Triple A: How Analytics, AI, and Algorithms Are Improving Inventory Management in Healthcare

Abstract

Background: Healthcare inventory management is critical for ensuring timely access to supplies and reducing stockouts. As supply chains grow more complex, algorithms, AI, and analytics techniques have emerged as tools for forecasting, tracking, classification, and procurement. Yet empirical validation across diverse contexts remains inadequate, and existing reviews treat these approaches as separate streams rather than an integrated system. Methods: To evaluate these capabilities, a systematic review of 64 peer-reviewed articles published between 2011 and 2025 was conducted using a descriptive and content analysis approach on the use of Triple A (Analytics, AI, and Algorithms) techniques in inventory frameworks across various healthcare contexts, such as hospitals, pharmaceutical supply chains, and humanitarian supply chains. Results: Integrating multiple Triple A approaches consistently outperforms single-method strategies, particularly with RFID and IoT tools. Key findings often overlooked are: emergency procurement and classification, which remain neglected despite the highest patient safety stakes, and key procurement drivers—organizational conditions, supplier reliability, and team capacity. Data quality, interoperability, and cybersecurity further constrain generalizability. Conclusions: Bridging these gaps requires integrated Triple A approaches rather than single methods. Phased implementation, cloud-based platforms, and privacy-by-design offer practical pathways for building resilience under real-world constraints.

1. Introduction

Healthcare inventory management, involving overseeing the purchase, storage, and distribution of medical supplies, has evolved significantly, transitioning from manual record-keeping and stock counting to technology-driven approaches that enhance accuracy, efficiency, and cost control. Earlier reliance on these manual methods often resulted in errors, inefficiencies, and higher costs, while undermining care quality. Poor evaluation of inventory systems has also been shown to affect the functionality of healthcare systems [1,2]. Neve and Schmidt [3] illustrated this in their study, using a simulation-based analysis of point-of-use (Just-in-Time) inventory management in US hospitals, particularly those operating stockless models. Their findings showed that inaccurate tracking of usage data can lead to persistent overstocking and frequent stockouts, disrupting supply availability and inflating operational costs. Commonly adopted corrective strategies, such as increasing safety stock or manual adjustments, often fail to address the root cause, highlighting how unreliable inventory data can erode both logistical efficiency and financial performance. As healthcare services expand, supply chains become more complex, making it increasingly challenging to ensure the accurate and timely availability of vital medical resources [4].

To address these challenges, organizations have increasingly turned to computational approaches spanning three broad categories: classical and advanced Algorithms, Artificial Intelligence (AI) and machine learning, and data Analytics. Despite this progress, many providers continue to hold excessive safety stock due to the underutilization of automated tools; probabilistic hybrid models have been recommended as a more effective alternative [5]. Lean principles have similarly been proposed to support waste reduction and continuous improvement, helping providers maintain timely access to supplies and improve patient satisfaction [2]. Recent advances in artificial intelligence, sophisticated algorithms, and predictive analytics have further transformed inventory practice, enabling more accurate demand forecasting, improved turnover, and more responsive supply operations [6]. These technologies generate actionable insights from large datasets, reducing costs while strengthening both operational performance and patient care delivery.

Although each of these approaches has been reviewed in isolation, their interdependence, and the system-level resilience that emerges from their integration, have not been sufficiently examined. This review introduces the Triple A framework as an organizing lens to synthesize how Algorithms, AI, and Analytics have been applied across the four core functions of healthcare inventory management: demand forecasting, tracking and control, classification, and procurement. By examining the individual and combined effects of these innovations, this study identifies gaps in the literature that could guide further improvements in agility, adaptability, and alignment within healthcare inventory management systems through data-driven strategies that enhance supply chain responsiveness, patient satisfaction, and health outcomes.

Conceptual Framework: The Triple A Approach

A central challenge in synthesizing this body of literature is definitional overlap: the terms algorithms, AI, and analytics are frequently used interchangeably or inconsistently across studies. This review adopts a unified Triple A framework that treats each component as analytically distinct while recognizing that their interaction, rather than any single method, produces resilient inventory systems.

Table 1. Triple A Framework: Definitions, Examples, and Distinguishing Features.

Category	Definition	Examples in Healthcare Inventory	Distinguishing Features
Algorithms	Step-by-step computational procedures with deterministic or stochastic logic applied to optimize inventory decisions	EOQ, MILP, genetic algorithms, stochastic programming, robust optimization	Rule-based; operate under defined constraints; typically, do not adapt based on new data without explicit re-optimization
Artificial Intelligence (AI)	Computational systems that learn patterns from data to make predictions, classifications, or autonomous decisions	Deep reinforcement learning, LSTM, Faster R-CNN, random forest, multi-agent RL	Data-driven; adaptive; capable of handling non-linear, high-dimensional inputs
Analytics	Systematic examination of data to extract insights, forecast outcomes, or support decisions across descriptive, predictive, and prescriptive levels	ABC-VED analysis, demand forecasting, regression, simulation, Lean Six Sigma, RFID-enabled real-time tracking	Insight-oriented; spans descriptive to prescriptive; often combined with algorithmic or AI methods
Triple A Integration	Deployment of two or more Triple A components within a unified inventory management function	IoT + neural network for tracking; stochastic programming + simulation for blood management; RL + predictive analytics for replenishment	Hybrid approaches often outperform single-method solutions; integration is the source of resilience

provides the operational definitions and distinguishing features used throughout this review.

Algorithms encompass step-by-step computational procedures, from classical mathematical models such as the Economic Order Quantity (EOQ) to advanced stochastic programs and robust optimization frameworks, that optimize inventory decisions under defined constraints. They are rule-based and may or may not learn from data. Artificial Intelligence (AI) refers to computational systems that learn patterns from data to make predictions, classifications, or autonomous decisions; representative methods include deep reinforcement learning, long short-term memory (LSTM) networks, and multi-agent reinforcement learning. Analytics denotes the systematic examination of data to extract insights, forecast outcomes, or support decisions; it spans descriptive analysis (what happened), diagnostic analysis (why it happened), predictive modeling (what will happen), and prescriptive modeling (what should be done). This taxonomy is not rigid, many deployed systems integrate all three components, but it provides the analytical scaffolding needed to identify where research investment is concentrated, where it is absent, and where hybrid integration yields the strongest outcomes.

2. Background and Theoretical Context

2.1. Algorithmic Approaches in Healthcare Inventory Management

Managing inventories in healthcare systems using algorithms has evolved over the years. As the demand for adequate healthcare rises, so does the complexity of healthcare inventory systems, necessitating the constant development of various methodologies to optimize inventory levels while minimizing costs and ensuring efficient delivery service. This evolution follows a clear progression, from simplified mathematical formulations to classical optimization models, and more recently toward AI-driven approaches capable of handling the dynamic, uncertain nature of real-world healthcare supply chains.

In earlier times, before the formulation of algorithms, mathematical formulations, including EOQ, were developed to optimize inventory systems. This model is often employed to target specific inventory management problems, such as the order quantity problem, but assumes simplified conditions such as constant demand and fixed lead times. While useful as a baseline, these assumptions render classical models inadequate for the dynamic demand patterns characteristic of healthcare environments. Some studies have therefore considered expanding basic models, such as EOQ, through integration with more complex algorithms to better fit real-life healthcare situations [7]. Most EOQ-based and classical models perform well under controlled assumptions but fail under dynamic demand, variable lead times, or multi-level supply chain structures, conditions that are the norm rather than the exception in healthcare settings.

Building on this foundation, in developing more efficient and optimized inventory systems, advanced algorithms have been considered across a range of healthcare inventory problems. For instance, an inventory system, such as vendor-managed inventory (VMI), simulated using a Mixed-Integer Linear Programming (MILP) model, was found to reduce stockouts while optimizing replenishment schedules that account for economic, environmental, and sustainability impacts [8]. Similarly, Rastegar et al. [9] applied a multi-objective MILP algorithm to a vaccine inventory and distribution supply chain to minimize cost and improve storage and distribution efficiency, a finding further validated by Uthayakumar and Priyan [10], who demonstrated that a mixed-integer nonlinear programming model enhances the storage and distribution of pharmaceutical supplies. Beyond linear programming, Yadav et al. [11] demonstrated the efficacy of genetic algorithms in blood inventory management by reducing cost and improving supply reliability, while Forcina et al. [12] developed a mathematical model to determine optimal target stocking levels, minimizing total inventory costs while meeting service-level constraints. Despite this progress, most of these models operate under static or simplified supply chain structures, typically two-level vendor–buyer relationships or single-facility scenarios, and do not inherently account for the uncertainty and variability in demand, lead times, or product perishability that characterize real-world healthcare inventory management.

To address this limitation, adding a stochastic layer to these optimization models is essential for accurately incorporating the uncertainty and variability inherent in real-world healthcare inventory systems, particularly regarding demand, lead times, and perishability. Studies demonstrate that stochastic programming and simulation methods, such as two-stage stochastic mixed-integer programming and Monte Carlo-based simulation, have been successfully applied to handle uncertainties in demand, shelf-life, and supply disruptions for perishable supplies like blood and pharmaceuticals [13,14,15,16,17]. However, the computational complexity and difficulty of implementing in real-world scenarios make the adoption of stochastic models challenging for many healthcare inventory systems.

Shifting from a stochastic inventory management approach, AI and machine learning (ML) models have been developed to overcome the limitations of previous approaches. Unlike their predecessors, AI/ML models do not rely on fixed assumptions about demand distributions or supply chain structure, making them inherently more adaptable to dynamic, real-world conditions. For instance, Deep Reinforcement Learning (DRL) can adapt to dynamic scenarios and varying demands when applied with a Proximal Policy Optimization approach to optimize inventory systems. The algorithm can continue learning from the current state of inventory systems to automate processes, reducing waste (cost reduction) and stockouts [18,19]. Beyond optimization, AI/ML algorithms have also been applied to pharmaceutical inventory management to improve stock keeping and drug classification. Pan et al. [20] showed that semi-supervised learning ML algorithms, such as Random Forest Classifier and Support Vector Machine, were used to identify and classify drugs for distribution and adequate record keeping. However, these applications tend to tackle narrow, specific tasks rather than healthcare inventory management as a whole. The integration of AI/ML into frameworks that can jointly manage demand forecasting, replenishment, and distribution within healthcare settings remains an underdeveloped area.

2.2. AI and Analytics in Healthcare Inventory Management

In recent times, the incorporation of AI and Analytics into inventory management has revolutionized the healthcare supply chain. These tools enhance healthcare inventory management by infusing accuracy, adaptability, and intelligence into supply chain operations, addressing long-standing challenges such as fluctuating patient demand, strict regulatory oversight, and limited budget allocations. AI leverages machine learning algorithms to forecast medical supply needs with greater precision, while analytics provides insights into consumption patterns and supply chain dynamics. The combination of these technologies reduces waste, enhances decision-making, and ensures that critical supplies are available when needed. For instance, predictive analytics informed by AI can anticipate demand surges for essential products, such as vaccines, surgical equipment, or blood units. These insights allow healthcare providers to proactively manage stock, avoid shortages, and minimize overstocking [21].

Furthermore, integrating AI technology with enterprise resource planning (ERP) systems enables healthcare providers to enhance their operational efficiency by automating inventory management and leveraging predictive analytics [22]. This integrated approach supports more agile and well-informed decisions when managing inventory, making it possible for healthcare organizations to respond efficiently to fluctuating demand while still upholding high standards of patient care.

2.2.1. Impacts of AI on Healthcare Inventory Management

The applications of AI in healthcare inventory management leverage machine learning, computer vision, and predictive analytics to address critical supply chain challenges. These technologies enable more accurate demand forecasting for medical supplies and equipment, preventing stockouts and overstocking. Beyond prediction, AI facilitates the integration of data across various departments for optimum operations, mitigating the risk of data siloing [23]. This enhances decision-making, ensuring adequate supply levels, thereby improves patient health outcomes.

Research showed that the AI-based approach enhances inventory management in healthcare by improving item identification, reducing errors, and streamlining operational workflows for more efficient restocking and tracking. In this regard, Riaz et al. [24] proposed a multi-scale attention network using a deep learning framework with feature refinement for medical item classification, achieving a high accuracy of 95% and a high mean average precision on a diverse healthcare inventory dataset that outperforms conventional models. A similar approach has been used in pharmaceutical inventory management to identify and classify medications in warehouses. Tavakoli et al. [25] in their study demonstrated the efficacy of an AI model, the Faster Region-based Convolutional Neural Network (Faster R-CNN), for identifying and counting pharmaceutical products to improve inventory accuracy, reduce human error, and enhance overall pharmacy warehouse management efficiency. Similarly, another study by Kalusivalingam et al. [23] demonstrated that integrating machine learning algorithms, such as long short-term memory (LSTM) networks, together with predictive analytics, is effective in enhancing inventory management by providing accurate demand forecasts, reducing operational costs, and improving the overall supply chain.

Despite these advancements, the use of AI in inventory management faces significant challenges. Implementing AI systems requires access to data for a robust model. Many organizations, including the health sector, struggle with inconsistent and fragmented datasets, resulting in less credible, biased, and unreliable AI models. Developing and implementing AI systems also requires substantial computational and financial resources, making it difficult for smaller health establishments to adopt. In addition, real-world inventory systems are dynamic and subject to frequent disruptions; however, most AI models trained on historical data may struggle to adapt to unpredictable changes or new scenarios. Integration with existing management systems can also be challenging due to compatibility issues, legacy infrastructure, and the adjustments of workflow. This is further complicated by privacy and ethical concerns, especially when managing sensitive operational or customer data. These challenges must be addressed to effectively harness the full potential of AI-powered inventory systems.

2.2.2. How Analytics Improves Inventory Management in Healthcare

Analytics, ranging from descriptive data analysis to advanced predictive and prescriptive modeling, have emerged as a transformative tool that enhances responsiveness, optimizes resource utilization, and informs real-time decision-making. Various studies and healthcare management systems have examined and incorporated these analytical tools and techniques to organize health-related data, including inventories of healthcare supplies to monitor availability, usage patterns, demand forecasting, and real-time stock levels for efficient resource allocation and waste reduction. Knowles et al. [26] used data analytics to optimize vascular surgery instrument trays, demonstrating that quantitative utilization analysis, statistical modeling, and cost–benefit analysis can yield substantial cost and efficiency savings. Similarly, Saha and Ray [27] used a mathematical and simulation-based model that integrates stochastic demand forecasting, regression analysis, and optimization techniques to demonstrate the dependability of analytics in inventory management and cost reduction. Analytics have also been applied in healthcare inventory management as it relates to the classification of order items in stock keeping. Kumar & Chakravarty [28] described this using the ABC (Always, Better Control)-VED (Vital, Essential, Desirable) strategy.

Predictive analytics drive inventory optimization, especially in the pharmaceutical industry, where accurate forecasting limits stockouts, reduces costs, and improves patient satisfaction. The combination of analytical tools can increase the precision of forecasting models, as demonstrated by Merkuryeva et al. [29], who combined a time-series forecasting model (autoregressive integrated moving average and exponential smoothing) with a machine learning technique (neural network) to forecast medication demand, thereby reducing stockouts, enhancing resilience to fluctuating demand, and improving overall supply chain efficiency. Likewise, O’Mahony et al. [30] demonstrated that techniques such as Lean Six Sigma, combined with analytics, aided in identifying inefficiencies in inventory frameworks to enhance clinical process improvement. In addition, Al Khatib et al. [4] highlighted that cloud-based analytics integrated with radio frequency identification (RFID) technology provides transparency in real time, enabling healthcare facilities to track and verify inventory conditions. This is especially valuable for products requiring careful handling, such as biologics or vaccines.

The use of analytics alone in healthcare inventory management may not be sufficient to drive the positive change necessary for optimum improvement. Its combination with other techniques enhances decision-making accuracy, improves responsiveness, and creates a more resilient inventory system [27]. Such hybrid approaches strengthen visibility across the supply chain and support sustainable, data-driven inventory management practices in healthcare organizations.

3. Materials and Methods

This review was conducted to rigorously analyze how AI, analytics, and algorithms, collectively called Triple A approaches, have been applied in healthcare inventory management. It was designed to synthesize findings into a coherent understanding of trends, limitations, and practical implications. To ensure transparency and rigor, the review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses PRISMA 2020 guidelines (Supplementary Materials), as shown in Figure 1 (while exact counts for intermediate screening stages were not retained, the final articles included for this study met all inclusion criteria). To better guide the processes undertaken in this study, the research questions below were developed. These questions informed the selection and synthesis of studies and facilitated the identification of patterns, theoretical positions, and practical guidance.

• How have Triple A approaches in healthcare inventory management evolved, and what patterns emerge when examined as an integrated system?
• Across which inventory functions and healthcare contexts are Triple A approaches most and least developed, and where do the most consequential gaps exist?
• What are the key implementation barriers to Triple A adoption, and what strategies and future directions can bridge the gap between demonstrated capability and real-world application?

3.1. Literature Sources

To better understand the dynamics and trends of technological innovation in healthcare inventory management, relevant articles were sourced from various databases. Research publications were retrieved from PubMed, ScienceDirect, Google Scholar search engine, Elsevier, and the Bielefeld Academic Search Engine (BASE) for peer-reviewed articles. The selected studies covered evidence published between 2010 and 2025 to capture the period during which AI and advanced analytical tools began to meaningfully enter healthcare inventory management practice, while also including foundational classical modeling studies that provide important comparative context.

Since the study aimed to synthesize technological trends and methodological approaches rather than evaluate intervention effectiveness, a formal risk-of-bias assessment was not conducted.

3.2. Keywords Identification, Inclusion, and Exclusion Criteria

In identifying suitable research articles focused on the application of AI, analytics, and Algorithms in healthcare inventory management, each database was searched using combinations of keywords like “AI “AND” Healthcare Inventory management”, “AI” “OR” “Predictive Analytics”, “Algorithms “AND” Healthcare Inventory Optimization”, “Analytics “AND” Healthcare Inventory Management”, “AI” “OR” “Machine Learning” “Inventory Optimization” “OR” “Inventory Control”. The keyword search focused on healthcare inventory management rather than on general industrial inventory and supply chain systems, as the main goal of this research is to review studies that examine technological advancements in healthcare inventory management. However, given the close operational relationship between supply chain and inventory management, studies addressing healthcare supply chains were included when their findings provided direct insights into inventory management methods.

Articles retrieved from the database searches were screened in two stages, as illustrated in Figure 1. In the first stage, the titles, abstracts, and keywords were reviewed to remove duplicates and clearly irrelevant studies. In the second stage, full texts were reviewed for all remaining articles to confirm eligibility against the predefined inclusion and exclusion criteria. To be included, studies had to be published between 2010 and 2025 in peer-reviewed journals or conference proceedings, written in English, and address the application of at least one Triple A approach in a healthcare inventory management context with sufficient methodological detail to allow meaningful assessment. Studies were excluded if they focused on non-healthcare inventory contexts, addressed supply chain management without direct relevance to inventory functions, or were reviewed articles and opinion pieces without original empirical or methodological content.

To further refine our focus, we conducted a keyword co-occurrence analysis following the approach described by Albayrak Ünal et al. [31]. This analysis helped identify studies specifically addressing the application of analytics, AI, and algorithms in healthcare inventory management.

3.3. Literature Selection

After refining the search in the databases, each article was carefully reviewed by examining the title, abstract, keywords, and the full text, focusing on the application of AI, analytics, and algorithms in healthcare inventory management. Based on relevance, 64 articles were identified and included in this study. These 64 articles were selected for their direct relevance to the study’s aim, focusing solely on healthcare inventory management. The remaining articles were excluded from the final review because they did not focus on healthcare applications. Instead, these studies addressed general inventory management, supply chain operations, or the application of these technologies in non-healthcare industries.

3.4. Synthesis Framework

Each included study was classified according to three dimensions: the Triple A approach applied (as defined in Table 1), the inventory management function addressed (demand forecasting, tracking and control, classification, procurement), and the healthcare context (pharmaceutical, hospital, emergency/humanitarian, general healthcare). These dimensions were identified from iterative reading of the included studies rather than defined in advance, producing a grounded taxonomy. The studies were then analyzed to identify patterns of concentration and consequential gaps across the literature.

4. Results and Discussion

A descriptive analysis, including categorizing the studies by year of publication and identifying the most frequently used words, is discussed in this section. This section also includes the content analysis of the articles used for this study. The analysis considers the type of models and strategies utilized in healthcare inventory management, the inventory functions they address, and the key factors that could disrupt the effectiveness of data-driven approaches in healthcare inventory systems.

4.1. Descriptive Analysis of the Articles

This descriptive analysis was based on the selected 64 articles, examining publication trends, major themes, and key areas of focus. Although the search was for studies published between 2010–2025, the studies found and included in this review were published between 2011 and 2025, with the majority between 2021 and 2025 (Figure 2). This distribution reflects a dynamic and rapidly evolving body of research on healthcare inventory management systems. Early studies focused on the use of inventory management strategies and basic optimization models for better control of hospital and pharmaceutical supplies [10,28,32,33,34]. They emphasized the development of inventory management strategies that could limit stockouts and reduce costs, without incorporating the constantly changing nature of demand. During this period, studies focused on classic inventory control models, point-of-use strategies, and other optimization strategies that shaped early approaches to decentralized healthcare management systems. While these approaches established important foundations, their primary orientation was toward cost efficiency and operational stability rather than adaptability under disruption.

The rise in technological advancement led to the expansion in methodological approaches such as stochastic programming models, big data forecasting, as well as simulation-based optimization between the years 2017 to 2020 [11,13,15,17]. These studies incorporated dynamic demand patterns, marking a transition from traditional classical quantitative models toward hybrid approaches that accounted for demand uncertainty and the perishability of medical supplies. This period is significant not only for its methodological diversification but because it introduced the first consistent evidence that combining approaches, pairing stochastic modeling with simulation, for instance, produced stronger outcomes than relying on any single method alone.

From the year 2021, marked by the post-COVID-19 era, there was a significant shift in technological advancement to boost global supply chain systems. Most studies from this time onward were geared towards the application of advanced analytics and algorithms, such as AI-driven and machine learning-enabled approaches, as well as blockchain applications for a more data-driven, intelligent, predictive, adaptive, and resilient healthcare inventory management systems [9,25,35,36]. Within this period, from 2024, some studies also considered data-driven approaches in optimizing medical inventory supplies for humanitarian logistics, such as emergency and disaster responses. For instance, Che et al. [37] developed a two-stage distributionally robust framework for pre-positioning emergency inventory under uncertain demand. However, a significant number of the studies within the 2023–2025 period further expanded the application of AI/ML for predictive analytics and inventory optimization, including healthcare supply chain management [19,21,24,38,39].

In relation to the themes and primary areas of focus, as shown in Figure 3, out of the 64 articles reviewed, 39% of the articles (25) focused on the optimization of pharmaceutical inventory and supply chain systems; 36% were on hospital inventory and operations management and supply chain (23); 10% articles (6) on healthcare issues related to emergency and disaster logistics & humanitarian supply chains; the other articles (10), which is 16%, were centered on general topics in healthcare-related inventory management. While pharmaceutical and hospital contexts together account for 75% of all reviewed studies, emergency and humanitarian settings, where inventory failures can directly cost lives, account for only 10% of the evidence base. This is not simply a quantitative imbalance; it points to a deeper assumption in the field that Triple A approaches are primarily tools for efficiency improvement in stable settings rather than resilience instruments applicable under crisis conditions. Similarly, classification receives the least research attention across all contexts and is absent from emergency and humanitarian studies, despite the critical importance of rapid supply prioritization during crises. These distributional patterns suggest that the field’s methodological progress has been concentrated in areas of relatively lower practical urgency, leaving the contexts and functions where need is greatest and most underserved.

The co-occurrence analysis, as shown in Figure 4 and

Table 2. Keyword Clusters.

Color Code *	Clusters	Keywords	Count
	1	Demand forecasting, predictive analytics, machine learning, logistics, AI in supply chain, deep reinforcement learning, supply chain optimization	7
	2	Inventory management, supply chain, big data analytics, pharmaceutical supply chain, two-stage stochastic programming	5
	3	Healthcare, inventory, quality assurance, real-time monitoring, supply chain management	5
	4	Hospital supply chain, internal logistics, inventory control, operations research	4
	5	Artificial intelligence, healthcare supply chain	2

* Clusters are coded based on the color of each cluster on the co-occurrence map in Figure 4.

, reveals that the majority of key areas in healthcare inventory management revolve around the use of advanced computational techniques such as big data analytics, AI/ML, and stochastic programming. The focus on pharmaceutical inventory systems considered the application of these tools in areas like drug classification, distribution, and perishable product handling [25,33,40]. Studies like those of Kalusivalingam et al. [18], Saha & Rathore [19], and Perez et al. [41] used advanced algorithms for inventory optimization, demand forecasting, and general resilience supply chain operations.

Within hospital settings, studies focused on using analytics to reduce stockouts while enhancing the cost savings of specific essential medical items and supplies. Knowles et al. [26] described the effectiveness of using data analytics to enhance the cost savings of vascular surgery instrument trays. Dillon et al. [15] explored the use of stochastic models to enhance blood supply inventory management, while Yadav et al. [11] employed a genetic algorithm. Earlier studies on hospital inventory management have discussed the use of inventory models and strategies for optimizing and controlling healthcare inventory systems. Bijvank & Vis [32] discussed the use of the point-of-use strategy, Uthayakumar and Priyan [10] utilized a mathematical programming model for both pharmaceutical and hospital inventory optimization.

Outside of the key areas in healthcare like hospitals and pharmaceuticals, a small number of studies considered the application of analytics for managing medical supplies in emergency and disaster scenarios [9,37,42,43,44,45]. These studies leverage analytics and algorithms for predictive analytics in demand planning, stock management, and logistics optimization. In addition, literature discussing the application of models, analytics, and advanced algorithms in general healthcare systems was analyzed. The research applied innovative techniques to streamline inventory operations and increase supply chain adaptability [23,24,46,47]. Across all four thematic areas, three analytical patterns emerge consistently. First, the effectiveness of any Triple A approach depends heavily on the healthcare setting, data availability, and supply chain structure, making any single-method prescription not methodologically sound. Second, hybrid approaches consistently yield stronger outcomes than single-method approaches, particularly in managing perishability, demand uncertainty, and stockout risk, yet procurement and emergency contexts continue to rely predominantly on single methods, suggesting that the evidence for hybrid superiority has not yet translated into research practice. Third, implementation barriers, especially data quality, system integration, and resource constraints, are reported across virtually every thematic area but are rarely accompanied by practical guidance, leaving a gap between what the research demonstrates is possible and what healthcare organizations can actually implement. The field’s challenge, as this pattern reveals, is not a shortage of sophisticated methods but a shortage of integrated, context-sensitive, and practically grounded approaches.

4.2. Content Review

The subsections below offer a detailed overview of how Triple A approaches are applied across the key inventory management functions identified in the literature, along with the limitations of these approaches.

4.2.1. Healthcare Inventory Models and Strategies

Effective inventory management in healthcare is critical due to the unique challenges posed by perishable products, demand fluctuations, and the necessity for high service levels. Consequently, various strategies and models have been employed to manage inventories (Table 3). These models and strategies have been utilized in diverse contexts and settings, such as pharmaceuticals, hospitals, and humanitarian supply chains. Discussed below is how these models and strategies are applied across different aspects of healthcare inventory systems.

Demand Forecasting

Demand forecasting plays a vital role in healthcare inventory management, particularly in pharmaceutical and hospital supply chains. Accurately predicting demand fluctuations and consumption patterns is essential for reducing costs, minimizing excess inventory, and preventing stockouts, thereby ensuring reliable service. It strengthens planning and decision-making processes. Analytical approaches to demand forecasting have evolved over time. Early research considered regression and stochastic models to estimate needs for perishable products, medical supplies, and critical equipment. Forecasting demand based on patient condition with a two-stage stochastic approach, modeled using a Markovian Decision Process (step-by-step decision process based on current situation rather than past events), can determine optimal inventory and production policies for perishable pharmaceuticals in hospital pharmacies [33]. Another study by Li et al. [13] on deteriorating medications extended this approach by optimizing order quantities and reorder points under stochastic lead times, with fixed shelf life, limited storage capacity, and minimum service levels in hospital and pharmacy settings. This creates a framework that captures real-world scenarios for pharmaceutical inventory management. However, such scenario-based models are typically tailored to specific hospital or pharmacy settings and thus have limited generalizability across diverse healthcare and pharmaceutical contexts.

Forecasting using regression relies on historical data to estimate medical supply usage and inventory levels. However, regression-based models differ in accuracy. Basic models like simple moving averages (SMA) and multiple linear regression models are likely to have higher forecasting errors compared with more advanced models, such as a symbolic regression model that uses genetic programming, especially in pharmaceutical settings [29]. This is because demand for medical supplies often follows a dynamic pattern rather than assuming a linear and predictable trend, making simple regression models less effective in capturing real fluctuations.

Advancing from basic models, the use of big data analytics and AI/ML for demand forecasting has become increasingly popular. This shift is driven by the ability of advanced algorithms to learn complex patterns in demand data and generate more reliable forecasts than traditional statistical approaches, such as the stochastic and regression models. For instance, big data and deep learning techniques can limit forecasting errors for emergency medical supplies through supply categorization and iterative training, but care should be taken to ensure the generalizability of models with limited validation across general healthcare settings, particularly when using emergency simulation data [35]. For pharmaceutical inventory systems characterized by regular demand fluctuations, AI-driven deep reinforcement learning models can effectively predict supply needs and utilization, accounting for demand dynamics and variable lead times for perishable medical items [38]. Deep reinforcement learning has also been applied to drug replenishment in hospital settings, reducing shortages by integrating real-time demand and pricing data [48]. Beyond forecasting, these tools support broader supply chain coordination, including the development and distribution of products such as vaccines [49]. However, computational demands, data integration challenges, and context-specific applicability remain barriers to wider adoption across diverse healthcare settings. Demand in emergency environments is not only unpredictable but potentially life-critical, making accurate forecasting more urgent than in any stable healthcare setting, yet the reviewed emergency studies address pre-positioning, rotation, and logistics coordination without proposing forecasting-specific solutions. Bridging this gap requires approaches designed for conditions of radical uncertainty, sparse historical data, and rapidly shifting demand compositions that bear little resemblance to the stable pharmaceutical environments where most forecasting research has been conducted.

Table 3. Healthcare inventory management models and strategies.

Strategy/Model	Triple A Category	Purpose	Limitations	Healthcare Context	Inventory Function	Key References
Economic Order Quantity (EOQ)	Mathematical Algorithm	Minimize total ordering and holding costs by identifying optimal order quantity	Assumes constant demand, fixed lead times, and stable costs; unsuitable for perishable drugs or dynamic demand	Pharmaceutical; Hospital	Procurement	[5,7]
Vendor-Managed Inventory (VMI)	Algorithm/Prescriptive Analytics	Optimize inventory levels via supplier-led replenishment; reduce stockouts and sustainability impacts	Requires strong data integration, supplier collaboration, and trust	Pharmaceutical; Hospital	Procurement; Tracking & Control	[8,50]
Point-of-Use Strategy	Descriptive/Prescriptive Analytics	Ensure immediate availability of critical items near point of care; reduce stockouts and wastage	Dependent on accurate usage tracking; performance deteriorates with poor data quality; high implementation cost	Hospital	Tracking & Control	[1,3,32]
ABC-VED Analysis	Descriptive/Diagnostic Analytics	Classify inventory by monetary consumption value (ABC) and clinical criticality (VED) to prioritize replenishment	Static classification; may not capture dynamic demand patterns or newly introduced products	Hospital; Pharmaceutical	Classification	[28]
Just-in-Time (JIT)	Predictive/Prescriptive Analytics	Minimize waste and holding costs; enable rapid demand response through lean replenishment	Vulnerable to supply disruptions; requires reliable suppliers and accurate demand signals	Hospital; Pharmaceutical	Procurement; Tracking & Control	[2]
Multi-Echelon Inventory Models	Mathematical Algorithm	Optimize inventory levels and flow across multiple supply chain layers under uncertainty	High data and computational complexity; difficult to implement across diverse healthcare networks	Hospital; Pharmaceutical	Procurement; Tracking & Control	[15,16,34,51,52]
Stochastic Programming	Mathematical Algorithm	Model and optimize inventory decisions under uncertainty in demand, supply, and lead times	High computational complexity; challenging real-world implementation; requires probability distribution assumptions	Hospital; Pharmaceutical; Emergency	Forecasting; Procurement	[13,14,15,16,17]
Robust Optimization	Mathematical Algorithm	Optimize inventory decisions under worst-case uncertainty without requiring explicit distributional assumptions	Conservative solutions may reduce efficiency under stable conditions; computationally intensive	Emergency; Humanitarian	Procurement; Forecasting	[37]
Linear Programming (MILP/MINLP)	Mathematical Algorithm	Optimize multi-objective inventory decisions, cost, service level, sustainability, under linear or nonlinear constraints	Assumes linearity or simplified relationships; may not capture real-world complexity	Pharmaceutical; Hospital	Procurement; Forecasting	[8,9,10,41]
Discrete-Event Simulation	Prescriptive Analytics	Test replenishment policy performance under various demand and supply scenarios; support complex system evaluation	Requires extensive data; may not capture all variabilities; requires model validation	Hospital; Pharmaceutical	Tracking & Control; Forecasting	[27,53,54]
Genetic Algorithm/Metaheuristic Optimization	Algorithm	Determine near-optimal reorder quantities, order points, and rotation policies for perishable and multi-product inventory	Computationally intensive for large-scale systems; solutions may not be globally optimal	Hospital; Pharmaceutical	Tracking & Control; Procurement	[11,12,55,56]
Two-Stage Stochastic/Markovian Decision Models	Algorithm/Predictive Analytics	Determine optimal inventory and production policies for perishable pharmaceuticals under patient-driven demand uncertainty	Tailored to specific settings; limited generalizability across diverse healthcare contexts	Pharmaceutical; Hospital	Forecasting; Procurement	[13,14,33]
Deep Reinforcement Learning (DRL)	AI/Machine Learning	Automate replenishment and optimize ordering decisions under dynamic demand and variable lead times for perishable items	High computational cost; requires large high-quality datasets; risk of model degradation without regular updates	Pharmaceutical; Hospital	Forecasting; Procurement	[19,38,48,49,57]
Multi-Agent Reinforcement Learning (MARL)	AI/Machine Learning	Coordinate ordering decisions across multiple hospital units by capturing demand dependencies	Complex to implement; requires real-time data sharing across units; high computational demand	Hospital	Procurement; Forecasting	[19]
LSTM/Neural Network Forecasting	AI/Predictive Analytics	Improve demand forecast accuracy for medical supplies by capturing non-linear temporal patterns	Requires large historical datasets; computational overhead; risk of overfitting on limited data	Hospital; Pharmaceutical	Forecasting	[18,29,35]
Semi-Supervised/Supervised ML Classification	AI/Machine Learning	Identify, classify, and prioritize medical supplies in pharmaceutical and hospital warehouses to improve inventory accuracy	Assumes data completeness; limited performance under fragmented or inconsistent datasets	Pharmaceutical; Hospital	Classification	[20,58,59]
Computer Vision (Faster R-CNN/Deep Learning)	AI/Machine Learning	Automate medical item identification, counting, and stock level optimization based on consumption patterns	High implementation cost; requires adequate lighting and camera infrastructure; limited scalability to legacy systems	Pharmaceutical; Hospital	Tracking & Control; Classification	[24,25,60]
Big Data Analytics & Predictive Forecasting	Analytics/AI	Reduce forecasting errors for medical and emergency supplies through large-scale data processing and pattern recognition	Generalizability concerns when trained on simulation or limited real-world data; requires robust data infrastructure	Pharmaceuticals; Hospital; Emergency	Forecasting	[21,35,38,40,46]
IoT-Based Asset Tracking Systems	Machine Learning & IoT/Prescriptive Analytics	Enable real-time tracking, condition monitoring (temperature, humidity, expiry), and automated alerts for perishable supplies	High device cost; interoperability issues; cybersecurity risks; reliance on sensor accuracy	Hospital; Pharmaceutical	Tracking & Control	[4,61,62]
Blockchain-Enabled Inventory Model	Blockchain with AI/Analytics	Provide secure, transparent, and tamper-proof tracking and traceability across supply chain nodes	High implementation cost; integration challenges with existing systems; scalability limits; data governance concerns	Pharmaceutical; Hospital	Tracking & Control	[36]
Hybrid IoT + AI (Neural Network/CNN)	AI/Machine Learning & IoT	Combine real-time sensor data with AI models for automated inventory control, replenishment, and anomaly detection	Complex integration; high infrastructure cost; real-world validation at scale remains limited	Hospital	Tracking & Control	[62]
Lean Six Sigma with Analytics	Prescriptive Analytics	Identify and eliminate inventory inefficiencies through structured process improvement combined with data analysis	Requires sustained organizational commitment; primarily a process tool rather than a predictive model	Hospital	Tracking & Control; Procurement	[26,30]
Group Purchasing Organization (GPO) Models	Algorithm/Prescriptive Analytics	Leverage collective buying power to reduce procurement costs, standardize ordering, and streamline contracting across member institutions	Effectiveness tied to supplier relationships and organizational trust; may limit flexibility for individual providers	Hospital	Procurement	[63,64]
Network-Flow/Multi-Echelon Procurement Optimization	Mathematical Algorithm	Determine optimal purchasing quantities from regular and outsourced suppliers across hospitals and distribution centers to reduce cost while managing perishability	Data-intensive; assumes stable supplier relationships; limited applicability in disrupted supply environments	Hospital; Pharmaceutical	Procurement	[41,52]
Emergency Supply Pre-Positioning Models	Algorithm/Analytics	Optimize pre-positioning and rotation of medical supplies for disaster and emergency response under uncertain demand	Highly context-specific; demand distributions in crises are often unknowable; limited real-world validation	Emergency; Humanitarian	Procurement; Forecasting	[37,42,43,44,45]
AI-ERP Integration	AI/Prescriptive Analytics	Automate inventory management and leverage predictive analytics within enterprise resource planning systems to improve operational efficiency	Integration complexity with legacy ERP systems; high implementation cost; change management requirements	Hospital; Pharmaceutical	Forecasting; Procurement; Tracking & Control	[6,22,39,47]
Cloud-Based Analytics Platforms	Analytics/AI	Enable scalable, accessible deployment of predictive and prescriptive analytics without on-premises infrastructure investment	Data privacy and security concerns; dependence on internet connectivity; vendor lock-in risks	General Healthcare; Pharmaceutical	Forecasting; Tracking & Control	[4,31,59]

Note: Articles were analyzed to determine the uses and impacts of the different tools and strategies on healthcare inventory management. Studies addressing multiple model types appear in more than one row. The Healthcare Context and Inventory Function columns reflect the coding framework described in Section 3.4. References in each row include all reviewed studies that applied that model type.

Table 3. Healthcare inventory management models and strategies.

Strategy/Model	Triple A Category	Purpose	Limitations	Healthcare Context	Inventory Function	Key References
Economic Order Quantity (EOQ)	Mathematical Algorithm	Minimize total ordering and holding costs by identifying optimal order quantity	Assumes constant demand, fixed lead times, and stable costs; unsuitable for perishable drugs or dynamic demand	Pharmaceutical; Hospital	Procurement	[5,7]
Vendor-Managed Inventory (VMI)	Algorithm/Prescriptive Analytics	Optimize inventory levels via supplier-led replenishment; reduce stockouts and sustainability impacts	Requires strong data integration, supplier collaboration, and trust	Pharmaceutical; Hospital	Procurement; Tracking & Control	[8,50]
Point-of-Use Strategy	Descriptive/Prescriptive Analytics	Ensure immediate availability of critical items near point of care; reduce stockouts and wastage	Dependent on accurate usage tracking; performance deteriorates with poor data quality; high implementation cost	Hospital	Tracking & Control	[1,3,32]
ABC-VED Analysis	Descriptive/Diagnostic Analytics	Classify inventory by monetary consumption value (ABC) and clinical criticality (VED) to prioritize replenishment	Static classification; may not capture dynamic demand patterns or newly introduced products	Hospital; Pharmaceutical	Classification	[28]
Just-in-Time (JIT)	Predictive/Prescriptive Analytics	Minimize waste and holding costs; enable rapid demand response through lean replenishment	Vulnerable to supply disruptions; requires reliable suppliers and accurate demand signals	Hospital; Pharmaceutical	Procurement; Tracking & Control	[2]
Multi-Echelon Inventory Models	Mathematical Algorithm	Optimize inventory levels and flow across multiple supply chain layers under uncertainty	High data and computational complexity; difficult to implement across diverse healthcare networks	Hospital; Pharmaceutical	Procurement; Tracking & Control	[15,16,34,51,52]
Stochastic Programming	Mathematical Algorithm	Model and optimize inventory decisions under uncertainty in demand, supply, and lead times	High computational complexity; challenging real-world implementation; requires probability distribution assumptions	Hospital; Pharmaceutical; Emergency	Forecasting; Procurement	[13,14,15,16,17]
Robust Optimization	Mathematical Algorithm	Optimize inventory decisions under worst-case uncertainty without requiring explicit distributional assumptions	Conservative solutions may reduce efficiency under stable conditions; computationally intensive	Emergency; Humanitarian	Procurement; Forecasting	[37]
Linear Programming (MILP/MINLP)	Mathematical Algorithm	Optimize multi-objective inventory decisions, cost, service level, sustainability, under linear or nonlinear constraints	Assumes linearity or simplified relationships; may not capture real-world complexity	Pharmaceutical; Hospital	Procurement; Forecasting	[8,9,10,41]
Discrete-Event Simulation	Prescriptive Analytics	Test replenishment policy performance under various demand and supply scenarios; support complex system evaluation	Requires extensive data; may not capture all variabilities; requires model validation	Hospital; Pharmaceutical	Tracking & Control; Forecasting	[27,53,54]
Genetic Algorithm/Metaheuristic Optimization	Algorithm	Determine near-optimal reorder quantities, order points, and rotation policies for perishable and multi-product inventory	Computationally intensive for large-scale systems; solutions may not be globally optimal	Hospital; Pharmaceutical	Tracking & Control; Procurement	[11,12,55,56]
Two-Stage Stochastic/Markovian Decision Models	Algorithm/Predictive Analytics	Determine optimal inventory and production policies for perishable pharmaceuticals under patient-driven demand uncertainty	Tailored to specific settings; limited generalizability across diverse healthcare contexts	Pharmaceutical; Hospital	Forecasting; Procurement	[13,14,33]
Deep Reinforcement Learning (DRL)	AI/Machine Learning	Automate replenishment and optimize ordering decisions under dynamic demand and variable lead times for perishable items	High computational cost; requires large high-quality datasets; risk of model degradation without regular updates	Pharmaceutical; Hospital	Forecasting; Procurement	[19,38,48,49,57]
Multi-Agent Reinforcement Learning (MARL)	AI/Machine Learning	Coordinate ordering decisions across multiple hospital units by capturing demand dependencies	Complex to implement; requires real-time data sharing across units; high computational demand	Hospital	Procurement; Forecasting	[19]
LSTM/Neural Network Forecasting	AI/Predictive Analytics	Improve demand forecast accuracy for medical supplies by capturing non-linear temporal patterns	Requires large historical datasets; computational overhead; risk of overfitting on limited data	Hospital; Pharmaceutical	Forecasting	[18,29,35]
Semi-Supervised/Supervised ML Classification	AI/Machine Learning	Identify, classify, and prioritize medical supplies in pharmaceutical and hospital warehouses to improve inventory accuracy	Assumes data completeness; limited performance under fragmented or inconsistent datasets	Pharmaceutical; Hospital	Classification	[20,58,59]
Computer Vision (Faster R-CNN/Deep Learning)	AI/Machine Learning	Automate medical item identification, counting, and stock level optimization based on consumption patterns	High implementation cost; requires adequate lighting and camera infrastructure; limited scalability to legacy systems	Pharmaceutical; Hospital	Tracking & Control; Classification	[24,25,60]
Big Data Analytics & Predictive Forecasting	Analytics/AI	Reduce forecasting errors for medical and emergency supplies through large-scale data processing and pattern recognition	Generalizability concerns when trained on simulation or limited real-world data; requires robust data infrastructure	Pharmaceuticals; Hospital; Emergency	Forecasting	[21,35,38,40,46]
IoT-Based Asset Tracking Systems	Machine Learning & IoT/Prescriptive Analytics	Enable real-time tracking, condition monitoring (temperature, humidity, expiry), and automated alerts for perishable supplies	High device cost; interoperability issues; cybersecurity risks; reliance on sensor accuracy	Hospital; Pharmaceutical	Tracking & Control	[4,61,62]
Blockchain-Enabled Inventory Model	Blockchain with AI/Analytics	Provide secure, transparent, and tamper-proof tracking and traceability across supply chain nodes	High implementation cost; integration challenges with existing systems; scalability limits; data governance concerns	Pharmaceutical; Hospital	Tracking & Control	[36]
Hybrid IoT + AI (Neural Network/CNN)	AI/Machine Learning & IoT	Combine real-time sensor data with AI models for automated inventory control, replenishment, and anomaly detection	Complex integration; high infrastructure cost; real-world validation at scale remains limited	Hospital	Tracking & Control	[62]
Lean Six Sigma with Analytics	Prescriptive Analytics	Identify and eliminate inventory inefficiencies through structured process improvement combined with data analysis	Requires sustained organizational commitment; primarily a process tool rather than a predictive model	Hospital	Tracking & Control; Procurement	[26,30]
Group Purchasing Organization (GPO) Models	Algorithm/Prescriptive Analytics	Leverage collective buying power to reduce procurement costs, standardize ordering, and streamline contracting across member institutions	Effectiveness tied to supplier relationships and organizational trust; may limit flexibility for individual providers	Hospital	Procurement	[63,64]
Network-Flow/Multi-Echelon Procurement Optimization	Mathematical Algorithm	Determine optimal purchasing quantities from regular and outsourced suppliers across hospitals and distribution centers to reduce cost while managing perishability	Data-intensive; assumes stable supplier relationships; limited applicability in disrupted supply environments	Hospital; Pharmaceutical	Procurement	[41,52]
Emergency Supply Pre-Positioning Models	Algorithm/Analytics	Optimize pre-positioning and rotation of medical supplies for disaster and emergency response under uncertain demand	Highly context-specific; demand distributions in crises are often unknowable; limited real-world validation	Emergency; Humanitarian	Procurement; Forecasting	[37,42,43,44,45]
AI-ERP Integration	AI/Prescriptive Analytics	Automate inventory management and leverage predictive analytics within enterprise resource planning systems to improve operational efficiency	Integration complexity with legacy ERP systems; high implementation cost; change management requirements	Hospital; Pharmaceutical	Forecasting; Procurement; Tracking & Control	[6,22,39,47]
Cloud-Based Analytics Platforms	Analytics/AI	Enable scalable, accessible deployment of predictive and prescriptive analytics without on-premises infrastructure investment	Data privacy and security concerns; dependence on internet connectivity; vendor lock-in risks	General Healthcare; Pharmaceutical	Forecasting; Tracking & Control	[4,31,59]

Note: Articles were analyzed to determine the uses and impacts of the different tools and strategies on healthcare inventory management. Studies addressing multiple model types appear in more than one row. The Healthcare Context and Inventory Function columns reflect the coding framework described in Section 3.4. References in each row include all reviewed studies that applied that model type.

Inventory Tracking and Control

The complexity of healthcare inventory systems requires adequate optimization to ensure proper replenishment and record keeping of medical supplies. Traditional methods, including manual counting and grouping of items, are often tedious, error-prone, and lack real-time visibility. Hence, more recently, the incorporation of advanced analytics and information and communication technologies (ICT) has been applied for real-time medical items tracking and classifying items based on their consumption and shelf life.

Generally, studies in inventory tracking and control often focus on design and evaluation of replenishment policies, classification, and real-tracking technologies for optimum stock keeping, maintaining service levels, and minimizing storage and holding costs. Early approaches include the use of point-of-use inventory control systems that improve replenishment efficiency, service level, and reduce stockouts. This technique improves the responsiveness of patient care as it ensures necessary supplies are near the point of care, especially for limited storage spaces with simple reorder policies, which are common in hospitals. However, their performance depends on accurately capturing inventory usage, since the model requires periodic manual counting of items used [32]. Hence, recent advancements have incorporated technology like RFID systems to improve precise usage tracking and mitigate record errors, boosting the effectiveness of point-of-use systems. Additionally, the performance of the point-of-use strategy can remain stable with moderate inaccuracies through adjusted base-stock levels and optimized counting policies, although performance deteriorates when data quality is poor [3]. Other early contributions to inventory tracking and control frameworks include the heuristic and metaheuristic optimization algorithms, such as genetic algorithms, determining reorder quantities, order points, and rotation policies, especially for perishable items and blood products [10,11,55,56].

In enhancing previous analytical models, the subsequent approach expanded to the use of IoT-based technologies, stochastic programming, and simulations for tracking and control. Cyber–physical systems (e.g., RFID) and IoT-based frameworks enhance real-time tracking and traceability of medical supplies, reducing human error [61]. This ICT system is efficient in handling perishable medical supplies using sensor analytics for real-time monitoring of temperature, humidity, location, and expiration dates to trigger automated alerts and replenishment via threshold-based algorithms and edge/cloud processing. However, despite their operational benefits, these systems involve high implementation costs and rely on basic algorithms that may not capture high complexities in some healthcare data, thereby reducing their scalability into legacy healthcare inventory systems.

The use of big data, computer vision, and AI-driven models (e.g., Faster R-CNN and deep learning-based computer vision) has aided the automation of replenishment, through medical item identification, counting, and optimizing stock levels based on consumption patterns [25,60]. The combination of these technologies often supports a robust, advanced technological system for inventory optimization. Zonayed et al. [62] demonstrated the effectiveness of combining IoT-based technology with neural network and convolutional AI models for healthcare inventory control. Policy-based models such as just-in-time, VMI, and related approaches complement these technological developments by reducing excess inventory and lowering holding costs, particularly when supported by adequate data sharing across supply chain partners [8,50]. The convergence of IoT sensing, AI-driven identification, and policy-based replenishment represents the current frontier of tracking and control capability in the reviewed literature.

However, a critical gap separates technical capability from operational reality. Real-world implementation studies validating these hybrid systems at scale remain scarce across the reviewed literature. Systems that perform well in controlled or pilot conditions consistently encounter resistance in routine practice due to staff training requirements, infrastructure costs, workflow disruption, and compatibility with legacy systems. This pattern suggests that the field has reached a point of diminishing returns from further technical development in tracking and control and is instead constrained by implementation readiness, the organizational capacity to deploy, integrate, and sustain advanced systems in complex healthcare environments. Implementation readiness is the most underexamined dimension in this body of research, and addressing it requires a shift in research design from technical evaluation toward longitudinal implementation studies that track adoption, adaptation, and sustained performance across real operational contexts.

Inventory Classification

Inventory classification is as important as every other healthcare optimization technique, since it aids in prioritization of medical supplies, especially in pharmaceutical and hospital warehouses, enhancing replenishment decisions, reducing waste, and ensuring the availability of critical items when needed. The ABC-VED technique, which classifies items based on monetary usage value and clinical criticality, represented the standard approach in early studies [28]. While effective at directing attention to high-priority supplies, this technique relies on static data that may not capture dynamic demand patterns or newly introduced products, potentially leading to misclassification.

The increasing demand for prescription drugs and other medical supplies has driven the expansion of the healthcare inventory system, thereby increasing the need for advanced analytical methods to optimize service levels and patient care through efficient inventory classification. As a result, big data and advanced analytics have been utilized to organize and classify medical items in hospitals and pharmaceutical inventory systems. ML models such as k-nearest neighbors (KNN), random forests, and other ML methods are employed to classify medical items in hospitals, helping reduce costs and waste [58]. Pan et al. [20] found that a semi-supervised learning model assists in identifying and classifying medical supplies within pharmaceutical inventory systems, leading to cost savings and improved inventory management while enhancing patient care. Similarly, Riaz et al. [24] demonstrated the effectiveness of AI-driven models with computer vision technology to optimize healthcare inventory systems. Interestingly, beyond medical item identification and classification, these advanced analytical techniques can also improve the classification of healthcare inventory data to enhance supply chain management. A data-encrypted, cloud-based system combined with various machine learning models was evaluated for classifying blood bank data, reducing human intervention and errors [59]. Although these advanced techniques are efficient, they are fraught with risks, especially the difficulty of integration into existing systems, high cost, and data security issues. Classification is the least studied inventory function among the studies reviewed and is absent from those addressing emergency and humanitarian contexts. There is a lack of application of Triple A approaches to supply classification in disaster or crisis scenarios, a notable absence given that several of the emergency studies [37,42,43,44,45] identified supply prioritization as a key operational challenge without proposing classification-specific solutions.

Procurement and Purchasing

In healthcare inventory management, the use of analytics and algorithms has enhanced procurement and purchasing through demand prediction, optimizing stock levels, improving order timing, and strengthening decision-making to ensure essential supplies remain available while minimizing waste and cost. These methods support hospitals, pharmacies, clinics, and other healthcare product inventory management firms in determining “what to order, how much to order, and when to place orders,” moving procurement from a reactive activity to a more proactive, data-driven process. Studies have shown that various algorithmic models have been developed and applied to optimize replenishment strategies and ordering policies for medical items across different healthcare networks. For instance, a multi-agent reinforcement learning framework can account for demand dependencies while coordinating ordering in hospitals [19]. Zwaida et al. [48] demonstrated that DRL models can be applied to automate refill decisions to prevent drug shortages. Similarly, using AI models (Q-learning and deep Q-networks) for optimizing perishable items in pharmaceutical inventory can optimize order timing and quantities, and reduce waste, thus minimizing total procurement cost [57]. Perlman and Levner [52] further showed that a network-flow optimization model for a multi-echelon healthcare inventory system can determine optimal purchasing quantities from regular and outsourced suppliers to reduce costs, while managing perishability across hospitals and distribution centers.

The use of algorithmic approaches to enhance the efficiency of procurement in healthcare systems may also be influenced by the structure and governance of purchasing processes. Group purchasing organizations (GPOs) transform hospital procurement by leveraging collective buying power. Instead of individual hospitals negotiating separately, GPOs centralize contracting, standardize ordering patterns, and negotiate pricing, thereby improving efficiency and reducing costs across member institutions [63]. Complementing this collaboration in enhancing purchasing and procurement processes, Rego et al. [64] developed a hybrid framework for integrated cooperative purchasing, demonstrating that coordinated procurement across multiple facilities can further optimize costs and streamline ordering decisions. Thus, their effectiveness is closely tied to the structure of purchasing relationships, clarity in supplier responsibilities, and the ability of procurement teams to integrate model outputs into practice. In procurement, the central challenge lies less in developing increasingly sophisticated ordering models and more in the conditions under which those models are actually used. Supplier reliability, data-sharing agreements, organizational trust, and procurement team capacity are repeatedly cited as the factors that determine whether an algorithmic or AI-driven procurement tool succeeds or fails in practice, yet these organizational factors are treated as secondary in most of the reviewed studies.

4.3. Limitations and Future Directions

This review provides valuable insights into how Triple A techniques can enhance healthcare inventory management, but several limitations must be acknowledged. The scope is limited to studies published between 2011 and 2025, which means that foundational pre-AI work that could offer additional comparative context is underrepresented. Access restrictions on certain databases may also have limited the evidence available. Most critically, most of the reviewed studies were conducted in pharmaceutical or hospital settings, while areas such as emergency and humanitarian inventory management are only sparsely covered. As a result, the generalizability of the findings across diverse healthcare systems and regions is constrained. This context specificity reflects a structural limitation in the evidence base: the absence of validated, transferable approaches to guide Triple A implementation across different healthcare settings, a gap that future research must actively address rather than simply acknowledge.

These limitations define a clear research agenda grounded in the review’s findings. The near-complete absence of Triple A applications in emergency and humanitarian inventory management, particularly in procurement and classification functions, represents a disconnect between research concentration and practical urgency identified across the reviewed studies. Given the demonstrated value of algorithmic procurement coordination in hospital settings [19,48,57] and the growing body of disaster logistics studies, future research should explicitly bridge these two streams by developing and testing Triple A approaches designed for emergency supply chain conditions, where the organizational, data, and time constraints are fundamentally different from those in stable healthcare environments. The contradiction identified in the classification studies [20,24,58,59], that advanced ML models assume data completeness that is absent precisely in the settings where accurate classification matters most, points to a different kind of research need: not more sophisticated models, but models designed from the outset to function reliably under conditions of incomplete, inconsistent, or rapidly changing data. Addressing this would require a deliberate shift in how classification research frames its validation criteria, prioritizing robustness under realistic data conditions over performance under idealized ones, a shift that the concentration of reviewed studies in well-resourced pharmaceutical settings has so far obscured.

The organizational barriers surfaced in the procurement studies [19,48,57,64], including supplier reliability, data sharing agreements, and procurement team capacity, suggest that research should move beyond model development toward the governance and organizational conditions under which Triple A procurement tools are actually adopted and sustained. The reviewed studies consistently demonstrate what these tools can achieve technically while largely leaving unexamined the question of what organizational conditions make that achievement possible in practice, a gap that reflects the broader tendency of the reviewed literature to evaluate Triple A tools in isolation from the institutional environments in which they must function.

Finally, the absence of longitudinal evidence across all four inventory functions means that Triple A adoption’s sustained impact on operational performance and patient outcomes remains undemonstrated. Virtually all reviewed studies measure outcomes at a single point in time or within bounded simulations. Longitudinal studies tracking performance across multiple inventory cycles would provide healthcare managers with the real-world validation needed to justify long term investment, confirming durable value beyond controlled conditions.

Implementation Barriers and Practical Guidance for Healthcare Managers

The reviewed literature consistently identifies data quality, system integration, and cybersecurity as barriers to Triple A adoption, but rarely moves beyond diagnosis to offer actionable guidance for the healthcare managers who ultimately determine whether these tools are adopted. Drawing directly on patterns observed across the studies reviewed, several practical strategies emerge.

Organizations with fragmented or inconsistent data, a challenge documented across pharmaceutical, hospital, and humanitarian contexts, should prioritize data governance reforms before committing to advanced AI model development, since model reliability is fundamentally contingent on input data quality. A phased implementation pathway, beginning with descriptive analytics and progressing toward predictive and prescriptive tools as data infrastructure matures, is more feasible than attempting wholesale system replacement, particularly for resource-constrained settings. Legacy system integration challenges, especially pronounced in hospital tracking and control applications [61,62], can be partially managed through middleware solutions and API-based architectures that allow new Triple A tools to interface with existing systems without requiring full infrastructure replacement. For smaller healthcare organizations where computational and financial resources are limited, cloud-based analytics platforms provide a more accessible entry point into Triple A adoption, reducing the capital expenditure associated with on-premises AI infrastructure.

Cybersecurity and data privacy concerns, particularly when handling sensitive patient and operational data, require privacy-by-design principles to be built into any Triple A implementation from the outset rather than treated as an afterthought. Future studies should explore how blockchain for secure data sharing can be practically integrated with existing Triple A systems, drawing on early evidence from vaccine supply chain applications [36]. Addressing these barriers is not purely a technical challenge; it is an organizational and governance challenge that requires healthcare managers to build the institutional capacity and cross-functional coordination needed to deploy new technological tools in ways that are both operationally sound and ethically responsible.

5. Conclusions

The growing complexity of healthcare supply chains demands resilient, data-driven inventory systems. This review synthesizes 64 studies published between 2011 and 2025 through the Triple A integration framework, revealing that the integration of algorithmic, AI, and analytics approaches across forecasting, tracking, classification, and procurement drives resilience more effectively than any single method. Three novel insights emerge: research concentrates in pharmaceutical and hospital contexts while emergency procurement and classification gaps persist where patient safety risks are highest; advanced ML classification models assume data completeness that is unavailable in the settings where classification decisions matter most; and procurement success hinges on organizational conditions, supplier reliability, data sharing, and team capacity, that remain peripheral to most reviewed studies. The Triple A framework unifies these findings, connecting emergency logistics to mainstream research and making explicit that resilience emerges from system integration rather than methodological competition. While data quality, system integration, and cybersecurity challenges persist, phased implementation strategies, cloud-based platforms, and privacy-by-design approaches offer practical pathways for healthcare organizations at different stages of Triple A readiness.