Predictive Healthcare Insights

Introduction
Healthcare systems face a major challenge with patient readmissions, leading to increased costs and reduced patient care quality. This project demonstrates an end-to-end workflow to predict readmissions and generate actionable insights using Python, SQL, Azure, and Power BI.

Problem Statement
Hospitals often struggle to identify which patients are at risk of returning after discharge. By predicting readmissions, hospitals can optimize resources, improve patient outcomes, and reduce operational costs.

Workflow Overview

1. Python Data Pipeline

   • Generate & clean synthetic healthcare data (hospital_admissions.csv)

   • Feature selection: age, length_of_stay, num_lab_tests, total_cost, prev_admissions

   • Train-test split with stratified sampling to maintain class balance

   • ML Pipeline: SimpleImputer + StandardScaler + RandomForestClassifier

   • Predictions & feature importance saved for dashboard use

   • 

2. SQL Analysis (MySQL)

   • Average Length of Stay per Patient

   • Readmission Rate

   • Cost Analysis

   • Gender-wise Readmission Rate

   • Monthly Admissions Trend

3. 4. Patients with High Readmission Risk (Window Function)

    SELECT patient_id,
           COUNT(*) AS total_admissions,
           SUM(readmitted) AS total_readmissions,
           ROUND(SUM(readmitted) * 100.0 / COUNT(*), 2) AS readmission_rate,
           RANK() OVER (ORDER BY SUM(readmitted) DESC) AS risk_rank
    FROM admissions
    GROUP BY patient_id
    ORDER BY readmission_rate DESC
    LIMIT 10;
   

   Monthly Admissions Trend

    SELECT DATE_FORMAT(admit_date, '%Y-%m') AS month,
           COUNT(*) AS admissions,
           ROUND(SUM(total_cost), 2) AS total_cost
    FROM admissions
    GROUP BY month
    ORDER BY month;
   

   Gender-wise Readmission Rate

    SELECT p.gender,
           COUNT(*) AS total_patients,
           SUM(a.readmitted) AS total_readmitted,
           ROUND(SUM(a.readmitted) * 100.0 / COUNT(*), 2) AS readmission_rate
    FROM admissions a
    JOIN patients p ON a.patient_id = p.patient_id
    GROUP BY p.gender
    ORDER BY readmission_rate DESC;
   

   Age Group vs Average Length of Stay

    SELECT CASE 
               WHEN p.age BETWEEN 0 AND 18 THEN '0-18'
               WHEN p.age BETWEEN 19 AND 35 THEN '19-35'
               WHEN p.age BETWEEN 36 AND 50 THEN '36-50'
               WHEN p.age BETWEEN 51 AND 65 THEN '51-65'
               ELSE '65+'
           END AS age_group,
           ROUND(AVG(a.length_of_stay), 2) AS avg_stay
    FROM admissions a
    JOIN patients p ON a.patient_id = p.patient_id
    GROUP BY age_group
    ORDER BY avg_stay DESC;
   

   Azure Blob Storage

   • CSV datasets stored securely in the cloud

   • Enables integration with Power BI for real-time dashboards

1. Power BI Dashboard

   • KPI Cards: Total Patients, Readmissions, Avg Length of Stay

   • Pie/Donut Charts: Readmission vs No Readmission, Gender Distribution

   • Bar/Line Charts: Readmission trends, Top 5 Costly Diagnoses

   • Patient-level predictions table with probabilities

2. • 

     Key Insights

     • Patients with longer stays or previous admissions are more likely to be readmitted

     • Readmission patterns by gender and cost correlate with length of stay

     • Interactive dashboards allow hospital management to proactively allocate resources

3. Conclusion
   This project provides a MAANG-level demonstration of healthcare analytics, combining Python, SQL, Azure, and Power BI. It shows how end-to-end analytics pipelines can deliver predictive insights and aid in data-driven decision making in healthcare.

4. Website-Link: https://tanmayportfolio52.wordpress.com/

5. Git hub- Link: https://github.com/Tanu272004