Risk factor dashboard
Cross-cutting views of the 997-factor atlas: which factors and groups confer the greatest risk, which are pharmacologically targetable, how evidence is distributed, and how much of this the risk calculators actually capture.
1 · Markers conferring the highest risk
Risk-magnitude score for individual factors with a categorical effect estimate (high-vs-low or presence-vs-absence). Bars with a faded/hatched fill rest on a single extracted estimate — wide confidence intervals, read with caution. Colored by category; hover for the effect size, estimate count, and reliability.
▸ Score = 100·ln(min(RR-equiv, 6))/ln(6), where the reported HR/OR/RR is treated as a fold-risk. 586 estimates were LLM-extracted from 136 factors' abstracts (meta-analyses preferred). typical_effect is the median of a factor's CATEGORICAL estimates only; per-SD/continuous estimates are retained in the record but excluded from this ranking. Several top-ranked factors (HIV, visceral adiposity, epicardial fat) rest on a single study — their high rank reflects one wide-CI estimate, not settled magnitude.
2 · Which groups carry the greatest risk
Median (bar) and maximum (marker) risk-magnitude score within each category, across comparable categorical risk factors. Ranks the factor families by the typical effect size of their members.
▸ Median over each category's categorical, risk-increasing factors with an extracted estimate. Categories with few estimates are noisier — n shown on hover.
3 · Which groups have drug targets
For every category, the share of its factors with an approved or investigational pharmacologic target, versus lifestyle-only or none. Shows where the therapeutic arsenal is deep (lipids, blood pressure, glycemic) and where risk is currently untreatable (demographic, psychosocial, environmental, much of the novel-biomarker frontier).
▸ 234 of 997 factors flagged with a target. 82 were curated directly from the ASCVD drug literature and trial interventions; the rest inherit a category-level default gated by modifiability, so read this at the group level rather than as per-factor drug facts.
4 · Evidence gradient by group
Grade composition of each category, from strong (A) to preliminary (D). Ranks groups by their count of strongly-evidenced (A/B) factors — the lipids, inflammatory, and comorbid families anchor the high-evidence end; environmental and novel biomarkers are mostly emerging.
▸ Grades from the atlas rubric: literature volume + meta-analyses + trials + genetic support + a genetic-causal-candidate bump. Ordered by A+B count.
5 · What the calculators actually measure
How many of the 17 cataloged risk calculators include each factor. A handful of traditional inputs (age, sex, smoking, blood pressure, cholesterol, diabetes) appear nearly universally; almost everything else in the atlas is used by few or none.
▸ Counted across all 17 calculators by mapping each input variable to its atlas factor. Top 20 shown.
6 · How broadly each calculator covers the atlas
Calculators ranked by the number of strongly-evidenced (A/B) atlas factors they include, split into strong vs other factors. QRISK3 is the broadest; Globorisk, followed by the SCORE2 family and Framingham, is the most parsimonious. Links go to each calculator's detail.
▸ Coverage = distinct atlas factors mapped from a calculator's inputs. Ranked by A/B factors covered, then total, then categories.
Full scoring definition: risk_magnitude_score = 100*ln(min(RR_equiv,6))/ln(6). RR_equiv = point estimate expressed as fold-risk >=1. Approximate: HR/OR treated as RR-equivalent. Compare only within per_sd_or_categorical=='categorical' (comparable_categorical=true); per-SD/per-unit flagged separately. NOTE: typical_effect and score for comparable_categorical factors are computed from categorical (high-vs-low / present-vs-absent) estimates ONLY; factors whose estimates are entirely per-SD/continuous are excluded from the categorical ranking.