This project is an exploratory data analysis of the Heart Disease UCI dataset, a clinical dataset containing records from 1,025 patients who were assessed for cardiovascular disease. The analysis was conducted using Python, SQL, and a set of visualization libraries to identify patterns and risk factors associated with heart disease presence.
The goal was not simply to plot data, but to ask meaningful questions and let the data answer them honestly — including where the answers are surprising.
- Python — Pandas, NumPy, Matplotlib, Seaborn
- SQL — SQLite via Python's sqlite3 library
- Jupyter Notebook
- Git and GitHub for version control
patient-data-analysis/
├── data/ Raw dataset (heart.csv) and SQLite database
├── notebooks/ Jupyter Notebook containing the full analysis
├── outputs/ All generated charts and the dashboard image
├── sql/ Standalone SQL queries used for data extraction
└── README.md This file
Source: Heart Disease UCI Dataset via Kaggle Records: 1,025 patients Features: 14 variables including age, sex, cholesterol, resting blood pressure, maximum heart rate, chest pain type, and a binary target variable indicating whether heart disease was diagnosed.
The dataset originates from a clinical setting, meaning all patients in the sample were already presenting with cardiovascular symptoms when assessed. This context matters when interpreting certain findings.
- What proportion of this patient population has heart disease?
- Does age differ meaningfully between patients with and without disease?
- Is there a visible gender pattern in disease prevalence?
- How strongly does cholesterol level separate the two groups?
- What does maximum heart rate reveal about heart health?
- Which variables are most correlated with each other in this dataset?
The dashboard above contains six visualizations produced from the dataset. Each chart addresses a specific analytical question and is discussed in detail under Key Findings below.
Disease prevalence 54.3 percent of patients in this dataset were diagnosed with heart disease, and 45.7 percent were not. The near-even split makes this a reasonably balanced dataset for analysis, reducing the risk of skewed conclusions.
Age and disease Patients diagnosed with heart disease tend to be older on average, clustering around the 55 to 60 age range. Patients without disease are more evenly distributed across ages, with a peak around 50 to 55. This confirms age as a genuine risk factor and aligns with established cardiovascular research.
Gender pattern Male patients outnumber female patients in this sample significantly. Among female patients, the proportion with heart disease is notably high. However, this should not be interpreted as women being more susceptible in general. Because this is a clinical dataset — collected from patients already showing symptoms — the female group in this sample was already a high-risk population by selection. This is a case where understanding the data collection context prevents a misleading conclusion.
Cholesterol One of the more counterintuitive findings is that cholesterol levels show very weak separation between the disease and no-disease groups. The median cholesterol values are nearly identical across both groups, and the distributions overlap heavily. This suggests that cholesterol alone, without other clinical context, is not a reliable standalone predictor of heart disease in this dataset.
Maximum heart rate This variable shows the clearest visual separation between the two groups. Across all ages, patients with heart disease consistently achieve a lower maximum heart rate during exercise compared to patients without disease. A healthy heart responds to physical stress by pumping significantly faster. When that capacity is reduced, it is often a sign of underlying cardiovascular damage. This finding is medically well-supported and stands out as the strongest signal in this dataset.
Correlation structure The correlation heatmap reveals several meaningful relationships. Maximum heart rate and ST depression (Oldpeak) are negatively correlated at -0.58, confirming that patients with higher exercise heart rates tend to show less cardiac stress. Age and the number of blocked major vessels carry a positive correlation of 0.30, which is consistent with the medical understanding that arterial blockage accumulates over time. Cholesterol shows near-zero correlation with most other variables, which is consistent with its weak predictive performance observed in the boxplot.
This dataset is relatively small at 1,025 records and comes from a specific clinical population. The findings describe patterns within this sample and should not be generalized as universal medical conclusions. A proper clinical study would require larger, more diverse, and carefully controlled data collection. The purpose of this project is analytical and educational.
git clone https://github.com/mdabrarfaiyaj/patient-data-analysis
cd patient-data-analysis
pip install -r requirements.txt
jupyter notebookOpen the notebook inside the notebooks folder and run all cells from top to bottom.
Heart Disease UCI Dataset https://www.kaggle.com/datasets/johnsmith88/heart-disease-dataset
Md Abrar Faiyaj
Graduate Biotechnology Student | BRAC University,Dhaka,Bangladesh
Junior Research Collaborator | ABCD Laboratory, Chittagong, Bangladesh
GitHub: https://github.com/mdabrarfaiyaj
LinkedIn: https://www.linkedin.com/in/md-abrar-faiyaj-559246381