AnalysisForPublicationRmd.Rmd

---
title: "DIMIS Analysis"
output:
  html_document:
    df_print: paged
  html: default
  pdf_document: default
date: "2024-11-27"
---
# load libraries
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = FALSE)
```

# Load Data and Data Cleaning
```{r}
# Load necessary libraries
library(readxl)
library(dplyr)
library(writexl)
library(ggplot2)

# Load the data
Master <- read_excel("DIMIS_Master.xlsx")

# Inspect data structure and column names
names(Master)
str(Master)
```
## Clean and Restructure Data
```{r}
# Clean and restructure data
Master <- Master %>%
  mutate(
    Domain = as.factor(Domain),
    ISO_LOINC = as.factor(ISO_LOINC),
    ISO_SCT = as.factor(ISO_SCT),
    NH_SCT = as.numeric(NH_SCT),
    MM_SCT = as.numeric(MM_SCT)
  )

# Check the updated structure
str(Master)

# Export cleaned data to Excel for quality check
write_xlsx(Master, "DIMIS_Master_Clean.xlsx")
```

*** 
# ISO-Scores of SCT Codes
## Percentages of ISO-Scores
```{r}
# Calculate counts and percentages for ISO_LOINC
summary_table_SCT <- Master %>%
  count(ISO_SCT) %>%
  mutate(Percentage = (n / sum(n)) * 100)

# Display the summary table
knitr::kable(summary_table_SCT, caption = "Counts and Percentages of ISO SCT Codes")
```
## Calculate ISO-Score
```{r}
# Ensure ISO_LOINC is converted to numeric if it's a factor
summary_table_SCT$ISO_SCT <- as.numeric(as.character(summary_table_SCT$ISO_SCT))

# Calculate the weighted sum of equivalence measures
weighted_sum_SCT <- sum(summary_table_SCT$ISO_SCT * summary_table_SCT$n)

# Calculate the total number of maps
total_maps_SCT <- sum(summary_table_SCT$n)

# Compute the average equivalence score
average_equivalence_score_SCT <- weighted_sum_SCT / total_maps_SCT

# Print the result
cat("The Equivalence Score is:", round(average_equivalence_score_SCT, 2), "\n")
```



*** 
# ISO-Scores of LOINC Codes
## Percentages of ISO-Scores
```{r}
# Calculate counts and percentages for ISO_LOINC
summary_table_LOINC <- Master %>%
  count(ISO_LOINC) %>%
  mutate(Percentage = (n / sum(n)) * 100)

# Display the summary table
knitr::kable(summary_table_LOINC, caption = "Counts and Percentages of ISO LOINC Codes")

```
## Calculate ISO-Score
```{r}
# Ensure ISO_LOINC is converted to numeric if it's a factor
summary_table_LOINC$ISO_LOINC <- as.numeric(as.character(summary_table_LOINC$ISO_LOINC))

# Calculate the weighted sum of equivalence measures
weighted_sum_LOINC <- sum(summary_table_LOINC$ISO_LOINC * summary_table_LOINC$n)

# Calculate the total number of maps
total_maps_LOINC <- sum(summary_table_LOINC$n)

# Compute the average equivalence score
average_equivalence_score_LOINC <- weighted_sum_LOINC / total_maps_LOINC

# Print the result
cat("The Equivalence Score is:", round(average_equivalence_score_LOINC, 2), "\n")
```

# ISO-Scores by Domain
*** 
Prepare SCT Data
## Cross-Table of ISO Scores and Domains
```{r}
# Generate a cross-table of ISO_SCT by Domain and calculate proportions
SCT_table <- table(Master$ISO_SCT, Master$Domain)
SCT_prob <- prop.table(SCT_table, 2)

# Convert the table to a data frame
SCT_dataframe <- as.data.frame.matrix(SCT_prob)

# Add rownames as a column for long-format conversion
SCT_dataframe$ISO_SCT <- rownames(SCT_dataframe)

```

## Convert Data to Long Format
```{r}
# Convert wide-format dataframe to long format
library(reshape2)
SCT_DataLong <- melt(
  SCT_dataframe,
  id.vars = "ISO_SCT",
  value.name = "proportion"
)

```

*** 
Prepare LOINC Data
## Cross-Table of ISO Scores and Domains
```{r}
# Generate a cross-table of ISO_LOINC by Domain and calculate proportions
LOINC_table <- table(Master$ISO_LOINC, Master$Domain)
LOINC_prob <- prop.table(LOINC_table, 2)

# Convert the table to a data frame
LOINC_dataframe <- as.data.frame.matrix(LOINC_prob)

# Add rownames as a column for long-format conversion
LOINC_dataframe$ISO_Loinc <- rownames(LOINC_dataframe)

```

## Convert Data to Long Format
```{r}
# Convert wide-format dataframe to long format
LOINC_DataLong <- melt(
  LOINC_dataframe,
  id.vars = "ISO_Loinc",
  value.name = "proportion"
)

```

## Prepare Both Datasets for Combined Dataset
```{r}
# Add a source column to each dataset
SCT_DataLong$Source <- "SNOMED"
LOINC_DataLong$Source <- "LOINC"
# all columns need to have the same name
colnames(SCT_DataLong)[colnames(SCT_DataLong) == "ISO_SCT"] <- "ISO_Code"
colnames(LOINC_DataLong)[colnames(LOINC_DataLong) == "ISO_Loinc"] <- "ISO_Code"

# Combine the two datasets
combined_data <- rbind(SCT_DataLong, LOINC_DataLong)
```
## Graph
```{r}
p <- ggplot(combined_data, aes(y = variable, x = proportion, fill = ISO_Code)) +
  geom_bar(stat = "identity") +
  
  # Customize axes and labels
  ylab("") +
  labs(
    fill = "ISO Score",
    title = "ISO Scores for SNOMED and LOINC Codes by DIMIS Domain"
  ) +
  
  # Add color palette
  scale_fill_brewer(palette = "BuPu") +
  
  # Adjust faceting to make the plots wider
  facet_wrap(~ Source, scales = "free_x", ncol = 2) +
  
  # Enhance plot aesthetics
  theme_minimal() +
  theme(
    plot.title = element_text(hjust = 0.5, size = 16, face = "bold"),
    axis.text.y = element_text(size = 8),  # Reduce y-axis text size
    strip.text = element_text(size = 14),  # Adjust facet label size
    legend.title = element_text(size = 12),
    legend.text = element_text(size = 10),
    panel.spacing = unit(1, "lines")  # Add spacing between facets
  )

```

```{r}
print(p)
```
```{r}
ggsave("~/Library/CloudStorage/OneDrive-Charité-UniversitätsmedizinBerlin/BIH/Repositories/DIMIS/graphPaper.png", plot = p, width = 10, height = 6, dpi = 300)
```

***
# Krippendorf
## SCT
```{r}
library(irr)
# Step 1: Remove rows where either NH_SCT or MM_SCT has the value -1
Kripp_clean_SCT <- Master %>%
  filter(NH_SCT != -1, MM_SCT != -1)

# Step 2: Calculate the sum of matches and non-matches
sum_matches <- sum(Kripp_clean_SCT$NH_SCT == Kripp_clean_SCT$MM_SCT)
sum_non_matches <- sum(Kripp_clean_SCT$NH_SCT != Kripp_clean_SCT$MM_SCT)

# Step 3: Calculate proportions
total_rows <- nrow(Kripp_clean_SCT)
proportion_matches <- sum_matches / total_rows
proportion_non_matches <- sum_non_matches / total_rows

# Step 4: Prepare data for Krippendorff's alpha
agreement_matrix_SCT <- as.matrix(Kripp_clean_SCT[, c("NH_SCT", "MM_SCT")])

# Step 5: Calculate Krippendorff's alpha
result_SCT <- kripp.alpha(agreement_matrix_SCT, method = "nominal")  # Use "interval" for numeric data

# Step 6: Create a summary table
summary_table <- data.frame(
  Metric = c("Matches", "Non-Matches", "Krippendorff Alpha"),
  Count = c(sum_matches, sum_non_matches, NA),  # NA for alpha, as it's not a count
  Proportion = c(proportion_matches, proportion_non_matches, result_SCT$value)
)

# Print the summary table
print(summary_table)
print(result_SCT)
```


## LOINC

```{r}
# Step 1: Remove rows where either NH_SCT or MM_SCT has the value -1
Kripp_clean_LOINC <- Master %>%
  filter(NH_LOINC != -1, MM_LOINC != -1)

# Step 2: Prepare data for Krippendorff's alpha
# Select only the two relevant columns and convert to a matrix
agreement_matrix_LOINC <- as.matrix(Kripp_clean_LOINC[, c("NH_SCT", "MM_SCT")])

# Step 3: Calculate Krippendorff's alpha (nominal or interval, depending on data type)
result_LOINC <- kripp.alpha(agreement_matrix_LOINC, method = "nominal")  # Use "interval" for numeric data

# Print the result
print(result_LOINC)
```

```{r}
library(irr)
# Step 1: Remove rows where either NH_SCT or MM_SCT has the value -1
Kripp_clean_LOINC <- Master %>%
  filter(NH_LOINC != -1, MM_LOINC != -1)

# Step 2: Calculate the sum of matches and non-matches
sum_matches <- sum(Kripp_clean_LOINC$NH_LOINC == Kripp_clean_LOINC$MM_LOINC)
sum_non_matches <- sum(Kripp_clean_LOINC$NH_LOINC != Kripp_clean_LOINC$MM_LOINC)

# Step 3: Calculate proportions
total_rows <- nrow(Kripp_clean_LOINC)
proportion_matches <- sum_matches / total_rows
proportion_non_matches <- sum_non_matches / total_rows

# Step 4: Prepare data for Krippendorff's alpha
agreement_matrix_LOINC <- as.matrix(Kripp_clean_LOINC[, c("NH_SCT", "MM_SCT")])

# Step 5: Calculate Krippendorff's alpha
result_LOINC <- kripp.alpha(agreement_matrix_LOINC, method = "nominal")  # Use "interval" for numeric data

# Step 6: Create a summary table
summary_table <- data.frame(
  Metric = c("Matches", "Non-Matches", "Krippendorff Alpha"),
  Count = c(sum_matches, sum_non_matches, NA),  # NA for alpha, as it's not a count
  Proportion = c(proportion_matches, proportion_non_matches, result_LOINC$value)
)

# Print the summary table
print(summary_table)
print(result_LOINC)
```