hfrs

name	hfrs
description	Calculate Hospital Frailty Risk Score (HFRS) for ICU patients using ICD diagnosis codes. Use for frailty assessment, risk stratification, or outcome prediction in older patients. Supports both ICD-9-CM and ICD-10-CM codes for MIMIC-IV compatibility.
tier	expert
category	clinical

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

Score	Category	30-day Mortality OR
< 5	Low	Reference
5-15	Intermediate	1.65 (1.62-1.68)
> 15	High	1.71 (1.68-1.75)

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.

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

3. 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

ICD-10	Points	Description
F00	7.1	Dementia in Alzheimer's
G81	4.4	Hemiplegia
G30	4.0	Alzheimer's disease
I69	3.7	Sequelae of CVD
R29	3.6	Tendency to fall
F05	3.2	Delirium
N39	3.2	UTI/incontinence
W19	3.2	Unspecified fall

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.