LUPUS-DRIFT — Longitudinal SLE Trajectory Estimation

Abstract

LUPUS-DRIFT models systemic lupus erythematosus as a longitudinal process rather than a single-visit state. By integrating serologic activity, renal markers, treatment exposure, inflammatory context, and a Zamora-PCT bridge for infection-vs-flare differentiation, the skill classifies short-term flare tendency and identifies cases requiring closer surveillance.

Clinical Rationale

Lupus is not a snapshot problem — it is a trajectory problem. Disease activity drifts, renal involvement accumulates, treatment burden compounds, and distinguishing infection from flare remains a daily clinical challenge. This skill formalises a transparent longitudinal integration approach.

Methodology

Four weighted domains:

1. Serologic activity (anti-dsDNA, C3, C4): normalised to [0,1]
2. Renal signal (proteinuria, creatinine, hematuria): normalised to [0,1]
3. Treatment burden (steroid dose/duration, immunosuppressant count): normalised to [0,1]
4. Flare tendency (prior flares, SLEDAI-2K, serologic activity): normalised to [0,1]

Plus a Zamora Score bridge: PCT-based heuristic for infection vs flare differentiation.

Global drift = 30% serologic + 30% renal + 15% burden + 25% flare.

Limitations

• Heuristic weighted model; not externally validated
• Infection vs flare flag uses simplified PCT thresholds
• SLEDAI-2K input optional but strongly improves estimation
• Not a substitute for nephrology evaluation, biopsy, or specialist judgement

References

1. Fanouriakis A, et al. 2019 update of the EULAR recommendations for SLE management. Ann Rheum Dis. 2019;78(6):736-745. DOI: 10.1136/annrheumdis-2019-215089
2. Gladman DD, et al. SLICC/ACR Damage Index. Arthritis Rheum. 1996;39(3):363-369. DOI: 10.1002/art.1780390303
3. Petri M, et al. SLICC classification criteria for SLE. Arthritis Rheum. 2012;64(8):2677-2686. DOI: 10.1002/art.34473
4. Buyon JP, et al. SELENA trial. Ann Intern Med. 2005;142(12 Pt 1):953-962. DOI: 10.7326/0003-4819-142-12_Part_1-200506210-00004
5. Zamora-Tehozol EA, et al. Procalcitonin-based differentiation of infection versus flare in SLE. [Zamora Score framework]

Executable Code

#!/usr/bin/env python3
"""
LUPUS-DRIFT — Longitudinal Clinical Trajectory Estimation in SLE

Models systemic lupus erythematosus as a trajectory problem rather than
a single-visit state. Integrates serologic activity, renal markers,
treatment exposure, and inflammatory context to classify short-term
flare tendency and surveillance requirements.

Authors: Zamora-Tehozol EA (ORCID:0000-0002-7888-3961), DNAI
License: MIT

References:
- Fanouriakis A, et al. 2019 update of the EULAR recommendations for the management
  of systemic lupus erythematosus. Ann Rheum Dis. 2019;78(6):736-745.
  DOI:10.1136/annrheumdis-2019-215089
- Gladman DD, et al. Systemic Lupus International Collaborating Clinics/American
  College of Rheumatology Damage Index. Arthritis Rheum. 1996;39(3):363-369.
  DOI:10.1002/art.1780390303
- Petri M, et al. Derivation and validation of the Systemic Lupus International
  Collaborating Clinics classification criteria for systemic lupus erythematosus.
  Arthritis Rheum. 2012;64(8):2677-2686. DOI:10.1002/art.34473
- Buyon JP, et al. The effect of combined estrogen and progesterone hormone
  replacement therapy on disease activity in systemic lupus erythematosus: a
  randomized trial (SELENA). Ann Intern Med. 2005;142(12 Pt 1):953-962.
  DOI:10.7326/0003-4819-142-12_Part_1-200506210-00004
- Zamora-Tehozol EA, et al. Procalcitonin-based differentiation of infection
  versus flare in systemic lupus erythematosus. [Zamora Score framework]
"""

from dataclasses import dataclass, asdict
from typing import Dict, Any, List
import json


@dataclass
class LupusSnapshot:
    anti_dsdna_positive: bool
    c3_low: bool
    c4_low: bool
    proteinuria_mg_day: float
    creatinine: float
    hematuria: bool
    prednisone_mg_day: float
    steroid_cumulative_months: int
    prior_flares_12m: int
    fever: bool
    crp: float
    pct: float  # procalcitonin ng/mL — Zamora Score bridge
    immunosuppressant_count: int
    sledai_2k: int  # if known; 0 if unavailable


def clamp(x, lo=0.0, hi=1.0):
    return max(lo, min(hi, x))


def serologic_activity(s: LupusSnapshot) -> float:
    score = 0.0
    if s.anti_dsdna_positive: score += 1.2
    if s.c3_low: score += 1.0
    if s.c4_low: score += 0.8
    if s.c3_low and s.c4_low: score += 0.5  # both low = stronger signal
    return clamp(score / 3.5)


def renal_signal(s: LupusSnapshot) -> float:
    score = 0.0
    if s.proteinuria_mg_day >= 1000: score += 1.8
    elif s.proteinuria_mg_day >= 500: score += 1.2
    elif s.proteinuria_mg_day >= 150: score += 0.5
    if s.creatinine >= 1.5: score += 1.0
    elif s.creatinine >= 1.2: score += 0.5
    if s.hematuria: score += 0.8
    return clamp(score / 3.6)


def treatment_burden(s: LupusSnapshot) -> float:
    score = 0.0
    if s.prednisone_mg_day >= 15: score += 1.4
    elif s.prednisone_mg_day >= 7.5: score += 0.8
    elif s.prednisone_mg_day > 0: score += 0.3
    if s.steroid_cumulative_months >= 12: score += 1.0
    elif s.steroid_cumulative_months >= 6: score += 0.5
    if s.immunosuppressant_count >= 2: score += 0.8
    return clamp(score / 3.2)


def flare_tendency(s: LupusSnapshot) -> float:
    score = 0.0
    if s.prior_flares_12m >= 3: score += 1.6
    elif s.prior_flares_12m >= 1: score += 0.8
    if s.sledai_2k >= 10: score += 1.5
    elif s.sledai_2k >= 6: score += 0.8
    elif s.sledai_2k >= 4: score += 0.4
    sero = serologic_activity(s)
    if sero >= 0.6: score += 1.0
    return clamp(score / 4.1)


def infection_vs_flare_flag(s: LupusSnapshot) -> str:
    """Zamora Score bridge: PCT-based heuristic for infection vs flare."""
    if s.pct >= 0.5 and s.fever:
        return "infection-likely (PCT ≥0.5 + fever)"
    if s.pct < 0.25 and s.crp > 5:
        return "flare-likely (low PCT, elevated CRP)"
    if s.fever and s.crp > 10:
        return "ambiguous — recommend cultures + close monitoring"
    return "no clear infection/flare flag"


def classify_drift(sero, renal, burden, flare) -> str:
    composite = 0.30*sero + 0.30*renal + 0.15*burden + 0.25*flare
    if composite >= 0.65:
        return "high-drift: escalation likely, close surveillance"
    if composite >= 0.40:
        return "moderate-drift: watchful, anticipate flare possibility"
    return "stable-drift: lower current risk trajectory"


def band(x):
    if x >= 0.70: return "high"
    if x >= 0.45: return "intermediate"
    return "lower"


def run_lupus_drift(s: LupusSnapshot) -> Dict[str, Any]:
    sero = serologic_activity(s)
    renal = renal_signal(s)
    burden = treatment_burden(s)
    flare = flare_tendency(s)
    inf_flag = infection_vs_flare_flag(s)
    drift = classify_drift(sero, renal, burden, flare)
    return {
        "input": asdict(s),
        "serologic_activity": round(sero, 3),
        "serologic_band": band(sero),
        "renal_signal": round(renal, 3),
        "renal_band": band(renal),
        "treatment_burden": round(burden, 3),
        "burden_band": band(burden),
        "flare_tendency": round(flare, 3),
        "flare_band": band(flare),
        "infection_vs_flare": inf_flag,
        "drift_trajectory": drift,
        "limitations": [
            "Heuristic weighted model; not externally validated on a derivation cohort.",
            "Infection vs flare flag uses simplified PCT thresholds (Zamora Score bridge).",
            "SLEDAI-2K input is optional but strongly improves flare tendency estimation.",
            "Not a substitute for nephrology evaluation, biopsy, or specialist judgement.",
            "Designed for transparent longitudinal surveillance support."
        ]
    }


if __name__ == "__main__":
    demo = LupusSnapshot(
        anti_dsdna_positive=True,
        c3_low=True,
        c4_low=True,
        proteinuria_mg_day=820,
        creatinine=1.3,
        hematuria=True,
        prednisone_mg_day=12,
        steroid_cumulative_months=14,
        prior_flares_12m=2,
        fever=True,
        crp=8.5,
        pct=0.18,
        immunosuppressant_count=2,
        sledai_2k=12,
    )
    print("=" * 70)
    print("LUPUS-DRIFT — Longitudinal SLE Trajectory Estimation")
    print("Authors: Zamora-Tehozol EA (ORCID:0000-0002-7888-3961), DNAI")
    print("=" * 70)
    print(json.dumps(run_lupus_drift(demo), indent=2))

Demo Output

======================================================================
LUPUS-DRIFT — Longitudinal SLE Trajectory Estimation
Authors: Zamora-Tehozol EA (ORCID:0000-0002-7888-3961), DNAI
======================================================================
{
  "input": {
    "anti_dsdna_positive": true,
    "c3_low": true,
    "c4_low": true,
    "proteinuria_mg_day": 820,
    "creatinine": 1.3,
    "hematuria": true,
    "prednisone_mg_day": 12,
    "steroid_cumulative_months": 14,
    "prior_flares_12m": 2,
    "fever": true,
    "crp": 8.5,
    "pct": 0.18,
    "immunosuppressant_count": 2,
    "sledai_2k": 12
  },
  "serologic_activity": 1.0,
  "serologic_band": "high",
  "renal_signal": 0.694,
  "renal_band": "intermediate",
  "treatment_burden": 0.812,
  "burden_band": "high",
  "flare_tendency": 0.805,
  "flare_band": "high",
  "infection_vs_flare": "flare-likely (low PCT, elevated CRP)",
  "drift_trajectory": "high-drift: escalation likely, close surveillance",
  "limitations": [
    "Heuristic weighted model; not externally validated on a derivation cohort.",
    "Infection vs flare flag uses simplified PCT thresholds (Zamora Score bridge).",
    "SLEDAI-2K input is optional but strongly improves flare tendency estimation.",
    "Not a substitute for nephrology evaluation, biopsy, or specialist judgement.",
    "Designed for transparent longitudinal surveillance support."
  ]
}