app.py

import streamlit as st
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

# ======================
# Load data
# ======================
@st.cache_data
def load_data():
    return pd.read_csv("data/lab_events.csv", parse_dates=[
        "collection_time","receipt_time","start_analysis_time","verification_time","report_time"
    ])

df = load_data()

# ------------------------
# Add SLA thresholds
# ------------------------
sla_map = {"CBC":120, "CMP":240, "PT/INR":60, "UA":120, "PathReview":2880}  # minutes
df["sla_min"] = df["test_code"].map(sla_map).fillna(240)
df["tat_total"] = (df["report_time"] - df["collection_time"]).dt.total_seconds()/60
df["sla_hit"] = (df["tat_total"] <= df["sla_min"]).astype(int)

# ======================
# App title
# ======================
st.title("🏥 Lab Turnaround Time Optimization")

# ======================
# Sidebar Filters
# ======================
st.sidebar.header("🔎 Filters")

test_filter = st.sidebar.multiselect(
    "Test Type", options=df["test_code"].unique(), default=list(df["test_code"].unique())
)

shift_filter = st.sidebar.multiselect(
    "Shift", options=df["shift"].unique(), default=list(df["shift"].unique())
)

priority_filter = st.sidebar.multiselect(
    "Priority", options=df["priority"].unique(), default=list(df["priority"].unique())
)

# Apply filters
filtered_df = df[
    df["test_code"].isin(test_filter) &
    df["shift"].isin(shift_filter) &
    df["priority"].isin(priority_filter)
]

# ======================
# Empty Data Handling
# ======================
if filtered_df.empty:
    st.warning("⚠️ No data available for the selected filters. Please adjust your choices.")
else:
    # ======================
    # KPI Calculations
    # ======================
    median_tat = filtered_df["tat_total"].median()
    p95_tat = np.percentile(filtered_df["tat_total"], 95)
    sla_rate = filtered_df["sla_hit"].mean() * 100

    st.subheader("Key Metrics")
    col1, col2, col3 = st.columns(3)
    col1.metric("Median TAT (min)", f"{median_tat:.1f}")
    col2.metric("95th Percentile TAT (min)", f"{p95_tat:.1f}")
    col3.metric("SLA Hit Rate", f"{sla_rate:.1f}%")

    # ======================
    # What-If Analysis
    # ======================
    st.subheader("⚙️ What-If Analysis")

    sla_slider = st.sidebar.slider("SLA Threshold (minutes)", min_value=60, max_value=600, step=30, value=240)
    pre_delay_factor = st.sidebar.slider("Pre-Analytical Delay Factor", min_value=0.5, max_value=1.5, step=0.1, value=1.0)

    # Staffing scenario toggle
    scenario = st.sidebar.radio(
        "Staffing Scenario",
        ["Baseline", "+1 Evening Tech", "Redistribute Night → Evening"]
    )

    # Adjust SLA and pre delay
    filtered_df["sla_min_custom"] = filtered_df["sla_min"] * (sla_slider / 240)

    # Apply pre-analytical delay factor
    filtered_df["tat_total_custom"] = filtered_df["tat_total"] * pre_delay_factor

    # Apply staffing adjustments (simple multipliers to total TAT as proxy)
    if scenario == "+1 Evening Tech":
        # assume ~10% reduction in evening TAT
        mask_evening = filtered_df["shift"] == "Evening"
        filtered_df.loc[mask_evening, "tat_total_custom"] *= 0.9

    elif scenario == "Redistribute Night → Evening":
        # assume evening improves slightly, night worsens slightly
        mask_evening = filtered_df["shift"] == "Evening"
        mask_night = filtered_df["shift"] == "Night"
        filtered_df.loc[mask_evening, "tat_total_custom"] *= 0.92
        filtered_df.loc[mask_night, "tat_total_custom"] *= 1.05

    # Recalculate SLA hit
    filtered_df["sla_hit_custom"] = (filtered_df["tat_total_custom"] <= filtered_df["sla_min_custom"]).astype(int)

    # KPIs
    median_tat_custom = filtered_df["tat_total_custom"].median()
    p95_tat_custom = np.percentile(filtered_df["tat_total_custom"], 95)
    sla_rate_custom = filtered_df["sla_hit_custom"].mean() * 100

    col1, col2, col3 = st.columns(3)
    col1.metric("Median TAT (What-If)", f"{median_tat_custom:.1f}")
    col2.metric("95th Percentile (What-If)", f"{p95_tat_custom:.1f}")
    col3.metric("SLA Rate (What-If)", f"{sla_rate_custom:.1f}%")

    st.markdown(f"**Scenario Applied:** {scenario}  |  SLA = {sla_slider} min  |  Pre-delay factor = {pre_delay_factor:.1f}x")


    # ======================
    # Grouped Tables & Charts
    # ======================

    # By Test
    st.subheader("TAT by Test")
    by_test = filtered_df.groupby("test_code").agg(
        n=("order_id","count"),
        tat_median=("tat_total","median"),
        tat_p95=("tat_total", lambda x: np.percentile(x,95)),
        sla_hit_rate=("sla_hit","mean")
    ).reset_index()

    st.dataframe(by_test.style.format({
        "tat_median":"{:.1f}",
        "tat_p95":"{:.1f}",
        "sla_hit_rate":"{:.1%}"
    }))

    fig, ax = plt.subplots(figsize=(8,4))
    ax.bar(by_test["test_code"], by_test["tat_median"], color="skyblue", label="Median TAT")
    ax.bar(by_test["test_code"], by_test["tat_p95"], color="lightcoral", alpha=0.6, label="95th pct TAT")
    ax.set_ylabel("TAT (minutes)")
    ax.set_title("TAT by Test")
    ax.legend()
    st.pyplot(fig)

    # By Shift
    st.subheader("TAT by Shift")
    by_shift = filtered_df.groupby("shift").agg(
        n=("order_id","count"),
        tat_median=("tat_total","median"),
        sla_hit_rate=("sla_hit","mean")
    ).reset_index()
    st.dataframe(by_shift)

    fig, ax = plt.subplots(figsize=(6,4))
    ax.bar(by_shift["shift"], by_shift["tat_median"], color="mediumseagreen")
    ax.set_ylabel("Median TAT (minutes)")
    ax.set_title("Median TAT by Shift")
    st.pyplot(fig)

    # By Priority
    st.subheader("TAT by Priority")
    by_priority = filtered_df.groupby("priority").agg(
        n=("order_id","count"),
        tat_median=("tat_total","median"),
        sla_hit_rate=("sla_hit","mean")
    ).reset_index()
    st.dataframe(by_priority)

    fig, ax = plt.subplots(figsize=(6,4))
    ax.bar(by_priority["priority"], by_priority["tat_median"], color="orange")
    ax.set_ylabel("Median TAT (minutes)")
    ax.set_title("Median TAT by Priority")
    st.pyplot(fig)

    # ======================
    # Time Trend
    # ======================
    st.subheader("Daily Median TAT Over Time")
    by_day = filtered_df.groupby(filtered_df["receipt_time"].dt.date).agg(
        tat_median=("tat_total","median"),
        sla_hit_rate=("sla_hit","mean")
    ).reset_index()

    fig, ax = plt.subplots(figsize=(10,4))
    ax.plot(by_day["receipt_time"], by_day["tat_median"], marker="o", label="Median TAT")
    ax.set_ylabel("Median TAT (minutes)")
    ax.set_xlabel("Date")
    ax.set_title("Daily Median TAT Trend")
    ax.grid(True)
    st.pyplot(fig)

    fig, ax = plt.subplots(figsize=(10,4))
    ax.plot(by_day["receipt_time"], by_day["sla_hit_rate"]*100, marker="o", color="purple", label="SLA Hit Rate")
    ax.set_ylabel("SLA Hit Rate (%)")
    ax.set_xlabel("Date")
    ax.set_ylim(0, 100)
    ax.set_title("Daily SLA Compliance Trend")
    ax.grid(True)
    st.pyplot(fig)

    st.markdown("✅ Use the sidebar to filter by test, shift, and priority.")