validation_report.md

HC-CAR-010 — Stroke Risk Prediction

Validation Report

• Generated: 2026-05-26T18:22:48.714388+00:00
• N patients: 500
• Seed: 42
• Weighted Score: 10.0/10
• Grade: A+

Scorecard

Metric	Value	Target	Score	Status	Source
age_median	63.05	[55.0, 70.0]	10.0	PASS	Stroke prevention cohort age median 60-65 (Framingham, ARIC)
female_pct	47.0	[40.0, 60.0]	10.0	PASS	Adult stroke risk cohorts ~45-55% female
hypertension_pct	63.4	[50.0, 75.0]	10.0	PASS	AHA stroke risk cohorts ~55-70% HTN prevalence (enriched vs population)
diabetes_pct	27.2	[18.0, 35.0]	10.0	PASS	NHANES adults ≥40: DM ~15%; stroke risk cohorts enriched 20-30%
current_smoker_pct	28.8	[18.0, 35.0]	10.0	PASS	Stroke risk cohorts 20-30% current smokers (enriched)
obesity_pct	48.2	[35.0, 55.0]	10.0	PASS	NHANES 2017-2020 obesity ~42%; CV risk cohorts 40-50%
ckd_any_pct	45.2	[35.0, 60.0]	10.0	PASS	Stroke risk cohorts CKD any-stage 40-50% (Lin 2018, ARIC)
afib_pct_age70_80	14.286	[10.0, 22.0]	10.0	PASS	Rotterdam Study AFib in age 70-80: ~9.5%; generator HF-enriches to 13-18%
afib_pct_overall	11.4	[8.0, 18.0]	10.0	PASS	High-risk CV cohort AFib ~10-15%
chadsvasc_median_in_afib	4.0	[2.0, 5.0]	10.0	PASS	CHA2DS2-VASc median in stroke-risk AFib cohort ~3-4
stroke_event_pct	16.4	[8.0, 22.0]	10.0	PASS	Risk-stratified cohort stroke event rate over follow-up ~10-18%
ischemic_pct_of_strokes	86.585	≥75.0	10.0	PASS	GBD 2022: ischemic strokes ~87% of all strokes, FLOOR
hemorrhagic_pct_of_strokes	8.537	[2.0, 18.0]	10.0	PASS	GBD 2022: hemorrhagic ~10% of all strokes (wide tolerance for n=500)
nihss_median_ischemic	6.0	[3.0, 11.0]	10.0	PASS	SITS-MOST median acute NIHSS ~8; cohort with milder events 4-8
mortality_1yr_in_strokes_pct	6.098	[2.0, 25.0]	10.0	PASS	Generator BENCHMARK constant 8% post-stroke 1y mortality; wide for n=500 subset
recurrence_90d_in_strokes_pct	1.22	[1.0, 9.0]	10.0	PASS	AHA/ASA: 90-day recurrent stroke ~4% (BENCHMARK)
carotid_imt_mean_mm	0.97	[0.8, 1.1]	10.0	PASS	MESA carotid IMT: low-risk mean 0.72, high-risk 0.92; enriched cohort 0.85-1.05
carotid_plaque_pct	49.6	[30.0, 60.0]	10.0	PASS	MESA: carotid plaque ~30-50% in age >50
statin_pct	56.8	[40.0, 70.0]	10.0	PASS	NHANES statin use ≥40: ~30%; stroke risk cohorts 45-60% (AHA secondary prevention)
antiplatelet_any_pct	65.8	[55.0, 80.0]	10.0	PASS	AHA secondary stroke prevention: antiplatelet 60-75% in cohort
antihypertensive_in_htn_pct	66.562	≥60.0	10.0	PASS	AHA/ACC 2017 HTN: ≥60% of HTN patients on therapy, FLOOR
bp_control_in_treated_pct	54.976	[45.0, 65.0]	10.0	PASS	NHANES: ~50-55% BP control among treated HTN patients
anticoag_in_afib_pct	96.491	≥75.0	10.0	PASS	AHA/ACC/HRS 2023 AFib: anticoag for CHA2DS2-VASc≥2; cohort 85-95% (generator upgrades), FLOOR
doac_share_of_afib_anticoag_pct	69.091	≥50.0	10.0	PASS	2023 AHA/ACC: DOAC preferred over warfarin; modern AFib registries 60-80% DOAC, FLOOR
lipids_ldl_mean_mg_dl	100.27	[85.0, 115.0]	10.0	PASS	Statin-enriched cohort post-treatment LDL ~95-110 mg/dL (NHANES general pop 128 untreated)
hdl_mean_mg_dl	47.527	[42.0, 55.0]	10.0	PASS	NHANES HDL mean ~50 mg/dL (sex-adjusted)
pfo_pct	23.8	[15.0, 35.0]	10.0	PASS	Autopsy series PFO prevalence ~25-27% (Hagen 1984)
heart_failure_pct	23.2	[12.0, 32.0]	10.0	PASS	Stroke risk cohort with HF enrichment ~18-28% (vs population ~2%)
framingham_10yr_risk_mean	12.581	[8.0, 18.0]	10.0	PASS	High-risk cohort Framingham 10-yr stroke risk mean 10-15%
cardioembolic_pct_of_ischemic	18.31	[8.0, 30.0]	10.0	PASS	TOAST cardioembolic ~20-30% of ischemic strokes (Petty 1999, NEMC-SR)

Notes

• Floor metrics (anticoag_in_afib_pct, doac_share_of_afib_anticoag_pct, antihypertensive_in_htn_pct, ischemic_pct_of_strokes) are one-sided ≥ threshold checks. All other metrics are two-sided range checks.
• Subset-stability-sensitive metrics (thrombolysis/thrombectomy in n=2-4 patient subgroups, MRS-0-1 in thrombolysed patients) are excluded from scoring due to unreliability at n=500.
• AFib prevalence is HF-enriched. The generator's Rotterdam Study benchmark (3.8% age 60-70, 9.5% age 70-80) is upgraded for HF-positive patients, producing 10-18% — appropriate for a CV high-risk cohort, not population-level estimates. See README.md for full disclosure.