daly-weighted-analysis

name: daly-weighted-analysis description: Phase 6 Commit ι requirement. When a model produces an allocation or policy recommendation, the report must include DALY-averted figures alongside cases-averted unless the modeler explicitly justifies their absence. DALYs (disability-adjusted life-years) are the standard metric for Global Fund / GBD / WHO health-allocation analyses; cases-averted alone treats a 6-month-old's averted infection identically to a 30-year-old's, and dramatically under-weights interventions that target high-mortality subpopulations (SMC for U5 children, vaccines for newborns, ART for advanced HIV). Trigger phrases include "DALY", "disability-adjusted life-years", "cost per DALY", "DALY-averted", "GBD weight".

DALY-Weighted Analysis

Why this skill exists

Across three malaria runs (1302, 2057, 0912), the modeler reported "cases averted" as the headline outcome metric. This is the wrong metric for Global Fund allocation decisions because:

1. Cases-averted treats all cases equal. A clinical malaria episode in a 6-month-old (mortality risk ~0.4%, life-years lost per death ~30, severe-malaria probability ~5%) carries far more public-health weight than the same clinical episode in a 25-year-old. SMC (U5-only) appears 2-3× more cost-effective under DALYs than under cases-averted; ITNs (all-age) appear similar under both. The optimal allocation differs.

2. GF / GBD / WHO use DALYs as the standard. A Gates Foundation modeler producing a Nigeria GC7 supplementary analysis WILL get asked "what is the cost per DALY averted?" If the report doesn't answer, it doesn't ship.

3. Cost-effectiveness thresholds are DALY-based. GiveWell's ~$5,000/life-saved threshold, the WHO-CHOICE 1×GDP-per-capita threshold for cost-effective interventions, and CEA thresholds in Cochrane reviews are all DALY-denominated. Cases-averted alone can't be compared to these references.

The Phase 6 Commit ι gate enforces: any run with decision_rule.md or *allocation*.csv must mention DALYs in report.md (or explicitly justify their absence).

When DALYs matter (high-stakes scenarios)

• Allocation decisions with age-targeted interventions (SMC, IPTp for pregnant women, RTS,S/R21 vaccine, paediatric ARVs)
• Allocation decisions with mortality differentiation between packages (e.g., severe-malaria treatment vs prevention)
• Cost-effectiveness comparisons against published benchmarks (Conteh 2021, GiveWell, GBD CEA tables)
• Anything destined for Global Fund / GAVI / PMI decision-makers

When DALYs add little

• Within a single age group / interventions with similar age profile (e.g., comparing standard vs PBO LLINs at the same coverage)
• Pure prevalence outcomes (PfPR change) without case-counting
• Rapid screening / resource-prioritization at the same stratum

In these cases the modeler MAY justify cases-averted as the headline in a paragraph in report.md §Methods, but should still report DALYs in a Limitations or Sensitivity section for completeness.

Computation: the DALY formula

For each averted case, DALYs averted = YLL (years of life lost) + YLD (years lived with disability). For an averted infection that would have led to:

• A non-fatal clinical episode → contributes YLD only:

  YLD = duration_in_years × disability_weight
  

• A fatal case → contributes YLL only:

  YLL = (life_expectancy_at_age - 0)  # discounted in some conventions
  

For a malaria allocation model, the typical aggregation is:

def compute_dalys_averted(cases_averted_by_age: dict, country_anchors: dict):
    """cases_averted_by_age: {"0-5m": N, "6-59m": N, "5-14y": N, "15+": N}
    country_anchors: see DISEASE_ANCHORS below."""
    total_dalys = 0.0
    for age_group, n_cases in cases_averted_by_age.items():
        a = country_anchors[age_group]
        # YLD: clinical episodes that don't become fatal
        n_clinical_nonfatal = n_cases * (1 - a["p_fatal"])
        yld = n_clinical_nonfatal * a["episode_duration_yrs"] * a["disability_weight"]
        # YLL: cases that become fatal
        n_fatal = n_cases * a["p_fatal"]
        yll = n_fatal * a["yll_per_death"]
        total_dalys += yld + yll
    return total_dalys

Disease-specific anchor tables

Malaria (P. falciparum, sub-Saharan Africa)

Source: GBD 2019 / Murray 2020 / IHME malaria estimates.

Notes:

• p_fatal includes both directly-attributable and indirect malaria deaths. Sub-Saharan African values from IHME malaria estimates 2020-2022.
• yll_per_death is undiscounted and uses Nigeria's life expectancy table (LE at birth ~63yr; LE at 5yr is higher due to surviving infancy).
• disability_weight from GBD 2019 disability-weights catalog:
  • Severe acute malaria (Diseases): 0.211
  • Moderate clinical: 0.137
  • Mild / semi-immune adult: 0.069

TB (drug-susceptible, sub-Saharan Africa)

HIV (with ART access)

Measles

Required output for Phase 6 Commit ι

When a model produces an allocation, the report MUST include AT LEAST ONE of:

1. DALY-averted column alongside cases-averted column in the primary results table (preferred, e.g., Table 7 / Table 10):

   | Package | LGAs | Cases Averted | DALYs Averted | $/Case | $/DALY |
   

2. A separate §Cost-Effectiveness section with DALY computation methodology and per-package cost-per-DALY estimates.

3. A justification paragraph in §Methods or §Limitations explaining why DALYs are not relevant for this analysis (must address: are interventions age-targeted? do packages differ in mortality risk? is the audience GF/GBD/WHO?).

The validator regex looks for \b(DALY|disability-adjusted) in report.md. If absent, fires daly_analysis_missing MEDIUM.

Worked example: Nigeria GC7 malaria

For the 2057/0912 Nigeria GC7 setup (PBO+SMC dominant in 218 LGAs; SMC eligible only in northern Sahel zones), the DALY-averted shift typically looks like:

SMC cost-per-DALY ($510) is far better than PBO ($1,430) — 3× advantage that's invisible under cases-averted ($150/case for SMC vs $80/case for PBO would suggest the opposite ranking).

This is the kind of recommendation shift that Global Fund / NMEP planners depend on. Without DALY weighting, the model is structurally biased toward all-age interventions over child-targeted ones.

Common pitfalls

1. Using global GBD anchors for a specific country. Nigeria's under-5 malaria CFR is higher than SSA average; use country-specific values where available (IHME GHDx country profiles). Document the source.

2. Forgetting age-stratified case attribution. If your model doesn't track cases by age group, you cannot compute proper DALYs without an assumption about the U5 share. Document the assumption.

3. YLL discounting choices. GBD uses a 3% annual discount rate on YLL by default. Some agencies (PMI, GiveWell) use undiscounted YLL. Document which convention you use; it can shift CE by 2×.

4. Disability weights vs life expectancy table mismatch. Disability weights are GBD-uniform; life expectancy comes from country tables. Use them consistently (don't mix GBD weights with WHO regional life tables, etc.).