Disaster Resilience Assessment of Pharmaceutical Warehouses in Türkiye: A Multi-Criteria and Long Short-Term Memory-Based Approach

The sustainability of the pharmaceutical supply chain after disasters is critically important to ensure the continuity of healthcare services and prevent loss of life during crises. Türkiye is geographically located in a region highly susceptible to various disasters such as earthquakes, floods, landslides, and epidemics. This study aims to comprehensively evaluate the resilience of pharmaceutical warehouses, which are key components of Türkiye’s pharmaceutical supply chain, against disasters. The research adopted both static and dynamic approaches by integrating Multi-Criteria Decision-Making (MCDM) methods with Long Short-Term Memory (LSTM)-based forecasting models. Thus, while conducting an up-to-date analysis of the resilience of pharmaceutical warehouses to current risks, the study also forecasted potential future environmental burdens. In the MCDM analysis, seventeen criteria were determined under five main categories: logistics challenges, inventory management, quality control, institutional coordination, and financial capacity. The importance weights of the criteria were calculated using the AHP, Fuzzy AHP, and CRADIS methods, while the risk preparedness levels of the warehouses were evaluated through the TOPSIS and LOPCOW methods. According to experts, the most critical factors were identified as “Incorrect Drug Distribution,” “Stolen or Counterfeit Medicines,” “Inter-Institutional Communication,” and “Transportation Difficulties.” In the dynamic analysis based on the Long Short-Term Memory (LSTM) model, one- and three-layer LSTM models were trained using data on the number of disasters that occurred between 2000 and 2024 and Türkiye’s healthcare expenditures. According to the LSTM model predictions obtained at the epoch with the lowest RMSE value, it is estimated that by 2030, at least 800,000 people will be affected by disasters, and healthcare expenditures will triple compared to current levels. These findings indicate that demographic and economic growth will create a significant gap between the existing infrastructure and the increasing future burden, suggesting that even low-risk warehouses may exhibit operational vulnerabilities. In conclusion, the proposed dual approach not only provides concrete forecasts for the future but also offers a quantitative assessment of the current infrastructure’s capacity to meet these emerging demands.