This project focuses on uncovering inefficiencies and bias within healthcare insurance decision-making processes. We analyzed patient demographics, diagnosis categories, and decision timelines to help healthcare organizations streamline operations, reduce delays, and promote fairness in medical determinations.
Objective: Transform raw healthcare review data into actionable insights and future-ready dashboards using Python, Power BI, and Machine Learning foundations.
- What demographic groups experience longer review times or higher rejection rates?
- Are certain diagnoses linked to slower processing or more overturns?
- How has the fairness of decisions evolved over time?
- What improvements can be made in review processes and healthcare communication?
| Stack | Tools Used |
|---|---|
| Data Cleaning & Prep | Python (Pandas, NumPy), Excel |
| Visualization | Matplotlib, Seaborn , Power BI |
| Data Modeling (next phase) | ML model for Determination prediction |
- Handled missing values, standardized diagnosis names, and extracted meaningful features.
- Engineered key variables:
DaysToReview,AgeGroup,DecisionType,DiagnosisType. - Created filtered and optimized datasets for dashboarding and future ML usage.
- Yearly trends of Rejections, Overturned vs. Upheld
- 2016 spike linked to Senate Bill 863 — uncovered through web research
- Visual drill-down into rejection reasons and gender/age-based discrepancies
- Mental disorders are most diagnosed, while infection diseases dominate ages 51–64
- Younger and underrepresented gender patients (11–20, ‘Others’ category) face more rejections — but show higher acceptance upon reapplication
- After 2019, overturned outcomes have increased, indicating improved fairness
- 📈 Overturned decisions are processed faster than upheld ones
- 🧑⚕️ After 2008, decision adoption takes longer than review — change in process?
- 🔁 Reapplication success is highest for underrepresented groups (age 11–20, gender "Others")
- 🧠 Mental health and infections dominate case diagnoses
- 📅 2016 anomaly tied to policy changes (Senate Bill 863)
- Transparent communication with youth and guardians during initial filing
- Tailored outreach programs for patients aged 51–64
- Conduct quality audits on decisions with high overturn likelihood
- Feedback loop to improve and refine decision-making protocols
- AI-Powered Recommendation Engine (Next Phase) for automated risk-based predictions
- 💻 Real-Time Monitoring: Embed dashboards for real-time decision insights
- 🤖 Predictive Modeling: Train ML models to predict
Determinationvariable using logistic regression, random forest, or XGBoost - 🔁 Automated Alerts: Flag potential unfair or delayed cases for manual re-evaluation
- Mental health and infectious diseases top the list of diagnosed conditions, especially in ages 51–64
- Fairness is improving post-2019, with more overturned decisions
- Timely processing often leads to more favorable outcomes (faster = overturned)
- Younger and underrepresented patients are more likely to face rejections, but also more likely to succeed on reapplication