Earlier Care Is Not Earlier Proof: The Measurement Loop MBG Needs for Mothers, Toddlers, and 3T Nutrition

MBG Watch · 2026-08-22

The premise

Indonesia’s Free Nutritious Meals program is being asked to work earlier in life and farther from the school gate.

On 7 August 2026, ANTARA reported that the government was sharpening MBG’s target toward pregnant women, breastfeeding mothers, toddlers, and communities in 3T areas — disadvantaged, frontier, and outermost regions. Coordinating Minister for Food Affairs Zulkifli Hasan said about 11.15 million pregnant women, breastfeeding mothers, and toddlers still had not received MBG benefits, and that 352 SPPG units in 3T areas across eight high-stunting provinces were being accelerated. BGN’s own February statement had already described MBG as “school meal plus,” because its priority group includes toddlers, pregnant women, and breastfeeding mothers before students.

That shift may be reasonable. The first 1,000 days are a real window of risk. The Ministry of Health’s 2025 release on SSGI 2024 says national stunting prevalence fell from 21.5 percent in 2023 to 19.8 percent in 2024, while also stressing that intervention must begin before birth, including attention to maternal nutrition, anemia, mid-upper-arm circumference, hemoglobin, iron tablets, micronutrients, and measurement quality at Posyandu.

But earlier service is not earlier proof.

A meal delivered to a toddler, a pregnant woman, or a remote household is still only a delivery record. It is not yet evidence that anemia risk fell, dietary diversity improved, wasting was caught sooner, a referral was completed, or birth and child-growth risks changed. MBG Watch has already argued this from several angles: “Why MBG Has No Measurable Nutrition Outcomes — The Corruption-Nutrition Pathway” showed the gap between reach claims and health outcomes; “Seen Without Being Watched” set the privacy boundary for beneficiary validation; “The Care Has to Travel” explained why non-school beneficiaries need a care-route record; “Before the Number Becomes a Fact” named the provenance standard for public claims; “From Dashboard to Guidance” asked for parent-understandable menu records; and “Year 2 Baseline” treated the second year as an evidence problem, not only an expansion problem.

The next step is to join those records. MBG needs a measurement loop that is useful enough for care, disciplined enough for public accountability, and private enough not to turn vulnerable families into a surveillance dataset.

Three records, not one

The simplest mistake would be to build one giant dashboard and call it evidence. MBG needs three linked but separated records.

The delivery record answers whether the promised food reached the intended group. For 3B and 3T beneficiaries, that means more than a kitchen output count. It should distinguish meals planned, cooked, dispatched, received at the handoff point, collected or delivered, refused, missed, spoiled, discarded, replaced, and reimbursed. It should also identify the route type without exposing the person: school, PAUD, Posyandu, Puskesmas, community drop point, home delivery, boat route, motorbike route, shelter route, or another care route.

The safety record answers whether the food stayed safe. It should carry batch time, holding condition, kitchen and route inspection status, complaint channel, illness reports, lab confirmation where relevant, corrective action, and restart criteria. This is the continuity with MBG Watch’s earlier food-safety work. A nutrition program that causes foodborne illness has not merely had an operational incident; it has harmed the health outcome it exists to improve.

The outcome record answers whether the beneficiary’s nutrition risk is changing. This record should not be built from MBG delivery data alone. It should connect, with privacy protections, to Indonesia’s existing maternal and child health infrastructure: Posyandu, Puskesmas, e-PPGBM, ASIK, and national surveys such as SSGI. The Ministry of Health’s SSGI 2024 data catalogue describes SSGI as a national cross-sectional survey of toddler nutritional status across 514 districts/cities, collecting nutrition-status and intervention indicators. It is a population evidence instrument. It is not the same thing as a monthly care record for every MBG recipient.

The three records should meet only where they need to. Delivery shows exposure. Safety shows harm or safe completion. Outcome shows health movement. A meal handoff should never be allowed to masquerade as a nutrition outcome.

A minimum outcome loop

For pregnant women, the minimum loop should start with a protected baseline inside the health system: pregnancy status, gestational period where needed for care, attendance at antenatal care, nutrition-risk flags already used by health workers, anemia screening or hemoglobin where available, mid-upper-arm circumference where used, receipt and adherence support for iron-folic-acid or other micronutrient supplementation, and referral status when risk is detected. The public does not need a pregnancy register. It needs aggregated evidence by district, kitchen cluster, or care route: how many eligible pregnant women were reached, how regularly, with what meal pattern, and what proportion completed the follow-up checks already recognized by health services.

For breastfeeding mothers, MBG should measure service continuity without pretending that a meal alone can prove infant outcome. Useful protected fields include service contact, breastfeeding support referral where appropriate, maternal nutrition-risk flags, meal receipt, missed meals, adverse events, and follow-up contact. Public reporting should be aggregated: coverage, missed-delivery rate, exception resolution, complaint closure, and linkage to health-service follow-up.

For toddlers and young children, MBG should not wait years for stunting headlines. WHO child-growth standards already provide the anthropometric backbone: length or height-for-age, weight-for-age, weight-for-length or weight-for-height, BMI-for-age, head circumference, arm circumference, and related standards. WHO and UNICEF child-feeding indicators also provide near-term diet measures. WHO defines minimum dietary diversity for children 6–23 months as consumption, during the previous day, of foods and beverages from at least five of eight food groups. WHO’s minimum acceptable diet indicator combines dietary diversity and age-appropriate meal frequency, with milk-feed requirements for non-breastfed children.

Those indicators are not a demand that MBG publish a child’s health record. They are a way to choose sensible aggregate measures: proportion of 6–23 month beneficiaries meeting minimum dietary diversity in a sampled check; proportion receiving age-appropriate meal frequency support; proportion with growth monitoring completed on schedule; proportion flagged for wasting, severe underweight, or growth faltering; proportion referred; and proportion whose referral was completed.

For 3T care routes, the minimum loop must include access friction. A kitchen can be technically open while a household remains unreachable. Public indicators should include planned versus completed route days, weather or road exceptions, cold/hot holding breaches, missed settlements, replacement deliveries, safe-discard volumes, referral delays, and the number of days between a risk flag and completed follow-up. In remote service, timeliness is not decoration. It is part of care.

Use what Indonesia already has

MBG should add a nutrition loop; it should not duplicate the whole health system.

Indonesia already has community health posts and health centers that carry much of the maternal and child monitoring function. A 2022 Frontiers in Public Health study describes Posyandu as integrated health service posts for maternal and child health where cadres conduct nutritional screening for children and pregnant women, with Puskesmas support. The same study found common practical constraints: data recording and cadre skills needed improvement; e-PPGBM existed for nutritional-status recording, but cadres often still recorded on paper and Puskesmas staff later entered data; double entry was inconvenient and time-consuming; data quality was not reassuring where skills and knowledge were thin.

UN Global Pulse’s assessment of e-PPGBM, republished by ADB SEADS, makes a similar point at system level. It describes e-PPGBM as the Ministry of Health’s application for recording and reporting community nutritional status, designed to identify and monitor nutritional data in a timely, systematic, and sustainable way. But it also identifies implementation gaps: unclear follow-up mechanisms, inefficient data use at district and provincial health offices, heavy reliance on Puskesmas nutritionists, large monthly data-entry burdens, error-prone systems, a data-action gap, and underreporting of interventions.

The Ministry of Health’s ASIK support pages show an attempted bridge. ASIK Mobile is used by health workers and cadres for outside-facility recording at Posyandu, including infant-toddler services, school-age and adolescent services, and pregnant women. The ASIK purpose page is plain about the operational problem: Posyandu recording has often been manual; Puskesmas nutrition staff may need to enter thousands of records monthly; input errors and delayed interpretation can slow treatment; ASIK is meant to simplify reporting, validate weight and height, and report cases needing referral.

For MBG, this means the measurement loop should be interoperable with health records rather than parallel to them. MBG can own delivery, menu, safety, kitchen, route, and exception data. Health services should own identifiable health assessment, clinical follow-up, and protected case management. Public accountability should receive aggregate joins: not “this child is wasted,” but “in this district and route type, this share of eligible toddlers received the intended meal exposure, this share had growth monitoring, this share triggered referral, and this share completed referral within the reporting window.”

That is enough to learn without exposing the child.

The privacy boundary

The privacy boundary should be bright.

Public MBG reporting should not include child-level identity, household identity, precise location, pregnancy status, health condition, attendance trace, biometric marker, face image, phone number, or a route pattern that can identify a household by inference. Indonesia’s Personal Data Protection Law already treats children’s data and health data as sensitive categories, as MBG Watch discussed in “Seen Without Being Watched.” WHO’s own data policy points in the same direction for public-health data: publish data stripped of personal identifiers, aggregate and analyze anonymized data, and use ethical and secure measures to protect privacy, confidentiality, and avoid stigmatization or exclusion.

A privacy-preserving MBG measurement loop would publish method, denominator, uncertainty, aggregate result, and correction trail. It would not publish the person.

This distinction matters most for 3B and 3T beneficiaries. A school count can sometimes be aggregated without exposing a child. A pregnant woman in a small village, a toddler with severe wasting on a remote route, or a breastfeeding mother receiving home delivery may be identifiable from very little information. The more vulnerable the group, the more careful the public record must be.

What should be public

A minimum public measurement loop could be published monthly or quarterly at district, kitchen-cluster, and care-route level, with suppression rules for small cells.

First, baseline context: eligible beneficiary estimate, data source, update date, and whether the denominator is official target, verified register, sampled estimate, or corrected count.

Second, dose and exposure: number and share receiving the intended meals, average meal days per beneficiary, missed-meal rate, replacement rate, and continuity across school holidays or route disruption.

Third, meal quality: menu compliance, age or pregnancy suitability, dietary-diversity contribution, local substitution rule, and parent or beneficiary guidance record.

Fourth, safe delivery: kitchen inspection status, route time, holding condition, incident count, complaint count, illness investigation count, and corrective action.

Fifth, follow-up measure: aggregate share of toddlers with recent growth monitoring, aggregate share of pregnant women with relevant nutrition checks where available, and sampled dietary-diversity or meal-frequency indicators for 6–23 month children.

Sixth, exception and referral: number of wasting, severe underweight, anemia-risk, non-consumption, missed-delivery, unsafe-food, or service-access flags; share referred; share completed; median days to completion.

Seventh, correction: what changed because the record showed a problem — menu adjustment, route change, kitchen suspension, cadre support, referral escalation, replacement meal, procurement correction, or public correction of a claim.

This is not a call for perfect causal attribution every month. It is a call for a visible learning loop. Public numbers can be modest and still be useful if they connect exposure, risk, follow-up, and correction.

What the evidence does not support

The current public record does not support a claim that expanding MBG to pregnant women, breastfeeding mothers, toddlers, and 3T regions will, by itself, reduce stunting or maternal-child nutrition risk.

It supports a narrower claim: the government has publicly prioritized these groups; Indonesia already has health-system instruments relevant to measurement; international indicators provide usable child-feeding and growth measures; and privacy-protecting aggregation is possible. The evidence also supports a warning: e-PPGBM and Posyandu systems have known implementation gaps, especially where data entry, follow-up, human resources, and intervention feedback are weak.

The missing proof is not ideological. It is operational. MBG has not yet made public a measurement design that shows how delivery data, safety data, and health outcome data will be linked without exposing beneficiaries.

The least-harm path

The least-harm path is not to pause early-life service until a perfect evaluation exists. It is also not to count meals and declare nutrition success.

A practical sequence would be this.

First, publish the measurement protocol before publishing outcome claims. Define which indicators are delivery, which are safety, and which are health outcomes. Define which are routine administrative indicators and which require sampling or health-service verification.

Second, pilot the loop in a small set of 3B and 3T districts before national claims are made. Test the data burden on cadres, Puskesmas staff, SPPG operators, and district health offices. If the loop requires double entry or produces no corrective action, it will become another reporting ritual.

Third, keep identifiable health data inside authorized health systems. MBG should receive only the minimum linkage needed to correct delivery and support care, with public reporting at aggregate level.

Fourth, measure correction, not only detection. A dashboard that flags anemia risk, wasting, missed meals, or unsafe delivery but cannot show referral completion and operational change is not a care loop. It is a warning light disconnected from the brake.

Fifth, label every public number by evidentiary state: target, registered beneficiary, served beneficiary, sampled dietary indicator, verified health follow-up, survey estimate, corrected figure, or unresolved claim.

That would not prove MBG’s long-term effect immediately. It would make the program learnable while it acts.

What I am uncertain about

I am uncertain how far BGN has already built maternal, toddler, and 3T outcome measurement into its internal systems. BGN’s public technical guide for pregnant women, breastfeeding mothers, and non-PAUD toddlers names accurate BKKBN beneficiary data, periodic updating, monitoring and evaluation, follow-up on findings, and feedback from beneficiaries and field implementers. It does not, in the public page I retrieved, specify the minimum outcome indicators, privacy-preserving linkage rules, or public aggregate reporting format.

I am also uncertain how much additional burden Posyandu cadres and Puskesmas nutritionists can absorb without weakening the work they already do. The e-PPGBM and ASIK evidence suggests that measurement quality depends on workflow, not only software.

The central conclusion is therefore deliberately narrow. Earlier care can be valuable. Earlier care is not earlier proof. For MBG’s 3B and 3T expansion, the public record should show a loop: baseline, exposure, meal quality, safe delivery, follow-up measure, exception, referral, correction, and aggregate update. Without that loop, the program can count more vulnerable beneficiaries while still leaving the question of benefit unanswered.