The High-Transmission Gate
Where the 30% PfPR cutoff came from, what the literature says "high transmission" should be, and how the malaria portfolio would change if the gate were set at 20 / 25 / 30 / 35 / 40%.
1. The gate is 30% PfPR, chosen with reference to Kenya's NMCP stratification (≥30% = high) and endorsed by the modeling/strategy team in a review meeting. It replaced an earlier 35% on 2025-10-01.
2. It is a deliberate, defensible choice — but not the single most widely-accepted one. The most-citable WHO line is 35% (WHO/PMI 2019); 30% follows national-program (Kenya/Tanzania) practice. The choice is also highly consequential: 35%→30% roughly doubles the population in the intensive package (91M → 209M).
3. The bigger question is regional fit. The Kenya/Tanzania precedent is East African, where transmission is far lower (median PfPR ~3–8%) and 30% is genuinely extreme. But most of the portfolio's impact is in West/Central Africa (median ~22–25%), where 30% is only an upper-third cutoff — so a West-African program might reasonably set "high" higher.
1 · The current choice and its basis
The gate lives in 1a_geo_coverage.Rmd: admins with 2020 MAP PfPR at or above the cutoff are classed high → the is_high flag → eligibility for the intensive package (age-extended vaccination, LAI chemoprevention, gene-limited tools). The current value is 30% PfPR, selected with reference to Kenya's national malaria program stratification (which uses ≥30% for its "high" stratum) and endorsed by the modeling and strategy team in a review meeting.
For the record, the version history is clean: the cutoff was 35% from the earliest pipeline (v8, 2023) and was moved to 30% on 2025-10-01 (commit 37fbefc) alongside the strategy-review work. The gate has only ever taken those two values in code.
This report does not relitigate that decision. It does two things the decision deserves: (1) place 30% against the published definitions of "high transmission," and (2) stress-test whether a threshold drawn from East-African programs is the right one for a portfolio whose impact is concentrated in West and Central Africa.
2 · What the literature calls "high transmission"
"High transmission" is defined against P. falciparum parasite prevalence (PfPR), most rigorously the age-standardized PfPR2-10. The authoritative anchors:
| Source | Metric | "High" / relevant threshold | Used for | Link |
|---|---|---|---|---|
| WHO / PMI — Framework for Evaluating Malaria Programs in Moderate- & Low-Transmission Settings (2019) | all-ages PfPR / API | High = PfPR ≥35%; Moderate 10–<35%; Low 1–10%; Very low <1% | Canonical WHO operational stratification of settings & interventions | WHO/PMI 2019 |
| WHO Guidelines for malaria — vaccine & chemoprevention (2022–2025) | PfPR / API | "Moderate-to-high" = PfPR >10% (indicative, not absolute) | Gates RTS,S/R21 vaccine & chemoprevention eligibility | WHO 2023 |
| Lysenko & Semashko (1968) classic endemicity — via Hay et al. 2008 (Lancet ID) | PfPR2-10 | Hypo <10%; Meso 11–50%; Hyper 51–75%; Holo >75% | Origin of all later PfPR cutoffs (historical, still cited) | Hay 2008 |
| Gething et al. 2011 — A new world malaria map (Malaria Journal); MAP endemicity classes | PfPR2-10 | Hypo 1–10%; Meso >10–50%; Hyper >50–75%; Holo >75% | MAP global endemicity surface; modern PfPR2-10 classes | Gething 2011 |
| Tanzania NMCP subnational stratification (Malaria Journal 2020) | PfPR5-16 (school-age) | High ≥30%; Moderate 5–<30%; Low 1–<5% | NMCP geographic targeting of intervention packages (IPTsc, vector mix) | Tanzania SNT 2020 |
| RTS,S Phase 3 — RTS,S Partnership (Lancet 2015) | PfPR2-10 | Efficacy falls ~60% → ~4% as PfPR rises ~10% → ~70% | Why add intensive tools at top end: efficacy lowest where burden highest | RTS,S 2015 |
Synthesis
- The most citable operational line is WHO's PfPR ≥35% = "high" (WHO/PMI 2019; reused in WHO subnational-tailoring exercises). Moderate is 10–<35%.
- The classic / MAP endemicity scheme has no 30–40% boundary at all: the mesoendemic band is a broad 10–50%, with the next break (meso→hyper) at 50%. So any cutoff in 30–40% is a programmatic choice inside mesoendemic, not a classical class edge.
- National programs that actually gate packages (Tanzania, Kenya) most often use ≥30% for their "high" stratum — the clearest operational precedent for the model's current value (Tanzania uses school-age PfPR5-16, a caveat).
- For gating vaccines/chemoprevention specifically, WHO triggers much lower — "moderate-to-high" = PfPR >10%.
- Defensible range ≈ 30–40%: 35% WHO-anchored, 30% NMCP-operational, 40% upper-meso/"high-risk". 25% and 20% lack any named anchor and should be treated as sensitivity-only.
- An intensive package at the top end is empirically motivated: RTS,S efficacy falls as PfPR rises, and the highest-PfPR admins carry the most clinical burden — both arguments for extra tools precisely where the gate sits.
| Gate | Literature anchor | Strength |
|---|---|---|
| 40% | Upper mesoendemic; "high-risk" research cutoff | Defensible (conservative) |
| 35% (model's original, pre-2025-10-01) | WHO-canonical "high" (WHO/PMI 2019; WHO SNT) | Strongest / most citable |
| 30% (model's current) | NMCP operational "high" (Tanzania/Kenya) | Defensible (operational) |
| 25% | No named WHO/NMCP/classical anchor (mid-meso) | Sensitivity-only |
| 20% | No named anchor; near vaccine ">10%" floor ×2 | Sensitivity-only |
3 · Does an East-African precedent fit a West-African portfolio?
The strongest caveat on 30% is not whether it is "high" in the abstract, but whether a threshold borrowed from Kenya/Tanzania transfers to where this portfolio actually acts. It largely does not, because East-African transmission is a different regime from West and Central Africa.
| Region | Median admin PfPR | Admins ≥ 30% | 30% is the… |
|---|---|---|---|
| West Africa | 24.5% | 35% | 65th pctile (upper third) |
| Central Africa | 22.3% | 38% | 62nd pctile |
| East Africa | 5.6% | 4% | 96th pctile (extreme top) |
| — Kenya | 2.6% | — | near-max |
| — Tanzania | 7.7% | 0% (max 13.5%) | above every admin |
Kenya and Tanzania — the precedent — are low-transmission by continental standards (median admin PfPR 2.6% and 7.7%; Tanzania never exceeds 13.5%). There, 30% genuinely marks the extreme top (the 96th percentile of East-African admins; above every Tanzanian admin). Apply the same 30% to West Africa — median 24.5% — and it becomes merely an upper-third cutoff (65th percentile). To flag the same top slice in West Africa that 30% flags in the East, you would need a gate of about 41%.
This matters because the gate's effect is overwhelmingly West/Central African. At 30%, the high-transmission population splits 60% West Africa, 28% Central, 12% East:
| PfPR gate | East Africa | Central Africa | West Africa |
|---|---|---|---|
| 20% | 47M | 97M | 297M |
| 25% | 33M | 75M | 213M |
| 30% | 25M | 58M | 126M |
| 35% | 17M | 48M | 26M |
| 40% | 9M | 38M | 9M |
Note the steep West-African drop between 30% and 35% (126M → 26M): a dense mass of West-African admins — Nigeria foremost — sits in the 30–35% band, so the regional-fairness question and the cost/impact sensitivity are the same question.
4 · How much the choice moves
Across the 18 IPM countries (315 admin-1 units, MAP PfPR 2020), the gate is extremely sensitive — both in admins and, more importantly, in population receiving the intensive package:
| PfPR gate | High admins | High pop (M) | % of admins | % of population |
|---|---|---|---|---|
| 20% | 137 | 440.0 | 43% | 52% |
| 25% | 103 | 321.7 | 33% | 38% |
| 30% | 75 | 209.1 | 24% | 25% |
| 35% | 45 | 91.1 | 14% | 11% |
| 40% | 29 | 55.6 | 9% | 7% |
The population effect is non-linear: dropping 35%→30% adds only 30 admins but more than doubles the covered population (91M→209M), because dense, high-burden admins — 13 of the 30 are in Nigeria — sit right in the 30–35% band.
5 · Distribution of prevalence by country
Why the gate placement matters so much: the admin-1 PfPR distribution is dense exactly through the 20–40% range, so small moves in the cutoff sweep many admins. Countries are ordered by median prevalence.
6 · The geographic shift
The same move seen spatially: the "high" zone (red, receiving the intensive package) contracts sharply as the gate rises from 20% to 40%.
7 · Population density vs prevalence
For impact, what matters is not just whether an admin is "high" but whether it is dense and high-burden. The upper-right cluster — dense, high-prevalence admins — is where the gate decision drives the most averted burden (and cost).
8 · Options for review
30% is a reasonable, team-endorsed working choice. The points below are about making it robust and transparent, not overturning it.
- Add the one-line citation in
1a_geo_coverage.Rmd(Kenya/Tanzania NMCP ≥30% precedent) so the basis travels with the code. - Report impact at 30% and 35% side-by-side. Because 35→30 doubles the intensive-package population (91M→209M), leadership should see the sensitivity explicitly rather than a single point estimate.
- Seriously weigh a region-aware or higher gate. Given that 60% of the high-transmission population is West African and 30% is only an upper-third cutoff there, a single continental gate tuned to East Africa likely over-classifies the highest-impact region. Candidates: a uniform 35% (WHO-canonical, more conservative in the West), or region-specific gates (e.g., 30% East / ~40% West) if the team is comfortable with heterogeneity.
- Keep 25% / 20% as sensitivity bounds only — no primary source anchors them.
- Note the metric mismatch: the model gates on all-ages MAP PfPR; WHO/MAP classes and the Kenya/Tanzania precedent use PfPR2-10 / PfPR5-16. Worth a sentence in the methods.