TruBridge Patient No-Show Analysis (EDA + dashboard)

Live dashboard: https://patient-noshow-dashboard.vercel.app/
Repo: https://github.com/ivan-711/Externship-TruBridge
Colab notebook: https://colab.research.google.com/drive/1Yb-MIOGAxo0x3LVX7mIhcRuF0lFKLEs_?usp=sharing
One-pager (PDF): one-pager-trubridge-noshow.pdf

Patient no-show analysis built from the Kaggle “Medical Appointment No Shows” dataset (May 2016).
I cleaned the raw data, engineered WaitingDays, ran basic statistical tests, trained a baseline logistic regression model, and published findings in a React dashboard.

---

What’s included

• Final Colab notebook (cleaning + EDA + stats + baseline model)
• Datasets
  • publicsafefinalnoshowdataset.csv (public-safe: removed PatientId and AppointmentId for the dashboard)
• React dashboard (Vite + React)

---

Key questions

• How does waiting time (lead time) relate to no-shows?
• Do SMS reminders correlate with no-show rate?
• Which age groups have higher no-show rates?
• Do no-shows shift week-to-week?

---

Data + target

• Source: Kaggle “Medical Appointment No Shows” (May 2016)
• Target: NoShow (0 = show, 1 = no-show)
• Key fields used: Age, SMS_received, ScheduledDay, AppointmentDay
• Feature engineered: WaitingDays = AppointmentDay - ScheduledDay (days)
• Cleaning rules (high level):
  • Dropped negative waiting days
  • Kept waiting days in a realistic window (0–365)

---

Results snapshot

• No-show rate: ~20%
• WaitingDays is higher for no-shows (mean show ≈ 8.75 days vs mean no-show ≈ 15.84 days; p-value < 0.001)
• SMS_received is associated with different no-show rates (chi-square statistic ≈ 1767.98, p-value < 0.001)
  Association only. This does not prove SMS causes no-shows.

---

Baseline model (logistic regression)

• Features: Age_normalized, WaitingDays_normalized, SMS_received
• Baseline accuracy (predict all “show”): ~0.7981
• Model accuracy: ~0.7960
• Takeaway: class imbalance + limited feature signal with this small feature set

---

Dashboard screenshots

Overview

Charts

Summary

---

Repo structure

• patient-dashboard/ → React + Vite dashboard app
• *.csv → cleaned datasets used for analysis + dashboard

---

Run the dashboard locally

cd patient-dashboard
npm install
npm run dev