Population health — Risk tiers
Anchor Health CooperativeA synthetic demo surface — "Anchor Health Cooperative" is a fictional, obviously-fake
HMO. No real payer, no PHI. Every number here traces to the vault: the risk score is
cs_member_risk_score (clinical diagnoses → a hand-curated chronic-only mapping);
cost is Σ fct_claim (status = paid) in the 2017+ window. The app carries no business
logic — the tiers and shares are computed in the risk_tier_summary mart.
Risk tiers · the sick-vs-well spread
Each member gets an HCC-style risk score from their chronic diagnoses, then falls into one of four tiers: none (0), low (up to 0.35), medium (up to 0.90), high (above 0.90). The score is computable for every patient; the cost side covers the members that resolve to a payer identity (and so carry claims).
High-risk members
High-tier share of cost
Members scored
Member-resolved paid
The four tiers
Member share is over the whole scored population; cost share is over the member-resolved subset (the members with claims). The two columns diverge on purpose — that gap is the population-health story.
Cost concentration by tier
Members are spread fairly evenly across the tiers, but paid dollars are not — they pile up in the high tier. A payer reads this to target care-management at the members where a marginal dollar of intervention buys the most.
cs_member_risk_score.risk_score (exact DECIMAL, so the
> 0.90 boundary is clean); cost = Σ
fct_claim.total_paid_amt where claim_status = 'paid'
in the 2017+ window, over the members that resolve through
cs_patient_member_same_as.