demo.Rmd

---
title: "Using R to derive robust insights from real-world healthcare data"
subtitle: "Workshop given at R/Pharma 2022"
output: github_document
---

```{r, include = FALSE}
# Install packages (utilizing renv)
source(here::here("setup/packages.R"))

# Load libraries
library(tidyverse)
library(pointblank)
library(survival)
library(gtsummary)
library(survminer)

# Load the robustrwd package
devtools::load_all(here::here("robustrwd"))

# Turn on messaging
be_noisy()
```

# Load data

* Delivery = 02
* ETL version = 02

```{r, results = "hide", message=FALSE}
tables_02 <- receive_delivery_02()

tables_02_etl_02 <- tables_02 %>%
  etl_02()
```

# QC

## Patients

```{r, echo = FALSE}
# Run on the patients data
patients_interrogation <- tables_02_etl_02$patients %>%
  create_agent(
    tbl_name = "Patients",
    label = "Patients data (Post ETL 02)"
  ) %>%
  expectations_patients() %>%
  interrogate()

# Look at tidy results
patients_interrogation %>%
  tidy_interrogation() %>%
  knitr::kable()

# Summarise into single pass / fail

patients_checks_pass <- patients_interrogation %>%
  tidy_interrogation() %>%
  pull(any_fail) %>%
  any() %>%
  `==`(FALSE)
```

## Inpatient

```{r, echo = FALSE}
inpatient_interrogation <- tables_02_etl_02$inpatient %>%
  create_agent(
    tbl_name = "Inpatient",
    label = "Inpatient data (Post ETL 02)"
  ) %>%
  expectations_inpatient() %>%
  interrogate()

# Look at tidy results
inpatient_interrogation %>%
  tidy_interrogation() %>%
  knitr::kable()

# Summarise into single pass / fail

inpatient_checks_pass <- inpatient_interrogation %>%
  tidy_interrogation() %>%
  pull(any_fail) %>%
  any() %>%
  `==`(FALSE)
```

## ORPP

```{r, echo = FALSE}
orpp_tbl <- tables_02_etl_02$patients %>%
  add_orpp_inpatient(inpatient_tbl = tables_02_etl_02$inpatient) %>%
  add_orpp_prescription(prescription_tbl = tables_02_etl_02$prescription)

orpp_interrogation <- create_agent(orpp_tbl,
  tbl_name = "ORPP cohort",
  label = "Patient level table"
) %>%
  expectations_orpp() %>%
  interrogate()

# Print result
orpp_interrogation %>%
  tidy_interrogation() %>%
  knitr::kable()

# Summarise into single pass / fail

orpp_checks_pass <- orpp_interrogation %>%
  tidy_interrogation() %>%
  pull(any_fail) %>%
  any() %>%
  `==`(FALSE)
```


# Cohort


## Attrition Table

```{r, echo = FALSE}
cohort_tbl <- orpp_tbl %>%
  filter(
    race_cd %in% c("White", "Black"),
    sex_ident_cd == 'Male',
    diabetes == TRUE,
    inpatient_payment_median >= 10000,
    prescription_rx_cost_median > 20,
    survival_years >= 18
  )

attrition_table <-
  create_attrition(
    orpp_tbl,
    "Race is white or black" = race_cd %in% c("White", "Black"),
    "Has Diabetes" = diabetes == TRUE,
    "Is Male" = sex_ident_cd == "Male",
    "Has at least 1000 inpatient costs" = inpatient_payment_median >= 10000,
    "Median prescription cost is > 50" = prescription_rx_cost_median > 20,
    "18 years of age or older" = survival_years >= 18
  )
```

```{r, echo = FALSE}
# Show the attrition table
attrition_table %>%
  knitr::kable()
```

## Attrition Plot

```{r, echo = FALSE, fig.width = 8, fig.height = 6}
# Plot an attrition chart
attrition_table %>%
  plot_attrition()

cohort_tbl <- orpp_tbl %>%
  filter(
    race_cd %in% c("White", "Black"),
    diabetes == TRUE,
    sex_ident_cd == "Male",
    inpatient_payment_median >= 10000,
    prescription_rx_cost_median > 20,
    survival_years >= 18
  )
```

# Analyses

## Survival

```{r, fig.width = 8, fig.height = 7}
# Fit the survival model
#  Follow-up time is survival_years
#  Status indicator is death_observed
#  Stratify results by sex_cd

fit <- survfit(Surv(survival_years, death_observed) ~ race_cd,
  data = cohort_tbl
)

# Create the final Kaplan-Meier curve
ggsurvplot(fit,
           conf.int = TRUE,
           surv.median.line = "hv") +
  labs(title = "Survival of patients from birth to death",
       subtitle = "Using CMS Medicare claims synthetic public use files",
       caption = glue::glue("Patients = {nrow(cohort_tbl)}\nPatients checks pass = {patients_checks_pass}\nInpatient checks pass = {inpatient_checks_pass}\nORPP checks pass = {orpp_checks_pass}"))
```