❤️‍🩹 EDA PROJECT 6:

Beyond the Average: Uncovering Geographic, Sex, and Race/Ethnicity Disparities in U.S. Heart Disease Mortality (2019–2021)

📌 Project Overview

This repository contains a comprehensive Exploratory Data Analysis (EDA) of heart disease mortality among U.S. adults (35+) for 2019–2021, using age-adjusted, spatially smoothed 3-year average mortality rates from the National Center for Health Statistics (NCHS) (data.gov). The analysis focuses on identifying geographic hotspots and disparities by sex and race/ethnicity, and assessing distributional characteristics of the mortality rates.

Primary aims:

• Identify geographic hotspots and county/state patterns of elevated heart disease mortality.
• Examine sex differences (male vs female) and geographic heterogeneity.
• Quantify disparities across race/ethnicity groups and subgroup hotspots (race × sex).
• Assess distributional properties of the mortality rates and implications for modeling.

---

🧰 Tools & Libraries

• Python (Jupyter / Colab)
• pandas, numpy — data wrangling
• matplotlib, seaborn — static visualization
• geopandas, plotly.express — mapping / choropleths
• scipy, statsmodels — basic statistical testing

---

🔍 Key Questions Addressed

1. Which counties and states have the highest age-adjusted heart disease mortality (2019–2021, smoothed 3-yr avg)?
2. How do male and female mortality rates compare across locations (paired comparisons and rate ratios)?
3. How do mortality rates vary by race/ethnicity, and which groups show the highest medians/IQRs?
4. Where are subgroup hotspots (race × sex) — e.g., Black males, AI/AN populations?
5. Are mortality rate distributions skewed or heavy-tailed, and do they require transformation for modeling?

---

🔍 Key Analysis Steps

1. Initial review — load data, inspect schema, data types, and sample rows.
2. Cleaning & normalization — snake_case columns, coerce data_value to numeric, create flags (smoothed, geographic level), remove 'Insufficient Data' rows, pad FIPS (location_id).
3. Filtering for analysis — extract county-level, smoothed, relevant stratifications (Overall, Male, Female, and race/ethnic groups).
4. Ranking & hotspot detection — rank counties/states by mortality rate and generate choropleth maps.
5. Paired comparisons by sex — merge male/female county rates for paired scatterplots and compute rate ratios/differences.
6. Race/ethnicity comparisons — group summaries (median, IQR) and boxplots by stratification2.
7. Distributional analysis — histograms, log-transforms, skewness/kurtosis to guide downstream modeling.

---

📊 Key Findings (summary)

• Geographic disparities & hotspots: The southeastern U.S. repeatedly emerges with the highest age-adjusted mortality (LA, KY, AL, MS, GA, SC, TN, OK). County-level choropleths confirm concentrated hotspots in the Southeast and parts of Appalachia.
• Sex differences: In most counties, male mortality > female mortality. Paired scatterplots show most counties above the y=x line; magnitude varies geographically.
• Race/ethnicity disparities: Median rates are highest among Black and Native Hawaiian / Other Pacific Islander groups, with Asian and “More than one race” showing lower medians. Boxplots reveal greater spread in certain groups (higher IQR).
• Subgroup hotspots: Race × sex maps show particularly severe hotspots for Black males/females in the Southeast. White males show elevated rates across the Southeast, Appalachia, and parts of the Midwest.
• Distributional characteristics: Original data_value is positively skewed (skew ≈ 0.73) with moderate kurtosis (≈1.52). Log-transform reduces skew but increases kurtosis; transformation choice should be driven by modeling goals and robustness considerations.

Public-health implication: The findings point to concentrated geographic and demographic inequities in heart disease mortality — targeted interventions and investments in the Southeast and among specific race/sex subgroups could yield high public-health benefit.

---

📸 Visual Highlights

Top 20 Counties (bar)	County Choropleth (smoothed rates)

Paired Male vs Female Scatter	Race/Ethnicity Boxplot
images/male_vs_female_scatter.png	images/Distribution_of_Heart-Disease_Mortality-Rate_by_Race-Ethnicity.png

Paired Male vs Female Scatter	Race/Ethnicity Boxplot
images/male_vs_female_scatter.png	images/race_boxplot.png

🗂 Repository Structure

eda-us-heart-disease-mortality-2019-2021/
├── notebooks/
│ └── EDA_Project_6_US_heart-disease_mortality_trend.ipynb
├── images/
│ ├── top20_counties.png
│ ├── choropleth_county.png
│ ├── male_vs_female_scatter.png
│ └── race_boxplot.png
├── data/ # optional: small sample or pointers (no raw NCHS files here)
├── README.md

---

📂 Dataset Access

You can access it here: 👉 Heart Disease Mortality Data Among US Adults (35+) – 2019-2021

---