README.qmd

---
title: "MOVER dataset: EPIC complications"
format: gfm
editor: visual
execute:
  echo: true
  warning: false
---

```{r}
library(tidyverse)
theme_set(theme_bw())
source("01_read_clean.R")
```


The post-operative complications dataset includes `r n_distinct(complications_orig$MRN)` patients who had `r n_distinct(complications_orig$LOG_ID)` surgeries.

The whole dataset includes 39685 patients who had 64354 surgeries, a discrepancy of 140 surgeries. Which is small considering the overall size. I wonder if these patients could have died without a recorded complication?

# Number of surgeries that resulted in any complication

```{r}
complications %>% 
  count(any_complication, name = "n_patients") %>% 
  knitr::kable()
```

# Number of complications per surgery

```{r}
complications %>% 
  count(n_complications, name = "n_patients") %>% 
  knitr::kable()
```

# Types of complications

```{r}
complications_all %>% 
  filter(comp_abbr != "None") %>% 
  #filter(comp_abbr != "Post-op AN") %>% 
  ggplot(aes(y = fct_rev(fct_infreq(comp_abbr)))) +
  geom_bar() +
  facet_wrap(~if_else(comp_abbr == "Post-op AN", "Big x axis scale", "Small x axis scale"),
             scales = "free", ncol = 1)
```

By far the most common complication is "AN Post-op Complications", which I've shortened to "Post-op AN" for plotting/brevity purposes. It seems to cover everything from "None" to "Death":

```{r}
complications_all %>% 
  filter(comp_abbr == "Post-op AN") %>% 
  count(comp_abbr, complication, sort = TRUE) %>%
  slice(1:20) %>% 
  knitr::kable()
```