STEROID-HYPERGLY: Transparent Risk Stratification for Glucocorticoid-Induced Hyperglycemia During Rheumatic and Autoimmune Disease Treatment

Authors: Dr. Erick Zamora-Tehozol, DNAI, RheumaAI
ORCID: 0000-0002-7888-3961

Abstract

Glucocorticoids remain indispensable across rheumatology and autoimmune medicine, but steroid-induced hyperglycemia is common, clinically consequential, and often missed when monitoring relies on fasting glucose alone. We present STEROID-HYPERGLY, an executable Python skill that estimates risk of glucocorticoid-induced hyperglycemia using baseline glycemic vulnerability, glucocorticoid exposure burden, and host susceptibility. The model incorporates HbA1c, fasting glucose, diabetes or prediabetes status, prednisone-equivalent dose, steroid duration, obesity, age, renal dysfunction, active infection, and pulse therapy. Outputs include a transparent risk score, categorical risk class, monitoring intensity recommendations, and safety alerts that emphasize steroid-aware timing of glucose checks. In demonstration cases, the tool classifies a low-dose polymyalgia rheumatica case as LOW risk, a prediabetic rheumatoid arthritis case receiving 25 mg/day prednisone as HIGH risk, and a high-dose systemic lupus erythematosus flare with diabetes, CKD, infection, and pulse therapy as VERY HIGH risk. This skill is intended to improve early monitoring, reduce missed dysglycemia, and support safer glucocorticoid use. It does not diagnose diabetes and does not replace local treatment protocols or endocrinology consultation.

Keywords: glucocorticoids, hyperglycemia, prednisone, rheumatology, autoimmune disease, diabetes risk, clinical decision support, DeSci

1. Clinical problem

Steroids are among the most effective short-term agents in rheumatology, but they impose a predictable metabolic cost. Hyperglycemia during glucocorticoid therapy is associated with infection, delayed healing, longer hospitalization, and treatment complexity. In practice, it is often underdetected because the glycemic pattern is not purely fasting. Patients receiving morning prednisone may have near-normal fasting glucose but substantial late-day excursions. This matters in rheumatic and autoimmune disease, where repeated steroid bursts, pulse therapy, and prolonged tapers remain common.

The clinical question is simple: who needs tighter monitoring from day 1? STEROID-HYPERGLY was designed to answer that question transparently.

2. Methodology

2.1 Design principles

The model follows five principles:

1. Baseline glycemia matters. HbA1c, fasting glucose, prediabetes, and known diabetes strongly modify risk.
2. Dose and duration matter. Higher prednisone-equivalent doses and longer exposure increase probability of clinically important dysglycemia.
3. Host susceptibility matters. Obesity, older age, CKD, and active infection amplify metabolic vulnerability.
4. Pulse therapy matters. Abrupt steroid loading can produce large glycemic excursions even when fasting values appear acceptable.
5. Monitoring should fit physiology. Steroid hyperglycemia is often post-prandial or afternoon-predominant, so fasting-only strategies miss cases.

2.2 Model structure

The skill computes three interpretable components:

• Baseline glycemic vulnerability — HbA1c, fasting glucose, diabetes status, family history, prior gestational diabetes
• Steroid exposure burden — prednisone-equivalent dose, duration, pulse therapy
• Host susceptibility — age, BMI, CKD/eGFR, active infection

A small interaction term increases risk when baseline dysglycemia and moderate/high steroid exposure coexist.

2.3 Output logic

The tool returns:

• Total score (0 upward, scaled for interpretability)
• Risk class: LOW, INTERMEDIATE, HIGH, VERY HIGH
• Monitoring plan matched to risk intensity
• Safety alerts reminding clinicians not to rely on fasting glucose alone
• Explicit limitations

3. Executable skill

3.1 Implementation

The implementation is standalone Python using only the standard library and is stored locally at:

skills/steroid-hypergly/steroid_hypergly.py

3.2 Demo output summary

Low-risk outpatient PMR -> LOW
Prediabetic RA on 25 mg/day prednisone -> HIGH
Very-high-risk SLE flare -> VERY HIGH

Representative very-high-risk signal:

total_score: 93.5
risk_class: VERY HIGH
alert: Do not rely on fasting glucose alone; steroid hyperglycaemia is frequently missed without afternoon/post-prandial checks.

4. Clinical use case

This skill is appropriate when clinicians are deciding whether a patient starting or escalating glucocorticoids needs:

• routine symptom-only follow-up,
• structured home glucose checks,
• day-1 intensive monitoring,
• or early treatment planning with diabetes support.

It is especially relevant in rheumatoid arthritis, systemic lupus erythematosus, vasculitis, polymyalgia rheumatica, inflammatory myopathies, and other autoimmune disorders where glucocorticoids remain part of flare management.

5. Limitations

1. This is an evidence-informed heuristic tool, not a prospectively validated prediction model.
2. It supports monitoring intensity; it does not diagnose diabetes.
3. Steroid formulation, administration timing, and inpatient context can change glycemic trajectories.
4. Local thresholds for pharmacologic treatment differ across settings and guidelines.
5. The score should support, not replace, specialist judgment and protocol-based diabetes care.

6. Why this solves a real problem

Steroid-induced hyperglycemia is common, but clinical workflows still miss it because the wrong glucose window is often checked. STEROID-HYPERGLY addresses the practical problem directly: identify who is vulnerable, identify when steroids are dangerous enough to justify structured monitoring, and generate simple, auditable recommendations before harm occurs.

References

1. Clore JN, Thurby-Hay L. Glucocorticoid-Induced Hyperglycemia. Endocr Pract. 2009;15(5):469-474. DOI: 10.4158/EP08331.RAR
2. Roberts A, James J, Dhatariya K, et al. Management of hyperglycaemia and steroid (glucocorticoid) therapy: a guideline from the Joint British Diabetes Societies (JBDS) for Inpatient Care group. Diabet Med. 2018;35(8):1011-1017. DOI: 10.1111/dme.13675
3. Mertens B, et al. A Practical Guide for the Management of Steroid Induced Hyperglycaemia in the Hospital. J Clin Med. 2021;10(10):2154. DOI: 10.3390/jcm10102154
4. Kwon S, Hermayer KL, Hermayer K. Glucocorticoid-Induced Hyperglycemia: A Neglected Problem. Endocrinol Metab (Seoul). 2024. DOI: 10.3803/EnM.2024.1951
5. Liu XX, Zhu XM, Miao Q, Ye HY, Zhang ZY, Li YM. Risk factors for the development of glucocorticoid-induced diabetes mellitus. Diabetes Res Clin Pract. 2014;105(3):363-372. DOI: 10.1016/j.diabres.2015.02.010

Executable Python skill

#!/usr/bin/env python3
"""
STEROID-HYPERGLY — Glucocorticoid-Induced Hyperglycemia Risk Stratification

Transparent clinical decision-support skill for estimating risk of steroid-
induced hyperglycemia in rheumatic and autoimmune disease before or during
systemic glucocorticoid therapy.

Authors: Dr. Erick Zamora-Tehozol (ORCID:0000-0002-7888-3961), DNAI, RheumaAI
License: MIT

References:
- Clore JN, Thurby-Hay L. Glucocorticoid-Induced Hyperglycemia.
  Endocr Pract. 2009;15(5):469-474. DOI:10.4158/EP08331.RAR
- Roberts A, James J, Dhatariya K, et al. Management of hyperglycaemia and
  steroid (glucocorticoid) therapy: a guideline from the Joint British
  Diabetes Societies (JBDS) for Inpatient Care group. Diabet Med.
  2018;35(8):1011-1017. DOI:10.1111/dme.13675
- Mertens B, et al. A Practical Guide for the Management of Steroid Induced
  Hyperglycaemia in the Hospital. J Clin Med. 2021;10(10):2154.
  DOI:10.3390/jcm10102154
- Kwon S, Hermayer KL, Hermayer K. Glucocorticoid-Induced Hyperglycemia:
  A Neglected Problem. Endocrinol Metab (Seoul). 2024.
  DOI:10.3803/EnM.2024.1951
- Liu XX, Zhu XM, Miao Q, Ye HY, Zhang ZY, Li YM. Risk factors for the
  development of glucocorticoid-induced diabetes mellitus. Diabetes Res Clin
  Pract. 2014;105(3):363-372. DOI:10.1016/j.diabres.2015.02.010
"""

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


@dataclass
class SteroidHyperglyInput:
    age: int
    bmi: float
    prednisone_mg_day: float
    steroid_days: int
    hba1c: float
    fasting_glucose: float
    known_diabetes: bool = False
    prediabetes: bool = False
    family_history_diabetes: bool = False
    prior_gestational_diabetes: bool = False
    egfr: float = 90.0
    ckd: bool = False
    active_infection: bool = False
    pulse_therapy: bool = False
    rheumatic_diagnosis: str = "RA"


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


def glucose_vulnerability(inp: SteroidHyperglyInput) -> float:
    score = 0.0
    score += clamp((inp.hba1c - 5.0) / 1.5) * 3.5
    score += clamp((inp.fasting_glucose - 90.0) / 35.0) * 2.5
    if inp.known_diabetes:
        score += 3.0
    elif inp.prediabetes:
        score += 1.5
    if inp.family_history_diabetes:
        score += 0.7
    if inp.prior_gestational_diabetes:
        score += 0.8
    return score


def steroid_exposure_burden(inp: SteroidHyperglyInput) -> float:
    dose = inp.prednisone_mg_day
    score = 0.0
    if dose >= 40:
        score += 2.5
    elif dose >= 20:
        score += 1.7
    elif dose >= 10:
        score += 0.9
    elif dose > 0:
        score += 0.3

    if inp.steroid_days >= 28:
        score += 1.2
    elif inp.steroid_days >= 14:
        score += 0.7
    elif inp.steroid_days >= 5:
        score += 0.3

    if inp.pulse_therapy:
        score += 1.0
    return score


def host_susceptibility(inp: SteroidHyperglyInput) -> float:
    score = 0.0
    if inp.age >= 75:
        score += 1.0
    elif inp.age >= 60:
        score += 0.7
    elif inp.age >= 45:
        score += 0.3

    if inp.bmi >= 35:
        score += 1.2
    elif inp.bmi >= 30:
        score += 0.8
    elif inp.bmi >= 25:
        score += 0.3

    if inp.ckd or inp.egfr < 45:
        score += 0.9
    elif inp.egfr < 60:
        score += 0.5

    if inp.active_infection:
        score += 0.8
    return score


def risk_score(inp: SteroidHyperglyInput) -> float:
    base = glucose_vulnerability(inp) + steroid_exposure_burden(inp) + host_susceptibility(inp)
    # mild interaction: high baseline glycaemia plus high steroid dose is worse than additive
    if (inp.hba1c >= 5.7 or inp.fasting_glucose >= 100 or inp.prediabetes) and inp.prednisone_mg_day >= 20:
        base += 0.8
    if inp.known_diabetes and inp.prednisone_mg_day >= 20:
        base += 1.0
    return round(base * 5, 1)


def classify(score: float) -> str:
    if score >= 70:
        return "VERY HIGH"
    if score >= 40:
        return "HIGH"
    if score >= 20:
        return "INTERMEDIATE"
    return "LOW"


def monitoring_plan(inp: SteroidHyperglyInput, score: float) -> List[str]:
    plan = []
    if score < 25:
        plan.append("Check capillary glucose or fasting glucose if symptoms emerge; reinforce diet and hydration.")
    elif score < 50:
        plan.append("Monitor post-lunch or late-afternoon glucose for the first 3-5 days after steroid start or escalation.")
        plan.append("Repeat monitoring after each dose increase and during taper if glycaemia was abnormal.")
    elif score < 75:
        plan.append("Structured home glucose monitoring recommended: pre-dinner and/or 2h post-lunch for at least 5-7 days.")
        plan.append("Consider early endocrinology or primary-care diabetes support if values exceed local thresholds.")
    else:
        plan.append("Daily glucose monitoring from day 1 is recommended; same-day treatment plan should be arranged.")
        plan.append("If inpatient, use steroid-aware insulin strategy or protocol-based correction rather than fasting-only checks.")

    if inp.prednisone_mg_day >= 20:
        plan.append("Prioritise afternoon/evening glucose checks because prednisone-type steroids often peak later in the day.")
    if inp.known_diabetes:
        plan.append("Existing diabetes therapy will usually need temporary intensification during steroid exposure.")
    return plan


def alerts(inp: SteroidHyperglyInput, score: float) -> List[str]:
    out = []
    if inp.hba1c >= 6.5 and not inp.known_diabetes:
        out.append("Baseline HbA1c is already in the diabetes range; this is not purely steroid-related risk.")
    if inp.prednisone_mg_day >= 40:
        out.append("High-dose glucocorticoids strongly increase post-prandial and afternoon hyperglycaemia risk.")
    if inp.pulse_therapy:
        out.append("Pulse therapy may produce abrupt glycaemic excursions even when fasting glucose is near normal.")
    if inp.egfr < 45 or inp.ckd:
        out.append("Renal dysfunction increases management complexity and may limit some glucose-lowering options.")
    if inp.active_infection:
        out.append("Concurrent infection can worsen insulin resistance and amplify steroid-related dysglycaemia.")
    if score >= 50:
        out.append("Do not rely on fasting glucose alone; steroid hyperglycaemia is frequently missed without afternoon/post-prandial checks.")
    return out


def run_steroid_hypergly(inp: SteroidHyperglyInput) -> Dict[str, Any]:
    score = risk_score(inp)
    return {
        "input_summary": {
            "diagnosis": inp.rheumatic_diagnosis,
            "age": inp.age,
            "bmi": inp.bmi,
            "prednisone_mg_day": inp.prednisone_mg_day,
            "steroid_days": inp.steroid_days,
            "hba1c": inp.hba1c,
            "fasting_glucose": inp.fasting_glucose,
            "known_diabetes": inp.known_diabetes,
            "prediabetes": inp.prediabetes,
            "egfr": inp.egfr,
            "active_infection": inp.active_infection,
            "pulse_therapy": inp.pulse_therapy,
        },
        "baseline_glycaemic_vulnerability": round(glucose_vulnerability(inp), 2),
        "steroid_exposure_burden": round(steroid_exposure_burden(inp), 2),
        "host_susceptibility": round(host_susceptibility(inp), 2),
        "total_score": score,
        "risk_class": classify(score),
        "monitoring_plan": monitoring_plan(inp, score),
        "alerts": alerts(inp, score),
        "limitations": [
            "Heuristic evidence-informed score; not yet prospectively validated.",
            "Designed for screening and monitoring intensity, not diagnosis of diabetes.",
            "Does not replace inpatient insulin protocols or endocrinology judgement.",
            "Steroid formulation, timing, and route can alter glycaemic patterns beyond this simplified model.",
            "Local treatment thresholds vary by guideline and care setting."
        ]
    }


if __name__ == "__main__":
    demos = [
        ("Low-risk outpatient PMR", SteroidHyperglyInput(age=42, bmi=23.8, prednisone_mg_day=7.5, steroid_days=10, hba1c=5.2, fasting_glucose=88, rheumatic_diagnosis="PMR")),
        ("Intermediate-risk RA", SteroidHyperglyInput(age=61, bmi=31.2, prednisone_mg_day=25, steroid_days=21, hba1c=5.9, fasting_glucose=104, prediabetes=True, family_history_diabetes=True, egfr=58, rheumatic_diagnosis="RA")),
        ("Very-high-risk SLE flare", SteroidHyperglyInput(age=68, bmi=34.6, prednisone_mg_day=60, steroid_days=30, hba1c=7.1, fasting_glucose=148, known_diabetes=True, ckd=True, egfr=38, active_infection=True, pulse_therapy=True, rheumatic_diagnosis="SLE")),
    ]

    print("=" * 72)
    print("STEROID-HYPERGLY — Glucocorticoid-Induced Hyperglycemia Risk")
    print("Authors: Dr. Erick Zamora-Tehozol, DNAI, RheumaAI")
    print("=" * 72)
    for label, demo in demos:
        result = run_steroid_hypergly(demo)
        print(f"\n--- {label} ---")
        print(json.dumps(result, indent=2))