- SQL & Data Analytics Project – Predictors of In-Hospital Mortality.
- Dataset Used : Kaggle
This project explores patient survival data to identify the key predictors of in-hospital mortality.
By analyzing demographic, clinical, and ICU-related attributes, the goal was to uncover actionable insights that can help healthcare professionals develop targeted interventions and evidence-based protocols.
- Clean and preprocess patient data for SQL analysis.
- Identify mortality trends across age, gender, ethnicity, ICU sources, comorbidities, and vital signs.
- Calculate mortality rates, probabilities, and risk factors using SQL queries.
- SQL / MySQL – Data cleaning, querying, aggregation, probability analysis
- Excel – Initial data cleaning and formatting
- Data Analysis – Identifying anomalies, trends, and correlations
- Data Cleaning: Replaced missing values (e.g., ethnicity) to ensure consistency.
- Mortality Trends: Calculated total hospital deaths (6.34%) and mortality rates by age group, ethnicity, gender, and ICU type.
- Comorbidity Impact: Found that diabetes, immunosuppression, and solid tumors had the highest mortality risk.
- Physiological Factors: Analyzed weight, BMI, and heart rate among deceased patients, linking cardiovascular indicators with outcomes.
- ICU Analysis: Discovered that patients in "Accident & Emergency" had the highest death count, while “Floor” had the highest death probability (11.76%).
- Age is a strong predictor: nearly 20% mortality in patients 70+ years old.
- Diabetes emerged as the top comorbidity linked with mortality (24.45% mortality rate).
- Elevated heart rate (average 115 bpm in deceased patients) was associated with higher risk.
- ICU length of stay strongly correlated with survival — shorter stays meant better outcomes.
This project demonstrates how SQL-based analysis can extract meaningful insights from healthcare data.
The results can help healthcare teams prioritize high-risk patients, design preventive interventions, and ultimately improve patient outcomes.