| import pandas as pd | |
| import os | |
| import random | |
| from faker import Faker | |
| def generate_advanced_data(n=1000): | |
| """ | |
| Generate synthetic patient data for hospital readmission prediction. | |
| Includes clinical and social features commonly used in care management models. | |
| """ | |
| fake = Faker() | |
| diagnoses = ['Diabetes', 'CHF', 'COPD', 'CKD', 'Depression'] | |
| data = [] | |
| for i in range(n): | |
| chronic = random.randint(0, 5) | |
| er_visits = random.randint(0, 4) | |
| adherence = random.choice(['Low', 'Medium', 'High']) | |
| prev_readmits = random.randint(0, 3) | |
| age = random.randint(45, 90) | |
| risk_score = ( | |
| chronic * 1.2 + | |
| er_visits + | |
| (0 if adherence == 'High' else 1) + | |
| prev_readmits + | |
| (1 if random.random() > 0.5 else 0) | |
| ) | |
| readmitted = 1 if risk_score > 5 else 0 | |
| record = { | |
| "Patient_ID": f"P{i+1:04}", | |
| "Age": age, | |
| "Gender": random.choice(["M", "F"]), | |
| "Chronic_Conditions": chronic, | |
| "Primary_Diagnosis": random.choice(diagnoses), | |
| "Num_ER_Visits": er_visits, | |
| "Last_Discharge_Days_Ago": random.randint(1, 60), | |
| "Previous_Readmissions": prev_readmits, | |
| "FollowUp_Scheduled": random.choice(["Yes", "No"]), | |
| "Medication_Adherence": adherence, | |
| "Language_Barrier": random.choice(["Yes", "No"]), | |
| "Housing_Instability": random.choice(["Yes", "No"]), | |
| "Caregiver_Support": random.choice(["Yes", "No"]), | |
| "Readmitted_30_Days": readmitted | |
| } | |
| data.append(record) | |
| return pd.DataFrame(data) | |
| def load_or_generate_data(path="data/patients.csv", n=1000): | |
| """ | |
| Load existing patient data if available, otherwise generate synthetic data. | |
| """ | |
| if os.path.exists(path): | |
| return pd.read_csv(path) | |
| df = generate_advanced_data(n=n) | |
| os.makedirs(os.path.dirname(path), exist_ok=True) | |
| df.to_csv(path, index=False) | |
| return df | |