Add visuals in app.py

app.py

@@ -1,6 +1,3 @@
-
-# streamlit_app.py
-
 import streamlit as st
 import pandas as pd
 import matplotlib.pyplot as plt
@@ -8,64 +5,74 @@ import seaborn as sns
 import plotly.express as px
 import plotly.graph_objects as go
 
-# ---------------------------
-# Function Definitions
-# ---------------------------
-
-def create_histogram(df):
-    """Creates a histogram for Age Distribution."""
-    fig, ax = plt.subplots(figsize=(5, 3.5))
-    sns.histplot(df['anchor_age'], bins=30, kde=True, color='skyblue', ax=ax)
-    ax.set_xlabel("Age")
-    ax.set_ylabel("Number of Admissions")
-    ax.set_title("Age Distribution")
-    plt.tight_layout()
-    st.pyplot(fig)
 
-def create_gender_bar_chart(df):
-    """Creates a bar chart for Gender Distribution."""
-    fig, ax = plt.subplots(figsize=(5, 3.5))
-    sns.countplot(data=df, x='gender', palette='pastel', ax=ax)
-    ax.set_title("Gender Distribution")
-    ax.set_xlabel("Gender")
-    ax.set_ylabel("Number of Admissions")
-    plt.tight_layout()
-    st.pyplot(fig)
-
-def create_stacked_bar_admission_race(df):
-    """Creates a stacked bar chart for Admission Types by Race."""
-    admission_race = df.groupby(['race', 'admission_type']).size().unstack(fill_value=0)
-    admission_race_percent = admission_race.div(admission_race.sum(axis=1), axis=0) * 100
+# Plot Function Definitions
 
-    admission_race_percent.plot(kind='bar', stacked=True, figsize=(8, 6), colormap='tab20')
-    plt.title("Admission Types by Race (%)")
-    plt.xlabel("Race")
-    plt.ylabel("Percentage of Admission Types")
-    plt.legend(title='Admission Type', bbox_to_anchor=(1.05, 1), loc='upper left')
+def create_gender_pie_chart(df):
+    """Creates a bar chart for Gender Distribution."""
+    gender_counts = df['gender'].value_counts().reset_index()
+    gender_counts.columns = ['Gender', 'Count']
+    fig_gender = px.pie(
+        gender_counts,
+        names='Gender',
+        values='Count',
+        hover_data=['Count'],
+        hole=0.3
+    )
+    st.plotly_chart(fig_gender, use_container_width=True)
+
+
+def create_race_pie_chart(df):
+    race_counts = df['race'].value_counts().reset_index()
+    race_counts.columns = ['Race Type', 'Count']
+    fig_race = px.pie(
+            race_counts,
+            names='Race Type',
+            values='Count',
+            hover_data=['Count'],
+            hole=0.3
+        )
+    st.plotly_chart(fig_race, use_container_width=True)
+    
+def create_insurance_pie_chart(df):
+    insurance_counts = df['insurance'].value_counts().reset_index()
+    insurance_counts.columns = ['Insurance Type', 'Count']
+    fig_insurance = px.pie(
+            insurance_counts,
+            names='Insurance Type',
+            values='Count',
+            hover_data=['Count'],
+            hole=0.3
+        )
+    st.plotly_chart(fig_insurance, use_container_width=True)
+
+def create_mortality_pie_chart(df):
+    #plt.figure(figsize=(6,3), facecolor='white')
+    total_admissions = df.shape[0]
+    labels = ['Survived', 'Died']
+    sizes = [total_admissions - df['hospital_expire_flag'].sum(),
+            df['hospital_expire_flag'].sum()]
+    colors = ['#66b3ff', '#ff6666']
+    explode = (0.1, 0)
+
+    plt.pie(sizes, explode=explode, labels=labels, colors=colors,
+            autopct='%1.1f%%', startangle=140,  textprops={'fontsize': 14})
+    plt.axis('equal') 
     plt.tight_layout()
     st.pyplot(plt.gcf())
-
-def create_los_by_race(df):
-    """Creates a box plot for Length of Stay by Race."""
-    fig, ax = plt.subplots(figsize=(6, 4))
-    sns.boxplot(data=df, x='race', y='los', palette='Pastel1', ax=ax)
-    ax.set_title("Length of Stay by Race")
-    ax.set_xlabel("Race")
-    ax.set_ylabel("Length of Stay (Days)")
-    ax.set_xticklabels(ax.get_xticklabels(), rotation=45)
-    plt.tight_layout()
-    st.pyplot(fig)
-
-def create_correlation_heatmap(df):
-    """Creates a correlation heatmap for numerical features."""
-    numerical_features = df[['anchor_age', 'los']]
-    corr_matrix = numerical_features.corr()
-
-    fig, ax = plt.subplots(figsize=(3.5, 3))
-    sns.heatmap(corr_matrix, annot=True, cmap='coolwarm', fmt=".2f", ax=ax)
-    ax.set_title("Correlation Heatmap")
-    plt.tight_layout()
-    st.pyplot(fig)
+    
+def create_admission_type_bar_chart(df):
+    admission_counts = df['admission_type'].value_counts().reset_index()
+    admission_counts.columns = ['Admission Type', 'Count']
+    fig_admission = px.bar(
+        admission_counts,
+        y='Admission Type',
+        x='Count',
+        color='Admission Type',
+        labels={'Count': 'Number of Admissions', 'Admission Type': 'Admission Type'},
+        hover_data=['Count']
+    )
+    st.plotly_chart(fig_admission, use_container_width=True)
 
 def create_time_series_heatmap(df):
     """Creates an admissions over time heatmap."""
@@ -81,16 +88,76 @@ def create_time_series_heatmap(df):
         y='admission_year',
         z='counts',
         histfunc='sum',
-        title='Admissions Over Time',
-        labels={'counts': 'Number of Admissions'},
-        color_continuous_scale='Blues'
+        labels={'counts': 'Number of Admissions', 'admission_month': 'Admission Month', 'admission_year': 'Admission Year'},
+        color_continuous_scale='rdbu'
     )
-
     fig.update_xaxes(categoryorder='array', categoryarray=month_order)
     fig.update_layout(yaxis=dict(autorange='reversed'))
     fig.update_traces(colorbar=dict(title='Admissions'))
     st.plotly_chart(fig, use_container_width=True)
 
+
+
+
+
+
+# def create_stacked_bar_admission_race(df):
+#     """Creates a stacked bar chart for Admission Types by Race."""
+#     admission_race = df.groupby(['race', 'admission_type']).size().unstack(fill_value=0)
+#     admission_race_percent = admission_race.div(admission_race.sum(axis=1), axis=0) * 100
+
+#     admission_race_percent.plot(kind='bar', stacked=True, figsize=(8, 6), colormap='tab20')
+#     plt.xlabel("Race")
+#     plt.ylabel("Percentage of Admission Types")
+#     plt.legend(title='Admission Type', bbox_to_anchor=(1.05, 1), loc='upper left')
+#     plt.tight_layout()
+#     st.pyplot(plt.gcf())
+
+# def create_los_by_race(df):
+#     """Creates a box plot for Length of Stay by Race."""
+#     fig, ax = plt.subplots(figsize=(6, 4))
+#     sns.boxplot(data=df, x='race', y='los', palette='Pastel1', ax=ax)
+#     ax.set_xlabel("Race")
+#     ax.set_ylabel("Length of Stay (Days)")
+#     ax.set_xticklabels(ax.get_xticklabels(), rotation=45)
+#     plt.tight_layout()
+#     st.pyplot(fig)
+
+# def create_correlation_heatmap(df):
+#     """Creates a correlation heatmap for numerical features."""
+#     numerical_features = df[['anchor_age', 'los']]
+#     corr_matrix = numerical_features.corr()
+
+#     fig, ax = plt.subplots(figsize=(3.5, 3))
+#     sns.heatmap(corr_matrix, annot=True, cmap='coolwarm', fmt=".2f", ax=ax)
+#     plt.tight_layout()
+#     st.pyplot(fig)
+
+
+def create_age_distribution_by_gender(df):
+    plt.figure(figsize=(12, 8))
+    sns.histplot(data=df, x='anchor_age', bins=30,
+                kde=True, palette='bright', hue='gender')
+    plt.xlabel('Age', fontsize=16)
+    plt.ylabel('Number of Admissions', fontsize=16)
+    plt.xticks(fontsize=16)
+    plt.yticks(fontsize=16)
+    plt.tight_layout()
+    st.pyplot(plt.gcf())
+
+
+def create_age_distribution_by_admission_type(df):
+    plt.figure(figsize=(12, 8))
+    sns.boxenplot(data=df, x='admission_type',
+                y='anchor_age', palette='Set3')
+    plt.xlabel('Admission Type', fontsize=16)
+    plt.ylabel('Age', fontsize=16)
+    plt.xticks(fontsize=16, rotation=45)
+    plt.yticks(fontsize=16)
+    plt.tight_layout()
+    st.pyplot(plt.gcf())
+
+
 def create_mortality_by_race(df):
     """Creates a bar chart for Mortality Rate by Race."""
     mortality_race = df.groupby('race')['hospital_expire_flag'].mean().reset_index()
@@ -98,7 +165,6 @@ def create_mortality_by_race(df):
 
     fig, ax = plt.subplots(figsize=(6, 4))
     sns.barplot(data=mortality_race, x='race', y='mortality_rate', palette='Set2', ax=ax)
-    ax.set_title("Mortality Rate by Race")
     ax.set_xlabel("Race")
     ax.set_ylabel("Mortality Rate (%)")
     ax.set_xticklabels(ax.get_xticklabels(), rotation=45)
@@ -112,7 +178,6 @@ def create_mortality_by_gender(df):
 
     fig, ax = plt.subplots(figsize=(6, 4))
     sns.barplot(data=mortality_gender, x='gender', y='mortality_rate', palette='Set3', ax=ax)
-    ax.set_title("Mortality Rate by Gender")
     ax.set_xlabel("Gender")
     ax.set_ylabel("Mortality Rate (%)")
     plt.tight_layout()
@@ -120,7 +185,6 @@ def create_mortality_by_gender(df):
 
 def create_mortality_by_age_group(df):
     """Creates a bar chart for Mortality Rate by Age Group."""
-    # Define age bins and labels
     bins = [0, 30, 50, 70, 90, 120]
     labels = ['0-30', '31-50', '51-70', '71-90', '91-120']
     df['age_group'] = pd.cut(df['anchor_age'], bins=bins, labels=labels, right=False)
@@ -130,7 +194,6 @@ def create_mortality_by_age_group(df):
 
     fig, ax = plt.subplots(figsize=(6, 4))
     sns.barplot(data=mortality_age, x='age_group', y='mortality_rate', palette='coolwarm', ax=ax)
-    ax.set_title("Mortality Rate by Age Group")
     ax.set_xlabel("Age Group")
     ax.set_ylabel("Mortality Rate (%)")
     plt.tight_layout()
@@ -148,7 +211,6 @@ def create_violin_age_race_mortality(df):
         palette='Set2',
         ax=ax
     )
-    ax.set_title("Age Distribution by Race and Mortality")
     ax.set_xlabel("Race")
     ax.set_ylabel("Age")
     ax.legend(title='Mortality', loc='upper right')
@@ -162,42 +224,15 @@ def create_heatmap_race_gender_mortality(df):
         columns='gender',
         values='hospital_expire_flag',
         aggfunc='mean'
-    ) * 100  # Convert to percentage
-
+    ) * 100  
+    
     fig, ax = plt.subplots(figsize=(8, 6))
     sns.heatmap(pivot_table, annot=True, fmt=".1f", cmap='YlOrRd', ax=ax)
-    ax.set_title("Mortality Rate by Race and Gender (%)")
     ax.set_xlabel("Gender")
     ax.set_ylabel("Race")
     plt.tight_layout()
     st.pyplot(fig)
 
-def create_parallel_coordinates(df):
-    """Creates a parallel coordinates plot for Demographics and Outcomes."""
-    # Select relevant numerical features
-    parallel_df = df[['anchor_age', 'los', 'hospital_expire_flag']].copy()
-
-    # Encode categorical variables numerically
-    parallel_df['race_code'] = df['race'].astype('category').cat.codes
-    parallel_df['gender_code'] = df['gender'].astype('category').cat.codes
-
-    # Create the parallel coordinates plot
-    fig = px.parallel_coordinates(
-        parallel_df,
-        color='hospital_expire_flag',
-        labels={
-            'anchor_age': 'Age',
-            'los': 'Length of Stay',
-            'hospital_expire_flag': 'Mortality',
-            'race_code': 'Race',
-            'gender_code': 'Gender'
-        },
-        color_continuous_scale=px.colors.diverging.Tealrose,
-        color_continuous_midpoint=0.5
-    )
-
-    fig.update_layout(title='Parallel Coordinates Plot of Demographics and Outcomes')
-    st.plotly_chart(fig, use_container_width=True)
 
 def create_treemap_race_mortality(df):
     """Creates a treemap for Race and Mortality."""
@@ -209,70 +244,12 @@ def create_treemap_race_mortality(df):
         path=['race', 'Mortality'],
         values='counts',
         color='Mortality',
-        color_discrete_map={'Survived':'#66b3ff','Died':'#ff6666'},
-        title='Treemap of Race and Mortality'
+        color_discrete_map={'Survived':'#66b3ff','Died':'#ff6666'}
     )
     fig.update_layout(margin = dict(t=30, l=0, r=0, b=0))
     st.plotly_chart(fig, use_container_width=True)
 
-def create_sankey_race_mortality(df):
-    """Creates a Sankey diagram for Race to Mortality Outcomes."""
-    sankey_df = df.groupby(['race', 'hospital_expire_flag']).size().reset_index(name='counts')
-
-    # Map 'hospital_expire_flag' to 'Mortality' status
-    sankey_df['Mortality'] = sankey_df['hospital_expire_flag'].map({0: 'Survived', 1: 'Died'})
-
-    # Create source and target labels
-    source = sankey_df['race'].tolist()
-    target = sankey_df['Mortality'].tolist()
-    values = sankey_df['counts'].tolist()
-
-    # Create a list of unique labels ensuring no duplicates
-    unique_races = sankey_df['race'].unique().tolist()
-    unique_mortality = sankey_df['Mortality'].unique().tolist()
-    labels = unique_races + unique_mortality
-
-
-    # Create a mapping from label to index for efficient lookup
-    label_to_index = {label: idx for idx, label in enumerate(labels)}
-
-    # Map source and target labels to their corresponding indices
-    source_indices = [label_to_index[s] for s in source]
-    target_indices = [label_to_index[t] for t in target]
-
-    # Optionally, define colors for different node types
-    # For example, races could have one color and mortality outcomes another
-    race_color = "#FFA07A"  # Light Salmon
-    mortality_color = "#20B2AA"  # Light Sea Green
-    node_colors = [race_color] * len(unique_races) + [mortality_color] * len(unique_mortality)
-
-    # Create the Sankey diagram
-    fig = go.Figure(data=[go.Sankey(
-        node=dict(
-            pad=15,
-            thickness=20,
-            line=dict(color="black", width=0.5),
-            label=labels,
-            color=node_colors
-        ),
-        link=dict(
-            source=source_indices,
-            target=target_indices,
-            value=values
-        )
-    )])
-
-    # Add title to the layout
-    fig.update_layout(
-        title_text="Sankey Diagram of Race and Mortality Outcomes",
-        font_size=10
-    )
-
-    st.plotly_chart(fig, use_container_width=True)
-
-# ---------------------------
 # Streamlit Application
-# ---------------------------
 
 # Set Streamlit page configuration
 st.set_page_config(
@@ -281,11 +258,10 @@ st.set_page_config(
     initial_sidebar_state="expanded",
 )
 
-# Title and Description
 st.title("MIMIC-IV ICU Patient Data Dashboard")
-st.markdown("""
-Explore the general feature distribution and outcome metrics of ICU patients from the MIMIC-IV dataset. Utilize the sidebar filters to customize the data view and interact with various visualizations to uncover patterns and insights.
-""")
+st.markdown('''
+Explore the general feature distribution and demographics related bias in ICU patients from the MIMIC-IV dataset. Utilize the sidebar filters to customize the data view'''
+)
 
 # Sidebar Filters
 st.sidebar.header("Filter Data")
@@ -293,10 +269,10 @@ st.sidebar.header("Filter Data")
 @st.cache_data
 def load_data():
 
-    admissions_df = pd.read_feather('data/admissions.feather')
-    patients_df = pd.read_feather('data/patients.feather')
+    admissions_df = pd.read_csv('data/admissions.csv')
+    patients_df = pd.read_csv('data/patients.csv')
     # diagnoses_icd_df = pd.read_csv('data/diagnoses_icd.csv')
-    pharmacy_df = pd.read_feather('data/pharmacy.feather')
+    # pharmacy_df = pd.read_csv('data/pharmacy.csv')
     # prescriptions_df = pd.read_csv('data/prescriptions.csv')
     # d_hcpcs_df = pd.read_csv('data/d_hcpcs.csv')
     # poe_detail_df = pd.read_csv('data/poe_detail.csv')
@@ -337,16 +313,14 @@ def load_data():
     "NATIVE HAWAIIAN OR OTHER PACIFIC ISLANDER":"NATIVES"}
 
     admissions_df['race'] = admissions_df['race'].map(race_map) 
-    # Merge admissions and patients data on 'subject_id'
+
     merged_df = pd.merge(admissions_df, patients_df, on='subject_id', how='left')
     
-    # Handle missing values by dropping rows with critical missing data
     merged_df = merged_df.dropna(subset=['anchor_age', 'gender', 'race', 'hospital_expire_flag'])
     
-    # Convert datetime columns
     merged_df['admittime'] = pd.to_datetime(merged_df['admittime'])
     merged_df['dischtime'] = pd.to_datetime(merged_df['dischtime'])
-    merged_df['deathtime'] = pd.to_datetime(merged_df['deathtime'], errors='coerce')
+    merged_df['deathtime'] = pd.to_datetime(merged_df['deathtime'], errors='coerce') 
     
     # Create derived features
     merged_df['los'] = (merged_df['dischtime'] - merged_df['admittime']).dt.days
@@ -419,6 +393,7 @@ filtered_df = add_sidebar_filters(merged_df)
 # Display Summary Statistics for Q1
 st.header("Summary Statistics")
 
+# Create four columns for metrics
 col1, col2, col3, col4 = st.columns(4)
 
 with col1:
@@ -446,54 +421,82 @@ st.markdown("---")
 tabs = st.tabs(["General Overview", "Potential Biases"])
 
 # Q1: General Overview
+
 with tabs[0]:
     st.subheader("General Feature Distribution and Outcome Metrics")
     
+    # Define the number of columns per row
     num_cols = 2
     
-    q1_plots = [
+    # Define all Q1 plots in a list with titles and plot-generating functions
+    q1_plots_2_col = [
         {
-            "title": "Age Distribution of ICU Patients",
-            "plot": lambda: create_histogram(filtered_df)
+            "title": "Gender Distribution",
+            "plot": lambda: create_gender_pie_chart(filtered_df)
         },
         {
-            "title": "Gender Distribution of ICU Patients",
-            "plot": lambda: create_gender_bar_chart(filtered_df)
+            "title": "Race Distribution",
+            "plot": lambda: create_race_pie_chart(filtered_df)
         },
         {
-            "title": "Admission Types by Race",
-            "plot": lambda: create_stacked_bar_admission_race(filtered_df)
+            "title": "Insurance Type Distribution",
+            "plot": lambda: create_insurance_pie_chart(filtered_df)
         },
         {
-            "title": "Length of Stay by Race",
-            "plot": lambda: create_los_by_race(filtered_df)
-        },
+            "title": "Mortality Rate of ICU Patients",
+            "plot": lambda: create_mortality_pie_chart(filtered_df)
+        }
+    ]
+    # Arrange Q1 plots in a grid layout
+    for i in range(0, len(q1_plots_2_col), num_cols):
+        cols = st.columns(num_cols)
+        for j in range(num_cols):
+            if i + j < len(q1_plots_2_col):
+                with cols[j]:
+                    st.subheader(q1_plots_2_col[i + j]["title"])
+                    q1_plots_2_col[i + j]["plot"]()
+    
+    num_cols = 1
+    
+    q1_plots_1_col = [
         {
-            "title": "Correlation Heatmap of Age and LOS",
-            "plot": lambda: create_correlation_heatmap(filtered_df)
+            "title": "Admission Type Count",
+            "plot": lambda: create_admission_type_bar_chart(filtered_df)
         },
         {
             "title": "Admissions Over Time",
             "plot": lambda: create_time_series_heatmap(filtered_df)
         }
     ]
-
-    for i in range(0, len(q1_plots), num_cols):
+    
+    # Arrange Q1 plots in a grid layout
+    for i in range(0, len(q1_plots_1_col), num_cols):
         cols = st.columns(num_cols)
         for j in range(num_cols):
-            if i + j < len(q1_plots):
+            if i + j < len(q1_plots_1_col):
                 with cols[j]:
-                    st.subheader(q1_plots[i + j]["title"])
-                    q1_plots[i + j]["plot"]()
+                    st.subheader(q1_plots_1_col[i + j]["title"])
+                    q1_plots_1_col[i + j]["plot"]()
 
 
-# Q2: Potential Biases from patient side
+# Q2: Potential Biases
 with tabs[1]:
     st.subheader("Analyzing Potential Biases Across Demographics")
     
+    # Define the number of columns per row
     num_cols = 2
     
+    # Define all Q2 plots in a list with titles and plot-generating functions
     q2_plots = [
+        
+        {
+            "title": "Age Distribution of ICU Patients",
+            "plot": lambda: create_age_distribution_by_gender(filtered_df)
+        },
+        {
+            "title": "Boxen Plot of Age Distribution by Admission Type",
+            "plot": lambda: create_age_distribution_by_admission_type(filtered_df)
+        },
         {
             "title": "Mortality Rate by Race",
             "plot": lambda: create_mortality_by_race(filtered_df)
@@ -517,13 +520,10 @@ with tabs[1]:
         {
             "title": "Treemap of Race and Mortality",
             "plot": lambda: create_treemap_race_mortality(filtered_df)
-        },
-        {
-            "title": "Sankey Diagram: Race to Mortality Outcomes",
-            "plot": lambda: create_sankey_race_mortality(filtered_df)
         }
     ]
     
+    # Arrange Q2 plots in a grid layout
     for i in range(0, len(q2_plots), num_cols):
         cols = st.columns(num_cols)
         for j in range(num_cols):
@@ -536,6 +536,6 @@ with tabs[1]:
 st.markdown("""
 ---
 **Data Source:** MIMIC-IV Dataset  
-**Project:** Investigating Biases in ICU Patient Data  
+**Project:** Fairness in ICU Patient Data  
 **Developed with:** Streamlit, Python  
 """)