Health Supply Chain

A health supply chain moves commodities — medicines, vaccines, contraceptives, diagnostics, consumables — from a national warehouse to the point of use. The information system that supports it (an LMIS, or eLMIS) is one of the oldest categories of health software, and one of the least well integrated with clinical systems.

That gap is the main architectural opportunity: clinical systems know what was dispensed, and logistics systems mostly know what was issued. Those are different numbers, and the difference is why forecasting is poor.

The core data

Entity	Notes
Product	The commodity, at a defined level of specificity — see below
Facility / storage location	From the facility registry, including warehouses and sub-stores
Stock on hand	Quantity at a location at a point in time
Batch / lot	With expiry date; essential for recalls and expiry management
Transaction	Receipt, issue, dispense, adjustment, loss, transfer, return
Requisition	A request from a facility upward
Consumption	What was actually used — distinct from what was issued
Cold chain	Equipment, temperature excursions, functional status

The transaction log is the model. Stock on hand is a derived value — computed from the transaction history, not stored as a mutable number. Systems that store stock as an editable field lose the ability to explain discrepancies, which is the main analytical use of the data.

The product catalogue

The shared list of commodities, and the master data that determines whether any of this interoperates.

The specificity problem. "Amoxicillin" is a substance. "Amoxicillin 250 mg dispersible tablet" is a product. "Amoxicillin 250 mg dispersible tablet, 100 tablet blister, manufacturer X" is a trade item. Different systems need different levels:

   Substance            ← clinical decision support, allergy checking
        │
   Clinical product     ← prescribing, dispensing, EMR medication list
        │
   Packaged product     ← procurement, stock counting
        │
   Trade item (GTIN)    ← scanning, traceability, recall

A catalogue that models only one level forces every other use to approximate. The minimum workable design holds the clinical product and the packaged product, with an explicit relationship between them and defined pack sizes.

Standards:

GS1 — GTIN for trade items, GLN for locations, and barcode standards. Increasingly required for traceability and recall.
ATC — WHO's Anatomical Therapeutic Chemical classification, useful for grouping and consumption analysis.
National drug codes / essential medicines list — usually the operative catalogue.
RxNorm — normalised clinical drug names; US-centric but a useful model.
FHIR — Medication, MedicationKnowledge, and the supply resources SupplyRequest, SupplyDelivery, InventoryReport.

The catalogue must be a governed, versioned shared service, not a spreadsheet per system. See terminology services.

Issued versus consumed

The distinction that makes the clinical–logistics integration worth building.

Issued — what the warehouse sent to the facility. Known to the LMIS.
Consumed — what was actually dispensed to patients. Known to the pharmacy or EMR.

Forecasting from issues perpetuates existing allocation patterns, including their errors: a facility that was over-supplied continues to be over-supplied, because its issues look like demand. Forecasting from consumption reflects actual need.

   EMR / pharmacy ──dispensing events──▶ ┌──────────────┐
                                         │  IOL         │
   LMIS ──stock, issues, receipts──────▶ │              │
                                         └──────┬───────┘
                                                ▼
                                    Consumption-based forecasting
                                    Stock-out analysis against
                                    clinical demand

The requirement this places on the architecture: medication coding must be consistent between the EMR and the LMIS, at a level that maps to the catalogue. This is the practical argument for a national drug terminology, and it is usually the blocking issue.

Requisition and distribution models

Model	How	Suits
Pull (requisition)	The facility requests what it needs	Facilities with capacity to forecast; variable demand
Push (allocation)	The higher level decides and sends	Campaigns, emergencies, weak facility capacity
Informed push / vendor-managed	The supplier reviews stock and resupplies to a target	Reduces facility burden; needs reliable stock visibility
Max/min	Resupply to a maximum when below a minimum	The common default; requires reliable consumption data

The information requirements differ substantially. Informed push depends on accurate, timely stock data from every facility — which is a demanding requirement, and the reason it is often adopted in principle and abandoned in practice.

Cold chain

Vaccines and some biologicals require continuous temperature control, and the data has a different character from stock data.

Equipment registry — refrigerators and freezers as devices, with location, model, functional status and capacity. Overlaps with the facility registry.
Temperature monitoring — continuous or interval readings from remote temperature monitoring devices; high volume, time-series.
Excursions — an alert with a defined response workflow, because the question "is this stock still usable?" needs an answer within hours.
Vaccine vial monitors and shake tests — the low-technology controls that remain the operative ones in many settings.

Temperature data is time-series and does not belong in the transactional LMIS database. A separate time-series store with alerting is the appropriate shape.

Interoperability requirements

An LMIS depends on:

Facility registry — delivery points must be the same facilities the clinical systems use, including sub-stores which the clinical systems may not distinguish
Product catalogue — shared, versioned, at the right levels of specificity
Clinical systems — for dispensing and consumption
HMIS — stock-out indicators are standard health system performance measures
Financial systems — procurement, budget, payment
Programme systems — immunisation campaigns have their own commodity flows

The two that are consistently underestimated are the facility registry alignment (sub-store granularity) and the product catalogue mapping.

Analytics that matter

Stock-out rate by product and facility — the headline indicator, and the one most susceptible to definitional dispute (stocked out on the day of reporting, or at any point in the period?)
Days of stock remaining, against consumption
Wastage — expiry, breakage, cold chain loss
Order fill rate — how much of what was requested was supplied
Reporting rate and timeliness — whether the data exists at all
Expiry risk — batches approaching expiry with insufficient consumption

Publish the definitions with the numbers. Stock-out rate computed two different ways produces two different national figures, and both get quoted.

Platforms

Platform	Notes	Licence
OpenLMIS	Open-source LMIS designed for low-resource public health supply chains	AGPL
DHIS2	Widely used for aggregate stock reporting; not a full LMIS but often the pragmatic starting point	BSD
eLMIS implementations	Several national systems, some derived from OpenLMIS	Varies
mSupply	Used in several countries including for remote and mobile stores	Commercial, with open components
Commercial ERP	Where national procurement is already ERP-based	Commercial

Verify current status before selection; see platform directory.

Checklist

References

OpenLMIS — https://openlmis.org/
GS1 healthcare standards — https://www.gs1.org/industries/healthcare
WHO ATC/DDD — https://atcddd.fhi.no/
WHO immunization supply chain resources — https://www.who.int/teams/immunization-vaccines-and-biologicals/essential-programme-on-immunization/supply-chain
FHIR Medication — https://hl7.org/fhir/medication.html

The core data​

The product catalogue​

Issued versus consumed​

Requisition and distribution models​

Cold chain​

Interoperability requirements​

Analytics that matter​

Platforms​

Checklist​

References​