cdss_engine.py

import pandas as pd
import numpy as np
import os

class CDSSCoreEngine:
    def __init__(self, data_path: str):
        """
        Initializes the CDSS Engine, ingests raw clinical CSV files,
        and normalizes naming schemas.
        """
        if not os.path.exists(data_path):
            raise FileNotFoundError(f"Target data file at {data_path} not found.")
        self.df = pd.read_csv(data_path)
        self._sanitize_headers()

    def _sanitize_headers(self):
        """Normalizes column names to standard snake_case formatting."""
        self.df.columns = [str(col).strip().replace(" ", "_").replace("/", "_") for col in self.df.columns]

    def process_clinical_audit(self, output_report_path: str):
        """
        Iterates across patient data entries, evaluates the 12 input features,
        calculates compliance metrics, and generates a structured audit report.
        """
        audit_records = []
        
        for idx, row in self.df.iterrows():
            # Extract identifiers and metadata safely
            patient_id = row.get('Patient_ID', f"PID_VAL_{idx + 1001}")
            
            # Phase 1: Input Variable Extraction (Fields 1-12 Logic)
            dx = row.get('Diagnosis') # 1=STEMI, 2=NSTEMI, 3=UA
            age_flag = row.get('Age_Group', 0) # 0=<75, 1==>75
            spo2 = row.get('SpO2', 95) # Continuous numeric percentage
            
            # Additional continuous and medical history variables
            sbp = row.get('SBP', 120)
            hr = row.get('HR', 75)
            troponin = row.get('Troponin', 1 if dx in [1, 2] else 0)
            ecg = row.get('ECG', 0)
            
            # Phase 2: Treatment & Management Variable Extraction
            reperfusion = row.get('Reperfusion_Strategy', 4) # 1=PPCI, 2=Fibrinolytics, 3=CABG, 4=MedMgmt
            p2y12 = row.get('P2Y12_Inhibitor', 0) # 0=None, 1=Clopidogrel, 2=Ticagrelor, 3=Prasugrel
            hbr_status = row.get('Bleeding_Risk_HBR', 0) # 0=Low, 1=High Bleeding Risk
            statin = row.get('Statin_Intensity', 0) # 0=None, 1=Low/Mod, 2=High
            ldl = row.get('Follow_Up_LDL', 85) # Continuous follow-up value
            ppi_status = row.get('PPI_Prescribed', 0) # Binary status field
            oac_status = row.get('Concurrent_OAC', 0) # Binary concurrent oral anticoagulant field
            
            # Specific ischemic and bleeding indicators (reconstructed from patient data for fine-grained rules)
            ischemicInstability = 1 if (sbp < 90 or hr > 100) else 0
            ischemicHeartFailure = 1 if (row.get('ischemicHeartFailure') == 1) else 0
            ischemicAngina = 1 if (row.get('ischemicAngina') == 1) else 0
            ischemicArrhythmia = 1 if (row.get('ischemicArrhythmia') == 1) else 0
            
            is_high_ischemic = (troponin == 1 or ischemicInstability == 1 or 
                                ischemicAngina == 1 or ischemicHeartFailure == 1 or 
                                ischemicArrhythmia == 1)
            
            # ARC-HBR components
            bleedingRenal = row.get('bleedingRenal', 0)
            bleedingAnemia = row.get('bleedingAnemia', 0)
            bleedingAF = row.get('bleedingAF', 0)
            bleedingStroke = row.get('bleedingStroke', 0)
            
            is_high_bleeding = (age_flag == 1 or hbr_status == 1 or bleedingRenal == 1 or 
                                bleedingAnemia == 1 or bleedingAF == 1 or bleedingStroke == 1)

            # Evaluation containers
            system_deviations = []
            predicted_best_regimen = []
            safety_alerts = []

            # --- SYSTEM VALIDATION MODULES ---
            
            # 1. Oxygenation Check
            if pd.notna(spo2) and spo2 < 90:
                safety_alerts.append("HYPOXEMIA ALERT: Patient SpO2 < 90%. Administer supplementary oxygen immediately.")

            # 2. Reperfusion Strategy vs Antiplatelet Mapping Rules
            if dx == 1: # STEMI Primary Classification
                if reperfusion == 2: # Fibrinolytic Therapy Path
                    if age_flag == 1 or is_high_bleeding:
                        predicted_best_regimen.append("Aspirin (162-325mg load) + Clopidogrel (75mg daily, no load) [Guideline Preferred]")
                    else:
                        predicted_best_regimen.append("Aspirin (162-325mg load) + Clopidogrel (300mg load) [Guideline Preferred]")
                    
                    if p2y12 in [2, 3]:
                        system_deviations.append("CRITICAL MISMATCH: Potent antiplatelets (Prasugrel/Ticagrelor) run concurrently with Fibrinolytics.")
                        safety_alerts.append("HEMORRHAGE RISK HIGH: Discontinue Prasugrel/Ticagrelor immediately to mitigate intracranial bleeding risk.")
                elif reperfusion == 1: # Primary PCI Path
                    predicted_best_regimen.append("Aspirin (162-325mg load) + Prasugrel (60mg load) OR Ticagrelor (180mg load)")
                    if p2y12 == 1 and not is_high_bleeding:
                        system_deviations.append("SUBOPTIMAL PROTECTION: Clopidogrel used in standard risk PPCI path.")
                        safety_alerts.append("EFFICACY WARNING: Upgrade to Ticagrelor or Prasugrel for superior protection against stent thrombosis.")
                    elif p2y12 == 0:
                        system_deviations.append("CRITICAL OMISSION: A P2Y12 inhibitor is required for all STEMI patients undergoing Primary PCI.")
            elif dx in [2, 3]: # NSTE-ACS Protocols (NSTEMI / Unstable Angina)
                if reperfusion == 4: # Conservative Medical Management
                    predicted_best_regimen.append("Aspirin (81mg daily) + Ticagrelor (90mg BID for 12 months)")
                    if p2y12 == 1 and not is_high_bleeding:
                        system_deviations.append("NON-PREFERENTIAL REGIMEN: Clopidogrel selected for standard risk medical management.")
                elif reperfusion == 1: # Invasive Strategy
                    predicted_best_regimen.append("Aspirin + Ticagrelor (or Prasugrel administered at angiography)")
                    if p2y12 == 1 and not is_high_bleeding:
                        system_deviations.append("SUBOPTIMAL EFFICACY: Clopidogrel is less effective than Ticagrelor/Prasugrel. Switch to a high-potency agent unless contraindicated or high bleeding risk exists.")

            # 3. NSTE-ACS Invasive Strategy Timing
            if dx in [2, 3]:
                if is_high_ischemic:
                    safety_alerts.append("CRITICAL: Immediate Invasive Strategy Required (Angiography within 2 hours) due to presence of High Ischemic Risk Indicators (cardiogenic shock, refractory angina, heart failure, or sustained VT/VF).")
                else:
                    predicted_best_regimen.append("Guideline Compliant: Early Invasive Strategy (Angiography within 24 hours) is indicated for standard-risk NSTE-ACS.")

            # 4. High Bleeding Risk (HBR) Custom Safe-Overrides
            if is_high_bleeding:
                if p2y12 == 3: # Prasugrel assigned to HBR
                    system_deviations.append("CONTRAINDICATION OVERRIDE: Prasugrel allocated within high bleeding risk profile.")
                    safety_alerts.append("STOP SIGN: Immediately discontinue Prasugrel. Substitute with Clopidogrel 75mg daily.")
                
                # Check for protective PPI therapy (HBR is a gastric protection trigger)
                if ppi_status != 1:
                    safety_alerts.append("GASTRIC GAP: Co-prescribe a protective PPI (e.g., Pantoprazole) to manage GI bleeding margins.")
                
                # HBR Rules A, B, C
                if oac_status == 1 or bleedingAF == 1: # Rule C
                    predicted_best_regimen.append("Convert to Dual Therapy (OAC + Clopidogrel 75mg) after 1-4 weeks; completely drop Aspirin.")
                elif row.get('minorBleeding') == 1 or row.get('minor_bleeding') == 1: # Rule B
                    predicted_best_regimen.append("P2Y12 Down-Titration: Transition from Ticagrelor/Prasugrel to Clopidogrel 75mg daily due to bleeding/Hb drop.")
                else: # Rule A
                    predicted_best_regimen.append("Short-Course DAPT: Transition to Ticagrelor Monotherapy after 1-3 months, completely dropping Aspirin.")
            else:
                # If not high bleeding, check other GI triggers (peptic ulcer, concurrent NSAIDs)
                has_gi_trigger = (row.get('pepticUlcer') == 1 or row.get('nsaid') == 1)
                if has_gi_trigger and ppi_status != 1:
                    safety_alerts.append("GASTRIC GAP: Co-prescribe a protective PPI (e.g., Pantoprazole) to manage GI bleeding margins.")

            # PPI Metabolic Optimization Check
            if p2y12 == 1 and ppi_status == 1:
                safety_alerts.append("PPI Metabolic Optimization: If Clopidogrel is prescribed, Pantoprazole is preferred over Omeprazole (prevents CYP2C19 inhibition).")

            # 5. Secondary Lipid Prevention Engine (Ref: 2025 Guideline Fig 5)
            if statin != 2:
                system_deviations.append("OMISSION FAULT: Patient not initiated on baseline High-Intensity Statin therapy.")
                predicted_best_regimen.append("Initiate High-Intensity Atorvastatin 40-80mg or Rosuvastatin 20-40mg.")

            if pd.notna(ldl):
                if ldl >= 70:
                    system_deviations.append("LIPID GOAL MISSED: Follow-up LDL tracking >= 70 mg/dL.")
                    predicted_best_regimen.append("Add Non-Statin Therapy: Append Ezetimibe 10mg daily immediately (Class 1).")
                elif 55 <= ldl < 70:
                    predicted_best_regimen.append("Therapeutic Optimization: Evaluate addition of Ezetimibe to reach target < 55 mg/dL (Class 2a).")

            # Final aggregation formatting
            audit_records.append({
                "Patient_ID": patient_id,
                "Audit_Status": "FAIL (Deviations Found)" if system_deviations else "PASS (Guideline Compliant)",
                "Identified_Deviations": " | ".join(system_deviations) if system_deviations else "None Identified",
                "Safety_Alert_Overrides": " | ".join(safety_alerts) if safety_alerts else "Patient Stable",
                "Predicted_Optimal_Management": " | ".join(predicted_best_regimen) if predicted_best_regimen else "Maintain Current Schema"
            })
            
        # Save the audit report
        output_df = pd.DataFrame(audit_records)
        output_df.to_csv(output_report_path, index=False)
        return output_df

if __name__ == "__main__":
    source_csv = "IHD data entry.xlsx - IHD.csv"
    target_report = "CDSS_Final_Clinical_Audit.csv"
    
    try:
        engine = CDSSCoreEngine(data_path=source_csv)
        results = engine.process_clinical_audit(output_report_path=target_report)
        print(f"CDSS Engine successfully processed {len(results)} client records.")
        print(f"Audit matrix written to: {target_report}")
    except Exception as e:
        print(f"Engine execution halted. Parameter Exception: {str(e)}")