CSIAF: Cross-Survey Indicator Alignment Framework
The Cross-Survey Indicator Alignment Framework (CSIAF) is the rubric the platform uses to decide, indicator by indicator, whether an SNMS measure can stand in for the corresponding national-survey indicator and on what terms. Each indicator is scored on six alignment dimensions against MIS 2021 (primary) and NDHS 2023-24 (secondary), and the resulting verdict is the single canonical reading used across the site.
See also: Methodology and FAQ for calibration, robustness, frequently asked questions, the canonical citation, and weighting summary.
Indicator ratings
Benchmarked to MIS 2021. No indicator qualifies as a face-value direct substitute: SNMS is an engaged-population phone survey, so the Validity dimension carries a Major selection difference for every indicator, correctable in principle by calibration but never zero.
U (Universe): Whether SNMS covers the same eligible group as the national survey, such as the same birth-recall window or age range, so that both are describing the same population.
N (Numerator): Whether a positive case is defined identically on both sides, such as what makes a net count as insecticide-treated, so that a "yes" carries the same meaning in each survey.
D (Denominator): Whether the count is divided by the same base population, such as every household member versus the occupants of a single net, so that the two rates are built like-for-like.
R (Recall): Whether respondents report over the same reference period, since a longer look-back window (five years rather than two) adds memory error and changes who is captured.
Md (Mode): Whether the collection method is comparable, since SNMS is answered by phone or self-completed online while the national surveys use an in-person interviewer who can inspect items such as a net, which can shift the answers.
V (Validity): Whether the respondents the survey reaches represent the whole population, which for SNMS is always a major difference because a phone survey reaches a more connected, wealthier, more health-active group, a selection that calibration reduces but cannot fully remove.
| Code | Indicator | U | N | D | R | Md | V | Verdict |
|---|---|---|---|---|---|---|---|---|
| I1 | Any net ownership | A | A | A | A | m | M | SA |
| Main reason: Construct aligned; calibrate for selection. | ||||||||
| I1b | ITN ownership | A | M | A | A | M | M | CO |
| Main reason: SNMS uses brand-confirmed LLIN; DHS uses inspection. Different numerator; read as a lower bound. | ||||||||
| I4 | Person ITN use | A | A | M | A | m | M | CO |
| Main reason: SNMS captures occupancy of one net only; an approximation. | ||||||||
| I4a | U5 ITN use | M | M | M | A | m | M | CO |
| Main reason: Measured as net use among under-5 households that own a net (a household-occupancy proxy), not per child, and on a different denominator from the DHS measure; read as complementary. | ||||||||
| I6 | ANC by skilled provider | M | A | A | m | m | M | SA |
| Main reason: Numerator aligned to doctor or nurse or midwife; SNMS uses a five-year birth window pending a recency question. | ||||||||
| ANC4 | ANC 4+ visits | M | A | A | m | A | M | SA |
| Main reason: Five-year window to harmonise; otherwise aligned. | ||||||||
| I7b | IPTp 3+ doses | M | A | A | m | A | M | SA |
| Main reason: Five-year window to harmonise; otherwise aligned. | ||||||||
| CM1 | Care-seeking for febrile U5 | A | A | A | A | m | M | SA |
| Main reason: Child path; construct aligned; calibrate for selection. | ||||||||
| CM2 | Diagnostic testing | A | A | A | A | m | M | SA |
| Main reason: Construct aligned; large level gap is selection, correctable by calibration. | ||||||||
| CM3 | ACT receipt | A | A | A | m | m | M | SA |
| Main reason: Denominator is febrile children who took an antimalarial; aligned. | ||||||||
| KB1 | Malaria messages exposure | m | A | A | A | m | M | SA |
| Main reason: Restricted to women; minor age-ceiling difference. | ||||||||
| Pv1 | SMC coverage | n/a | n/a | n/a | n/a | n/a | n/a | NEW |
| Main reason: No MIS or DHS comparator; SMC scaled after 2021. | ||||||||
| Pv2 | R21 vaccine (any dose) | n/a | n/a | n/a | n/a | n/a | n/a | NEW |
| Main reason: No comparator; R21 introduced after both surveys. | ||||||||
| KB2 | Seasonal-only misconception (higher is worse) | n/a | n/a | n/a | n/a | n/a | n/a | NEW |
| Main reason: Share who believe malaria occurs only in the rainy season; incorrect, since transmission continues year-round in much of Nigeria. Sproxil knowledge measure; no comparator. | ||||||||
Summary
- Substitute with adjustment: 8 (I1, I6, ANC4, I7b, CM1, CM2, CM3, KB1)
- Complementary signal: 3 (I1b, I4, I4a)
- New Sproxil measure: 3 (Pv1, Pv2, KB2)
- Total: 14
How to read the values
- Substitute with adjustment means the SNMS construct matches the survey, and the calibrated SNMS value is the comparable figure. The unweighted headline is the engaged-population reading and should not be compared to the survey at face value.
- Complementary signal means read the SNMS value as additional information, not as a replacement for the survey indicator.
- Two indicators carry instrument limits: U5 ITN use is not measurable in SNMS, and person ITN use is approximate.
- ANC by skilled provider, ANC 4+, and IPTp 3+ use a five-year birth window in SNMS until a birth-recency question is added; the reference surveys use a two-year window.
Subgroup concordance (kernel)
Alongside each verdict, the indicator pages report a kernel concordance figure. It is a descriptive cross-check, not an input to the verdict, which is set by the six alignment dimensions above. The kernel asks how closely the SNMS estimate tracks MIS 2021 within matched population subgroups rather than at the national level alone.
Respondents on each side are grouped into cells defined by geopolitical zone and national wealth quintile. The kernel is the share of those cells in which the SNMS estimate falls within 10 percentage points of the MIS estimate, counting only cells that hold at least 30 respondents on each side (a stricter floor than the 25-respondent minimum used for displayed cell values, chosen to keep the subgroup comparison stable).
One limitation governs how the figure should be read. Because the SNMS sample is concentrated in the upper wealth strata (the Methodology and FAQ shows only 57 of 80,269 respondents in the poorest national fifth), almost every cell that clears the minimum size sits in the upper three wealth quintiles. The kernel therefore measures agreement in the wealth strata SNMS actually covers, not across the full national wealth distribution. This also explains why an indicator with a large national level gap can still show moderate subgroup concordance: the national gap is driven by the poorer strata SNMS barely samples, while in the wealthier strata the two surveys can read similarly.
| Code | Indicator | Kernel concordance | Comparable cells |
|---|---|---|---|
| I1 | Any net ownership | 54.2% | 13 of 24 |
| I1b | ITN ownership | 4.2% | 1 of 24 |
| I4 | Person ITN use | 45.8% | 11 of 24 |
| I6 | ANC by skilled provider | 52.6% | 10 of 19 |
| ANC4 | ANC 4+ visits | 63.2% | 12 of 19 |
| I7b | IPTp 3+ doses | 68.4% | 13 of 19 |
| CM1 | Care-seeking for febrile U5 | 42.9% | 6 of 14 |
| CM2 | Diagnostic testing | 0.0% | 0 of 14 |
| CM3 | ACT receipt | 83.3% | 5 of 6 |
| KB1 | Malaria messages exposure | 21.1% | 4 of 19 |
| I4a | U5 ITN use | Pending | Pending |
The three new Sproxil measures (SMC coverage, R21 vaccine, and the seasonal-only misconception) have no MIS comparator, so no kernel is computed for them. Figures resting on few cells, notably the case-management indicators, should be read as indicative; febrile children are scarce in the upper wealth strata, which is why those cell counts are small.
Scope and roadmap
This edition benchmarks SNMS against MIS 2021, with NDHS 2023-24 as a secondary reference. A DHS-primary edition of the framework, with the SNMS values recomputed on the DHS definitions and independently verified, is in development and will supersede the secondary-reference treatment when ready.