# TB-SCREEN

**Tuberculosis Screening and Latent TB Reactivation Risk Stratification Before Biologic Therapy in Rheumatic Diseases with Monte Carlo Uncertainty Estimation**

## Authors
Erick Adrián Zamora Tehozol, DNAI, RheumaAI

## Overview
TB-SCREEN is an agent-executable clinical decision support tool that stratifies tuberculosis reactivation risk in patients with autoimmune rheumatic diseases initiating biologic or targeted synthetic DMARD therapy. It integrates TST/IGRA results, chest radiography findings, epidemiologic exposure history, immunosuppression burden, and biologic-specific TB risk profiles to generate a composite risk score (0–100) with Monte Carlo confidence intervals.

## Clinical Problem
Biologic therapies — particularly TNF-α inhibitors — carry significant risk of reactivating latent tuberculosis infection (LTBI). The WHO estimates one-quarter of the global population has LTBI. Reactivation rates with anti-TNF agents range from 4–40× baseline depending on the agent and population. Current screening relies on clinician judgment integrating multiple data streams (TST, IGRA, CXR, exposure history, immunosuppression level). TB-SCREEN systematizes this assessment.

## Scoring Domains (10 domains, max 100)
1. **TST Result** (0–12): Induration mm, immunosuppression-adjusted cutoffs (≥5mm for immunosuppressed per ATS/IDSA)
2. **IGRA Result** (0–15): QuantiFERON/T-SPOT, indeterminate handling, discordance with TST
3. **Chest Radiograph** (0–12): Apical scarring, calcified granulomas, cavitation, lymphadenopathy
4. **Epidemiologic Risk** (0–10): Endemic country birth/residence, household contact, congregate settings, healthcare worker
5. **Immunosuppression Burden** (0–12): Current GC dose, csDMARD count, prior cyclophosphamide/rituximab
6. **Biologic TB Risk Profile** (0–15): Agent-specific risk (infliximab > adalimumab > etanercept; JAKi intermediate; abatacept/tocilizumab moderate; rituximab high)
7. **HIV/Comorbidity** (0–8): HIV status, diabetes, CKD, malnutrition, silicosis
8. **Prior TB History** (0–8): Treated vs untreated, completeness of prior LTBI therapy
9. **Age & Demographics** (0–4): Age >65, pediatric, recent immigration (<5 years)
10. **Laboratory Markers** (0–4): Lymphopenia, low CD4, hypoalbuminemia

## Risk Categories
- **0–15**: Low Risk → Proceed with biologic, standard monitoring
- **16–35**: Moderate Risk → Consider LTBI treatment before biologic, recheck IGRA in 1–3 months
- **36–55**: High Risk → LTBI treatment mandatory before biologic initiation, 1–2 month lead time
- **56–100**: Very High Risk → Infectious disease consultation required, active TB workup (sputum AFB/culture, CT chest), defer biologic until TB ruled out or treatment completed

## Evidence Base
- Winthrop KL et al. Tuberculosis and other opportunistic infections in tofacitinib-treated patients with rheumatoid arthritis. Ann Rheum Dis 2016;75:1133–8.
- Keane J et al. Tuberculosis associated with infliximab, a TNF-α–neutralizing agent. N Engl J Med 2001;345:1098–104.
- Singh JA et al. 2015 ACR Guideline for the Treatment of RA. Arthritis Rheumatol 2016;68:1–26.
- Lewinsohn DM et al. Official ATS/IDSA/CDC Clinical Practice Guidelines: Diagnosis of TB in Adults and Children. Clin Infect Dis 2017;64:e1–e33.
- Cantini F et al. Risk of tuberculosis reactivation in patients with rheumatoid arthritis, ankylosing spondylitis, and psoriatic arthritis receiving non-anti-TNF-targeted biologics. Mediators Inflamm 2017;2017:8909834.
- Solovic I et al. The risk of tuberculosis related to TNF antagonist therapies: a TBNET consensus statement. Eur Respir J 2010;36:1185–206.
- WHO Guidelines on the management of LTBI. Geneva: WHO; 2015.

## Usage
```bash
python3 tb_screen.py
```

## Output
JSON with composite score, risk category, domain breakdown, Monte Carlo 95% CI, and clinical recommendations including specific LTBI treatment regimens and biologic timing guidance.



## Executable Code

```python
#!/usr/bin/env python3
"""
TB-SCREEN: Tuberculosis Screening and Latent TB Reactivation Risk Stratification
Before Biologic Therapy in Rheumatic Diseases with Monte Carlo Uncertainty Estimation

Authors: Erick Adrián Zamora Tehozol, DNAI, RheumaAI
License: MIT
"""

import json, math, random, statistics
from dataclasses import dataclass, field, asdict
from typing import Optional, List

# ── Domain 1: TST Result (0–12) ──────────────────────────────────────────────
def score_tst(induration_mm: float, immunosuppressed: bool = True) -> dict:
    """TST scoring with immunosuppression-adjusted cutoffs (ATS/IDSA 2017)."""
    cutoff = 5 if immunosuppressed else 10
    if induration_mm < cutoff:
        score = 0
    elif induration_mm < cutoff + 5:
        score = 6
    elif induration_mm < cutoff + 10:
        score = 9
    else:
        score = 12
    return {"domain": "TST Result", "score": score, "max": 12,
            "detail": f"{induration_mm}mm (cutoff ≥{cutoff}mm)"}

# ── Domain 2: IGRA Result (0–15) ─────────────────────────────────────────────
def score_igra(result: str, tst_positive: bool = False) -> dict:
    """IGRA scoring: positive/negative/indeterminate/not_done."""
    r = result.lower().strip()
    if r == "positive":
        score = 15
    elif r == "indeterminate":
        score = 8 if tst_positive else 5
    elif r == "negative":
        score = 2 if tst_positive else 0  # discordance penalty
    else:  # not_done
        score = 3
    return {"domain": "IGRA Result", "score": score, "max": 15, "detail": result}

# ── Domain 3: Chest Radiograph (0–12) ────────────────────────────────────────
CXR_FINDINGS = {
    "normal": 0, "calcified_granuloma": 4, "apical_scarring": 6,
    "fibronodular": 8, "lymphadenopathy": 7, "cavitation": 12,
    "pleural_thickening": 3, "miliary_pattern": 12
}

def score_cxr(findings: List[str]) -> dict:
    """CXR scoring — takes worst finding."""
    scores = [CXR_FINDINGS.get(f.lower().strip(), 0) for f in findings]
    score = min(max(scores) if scores else 0, 12)
    return {"domain": "Chest Radiograph", "score": score, "max": 12,
            "detail": ", ".join(findings) if findings else "normal"}

# ── Domain 4: Epidemiologic Risk (0–10) ──────────────────────────────────────
def score_epidemiology(endemic_country: bool = False, household_contact: bool = False,
                       congregate_setting: bool = False, healthcare_worker: bool = False) -> dict:
    score = 0
    details = []
    if endemic_country:
        score += 5; details.append("endemic country")
    if household_contact:
        score += 4; details.append("household TB contact")
    if congregate_setting:
        score += 2; details.append("congregate setting")
    if healthcare_worker:
        score += 2; details.append("healthcare worker")
    score = min(score, 10)
    return {"domain": "Epidemiologic Risk", "score": score, "max": 10,
            "detail": ", ".join(details) or "none"}

# ── Domain 5: Immunosuppression Burden (0–12) ────────────────────────────────
def score_immunosuppression(gc_dose_mg: float = 0, csdmard_count: int = 0,
                            prior_cyclophosphamide: bool = False,
                            prior_rituximab: bool = False) -> dict:
    score = 0
    details = []
    # GC dose
    if gc_dose_mg >= 20:
        score += 5; details.append(f"pred {gc_dose_mg}mg (high)")
    elif gc_dose_mg >= 10:
        score += 3; details.append(f"pred {gc_dose_mg}mg (moderate)")
    elif gc_dose_mg > 0:
        score += 1; details.append(f"pred {gc_dose_mg}mg (low)")
    # csDMARDs
    if csdmard_count >= 3:
        score += 3; details.append(f"{csdmard_count} csDMARDs (triple)")
    elif csdmard_count >= 2:
        score += 2; details.append(f"{csdmard_count} csDMARDs")
    elif csdmard_count == 1:
        score += 1; details.append("1 csDMARD")
    # Prior heavy immunosuppression
    if prior_cyclophosphamide:
        score += 3; details.append("prior cyclophosphamide")
    if prior_rituximab:
        score += 2; details.append("prior rituximab")
    score = min(score, 12)
    return {"domain": "Immunosuppression Burden", "score": score, "max": 12,
            "detail": ", ".join(details) or "minimal"}

# ── Domain 6: Biologic TB Risk Profile (0–15) ────────────────────────────────
BIOLOGIC_TB_RISK = {
    "infliximab": 15, "adalimumab": 12, "certolizumab": 10, "golimumab": 10,
    "etanercept": 5, "rituximab": 9, "abatacept": 7, "tocilizumab": 7,
    "sarilumab": 7, "tofacitinib": 8, "baricitinib": 8, "upadacitinib": 8,
    "filgotinib": 7, "secukinumab": 3, "ixekizumab": 3, "guselkumab": 3,
    "belimumab": 4, "anifrolumab": 5, "none": 0
}

def score_biologic(agent: str) -> dict:
    key = agent.lower().strip()
    score = BIOLOGIC_TB_RISK.get(key, 6)
    return {"domain": "Biologic TB Risk Profile", "score": min(score, 15), "max": 15,
            "detail": agent}

# ── Domain 7: HIV/Comorbidity (0–8) ──────────────────────────────────────────
def score_comorbidity(hiv: bool = False, diabetes: bool = False,
                      ckd: bool = False, malnutrition: bool = False,
                      silicosis: bool = False) -> dict:
    score = 0; details = []
    if hiv: score += 4; details.append("HIV+")
    if diabetes: score += 2; details.append("DM")
    if ckd: score += 2; details.append("CKD")
    if malnutrition: score += 2; details.append("malnutrition")
    if silicosis: score += 2; details.append("silicosis")
    score = min(score, 8)
    return {"domain": "HIV/Comorbidity", "score": score, "max": 8,
            "detail": ", ".join(details) or "none"}

# ── Domain 8: Prior TB History (0–8) ─────────────────────────────────────────
def score_tb_history(prior_active_tb: bool = False, prior_ltbi_treatment: bool = False,
                     ltbi_treatment_complete: bool = False) -> dict:
    score = 0; details = []
    if prior_active_tb:
        score += 5; details.append("prior active TB")
    if prior_ltbi_treatment and ltbi_treatment_complete:
        score += 1; details.append("LTBI treated (complete)")
    elif prior_ltbi_treatment and not ltbi_treatment_complete:
        score += 4; details.append("LTBI treated (incomplete)")
    score = min(score, 8)
    return {"domain": "Prior TB History", "score": score, "max": 8,
            "detail": ", ".join(details) or "no prior TB"}

# ── Domain 9: Age & Demographics (0–4) ───────────────────────────────────────
def score_demographics(age: int, recent_immigrant: bool = False) -> dict:
    score = 0; details = []
    if age >= 65:
        score += 2; details.append("age ≥65")
    if age < 5:
        score += 2; details.append("pediatric <5")
    if recent_immigrant:
        score += 2; details.append("recent immigrant <5yr")
    score = min(score, 4)
    return {"domain": "Age & Demographics", "score": score, "max": 4,
            "detail": ", ".join(details) or f"age {age}"}

# ── Domain 10: Laboratory Markers (0–4) ──────────────────────────────────────
def score_labs(lymphocyte_count: Optional[float] = None,
               cd4_count: Optional[float] = None,
               albumin: Optional[float] = None) -> dict:
    score = 0; details = []
    if lymphocyte_count is not None and lymphocyte_count < 1.0:
        score += 2; details.append(f"lymphopenia {lymphocyte_count}")
    if cd4_count is not None and cd4_count < 200:
        score += 2; details.append(f"CD4 {cd4_count}")
    if albumin is not None and albumin < 3.0:
        score += 1; details.append(f"albumin {albumin}")
    score = min(score, 4)
    return {"domain": "Laboratory Markers", "score": score, "max": 4,
            "detail": ", ".join(details) or "within normal limits"}


# ── Monte Carlo Simulation ────────────────────────────────────────────────────
def monte_carlo(domain_scores: list, n: int = 10000, seed: int = 42) -> dict:
    rng = random.Random(seed)
    composite = sum(d["score"] for d in domain_scores)
    simulated = []
    for _ in range(n):
        total = 0
        for d in domain_scores:
            noise = rng.gauss(0, d["max"] * 0.08)
            perturbed = max(0, min(d["max"], d["score"] + noise))
            total += perturbed
        simulated.append(total)
    simulated.sort()
    ci_low = round(simulated[int(0.025 * n)], 1)
    ci_high = round(simulated[int(0.975 * n)], 1)
    return {"composite": round(composite, 1), "ci_95_low": ci_low, "ci_95_high": ci_high,
            "mean": round(statistics.mean(simulated), 1),
            "sd": round(statistics.stdev(simulated), 1), "n_simulations": n}


# ── Risk Classification ──────────────────────────────────────────────────────
def classify(score: float) -> dict:
    if score <= 15:
        return {"category": "Low", "action": "Proceed with biologic. Standard TB monitoring.",
                "ltbi_treatment": "Not indicated unless IGRA/TST positive."}
    elif score <= 35:
        return {"category": "Moderate",
                "action": "Consider LTBI treatment before biologic. Recheck IGRA in 1-3 months.",
                "ltbi_treatment": "Isoniazid 300mg daily × 9 months OR Rifampin 600mg daily × 4 months."}
    elif score <= 55:
        return {"category": "High",
                "action": "LTBI treatment mandatory before biologic initiation. 1-2 month lead time recommended.",
                "ltbi_treatment": "Isoniazid + Rifapentine weekly × 12 weeks (3HP) OR Isoniazid 9 months. Start biologic after ≥1 month of LTBI therapy."}
    else:
        return {"category": "Very High",
                "action": "Infectious disease consultation required. Active TB workup: sputum AFB ×3, mycobacterial culture, CT chest. Defer biologic until TB ruled out or treatment completed.",
                "ltbi_treatment": "If active TB: standard 4-drug regimen (RIPE) × 6-9 months. Do NOT start biologic until ≥2 months intensive phase completed with culture conversion."}


# ── Main Assessment ──────────────────────────────────────────────────────────
def tb_screen(patient: dict) -> dict:
    domains = [
        score_tst(patient.get("tst_mm", 0), patient.get("immunosuppressed", True)),
        score_igra(patient.get("igra", "not_done"),
                   patient.get("tst_mm", 0) >= (5 if patient.get("immunosuppressed", True) else 10)),
        score_cxr(patient.get("cxr_findings", ["normal"])),
        score_epidemiology(patient.get("endemic_country", False),
                          patient.get("household_contact", False),
                          patient.get("congregate_setting", False),
                          patient.get("healthcare_worker", False)),
        score_immunosuppression(patient.get("gc_dose_mg", 0),
                               patient.get("csdmard_count", 0),
                               patient.get("prior_cyclophosphamide", False),
                               patient.get("prior_rituximab", False)),
        score_biologic(patient.get("biologic", "none")),
        score_comorbidity(patient.get("hiv", False), patient.get("diabetes", False),
                         patient.get("ckd", False), patient.get("malnutrition", False),
                         patient.get("silicosis", False)),
        score_tb_history(patient.get("prior_active_tb", False),
                        patient.get("prior_ltbi_treatment", False),
                        patient.get("ltbi_treatment_complete", False)),
        score_demographics(patient.get("age", 50),
                          patient.get("recent_immigrant", False)),
        score_labs(patient.get("lymphocyte_count"), patient.get("cd4_count"),
                  patient.get("albumin")),
    ]
    mc = monte_carlo(domains)
    risk = classify(mc["composite"])
    return {
        "tool": "TB-SCREEN",
        "version": "1.0.0",
        "composite_score": mc["composite"],
        "max_possible": 100,
        "risk_category": risk["category"],
        "clinical_action": risk["action"],
        "ltbi_treatment": risk["ltbi_treatment"],
        "monte_carlo_95CI": [mc["ci_95_low"], mc["ci_95_high"]],
        "mc_mean": mc["mean"],
        "mc_sd": mc["sd"],
        "domains": domains,
        "biologic_specific_note": _biologic_note(patient.get("biologic", "none")),
        "references": [
            "Keane J et al. NEJM 2001;345:1098-104",
            "Lewinsohn DM et al. Clin Infect Dis 2017;64:e1-e33 (ATS/IDSA/CDC)",
            "Solovic I et al. Eur Respir J 2010;36:1185-206 (TBNET)",
            "Winthrop KL et al. Ann Rheum Dis 2016;75:1133-8",
            "Cantini F et al. Mediators Inflamm 2017;2017:8909834",
            "WHO LTBI Management Guidelines 2015",
            "Singh JA et al. Arthritis Rheumatol 2016;68:1-26 (ACR 2015)"
        ]
    }

def _biologic_note(agent: str) -> str:
    notes = {
        "infliximab": "Highest TB reactivation risk among TNFi (monoclonal Ab, complement-mediated macrophage apoptosis). Mandatory LTBI screening. Granuloma dissolution risk.",
        "adalimumab": "High TB risk (monoclonal Ab). Screen before initiation. Consider etanercept if TB risk factors present.",
        "etanercept": "Lower TB risk vs monoclonal TNFi (soluble receptor, preserves granuloma integrity). Still requires screening.",
        "tofacitinib": "JAK inhibitor — intermediate TB risk. Impairs IFN-γ signaling critical for TB containment.",
        "rituximab": "B-cell depletion impairs granuloma maintenance. Screen for LTBI. Delayed immune reconstitution prolongs risk window.",
    }
    return notes.get(agent.lower().strip(), f"Standard LTBI screening required before {agent}.")


# ── Demo Scenarios ────────────────────────────────────────────────────────────
if __name__ == "__main__":
    scenarios = [
        {
            "name": "Scenario 1: RA patient, Mexico-born, starting infliximab",
            "patient": {
                "age": 52, "tst_mm": 12, "igra": "positive",
                "cxr_findings": ["calcified_granuloma"], "endemic_country": True,
                "gc_dose_mg": 10, "csdmard_count": 2, "biologic": "infliximab",
                "immunosuppressed": True
            }
        },
        {
            "name": "Scenario 2: Young RA, US-born, etanercept, all negative",
            "patient": {
                "age": 34, "tst_mm": 0, "igra": "negative",
                "cxr_findings": ["normal"], "endemic_country": False,
                "gc_dose_mg": 5, "csdmard_count": 1, "biologic": "etanercept",
                "immunosuppressed": True
            }
        },
        {
            "name": "Scenario 3: SLE, HIV+, prior TB, rituximab, endemic",
            "patient": {
                "age": 45, "tst_mm": 18, "igra": "positive",
                "cxr_findings": ["apical_scarring", "lymphadenopathy"],
                "endemic_country": True, "household_contact": True,
                "gc_dose_mg": 20, "csdmard_count": 2,
                "prior_cyclophosphamide": True, "biologic": "rituximab",
                "hiv": True, "diabetes": True,
                "prior_active_tb": True, "prior_ltbi_treatment": True,
                "ltbi_treatment_complete": False,
                "lymphocyte_count": 0.6, "cd4_count": 180, "albumin": 2.8,
                "immunosuppressed": True
            }
        }
    ]

    for s in scenarios:
        print(f"\n{'='*70}")
        print(f"  {s['name']}")
        print(f"{'='*70}")
        result = tb_screen(s["patient"])
        print(f"  Composite Score: {result['composite_score']}/100")
        print(f"  Risk Category:   {result['risk_category']}")
        print(f"  95% CI:          [{result['monte_carlo_95CI'][0]}, {result['monte_carlo_95CI'][1]}]")
        print(f"  Action:          {result['clinical_action']}")
        print(f"  LTBI Treatment:  {result['ltbi_treatment']}")
        print(f"  Biologic Note:   {result['biologic_specific_note']}")
        print(f"  Domains:")
        for d in result["domains"]:
            print(f"    {d['domain']:30s} {d['score']:5.0f}/{d['max']} — {d['detail']}")
        print()

```


## Demo Output

```
lt                        0/15 — negative
    Chest Radiograph                   0/12 — normal
    Epidemiologic Risk                 0/10 — none
    Immunosuppression Burden           2/12 — pred 5mg (low), 1 csDMARD
    Biologic TB Risk Profile           5/15 — etanercept
    HIV/Comorbidity                    0/8 — none
    Prior TB History                   0/8 — no prior TB
    Age & Demographics                 0/4 — age 34
    Laboratory Markers                 0/4 — within normal limits


======================================================================
  Scenario 3: SLE, HIV+, prior TB, rituximab, endemic
======================================================================
  Composite Score: 80/100
  Risk Category:   Very High
  95% CI:          [74.2, 83.1]
  Action:          Infectious disease consultation required. Active TB workup: sputum AFB ×3, mycobacterial culture, CT chest. Defer biologic until TB ruled out or treatment completed.
  LTBI Treatment:  If active TB: standard 4-drug regimen (RIPE) × 6-9 months. Do NOT start biologic until ≥2 months intensive phase completed with culture conversion.
  Biologic Note:   B-cell depletion impairs granuloma maintenance. Screen for LTBI. Delayed immune reconstitution prolongs risk window.
  Domains:
    TST Result                        12/12 — 18mm (cutoff ≥5mm)
    IGRA Result                       15/15 — positive
    Chest Radiograph                   7/12 — apical_scarring, lymphadenopathy
    Epidemiologic Risk                 9/10 — endemic country, household TB contact
    Immunosuppression Burden          10/12 — pred 20mg (high), 2 csDMARDs, prior cyclophosphamide
    Biologic TB Risk Profile           9/15 — rituximab
    HIV/Comorbidity                    6/8 — HIV+, DM
    Prior TB History                   8/8 — prior active TB, LTBI treated (incomplete)
    Age & Demographics                 0/4 — age 45
    Laboratory Markers                 4/4 — lymphopenia 0.6, CD4 180, albumin 2.8


```