In cas-bioinf/covid19retrospective: What the Package Does (One Line, Title Case)

devtools::load_all()
library(tidyverse)
theme_set(cowplot::theme_cowplot())
data <- read_data_for_analysis()

data$patient_data %>% group_by(hospital_id) %>%
  summarise(n_patients = n())

nrow(data$patient_data)
data$breathing_data %>% group_by(patient_id) %>% summarise(n_days_data = n(), .groups = "drop") %>%
  ggplot(aes(x = n_days_data)) + geom_histogram(binwidth = 5)

# disease_levels <- c("Discharged", breathing_levels, "Death")
# outcome_data <- data$patient_data %>% 
#   filter(outcome %in% c("Discharged", "Death")) %>%
#   transmute(patient_id = patient_id, 
#             hospital_id = hospital_id,
#             day = last_record,
#             breathing = as.character(outcome))
#                                                 
# data_for_plot <- data$breathing_data %>% 
#   mutate(breathing = as.character(breathing)) %>%
#   rbind(outcome_data) %>%
#   mutate(breathing = factor(breathing, levels = disease_levels, ordered = TRUE))
# 
# if(any(is.na(data_for_plot$day)) || any(is.na(data_for_plot$breathing))) {
#   stop("Problem")
# }

data$breathing_data %>% ggplot(aes(x = day, y = patient_id, fill = breathing)) + 
  geom_tile() + 
  scale_fill_viridis_d(direction = -1) +
  theme(axis.text.y = element_blank()) +
  facet_wrap(~hospital_id, scales = "free")

data$patient_data %>% filter(hospital_id == "QKuFp") %>% select(first_day_invasive, last_day_invasive)

data$breathing_data %>% group_by(hospital_id, patient_id) %>%
  summarise(first = min(day, na.rm = TRUE)) %>%
  filter(first > 0)

wide <- prepare_marker_data_wide(data)
wide  %>% ggplot(aes(x = day, y = patient_id, fill = took_convalescent_plasma)) + 
  geom_tile() + 
  scale_fill_viridis_d(direction = -1) +
  theme(axis.text.y = element_blank()) +
  facet_wrap(~hospital_id, scales = "free")

wide  %>% ggplot(aes(x = day, y = patient_id, fill = breathing)) + 
  geom_tile() + 
  scale_fill_viridis_d(direction = -1) +
  theme(axis.text.y = element_blank()) +
  facet_wrap(~hospital_id, scales = "free")

data$patient_data %>% filter(is.na(last_record))

data$patient_data %>% group_by(hospital_id) %>% summarise(count = n())

Where do we have measurements for the individual markers (tiles show where we have breathing data, study sites are coded via fill color, crosses show available data).

all_markers <- data$marker_data %>% group_by(marker) %>% summarise(n_obs = n(), .groups = "drop") %>%
  arrange(desc(n_obs)) %>% filter(n_obs > 10) %>% pull(marker)

breathing_data_ordered <- data$breathing_data %>%  mutate(patient_id = factor(patient_id) %>% fct_reorder(hospital_id, .fun = unique))

n_markers <- length(all_markers)
step_size <- 4
for(i in 1:ceiling(n_markers / step_size)) {
  markers_to_show <- all_markers[((i - 1) * step_size + 1):min(n_markers, i * step_size)]
  print(
    data$marker_data %>%
      filter(marker %in% markers_to_show)  %>%
      ggplot(aes(x = day, y = patient_id)) +
        geom_tile(data = breathing_data_ordered, aes(fill = hospital_id)) +
        geom_point(shape = "cross") +
        theme(axis.text.y = element_blank()) +
        scale_fill_brewer(type = "qual", palette = 2, guide = FALSE) +
        facet_wrap(~marker, nrow = 1, scales = "free_x")

  )
}