hannesill/m4/hfrs

Hospital Frailty Risk Score (HFRS)

Calculate HFRS from ICD diagnosis codes. Based on Gilbert et al. (2018) Lancet methodology with 109 weighted ICD-10 codes.

When to Use This Skill

• Assessing frailty in ICU or hospitalized patients
• Risk stratification for older adults
• Predicting 30-day mortality, prolonged length of stay, or readmission
• Research requiring a validated frailty metric from administrative data

Methodology

Per the original publication, HFRS is calculated using diagnoses from:

1. The current (index) admission
2. All emergency admissions in the preceding 2 years

Each unique ICD code is counted only once, even if it appears in multiple admissions.

Risk Categories

Critical Implementation Notes

1. ICD-9 mappings are NOT part of the original publication. They have been added for MIMIC-IV compatibility, which contains both ICD-9 and ICD-10 coded admissions. Use ICD-9 mappings with appropriate caution.

1. Emergency admission types in MIMIC-IV: The SQL implementation considers EMERGENCY, URGENT, and EW EMER. as emergency admissions.

1. Code matching uses 3-character prefix for ICD-10 (e.g., F05 matches F05.0, F05.1). ICD-9 uses variable-length matching.

Top Contributing Codes

Full 109-code list: references/hfrs_icd_codes.csv

Example Queries

Calculate HFRS for all admissions

-- See scripts/calculate_hfrs.sql for the complete query
-- Key tables used:
-- - mimiciv_hosp.admissions (admission dates and types)
-- - mimiciv_hosp.diagnoses_icd (ICD codes)
 
-- Simplified example: score a single admission
SELECT
    a.subject_id,
    a.hadm_id,
    SUM(w.points) AS hfrs_score
FROM mimiciv_hosp.diagnoses_icd d
JOIN mimiciv_hosp.admissions a ON d.hadm_id = a.hadm_id
JOIN hfrs_weights w ON d.icd_code LIKE w.icd_prefix || '%'
    AND d.icd_version = w.icd_version
GROUP BY a.subject_id, a.hadm_id;

Python: ad-hoc calculation

from scripts.calculate_hfrs import calculate_hfrs
diagnoses = [
    {'icd_code': 'F05', 'icd_version': 10},  # Delirium
    {'icd_code': '2900', 'icd_version': 9},   # Dementia (ICD-9)
    {'icd_code': 'R29', 'icd_version': 10},   # Falls
]
result = calculate_hfrs(diagnoses)
print(f"HFRS: {result['score']} ({result['risk_category']} risk)")

References

• Gilbert T, et al. "Development and validation of a Hospital Frailty Risk Score focusing on older people in acute care settings using electronic hospital records: an observational study." Lancet. 2018;391:1775-82.