Decision-Making Library
Operational Decision-Making & Override Protocol Library Version 2.0 - Post-Validation Revision
Introduction & Rationale
This library provides a standardized protocol for the most critical and recurring decisions in Uganda's health supply chain. It is built directly from the scenarios and challenges documented in the 2025 Baseline Assessment.
Core Philosophy
The system informs, the health worker decides.
This ensures that local knowledge and clinical judgment are preserved while leveraging predictive analytics where beneficial.
Core Decision-Making Protocol
Every decision scenario follows a standard structure:
Trigger
What event or condition initiates this decision?
System Recommendation
What does the forecasting system (Tier 1, 2, or 3) suggest?
Decision Authority
Who is empowered to make the final call?
Override Process
How does the user accept, modify, or reject the suggestion?
Required Data Inputs
What information is needed for an informed decision?
Expected Outcome
What is the goal of this decision?
Foundational Principles
Local Edits Always Win
CRITICAL DESIGN PRINCIPLE
In all decision scenarios, when conflicts arise between system recommendations and user adjustments, the facility worker's local edit is preserved.
This policy reflects the technical validation finding that health workers on the ground have contextual knowledge algorithms cannot capture.
System logs conflicts for review by district officers
Cannot retroactively override facility decisions without explicit facility consent
Builds trust and acknowledges frontline workers have ground truth
Tier 2 Forecasts: Advisory Only
When Tier 2 hierarchical statistical forecasts are mentioned in use cases, these are informational only:
Generated by district-level models and pushed to facilities during sync
Provide a "second opinion" for comparison purposes
Do not replace the operational Tier 1 rule-based forecast
Viewable in a separate "District Insights" section of the mobile app
Use Case Library
UC-01: Routine Weekly Ordering with Storage Constraints
Baseline Context: Facilities with inadequate storage capacity experienced the most stockouts (r = -0.695).
Trigger
Weekly inventory count completed in offline app
System Recommendation
Tier 1 forecast calculates order, visibly capped by storage capacity (e.g., 70% cap for inadequate storage)
Decision Authority
Facility Store Manager/In-Charge
Data Inputs
Current stock, last 3-month consumption, storage capacity rating
Expected Outcome
Order balancing predicted demand with physical and financial realities
Override Process:
Adjust Up
Temporary storage arranged
Adjust Down
Budget constraint or lower consumption trend
UC-02: Disease Outbreak Response
Baseline Context: During disease outbreaks or seasonal surges, consumption may far exceed AI forecasts.
Trigger
DHIS2 surveillance reports >50% increase in disease cases OR user manually activates "Outbreak Mode"
System Recommendation
Tier 1 applies pre-set multiplier (e.g., 2x for malaria). Alert: "Outbreak Mode Active: Suggested order doubled to 2,000 Coartem"
Decision Authority
Facility Clinical Officer/In-Charge
Data Inputs
DHIS2 surveillance data, facility outbreak reports, current stock levels
Expected Outcome
Rapid response preventing stockouts during health emergencies
Override Process:
User can adjust multiplier (1.5x, 3x, etc.) based on outbreak scale/severity.
Mandatory Reasons (Dropdown):
Malaria Outbreak
Cholera Outbreak
Pneumonia Surge
Measles Outbreak
Other Clinical Judgment
UC-03: Response to Delivery Delays & Road Inaccessibility
Baseline Context: "From May to September, we have the rainy season... blocks access to the facility" - baseline respondent
Trigger
Delivery marked late in WFP LESS OR user reports road impassable due to weather
System Recommendation
Tier 3 model suggests redistribution options from nearest accessible facility with surplus stock
Decision Authority
District Supply Officer
Data Inputs
LESS delivery status, weather data, stock levels of nearby facilities
Expected Outcome
Minimized stockout duration for facilities cut off by logistics challenges
Override Process:
District officer reviews redistribution suggestion and can approve, modify quantities, or reject based on ground logistics knowledge.
Mandatory Reasons:
Roads impassable
Redistribution approved
Alternative route available
UC-04: Mid-Cycle Budget Reallocation
Baseline Context: "Budget information was unavailable at most spoke facilities and budget management is a serious issue here" - baseline respondent
Trigger
Mid-year budget top-up received from district or donor
System Recommendation
Central system recalculates facility budget ceilings and pushes updated, budget-aware Tier 1 forecasts upon next sync
Decision Authority
District Health Officer/Finance Officer
Data Inputs
Updated budget allocation, facility utilization rates, stockout risk scores
Expected Outcome
Transparent and equitable budget utilization responding to changing priorities
Override Process:
District officer can manually adjust auto-allocated budgets to prioritize high-need or outbreak-affected facilities.
Mandatory Reasons:
Donor funding received
Reprioritization due to outbreak
Unspent funds reallocated
UC-05: Manual Correction of Forecasting Errors
Baseline Context: Data or forecasting errors were cited as a key reason for human override.
Trigger
User identifies clearly erroneous forecast
System Recommendation
Tier 1 forecast displayed as usual
Decision Authority
Any authorized facility user
Data Inputs
User's local knowledge, current stock ledger, recent consumption data
Expected Outcome
Data quality improvement and prevention of stockouts due to system errors
Examples of Erroneous Forecasts:
Predicting zero quantities for essential items
Quantities 50% higher/lower than recent consumption without seasonal justification
Recommendations based on outdated stock data
Override Process:
User directly edits forecast quantity to correct value.
Mandatory Reasons:
Data entry error
System forecasting error
Changed consumption pattern
UC-06: Nutrition Program Enrollment Surge
Baseline Context: WFP stakeholder feedback emphasized nutrition supply adequacy for vulnerable populations including refugees and climate-affected communities. Persistent stockouts require improvisation and risk program effectiveness.
Trigger
Nutrition Appointment Platform reports 500+ new beneficiaries enrolled
System Recommendation
Tier 3 district model adjusts demand forecast for nutrition commodities (RUTF, micronutrient supplements, ORS) based on enrollment data
Decision Authority
District Supply Officer in coordination with nutrition program officers
Data Inputs
Nutrition Appointment Platform enrollment data, nutrition commodity stock levels, facility capacity assessments, WFP LESS delivery schedules
Expected Outcome
Proactive stock allocation prevents shortages during program scaling. Supply adequacy indicators pushed back to Nutrition Appointment Platform
Override Process:
District officer reviews allocation recommendations across facilities and can adjust based on:
Geographic distribution of beneficiaries
Facility storage capacity for nutrition commodities
Existing stock levels
Delivery logistics to refugee settlements or remote areas
Mandatory Reasons:
Nutrition program expansion
Refugee influx
Seasonal malnutrition spike
Redistribution for equitable access
Integration, Monitoring & Learning
Integration
These use cases are embedded within the offline app and district dashboard workflows, following the Integration Maturity Model (Level 2: Opportunistic Sync).
Monitoring
All overrides are logged with:
User ID
Timestamp
Original value
Final value
Reason
Logs serve two purposes:
Learning
Identify systematic forecasting errors and improve Tier 1 rules and Tier 3 models
Accountability
Provide transparent audit trail for procurement decisions
No-Blame Principle
High override rates are a flag for system improvement, not user penalization.
Override frequency above 50% for a specific commodity at a facility triggers investigation of forecasting assumptions, not disciplinary action.
Value-for-Money (VfM) Assessment Framework
This framework evaluates ROI using the standard 4E model, separating hard financial savings from operational efficiency gains.
Key VfM Indicators
Economy (Spending Less)
Reduction in expiries (UGX), Emergency logistics cost reduction, Deployment cost per facility
Efficiency (Spending Well)
Staff hours saved per ordering cycle, Sync reliability (target: >90%)
Effectiveness (Spending Wisely)
Stockout duration reduction (baseline: 120 days), Forecast accuracy improvement (MAPE)
Equity (Spending Fairly)
Variance reduction in stockout days between RRH and refugee-hosting facilities
Data Sources
Hard Costs
NMS Invoices (expiry values), District Fuel Logs (transport costs), Project Financial Reports
Soft Metrics
System Activity Logs (sync rates, processing time), DHIS2 (stockout duration)
Sample VfM Calculation
Total Annual Cost
UGX 50 million (maintenance, server, support)
Drug Expiry Reduction
UGX 80 million saved
Logistics Savings
UGX 20 million (fuel/per diems)
Total Hard Savings
UGX 100 million
Financial ROI
(100M - 50M) / 50M = 100%
Governance
Quarterly Technical Reviews: MAPE, sync rates reviewed to fix model drift
Annual Financial Review: MoH and partners evaluate if system pays for itself in saved commodities
Document Information
Version
2.0 (Post-Technical Validation)
Date
November 2025
Authors
Gideon Abako, Timothy Kavuma (IFRAD)
Validator
Ojok Ivan, Kyambogo University
License
CC BY 4.0
Last updated