Health Insurance Claims

Health Insurance Claims Analysis with Python and Jupyter Notebook, analyzing demographic and behavioural risk factors (Age, BMI, Smoking status) impacting medical insurance charges.

Notebook Link

---

Table of Contents

• Overview
• Dataset
• Technologies Used
• Installation
• Usage
• Analysis & Visualizations
• Conclusion
• Credits
• License

---

Overview

• The dataset was chosen to assess heath insurance underwriting.
• Python in Jupyter Notebook was used for the analysis.
• The objective was to determine segmentation, risk scoring, data insights and scenario testing

---

Dataset

• The dataset is Medical Cost Personal Datasets from Kaggle
• Size of the dataset is 1338 rows and 7 columns
• The analysis with this dataset is duplicated in Power PI and Tableau Desktop

Dataset: insurance.csv

---

Technologies Used

• Languages & Libraries: Python, Pandas, NumPy, Matplotlib, Seaborn, sklearn
• Tools: Jupyter Notebook, VS Code, Git, GitHub

---

Installation

Step-by-step instructions to set up the project locally:

# Clone the repository
git clone git clone https://github.com/Kurodataio/Health-Insurance-Claims.git

# Navigate to the project folder
cd Health-Insurance-Claims

# Launch Jupyter Notebook
jupyter notebook

Usage

Instructions for using the project:

1. Open the main notebook (health-insurance-claims.ipynb)
2. Run ALL cells or each cell sequentially to reproduce the analysis
3. Visualizations and results will be generated automatically

---

Analysis & Visualizations

• Age vs. Claims Cost (Smoker Highlighted)

  • 3 clear bands with linear relationship between age and claims cost
  • Bottom blue non-smokers band has lowest age to claims cost
  • Middle non-smokers and smokers
  • Top orange smokers band has the gighest agr to claims cost
  • The is a clear positive linear relationship between age and claims cost
  • Smoking increases the cost of caims

• BMI vs. Claims Cost

  • For non-smokers BMI is minor impact on costs. Most claims costs for non smokers are below $15k
  • For smokers, there is significant BMI impact beyond BMI of 30
  • BMI minimal impact on claims costs for non-smokers however for BMI over 30 there is an sudden and increasing impact on costs

• Claims Cost by Smoker Status

  • The average claims cost between smokers and non-smokers claims cost is significant
  • Non-smokers have a median cost of $7,500
  • Smokers have a mediam cost of $34,500
  • There is no overlap between the smoking and non-smoking groups
  • The 75th percentile for non-snokers is lower than the 25th percentile for smokers
  • For non smokers the claims cost are between $4,000 and $11,500
  • For smokers the claims cost are between $21,000 to $41,000
  • Smokers have a wider range (variance) of claims costs
  • Non smokers have many outlying claims around $23,000 up to $37,000
  • This plot shows high outliers for non-smokers are within the typical claims cost for smokers

• Feature_Importance

• Claims Cost by Dependents

  • The average (median) claims cost for all groups is $8,000 and $11,000
  • 0 to 2 dependents: The median is around $8,500–$9,500
  • 3 to 4 dependents: The mdian peaks at $11,000 for 4 dependents
  • 5 dependents: The median is under $9,000
  • The group with 0 dependents has outliers with the highest cost of $60,000+
  • The group with 5 dependents has outliers with the lowest cost at under $20,000
  • The number of dependents does not dictate high claims cost. In fact it seems the opposite.

• Claims Cost by Gender

  • Both genders have an average claim cost of $9,300.
  • Gender has no significance to median claim costs
  • Both genders have a long tail of outliers. Beyond $30K for Women and beyond $40k for men

• Claims Cost by Region

  • Southwest, Southeast, Northwest have a median claims cost around $9k to $9.3k
  • Northeast has a higher median claims cost around $10k.
  • Region is a not a significant determinant of claims cost.
  • The Southeast has the highest range of claims costs. It also has the highest outlier costs
  • The outliers can be associated with risk factors such as smoking and high BMI

• Average BMI by Region

  • Highest BMI Region: southeast (Mean: 33.36)

• Percentage of Smokers by Region

  • Highest Smoker Region: southeast (25.0%)

• The plot of Average BMI and smokers confirms why the Southeast has the higher claims costs

• Claims Cost Distribution by Smoking Status

  • The non-smokers show right-skewed distribution
  • The highest frequency (density) of individuals show claims cost between $2,000 and $10,000.
  • The data shows that most of the minor claims are by non-smokers
  • The data shows two peaks for smokers. It is bimodal.
  • The first peak cluster is between $15,000–$25,000.
  • The second peak cluster is between $35,000–$50,000.

• Correlation_Heatmap

  • The positive correlation between smoker and claims cost is 0.79
  • The positive correlation between age and claims cost is 0.30 and 0.20 for bmi and claims cost.
  • The heatmap does not show the interaction effect between BMI and smoking which is associated with higgher costs.
  • The heatmap clearly shows the following:
    • Smoker status (Highest correlation)
    • Age (High correlation)
    • BMI (Moderate correlation)

• Distribution of Claims Cost

  • The distribution of claims is skewed to the right (positive) skewed
  • Most of the insurance claims are low to medium in costs
  • The distribution is bimodal, primary and secondary peaks
  • The two peaks suggest sub populations of claims cost
  • Mode Claims Cost is $1639.56
  • Median Claims Cost is $9382.03
  • Mean Claims Cost is $13270.42

• Loss_Ratio_Distribution

  • The Loss Ratio (LR) represents the ratio of losses (claims paid) to premiums earned
  • Loss Ratio (LR) = Total Losses (L) / Total Premium (P)
  • A loss ratio of 1.0 (or 100%) means the company is paying out exactly what it collects in premiums.
  • Most of the small and medium claims are profitable.
  • The high claims costs are not profitable
  • Adjusting the slope coefficient to cover higher costs penalizes the lower claims with higher premiums

• Loss Ratio Distribution: Current vs Proposed Pricing

  • The risk based pricing is a flattened curve and spreads the risk across the x-axis way past 60.0 and 80.0
  • This plot shows the proposed model heavily underprices a large segment of the portfolio.
  • The proposed risk based pricing needs corrective changes

---

Conclusion

• Smoking is the biggest risk feature for high claims cost. This single driver puts an individual into a higher cost bracket regardless of age or BMI.
• Obesity (BMI $\ge$ 30) acts as a compounding multiplier when associated with smoking.
• Smoking and BMI combined are associated with the highest claims cost
• The Southeast region has the highest levels of smoking and obesity (high BMI) rates. Remediation could be targeted at this region as a priority.
• The southeast region has the highest claims cost, which is associated with the highest number of smokers and obese people.
• Further data and analysis is required to validated the surge in claims cost associated with smoking and BMI.
  • What are the actual illnesses associated with high BMI and smoking?
  • Are medical protocols and procedures initiated based on BMI and smoking signals or the actual presence of disease?

---

Acknowledgements & Credits

• Dataset Source: Kaggle → Datasets
• Google Gemini Link
• CoPilot Link

---

License

This project is licensed under the MIT License

---