Predicting Medical Insurance Charges Using Regression Analysis

Ayberk Kocakır

Economic Question & Motivation

Research Question

Which individual characteristics are associated with higher medical insurance charges?

Why is this important?

Healthcare costs affect households
Important for insurance pricing
Useful for healthcare policy

Dataset

Medical Cost Personal Dataset (Kaggle)

1,338 observations
Outcome variable: charges

Predictors:

Age
Sex
BMI
Children
Smoking Status
Region

Probability Analysis

Distribution of Medical Insurance Charges

Distribution is positively skewed
A small group has very high healthcare costs

Smoking Status Analysis

Charges by Smoking Status

Smokers have substantially higher charges
Smoking appears to be a strong predictor

Modeling Approach

Model 1

charges ~ age + bmi + children

Model 2

charges ~ age + bmi + children + smoker + region + sex

Data Split:

80% Training
20% Test

Model Comparison

Model Performance

Model	RMSE	R²
Model 1	11,323	0.113
Model 2	6,263	0.733

Model 2 was selected because it has lower prediction error and higher explanatory power.

Economic Interpretation

Variable	Effect on Charges
Smoker	+24,199
Age	+262
BMI	+318

Key Finding:

Smoking status is the strongest predictor.

Limitations

No income information
No education data
No medical history
Limited sample size

Future Improvement:

Add socioeconomic variables
Add medical history variables

Conclusion

Key Findings

Smoking is the strongest predictor
Age and BMI increase healthcare costs
Model 2 performs significantly better
Lifestyle factors matter

Thank you for listening.

Questions?