Significant gaps remain in linking national diagnostic and treatment data to inventory levels, limiting the ability of programs to anticipate shortages [9]. Current insights are often constrained by the recall bias of household surveys or the lack of real-time consumption data from facility-level assessments. Geographic access also remains a primary barrier; spatial analysis indicates that while high concentrations of health centers improve access to ACTs and mosquito nets, clusters of high malaria incidence persist in underserved regions [10].

Financial sustainability further complicates the supply chain. The Ugandan government allocates only 10% of the total malaria budget, leaving households to bear 67% of the cost [11]. Consequently, poorer regions with higher incidence rates frequently experience alarming shortages at lower-level healthcare facilities, perpetuating a cycle of poverty and disease. Currently, 20% of public facilities in Uganda experience stock-outs, leaving approximately 33% of essential medicines unavailable [1]. These shortages force vulnerable populations to seek alternatives in informal markets, where substandard or falsified antimalarials are prevalent [12]. Improving inventory management is therefore essential not only for clinical outcomes but also for reducing the economic burden on low-income households [1][13].

Geographic Information Systems (GIS) have emerged as a critical tool for improving stock visibility across Uganda’s diverse regions [14]. By mapping medicine availability in real-time, health authorities can identify regional shortages and facilitate emergency redistributions, particularly in remote areas [14]. Complementing GIS, AI-powered forecasting tools enable data-driven decision-making by analyzing consumption trends, weather patterns, and outbreak alerts to predict future demand. Unlike manual forecasting, machine learning can detect complex patterns and trigger proactive countermeasures—such as automated resupply alerts or reallocations—before a stock-out occurs [3][15]. Integrating these technologies makes the supply chain more agile, reducing health inequities and upholding the right to health for at-risk populations [16]. However, a fully operational, digitized surveillance and evaluation (SME) system that incorporates climate and seasonality models into inventory management is still lacking [9].

Phase I established a historical baseline utilizing 260 weekly malaria surveillance reports spanning January 2020 through December 2024, sourced from the Uganda Ministry of Health (MOH) Knowledge Management Portal [4][5][17][18]. To ensure the structural integrity of the Space-Time Cube required for Phase II, 133 of Uganda’s 146 districts were included based on a a ≥ 90% reporting completeness threshold. To mitigate potential Selection Bias, a comparative analysis confirmed that the 13 excluded districts demonstrated no statistically significant differences in baseline Test Positivity Rates (TPR) compared to adjacent included districts (p > 0.05). Geographic Bias was minimized as exclusions were distributed across the West Nile, Karamoja, and Central regions, suggesting data was Missing at Random (MAR). Missing entries (<5% total volume) were addressed via linear interpolation; sensitivity analysis confirmed this method preserved seasonal peaks, with variance remaining within a +2% margin of error compared to five-year historical means [19][20].

Quantitative processing was performed using Python (Pandas/NumPy), with longitudinal trends visualized via Matplotlib to identify systemic “mismatch” patterns. This workflow provided the validated foundation for Phase II, which utilized ArcGIS Pro to conduct an Emerging Hot Spot Analysis (EHSA). The primary outcome variable was defined as Weeks of Stock on Hand (SOH). Understock (<8 weeks) and overstock (>52 weeks) thresholds were operationalized according to the MOH Health Sector Strategic and Investment Plan (HSSIP) [21].

The EHSA utilized a neighborhood of eight nearest neighbors to reflect the average administrative adjacency of Ugandan districts, ensuring identified hotspots represent regional clusters rather than isolated anomalies [2][6]. A two-month temporal window was selected to align with the MOH bimonthly emergency resupply cycles, allowing the model to distinguish between transient logistical delays and sustained trends [16][22]. Statistical significance was determined using the Mann-Kendall trend test to evaluate Z-score trajectories across 15 monthly time steps (p < 0.05) [10][16][23].

For the Emerging Hot Spot Analysis (EHSA), a neighborhood of eight nearest neighbors was selected to reflect the average administrative adjacency of Ugandan districts, ensuring that identified hotspots represent regional clusters rather than isolated outliers [6]. A two-month temporal window was utilized to align with the bimonthly emergency resupply cycles of the Ministry of Health, allowing the model to distinguish between transient logistical delays and sustained trends. Within this framework, the Mann-Kendall trend test evaluated the statistical significance of Z-score trajectories across 15 time steps to categorize spots as “persistent” or “emerging” at p < 0.05 [10][16]. Malaria burden was incorporated using TPR data sourced from official MOH weekly reports [4][5]. Districts were classified as “High-Burden” if TPR consistently exceeded 50%. This measure was linked to inventory data to calculate a Demand-Weighted Mismatch (DWM) score for each district (i), operationalized as:

A positive DWM score identifies critical priority zones where transmission pressure outpaces available inventory [10][20].

To evaluate the long-term stability of the supply chain, a continuous time-series analysis was performed across the 260-week study period (2020-2024). As illustrated in Figure 5, the national stock levels exhibit a cyclical ‘Pendulum Effect.’ Rather than a stable equilibrium, the data show high-frequency oscillations where acute understocking during transmission peaks is followed by sharp, aggregate over-corrections. By plotting the full weekly dataset, we confirm that the 79% overstock observed in late 2024 was not an isolated event but the culmination of a multi-year pattern of systemic volatility.

The early study period was defined by a paradoxical surplus of treatment and a deficit of diagnostics. In August 2020, 24% of districts were overstocked with ACTs (holding > 52 weeks of supply), yet 27% of these same districts faced critical understocking of RDTs (<8 weeks). This indicates a failure in “bundle-based” logistics, where patients could be treated but not accurately diagnosed. By January 2021, ACT overstocking peaked at 31.6% of districts, signaling a significant risk of pharmaceutical expiry and resource tie-up in low-burden areas.

Figure 6.Chronic understocking hotspots in high burden Ugandan Districts (2024).

The chart is sorted by TPR to emphasize the areas with the highest malaria burden relative to their supply challengesFigure 6.

Butebo shows the highest peak TPR at 66.3%, driven primarily by an RDT gap (Diagnostic Deficit).

Kole (Lango) follows with 65.1%, attributed to an ACT gap (Demand-Driven Depletion).

Districts like Maracha and Alebtong face a combined ACTs & RDTs gap, leading to Total Supply Failure and Seasonal Vulnerability, respectively.

The 2022 nationwide malaria upsurge exposed the fragility of the “pull” system. During this period, the number of districts experiencing frequent ACT stockouts increased exponentially from 5 in 2017 to 85 in 2022. RDT understocking reached its highest level in the study period, with 39% of districts reporting < 8 weeks of supply. Even during the 2022 crisis, a “distribution ghost” remained: while 23% of districts were understocked, 49% of districts simultaneously held overstocks of ACTs. This proves that the crisis was not a national shortage of medicine, but a geographic maldistribution that manual logistics could not rectify.

The use of GIS mapping identified distinct “Hotspots of Inequality.” In the Northern and Eastern regions, districts such as Maracha, Kole, and Serere consistently operated in “Emergency Mode.” Maracha documented 0 weeks of stock during peak transmission, representing a total service delivery collapse. In contrast, districts in the Southwestern highlands, such as Kabale and Rubirizi, maintained over 20 months of ACT supply Figure 7.

Figure 7.Overstocked commodity hotsports (Low Risk Districts).

Red (Expiry & Wastage): Kabale faces the most critical risk with over 20 months of stock on hand. Orange (Low TPR Inefficiency): Kisoro has a significant surplus (>18 months) that does not align with its local malaria positivity rates. Green & Blue (Operational Misalignment): Districts like Sheema and Wakiso have surpluses driven by routine supply issues or population density mismatches.

A critical analytical divergence exists between Acute Severity and Statistical Persistence. While the descriptive data identifies Northern and Eastern districts (e.g., Maracha, Kole, and Napak) as having the highest transmission burden, these areas frequently fall into the “No Pattern Detected” category within the EHSA framework. This classification is indicative of logistical volatility rather than stability. These regions experience “logistical chaos”—extreme oscillations between total service delivery collapse (zero stock) and emergency bulk replenishment. This high-variance signal lacks the linearity required for trend detection, effectively masking the severity of the crisis in a standard temporal model.

In contrast, the Persistent Hot Spots identified in the Central region (e.g., Kamuli, Kayunga, and Mukono) represent a different systemic failure. These locations exhibit a stable, chronic understocking trend that reaches statistical significance (p < 0.05). This suggests a structural mismatch in the “pull” system within peri-urban settings; routine allocations consistently fail to account for high patient volumes, yet these steady deficits rarely reach the catastrophic thresholds that trigger the emergency “shock” resupplies observed in the Northern “Red Zones in Figure 8.

Figure 8. Map of Uganda showing emerging hot spot analysis for untreated cases due to days out of stock.

Technical Analysis Report: Space-Time Commodity Imbalances

The findings confirm a high degree of spatial-temporal volatility within the national supply chain. Of the 133 locations analyzed, 68 (51%) exhibited statistically significant trends. The identification of Persistent Hot Spots combined with a high frequency of Sporadic fluctuations underscores a dual systemic challenge: entrenched understocking in specific high-burden clusters and inconsistent supply reliability across the broader network.

The analytical model was constructed using 1-month temporal intervals to provide high-fidelity tracking of stock shifts across 15 distinct time steps, covering the critical transition period through the end of 2025. By evaluating each location against its eight nearest neighbors over a 2-month temporal window, the methodology employed a neighborhood analysis that ensured identified “Hot Spots” reflected significant regional geographic clusters rather than isolated anomalies

1) Institutionalize an AI-GIS-Driven National Redistribution Policy

The Ministry of Health should institutionalize a national AI-GIS-enabled redistribution framework that integrates real-time inventory surveillance, malaria burden indicators, and predictive analytics to guide dynamic redistribution of malaria commodities across districts. Such a policy would allow automated identification of understocked “Hot Spots” and overstocked “Cold Spots,” enabling rapid lateral transfers before facilities reach critical shortages. By shifting from reactive procurement to predictive redistribution, Uganda can reduce geographic inequities, minimize expiries, and improve supply chain resilience during seasonal malaria surges.

2) Enforce Mandatory Diagnostic-Treatment Commodity Bundling

The Ministry of Health should adopt a mandatory commodity bundling policy requiring synchronized distribution of ACTs and RDTs across all levels of the health system. This policy would address the “diagnostic-treatment contradiction” identified in the study, where treatment commodities remained available while diagnostic supplies were critically understocked. Enforcing electronic linkage between ACT and RDT dispatches would improve rational malaria case management, reduce presumptive treatment, and strengthen accountability within the national malaria supply chain.

The four Persistent Hot Spots identified in this study represent chronic failures where routine resupply has been ineffective for over 12 months. An immediate multi-disciplinary audit of these zones is required to assess health center storage capacity, power stability for digital reporting, and the accuracy of facility-level data entry. This ensures that underlying barriers to resilience are addressed beyond mere inventory replenishment.

The proposed framework

This framework demonstrates how an integrated AI-GIS system enhances supply chain resilience by transitioning from traditional reactive distribution to a dynamic redistribution model driven by real-time demand.

Phase 1: Diagnostic Baseline. The framework first identifies core systemic vulnerabilities: inventory instability resulting from the bullwhip effect in forecasting, and critical “bundle mismatches” between diagnostic (RDT) and treatment (ACT) commodities. These inefficiencies historically manifest as simultaneous stockouts and overstocking across varied geographic clusters.

Phase 2: Operational Integration. The framework then synthesizes GIS spatial mapping, real-time inventory metrics, and epidemiological trends into a centralized AI-GIS Redistribution Hub. By utilizing Emerging Hot Spot Analysis (EHSA), the system distinguishes between chronic surpluses (Cold Spots) and acute shortages (Hot Spots), facilitating dynamic lateral transfers to optimize commodity alignment without requiring national-level resupply.

The proposed logic for this dynamic redistribution is visualized in the AI-GIS Framework (see Figure 9).

Figure 9. Streamlined supply chain framework: Reactive to dynamic AI-GIS control.

While this study provides high-resolution insights, several limitations exist. The analysis relies on secondary surveillance data from the MOH Knowledge Management Portal, subject to district-level reporting accuracy [4][5]. Second, although the 133 districts met a high reporting threshold, the absence of granular facility-level data and stock buffers from the private health sector means the findings primarily reflect public-sector logistics. Furthermore, EHSA does not account for non-spatial logistical shocks such as infrastructure failure or fuel price volatility [2][6].