---
title: "Discussion on \" Hospitals Are Penalised Every Time a Diabetic Patient Comes Back. I Built the Model That Predicts Who Will.\" | Hashnode"
description: "Discussion on \" Hospitals Are Penalised Every Time a Diabetic Patient Comes Back. I Built the Model That Predicts Who Will.\". How I used XGBoost, threshold calibration, SHAP, and causal thinking to predict 30-day diabetic readmissions — and why accuracy was the last metric I cared about"
url: "https://hashnode.com/posts/when-the-model-says-go-home-and-the-patient-comes-back-building-a-hospital-readmission-predictor-that-actually-works/69a6eec656428acc6fe9fc42"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Hospitals Are Penalised Every Time a Diabetic Patient Comes Back. I Built the Model That Predicts Who Will.](https://tesskamau.hashnode.dev/when-the-model-says-go-home-and-the-patient-comes-back-building-a-hospital-readmission-predictor-that-actually-works?utm_source=hashnode&utm_medium=feed)

How I used XGBoost, threshold calibration, SHAP, and causal thinking to predict 30-day diabetic readmissions — and why accuracy was the last metric I cared about

Mar 3 · 21 min read · tesskamau.hashnode.dev  
Read the article

## Written by

[Tee Wanjiru](https://hashnode.com/@TessKamau)

@TessKamau

  
[View profile](https://hashnode.com/@TessKamau)

Data Scientist

## Responses

Markdown supported · ⌘↵ to post

No responses yet.
