---
title: "Fairlabs Dashboard"
header-includes:
  - \usepackage{comment}
output:
  flexdashboard::flex_dashboard:
    css: style.css
    theme: cosmo
    orientation: rows
    vertical_layout: scroll
    source_code: embed
---
```{r, include=FALSE}
library(flexdashboard)
library(tidyverse)
library(plotly)
library(knitr)
library(janitor)
library(reshape2)
require(scales)
library(glue)
library(gtsummary)
library(htmltools)
library(lubridate)
```

```{r}
# cleans raw data based on user defined input columns

clean_data <- function(df, input){

  # Pull user defined columns

  # dates
  delivery_date_in <- input[input$column_out=='delivery_date','column_in']
  cps_reporting_date_in <- input[input$column_out=='cps_reporting_date','column_in']
  uds_collection_date_in <- input[input$column_out=='uds_collection_date','column_in']
  uds_test_in <- input[input$column_out=='uds_test','column_in']
  maternal_birth_date_in <- input[input$column_out=='maternal_birth_date','column_in']
  # static QI intervention date from dictionary 
  intervention_date_in <- as.Date(input[input$column_out=='intervention_date','column_in'])

  # filled if non-THC and THC are stored as individual drug columns
  non_thc_col <- strsplit(input[input$column_out=='non_thc_cols','column_in'], split ='\\\\t')[[1]]
  thc_col_in <- input[input$column_out=='thc_col','column_in']
  # filled if non-THC and THC are stored as true/false 
  non_thc_detect_in <- input[input$column_out=='non_thc_detect','column_in']
  thc_detect_in <- input[input$column_out=='thc_detect','column_in']

  maternal_age_in <- input[input$column_out=='maternal_age','column_in']
  maternal_race_in <- input[input$column_out=='maternal_race','column_in']

  order_indication_in <- input[input$column_out=='order_indication','column_in']
  # Used to determine which order indication is for non-THC drug use
  ord_indict_non_thc_in <- strsplit(input[input$column_out=='ord_indict_non_thc','column_in'], split ='\\\\t')[[1]]

  # Checks if each column exist, and format for summary and visual uses
  if (delivery_date_in == "") {
    df$delivery_date <- NA
    df$delivery_qtr <- NA
  } else {
    df$delivery_date <- as.Date(df[[delivery_date_in]])
    df$delivery_qtr <- floor_date(as.Date(format.Date(df$delivery_date, '%Y-%m-01')), unit = "quarter")
  }

  if (cps_reporting_date_in == "") {
    df$cps_reporting_date <- NA
    df$cps_report <- NA
  } else {
    df$cps_reporting_date <- as.Date(df[[cps_reporting_date_in]])
    df$cps_report <- if_else(is.na(df$cps_reporting_date), FALSE, TRUE)
  }

  if (is.na(intervention_date_in)) {
    df$pre_post_QI <- NA
  } else {
    df <- df %>%
      mutate(pre_post_QI = factor(if_else(delivery_date >= intervention_date_in, 'Post', 'Pre'), 
                                  levels = c('Pre', 'Post')))
  }

  if ((uds_collection_date_in == "" & uds_test_in != "")) {
    df$uds_collection_date <- NA
    df$uds_test <- df[[uds_test_in]]
  } else if (uds_collection_date_in != "") {
    df$uds_collection_date <- as.Date(df[[uds_collection_date_in]])
    df$uds_test <- if_else(is.na(df$uds_collection_date), FALSE, TRUE)
  } else {
    df$uds_collection_date <- NA
    df$uds_test <- NA
  }

  if (thc_detect_in != "" & thc_col_in == "") {
    df$thc_detect <- df[[thc_detect_in]]
  } else if (thc_col_in != "") {
    df$detected_tetrahydrocannabinol <- df[[thc_col_in]]
    df$thc_detect <- if_else(df$detected_tetrahydrocannabinol==1, TRUE, FALSE)
  } else {
    df$thc_detect <- NA
  }

  if (non_thc_detect_in != "" & is_empty(non_thc_col)) {
    df$non_thc_detect <- df[[non_thc_detect_in]]
  } else if (!is_empty(non_thc_col)) {
    df <- df %>%
      mutate(non_thc_detect = if_else(rowSums(df[non_thc_col]) > 0, TRUE, FALSE))
  } else {
    df$non_thc_detect <- NA
  }

  if (maternal_age_in != "") {
    df$maternal_age <- df[[maternal_age_in]]
    df <- df %>%
      mutate(age_group = factor(case_when(maternal_age < 25 ~ 'Under 25', 
                                maternal_age < 30 ~ '25 - 30', 
                                maternal_age < 34 ~ '30 - 34', 
                                TRUE ~ 'Over 34'), levels=c('Under 25','25 - 30','30 - 34','Over 34')))
  } else if (maternal_birth_date_in != "") {
    df$maternal_birth_date = as.Date(df[[maternal_birth_date_in]])
    df$maternal_age <- floor(as.numeric(difftime(delivery_date, maternal_birth_date, units = 'days'))/365.25)
    df <- df %>%
    mutate(age_group = factor(case_when(maternal_age < 25 ~ 'Under 25', 
                              maternal_age < 30 ~ '25 - 30', 
                              maternal_age < 34 ~ '30 - 34', 
                              TRUE ~ 'Over 34'), levels=c('Under 25','25 - 30','30 - 34','Over 34')))
  } else {
    df$maternal_age <- NA
    df$age_group <- NA
  }

  if (maternal_race_in != "") {
    df$maternal_race <- df[[maternal_race_in]]
    race_above_thresh <- df %>%
      group_by(maternal_race) %>% summarise(total = n()) %>%
      filter(total > 2000)

    df <- df %>%
      mutate(race_group = case_when(
        maternal_race %in% race_above_thresh$maternal_race ~ maternal_race,
        TRUE ~ 'Other'))
    ref_group <- df %>% group_by(race_group) %>% summarise(total = n()) %>% filter(total == max(total))
    df$race_group <- relevel(
      factor(df$race_group),
      ref=ref_group$race_group)
  } else {
    df$maternal_race <- NA
    df$race_group <- NA
  }

  if (order_indication_in == "") {
    df$ord_indict_non_thc <- NA
  } else {
    df$order_indication <- df[[order_indication_in]]
    df$ord_indict_non_thc <- if_else(df$order_indication %in% ord_indict_non_thc_in, TRUE, FALSE)
  }

  df
}

```

```{r}
data <- read.csv("/input.csv")
input <- read.csv("/dict.txt", sep = '\t')

race_pal <- setNames(object = c("#e0aae8","#bae4bc","#7bccc4", '#2b8cbe'), nm = c('Black or African American', 'White', 'Other', 'Asian'))
age_pal <- setNames(object = c("#e0aae8","#bae4bc","#7bccc4", '#2b8cbe'), nm = c('Under 25', '25 - 30', '30 - 34', 'Over 34'))
intervention_date <- as.Date(input[input$column_out=='intervention_date','column_in'])

data <- clean_data(data, input)
```


```{r}
# data summary

## demographic parity
demo_sum <- data %>%
  group_by(race_group,age_group,delivery_qtr) %>%
  summarise(total = n(),
         perc_total = round(sum(uds_test, na.rm = TRUE)/n(), digits = 2))

## predictive parity
pred_non_thc_sum <- data %>%
  filter(uds_test) %>%
  group_by(race_group,delivery_qtr) %>%
  summarise(total = n(),
         perc_total = round(sum(non_thc_detect, na.rm = TRUE)/n(), digits = 2))

pred_thc_sum <- data %>%
  filter(uds_test) %>%
  group_by(race_group,delivery_qtr) %>%
  summarise(total = n(),
         perc_total = round(sum(thc_detect, na.rm = TRUE)/n(), digits = 2))

## equalized odds depends on order indication data
if (sum(is.na(data$ord_indict_non_thc)) == nrow(data)) {
  equal_odds_wide <- NA
  # if no data available, display error message
  error_ord <- "The data set is missing order indication data. The dashboard cannot create this visual or table."
} else {
  error_ord <- NA
  ## equalized odds
  equal_odds <- data %>%
    filter(uds_test) %>%
    group_by(race_group,age_group,pre_post_QI,ord_indict_non_thc,non_thc_detect) %>%
    count() %>%
    ungroup() %>%
    mutate(ord_ind_detect = factor(case_when(ord_indict_non_thc & non_thc_detect ~ 'TP',
                                      ord_indict_non_thc & !non_thc_detect ~ 'FP',
                                      !ord_indict_non_thc & !non_thc_detect ~ 'TN',
                                      !ord_indict_non_thc & non_thc_detect ~ 'FN')))

  equal_odds_wide <- pivot_wider(equal_odds%>%
    select(c("race_group","age_group","pre_post_QI",'ord_ind_detect', 'n')), names_from = ord_ind_detect, values_from = n) %>%
    mutate(FPR = (FP/(FP+TN)), 
           TPR = (TP/(TP+FN)), 
           ratio = FPR/TPR)
}

## equal outcome and group benefit depends on cps report data
if (sum(is.na(data$cps_report)) == nrow(data)) {
  equal_out <- NA
  group_benefit_wide <- NA
  # if no data available, display error message
  error_cps <- "The data set is missing CPS report data. The dashboard cannot create this visual or table."
} else {
  error_cps <- NA
  ## equal outcomes
  equal_out <- data %>%
    group_by(uds_test, race_group, non_thc_detect,cps_report,delivery_qtr) %>%
    count() %>%
    ungroup() %>%
    group_by(uds_test, race_group,delivery_qtr) %>%
    mutate(total = sum(n),
           perc_total = round(n/total, digits = 2)) %>%
    ungroup()

  ## group benefit equality
  group_benefit <- data %>%
    filter(uds_test) %>%
    group_by(race_group,age_group,pre_post_QI,cps_report,non_thc_detect) %>%
    count() %>%
    ungroup() %>%
    mutate(cps_detect = factor(case_when(cps_report & non_thc_detect ~ 'TP',
                                         cps_report & !non_thc_detect ~ 'FP',
                                         !cps_report & !non_thc_detect ~ 'TN',
                                         !cps_report & non_thc_detect ~ 'FN')))

  group_benefit_wide <- pivot_wider(group_benefit %>%
    select(c("race_group","age_group","pre_post_QI","cps_detect","n")), names_from = cps_detect, values_from = n) %>%
    mutate(ratio = ((TP+FP)/(TP+FN)))
}
```

```{r, child = 'visuals.rmd'}
```

```{r, child = 'section.rmd'}
```

```{r, child = 'demographics.rmd'}
```