About this project
End-to-end analysis of about 84,000 synthetic hospital visits (2023 to 2025) in Python, from cleaning and exploratory analysis to correlation analysis and hypothesis testing. Because bill amounts and length of stay are heavily skewed, I used non-parametric methods throughout: Spearman and partial correlations, Chi-square tests with Cramér's V, and Mann-Whitney U and Kruskal-Wallis tests. Effect sizes were reported next to p-values so that significance in a large sample was not mistaken for practical importance. The analysis shows what drives admission, mortality, cost and repeat visits, and what does not.
Key takeaways
Admission and mortality followed clinical severity, mainly diagnosis and triage level, and not insurance, distance, payment method or gender. HIV/AIDS drove most repeat care, with a 38% repeat-visit rate. Monthly visits more than doubled in June 2024, putting the heaviest strain on Surgery, ICU/HDU and Emergency.