr::mutate(diagnosis_date = as.Date(diagnosis_date,
format = "%m/%d/%y"))
# Format the date variable and keep only the earliest treatment date for each patient
patients_early_treatment <- patients_treatment |>
mutate(treatment_date = as.Date(treatment_date,
format = "%m/%d/%y"))|>
group_by(patient_id) |>
arrange(treatment_date) |>
slice(1) |>
ungroup()
# Merge the datasets to have the diagnosis date and treatment date in the same dataset
all_data <- patients_early_diagnosis|>
left_join(patients_early_treatment, by = "patient_id")
# Calculate the time from diagnosis to treatment for each patient
all_data <- all_data |>
mutate(time_to_treatment = as.numeric(difftime(treatment_date, diagnosis_date, units = "days")))
# Calculate the mean, median, minimum and maximum time to treatment for each cancer group
time_to_treatment_summary <- all_data |>
group_by(class_diagnosis) |>
summarise(
mean_time = floor(mean(time_to_treatment, na.rm = TRUE)),
median_time = floor(median(time_to_treatment, na.rm = TRUE)),
min_time = floor(min(time_to_treatment, na.rm = TRUE)),
max_time = floor(max(time_to_treatment, na.rm = TRUE))
)
time_to_treatment_summary# A tibble: 3 × 5
class_diagnosis mean_time median_time min_time max_time
<chr> <dbl> <dbl> <dbl> <dbl>
1 Both 8 7 7 11
2 Breast Cancer 5 5 -3 20
3 Colon Cancer 30 4 0 304Table 2. Mean, Median, Minimum and Maximum time to treatment for each cancer group
| Type of cancer | Mean (days) | Median (days) | Minimiun (days) | Maximum(days) |
|---|---|---|---|---|
| Breast cancer | 5 | 5 | -3 | 20 |
| Colon cancer | 30 | 4 | 0 | 304 |
| Both | 8 | 7 | 7 | 11 |
# Identify the first-line treatment regimens for each cancer group
first_line_treatments <- all_data |>
filter(!is.na(treatment_date)) |>
group_by(class_diagnosis) |>
summarise(first_line_regimens = list(unique(drug_code)))
first_line_treatments# A tibble: 3 × 2
class_diagnosis first_line_regimens
<chr> <list>
1 Both <chr [2]>
2 Breast Cancer <chr [3]>
3 Colon Cancer <chr [3]> Table 3. First-line treatment regimens for each cancer group
| Types of cancer | First-line regimens |
|---|---|
| Breast Cancer | (A, B, C) |
| Colon Cancer | (A, D, B) |
| Both | (C, A) |
# Keep only the patient's first diagnosis
patients_first_diagnosis <- patients_diagnosis |>
group_by(patient_id) |>
arrange(diagnosis_date) |>
slice(1) |>
ungroup()
# Merge the demographic data with the diagnosis data to have all relevant information in one dataset and caluclated age at diagnosis
demographics_diagnosis <- patients_demographics |>
left_join(patients_first_diagnosis, by = "patient_id") |>
mutate(diagnosis_date = ymd(diagnosis_date),
birth_date = ymd(birth_date),
age = floor(as.numeric(difftime(diagnosis_date, birth_date, units = "days")) / 365.25))
# Generate the table using gtsummarry package
tble <- demographics_diagnosis |>
select(Age = age, Sex = birth_sex, Region=region, Grade = stage_dx, diagnosis) |>
gtsummary::tbl_summary(
by = diagnosis,
statistic = gtsummary::all_categorical() ~ "{n} ({p}%)",
percent = "column",
missing = "no"
) |>
gtsummary::add_overall(last = TRUE) |>
gtsummary::modify_header(label ~ "**Types of cancer**") |>
gtsummary::bold_labels()
tble| Types of cancer | Breast Cancer N = 321 |
Colon Cancer N = 151 |
Overall N = 471 |
|---|---|---|---|
| Age | 47 (33, 62) | 35 (33, 73) | 45 (33, 63) |
| Sex | |||
| F | 27 (84%) | 10 (67%) | 37 (79%) |
| M | 5 (16%) | 5 (33%) | 10 (21%) |
| Region | |||
| Mid-west | 12 (38%) | 1 (6.7%) | 13 (28%) |
| Northeast | 8 (25%) | 5 (33%) | 13 (28%) |
| South | 7 (22%) | 6 (40%) | 13 (28%) |
| West | 5 (16%) | 3 (20%) | 8 (17%) |
| Grade | |||
| I | 7 (22%) | 2 (13%) | 9 (19%) |
| II | 12 (38%) | 4 (27%) | 16 (34%) |
| III | 9 (28%) | 2 (13%) | 11 (23%) |
| IV | 4 (13%) | 7 (47%) | 11 (23%) |
| 1 Median (Q1, Q3); n (%) | |||