Deep Learning for Supply Chain Risk Management in Health Systems (DEFUSE)

Submitted to SCAN Health Design Competition 2020

The COVID-19 pandemic highlighted the criticality for health systems to easily identify primary personal protection equipment (PPE) and equivalent substitute products. There remains a critical need for some PPE, vital testing supplies (including kits), chemicals, reagents, and the physical materials needed for collecting and transporting samples. Health systems, OEMs, pharmaceutical companies, distributors, and other major stakeholders in healthcare supply chains are subject to many risks from unreliable sources of key raw materials and the excipients used to synthesize PPE and medicines.

Risks of PPE supply disruptions and aberrations in quantity, quality, and timely delivery to the right place and customers need to be addressed. The consolidation of generic manufacturers and a move towards global supply chains have decreased the number of facilities capable of manufacturing quality PPE. Moreover, countries have complex political and contentious trade relationships, which could threaten healthcare supply chains in the future.

The vulnerabilities and often manual processes that health systems go through to identify equivalent products are a concern for the safety and security of patients globally. There are a number of areas in which health systems’ current supply chain management (SCM) practices and vulnerability and risk (V&R) assessments are inadequate to promote agility and resiliency. These areas are as follows:

  1. Existing SCM tools focus more on supplier selection and pay less attention to V&R prediction and SC disruption mitigation;
  2. There is primary reliance on human knowledge and rudimentary metrics for identifying equivalent products due to a lack of effective decision support tools and standardized processes associated V&R assessments; and
  3. Existing methods do not support end-to-end visibility and primarily consider first tier suppliers, without addressing issues regarding second tier suppliers and beyond.

Therefore, health systems are seeking more advanced supply chain V&R assessment solutions, which can effectively assess and predict the V&Rs of PPE supply through advanced analytics (such as machine learning algorithms) using both internal and external data sources. Our proposed approach to address Eastern Health’s articulated challenges identifies alternative providers of primary PPE, as well as equivalent products (from the same and different suppliers), based on adherence to supply chain (not just healthcare) leading practices that help mitigate potential supply network risks as identified by our solution. Easy access to information about recommended equivalent products and alternative sources of supply will improve SCM efficiency, reduce staff burden, and potentially reduce costs in the face of potential V&Rs and support high fidelity and timely decisions.

To address this critical need, we propose the Deep Learning for Supply Chain Risk Management in Health Systems (DEFUSE) solution by leveraging our supply chain risk prediction and mitigation tool. Our team includes a solution provider, a health system, and an academic partner. DEFUSE will enable Eastern Health to determine equivalency of product specifications to achieve reliable, accurate, and standardized product data in an efficient manner to overcome supply shortages.

Based on internal and publicly available external data, DEFUSE can (1) determine the country of origin of PPE raw materials, components, and final products used by health systems in Canada and the USA, (2) analyze factors that contribute to SC V&R, (3) recommend alternative sources of supply that are not as susceptible to the risk factors, and (4) apply the V&R assessment to identifiable equivalent products to ensure they have a lower risk profile. The key features of DEFUSE are: 1) Automated data source exploration and data collection and intelligent information aggregation; 2) Artificial Intelligence (AI) aided SC V&R analysis, prediction, and mitigation; and 3) Flexible integration with existing MMIS, ERP, and clinical systems. DEFUSE is technology agnostic; this promotes product data standardization and harmonization due to its ability to relay data with existing systems. Because DEFUSE can leverage data in the public domain and handle data in various formats, it does not require health systems to have modern IT infrastructure.

During this effort, our health system partner will provide PPE lists, equivalent products lists, supplier data, FDA medical device approved facilities, product data, and specifications for determining product equivalency. Criteria for identifying equivalent products will be facility adjustable to ensure compliance with organizational and government guidelines. DEFUSE is conducive to use by novice and expert procurement and clinical personnel. The system will also handle product performance feedback, where appropriate, to adjust future recommendations based on product performance.

DEFUSE provides a unified automated framework for V&R analysis, assessment, prediction, and mitigation with novel modules:

  1. Data management module imports user data, automatically collects raw data from multiple data sources, improves data quality, merges and processes raw data to ascertain ground truth to facilitate further analysis;
  2. Data analytics module analyzes supply chain data using advanced machine learning techniques including deep learning, with a primary focus on deep learning, to predict future V&Rs;
  3. V&R identification module identifies V&Rs based on health systems’ requirements of acceptable fragility and criticality, and discovers/predicts new V&Rs from data analysis; and
  4. V&R mitigation module recommends alternative sourcing approaches to mitigate V&Rs to support improved efficiency and assurance of supply.

DEFUSE uses relevant, novel data sources such as stock market data, product reviews, social media feeds, web trends, and company profiles along with traditional supplier and consumption data. Through data crosschecking and correlation, DEFUSE ensures V&R details are relevant. As evidence of engaging key stakeholders, we have tested a preliminary version of DEFUSE with a large government entity and received positive feedback from a U.S. health system about the potential of DEFUSE to positively impact procurement activities and service delivery.

As a whole, DEFUSE offers advanced actionable intelligence and manages V&Rs timely and accurately compared to standard tools. The outcome of this effort will be a software tool, based on an open system architecture, that readily integrates with third party systems and demonstrates detection/prediction PPE V&Rs using a realistic health system scenario. DEFUSE addresses two prevalent challenges exacerbated by COVID-19, PPE procurement and assurance of supply and SC disruption. We welcome the opportunity to share more information about this transformative approach to addressing Eastern Health and the world’s common problem.