AI: Opportunities, threats meet, tease healthcare supply chain

To adapt a refrain from the cartoon philosopher Homer Simpson, “AI – the cause of and solution to ALL of supply chain’s problems.”

For some (e.g., entrepreneurs, innovators and opportunists, etc.), AI represents shortcuts and workarounds to be more efficient at work and either enjoy more recreational time or focus on other expedient tasks courtesy of the additional time.

For others (e.g., conspiracy theorists, cynics and skeptics, etc.), artificial intelligence (AI) calls to mind robots besmirching humans through deep fake audio, video and text or even replacing humans for jobs and revenue and taking over the world.

Still for others (e.g., forward-thinking dreamers and open-minded schemers), AI represents the first-generation leap to Star Trek-like automated decision support with automatic universal translators (no need to take language courses in high school and college to graduate!) that should accurately second-guess even the second-guessers on a global scale.

Recently, the U.S. Food and Drug Administration jumped into the frenzy by creating a new Digital Health Advisory Committee “to explore the complex, scientific and technical issues related to digital health technologies, such as artificial intelligence.”

At least one study posits that AI likely will reach a half-a-trillion-dollar industry by 2027, giving additional meaning to “generative” AI. That’s within three years and represents a faster market stickiness (in terms of prospective adoption and implementation) and predictive revenue trajectory than the internet and online exchanges, internet-of-things (IoT) and blockchain achieved during their hype cycles.

But beyond the hype, pomp and circumstance, does AI amount to anything more than a decision-support tool that requires training for the end user as well as the application itself? Is it a generative relative of blockchain? A veritable electronic Swiss Army Knife for healthcare supply chain? A “shadow” database/virtual professional assistant not unlike Amazon’s Alexa, Apple’s Siri, Google’s Assistant with Bard or Microsoft’s Cortana?

Healthcare Purchasing News reached out to numerous corporate executives on the supplier and provider segments of the industry to gauge their anticipation, application and prognostication of AI as it relates to supply chain operations. The wide-ranging exploration spanned the promise of process effectiveness and efficiency to the precariousness of potential abuse, control and cybersecurity dangers.

Optimists foresee much promise; pessimists highlight areas of concern that if left unchecked could have considerable repercussions. Overall, optimism, which is shared here and on HPN Online, seems to surpass pessimism, which can be found on HPN Online.

Several experts salute the promise that AI brings to supply chain, particularly around data collection and number crunching.

“Artificial intelligence will be able to make use of large amounts of data to make accurate predictions around usage and supply availability,” said Jack Koczela, director, Sourcing & Transformation, Supply Chain, Froedtert Health. “It can also comb through data from different sources, such as clinical systems and business systems, to create insights that we have not been able to process at scale in the past. For me, this is one of the most exciting applications of artificial intelligence. As we move into a clinically integrated supply chain, we have many small, manual models that indicate we could see significant improvement if we could just get through all the data that is available at our fingertips.”

Archie Mayani, chief product officer, GHX, marvels at the past decade of development and growth of interest in AI, particularly for supply chain applications.

“Ten years ago, the healthcare industry was just beginning to talk about AI,” she recalled. “Now, it is beginning to transform healthcare and, more specifically, the supply chain. As the industry continues to use generative AI to sort and manage the influx of data, supply chain management processes will become more predictive, effective and efficient. If used correctly, AI has the potential to refocus healthcare teams on higher-value work such as supply chain resiliency and aid in future crisis mitigation, waste reduction, and cost-containment efforts that will help hospitals reinvest funds into patient care. In the era of smart supply chains, AI is the prescription for optimizing, not just organizing.”

But another supply chain expert offers a profound observation seasoned with a pinch of irony that suits today’s workforce.

“AI can help us make better use of scarce labor and rid ourselves of repetitive low cognitive work,” noted Joe Dudas, division chair, Supply Chain Innovation and Strategy, Mayo Clinic. “It can change our jobs to be more human.”

Accentuate the obvious

The mere notion of AI motivates many experts to list a bevy of obvious applications that can benefit supply chain operations. A dozen experts share their insights with HPN that at times converge on common themes.

“Proper inventory management is a core competency for efficient supply chain management and an area where AI applications will continue to emerge for the foreseeable future,” assured Vasco Kollokian, director, Innovation and AI, Tecsys. “By analyzing historical supplies consumption data and using machine learning for demand forecasting, AI can predict accurate levels of inventory to ensure optimal stock levels.

“Once optimal levels have been established for a particular healthcare provider network, maintaining suitable levels of inventory through location-specific replenishment becomes another clear application for AI,” he continued. “Leveraging dynamic real-time inventory visibility through IoT-based technologies, AI-driven processes can optimize the minimum and maximum inventory counts and automatically replenish to match various hospitals’ consumption-driven needs for specific locations, such as operating rooms, catheter labs, nursing floors and in-hospital pharmacies.”

Kollokian further believes that AI will interface or integrate with materials management information systems (MMIS), enterprise resource planning (ERP) systems, item masters and chargemasters, regardless of brand. “ERPs typically have control over item masters,” he indicated. “If visibility into fine-grained location-specific consumption and replenishment information is provided by other, more specialized, vendors, then integration is often necessary.”

The root of such connectivity varies between built-in product programming, “bolt-on” software application or online download.

“Usually, integration APIs are provided by vendors to accommodate information exchange among disparate supply chain management solution systems,” Kollokian explained. “For example, order quantities in ERPs typically are at the ‘pallet’ level, yet Automated Dispensing Cabinets (ADC) replenish at ‘each’ or ‘box’ level. Such system disparity often results in fragmented information fabric, requiring a ‘bolt-on’ approach. Whether vendors agree to provide such integration (usually considered competitive) is an orthogonal matter.”

Demand planning rules

The global pandemic elevated predictive demand planning to the forefront of “must-have” capabilities on automated wish lists, and Kollokian agrees that it’s ripe for AI.

“While some demand spikes are quite apparent, like seasonal demand for flu vaccines, others are less so,” he indicated. “For example, states not mandating motorcycle helmets are likely to see increased head trauma cases resulting from motorcycle accidents, prompting the need for proactive prior stock-up. AI can detect epidemiological healthcare patterns for less apparent trends without reliance on tribal knowledge.”

John Freund, president and CEO, Jump Technologies, zeroes in on demand forecasting as the most obvious use. “AI algorithms can analyze historical sales data, patient demographics and other relevant factors to predict future demand more precisely than traditional methods,” he told HPN. “This can help healthcare providers optimize inventory levels, reducing costs and improving access to essential medications and supplies.”

Connectivity throughout multiple systems for supply chain is essential, according to Freund.

“The demand forecasting engine needs consumption information so the inventory management tool would be needed either from the ERP system or third-party system like JumpStock,” he indicated. “It also needs vendor performance information at the product level, so it needs receipt data either from the ERP system or third-party system. It needs census data, so it needs data from the EHR. There may also be a variety of other third-party data sets that are needed to do this properly. An AI-based demand forecasting engine will need data from a variety of sources.”

Expanding supply chain automation and optimization beyond current capabilities is another valuable contribution, Freund posits.

“AI can automate various tasks throughout the healthcare supply chain, from procurement and ordering to warehousing and distribution,” he said. “For example, AI-powered chatbots can handle routine customer inquiries, while robotic process automation (RPA) bots can automate repetitive tasks such as data entry and order processing. This can free up human workers to focus on more complex tasks, improve efficiency, and reduce errors.”

But AI doesn’t necessarily duplicate other pre-existing automated tasks, he insists.

“Today in many hospitals requisitions flow down to purchasing, who in turn place those orders with vendors, in some cases considering multiple sources for a given product,” Freund explained. “AI can automate that entire process, taking the human element out of it. Think of the number of calls to purchasing looking for order status. AI chatbots can analyze data on product defects and customer complaints to identify potential quality issues. Chatbots can [help] hospitals with regulatory requirements by tracking and reporting recall levels. There is a wide variety of ways that AI can work here.”

Arnold Chazal, CEO, VUEMED Inc., remains solidly behind AI as creating more accurate predictive models for supply re-ordering and replenishment and balance for inventory management. “AI can help ensure that proper orders with the adequate quantities are placed in time to support the clinical work regardless of the cyclical or seasonal variations and take into account the greater supply chain challenges such as backorders and recalls,” he noted.

“AI can optimize even more dynamically the on-hand inventory to meet the needs of each individual clinician by avoiding stock-outs while at the same time avoiding over-stocking, Chazal continued. “AI can do this by learning to predict what each clinician will actually need and use for each type of procedure, taking into account the specifics of anatomy and disease stage of patients to be treated based on their medical records. This application will definitely be helpful for addressing the case cart and preference cards disconnects with the clinical requirements once the patient is being treated, thus ensuring the preparation of a case cart that is more adapted to the physician’s practice and habits and the patient being treated. It will advance customized care to benefit both the patient and the clinicians.”

AI can enable supply demand forecasting for providers and suppliers alike, observes Tom Redding, senior managing director, Healthcare, St. Onge Co.

“Unfortunately, the healthcare supply chain primarily has a transactional relationship with their health system,” Redding lamented. “There is a significant potential to gain insights from transactional history to inform future purchasing decisions. Additionally, it may inform any contract negotiations with their suppliers to ensure they can meet the organization’s total demand requirements and the anticipated demand variation. Health systems are stuck in the allocation model, which drives them to seek other product options during peak periods.”

AI can assist supply chain operations within clinics, surgery centers and other non-acute facilities, too, Redding stresses.

“With more ‘low-demand’ non-acute facilities coming online in the coming years, there’s a need to continually monitor demand and adjust the available units of measure,” he noted. “There are opportunities to right-size the required units of measure. Manufacturers spend a considerable amount of time trying to determine the packaging size to optimize their costs but that doesn’t mean it meets the needs of their customers.”

GHX’s Mayani sees AI improving demand planning and forecasting overall based on data access. “AI can analyze historical and current inventory utilization data to help a hospital more accurately predict the demand for various medical supplies and equipment,” she said. “With the help of AI, high-performing hospital supply chains have an opportunity to reduce supply expenses as well as ensure the right product is available at the right time for the right procedure to support better clinical outcomes.”

Yet AI makes sense for improving clinical inventory management at the point-of-use level as well, according to Mayani. “As healthcare decision-makers consider inventory levels and their related inventory systems, inaccuracy at the point-of-use can create significant operational hurdles when trying to either reduce costs or maintain high service levels. AI models can help drive supply capture at the point-of-use and automate and streamline these workflows with seamless and interoperable data with electronic medical records.”

St. Onge’s Redding foresees AI driving inventory management optimization. “In most cases today, the process to manage clinical inventories is still a predominately manual and an arduous process,” he noted. “AI can provide a real-time monitoring and provide adjustments to how much inventory is needed and make decisions on when to order supplies to eliminate guess work.”

If anything, AI can help the healthcare industry speed up demand planning communications, according to Josh Wolfe, senior vice president, Inventory Management, Medline. “In an industry where data sharing and integration between trading partners has lagged, AI could help provide more accurate demand signals, resulting in fewer supply disruptions,” he noted.

Bruce Lieberthal, vice president and Chief Innovation Officer, Henry Schein, agrees on demand forecasting’s inherent benefits. “AI can be used to consider the myriad of factors affecting demand to help create forecasting models that are more accurate and cost-effective,” he said. “This is important so that healthcare companies can reduce stockouts or over-ordering.”

Maintaining equipment

Predictive maintenance for medical equipment is another function that AI could manage, according to JumpTech’s Freund. “Medical equipment is expensive and critical for patient care,” he indicated. “AI-powered predictive maintenance systems can analyze sensor data from medical devices to identify potential malfunctions before they occur. This proactive approach can help prevent costly downtime, extend the lifespan of equipment, and ensure that critical devices are always available for patient care.”

This extends a step further than the real-time maintenance programs offered by several original equipment manufacturers that automatically alert the OEM when something goes wrong onsite at the provider facility to minimize operational downtime.

“AI algorithms can go deeper than purely maintenance data,” Freund assured. “For example, AI could study sensor data from machines used to coat the joints with a biocompatible material. The algorithms can detect defects in the coating, such as air bubbles or uneven thickness. AI can be used to optimize the coating process by adjusting the machines' speed, improving the joint's quality or reducing recalls.”

For James Kollai, senior director, Client Solutions, PartsSource, AI reaches into the parts and inventory management aspects of equipment maintenance. “AI could optimize inventory management by predicting which parts are likely to fail and when,” he said. “This would then help clinical engineering and healthcare technology management teams ensure that the right components are available when needed without excessive stockpiling.”

PartsSource strives to integrate AI in critical decision support for dual and digital marketplace of providers and suppliers, according to Kollai. “One of the more obvious applications where we have used AI is through lead time predictions,” he noted. “In recent months, we have established a process leveraging our industry leading supply chain and transactional data sets along with a proprietary AI to predict lead times for a large portion of items purchased from our online marketplace. Approximately 90% or more vendors do not provide inventory feeds with accurate shipping information, which is crucial to customers who need parts to repair their medical equipment. Our AI can predict shipping times based on data from previous purchases by customers from the supplier to accurately fill in those inventory gaps.

“Additionally, PartsSource uses AI to help inform our Guaranteed Stock offering for our customers,” Kollai added. “Our teams are currently working on integrating AI for forward stocking, also called local stocking, options to minimize equipment downtime.”

Accentuate the alternatives

In addition to demand planning, AI can help manage clinically acceptable alternative products and sourcing, which attained a fever pitch during the COVID-19 pandemic.

“Given the new normal of supply disruptions, the ability to efficiently and accurately recommend clinically viable alternate products is a necessity in today’s healthcare supply chain,” indicated Medline’s Wolfe. “AI provides some exciting opportunities to meet that need.”

GHX’s Mayani contends AI can compile a list of substitutes to help avoid backorders and shortages, effectively serving as an extension of demand planning. “AI has the ability to offer more efficient and effective ways to manage supply chain disruptions and shortages,” she indicated. “Machine learning (ML) and AI algorithms can easily analyze historical data and market trends to predict potential backorders and make recommendations on product alternatives and substitutes based on evidence-based clinical outcomes. This not only helps limit overstocking and reduce waste, but also helps drive value-based care upstream and across the continuum of care.”

Carl Natenstedt, CEO, Z5 Inventory, concurs. “AI can be used to automate the ‘alternate sourcing scurry’ that occurs during backorders and product shortages,” he said. “AI can learn from past product shortage resolutions to mimic the human sourcing function and identify alternate sources of the same or similar product and place emergency orders.”

Evaluating supplier service

AI can assist in making considerable progress with supplier performance measures and programs, St. Onge’s Redding suggests. “Health systems will gravitate to AI to inform their decision-making process on which suppliers will get their business in the future,” he said. “Health systems can’t simply look at the cost of the product only. They will need to take into consideration the total cost of the relationship. If the health system takes into the consideration all their internal time and effort to resolve issues with a supplier, substituting supplies from a different supplier potentially at a higher cost, expediting freight costs and many other factors, the supplier relationship may not be worth all the effort.”

Product access and availability are the triggers, Redding insists.

“Every health system will share without pause that product availability is their biggest challenge daily; it was a challenge prior to the pandemic, and it has accelerated since the pandemic,” he noted. “Health systems will need leverage AI to continually monitor supplier performance and gain insights into the health of their suppliers. Creating a reliable, efficient and resilient supply chain starts with collaboratively sharing data/information between the health system and suppliers to ensure there is end-to-end visibility and each party can inform their internal decisions. Too often, we see reluctance to share data/information because there is concern that the other party may take advantage of the situation, either from a cost savings point of view and/or shifting away from the supplier because of lack of internal performance.”

Homing in on the granular

Demand management and monitoring warehouse performance top the list for George S. Godfrey, chief supply chain officer and corporate vice president, Financial Shared Services, Baptist Health South Florida, but he takes a cue from Amazon Prime, FedEx and UPS and points to delivery calendars for inbound and outbound shipments to manage “real-time scheduling that adjusts to traffic delays, urgency of goods, etc.”

Z5 Inventory’s Natenstedt sees benefits to automated and constant PAR area optimization and physician preference card management and maintenance from AI deployment, too.

“AI can learn from each replenishment assessment the consumption patterns of each specific PAR area and apply intelligent adjustments on a near-daily basis to streamline and reduce investment in inventory,” he indicated. “AI can learn from past case records and post-case waste and returned product assessments to constantly update and adjust the items on preference cards and their classifications to reduce product waste in the OR.”

Editor’s Note: For additional exclusive coverage of AI’s not-so-obvious uses, potential dangers and specific areas where AI is and will make clinical, financial and operational inroads relating to supply chain, visit HPN Online. Back in November 2023, HPN conducted an AI Virtual Forum, “Can AI Ease Current Healthcare Burdens?” To view it, click here: https://endeavor.swoogo.com/2023HPNVirtualForum/todays-landscape.

Pulling back the curtain on second-thought AI benefits for healthcare supply chain

What if AI can help with or manage this …

Call the obvious benefits that artificial intelligence (AI) brings to the healthcare supply chain a bushel of low-hanging fruit that contains such juicy morsels as demand planning, inventory forecasting and optimization, clinically equivalent product alternative/substitute generator, product and service consumption tracking, equipment maintenance and supplier performance review as well as more granular subsegments that include PAR level oversight, physician preference card management and freight/shipping auditor.

So says a dozen corporate supply chain executives in the supplier and provider segments of the industry with whom Healthcare Purchasing News spoke about one of the hottest topics simmering this year and last.

Move beyond the overt, however, and query them about some of the not-so-obvious benefits that AI brings to the healthcare supply chain – the harder-to-reach fruit higher in the tree – and the second-thought observations reach into such areas as personalized medicine, clinical decision support, demand matching and customer contact and service for internal departments.

Bruce Lieberthal, vice president and Chief Innovation Officer, Henry Schein, foresees useful applications for precision medicine. “AI can be used to match patients with customized therapeutic plans for more effective treatment,” he said. “This also helps enable individuals conducting drug trials to better match participants to studies.”

John Freund, president and CEO, Jump Technologies, homes in on using AI for what he calls the “patient-centric supply chain.

“AI can be used to optimize the supply chain around the needs of individual patients,” he indicated. “By analyzing patient data, including medical history, medication use and upcoming procedures, AI algorithms can predict the specific supplies needed for each patient. This can help ensure the right supplies are available at the right time, reducing the risk of delays or shortages.”

This will have to involve integration between the materials management information system (MMIS) or enterprise resource planning (ERP) system with the electronic health record (EHR) and electronic medical record (EMR). In fact, Freund envisions more consolidation between the EHR and EMR.

Arnold Chazal, CEO, VUEMED Inc., sees value in the concept as well.

“Based on analyses of medical records and individual physicians’ modes of practice, AI would be useful for making recommendations about the usage of a combination of medical products and implants that is closely tailored to each specific case, thereby avoiding availability issues and problems related to out-of-stock and backordered items,” he said.

Supply chain also could use AI as a clinical procedure tracer, according to Chazal. “Analyze patient outcomes and link them to their original diagnosis and treatment, including the choice of the combination of implants and medical products,” he noted. “This would enable hospitals to determine which winning combination has performed the best in terms of total cost of the procedure to the outcome, which could impact purchasing patterns, PAR levels and on-hand product variation, thus streamlining orders and rationalizing the composition of inventory.

Is the time ripe for a “SCarlet?”

If Amazon has Alexa, then what about Supply Chain accessing a similar conceit, maybe something like “SCarlet,” an AI-based operations support specialist?

“Healthcare professionals frequently need to look up medical surgical items in their supply chain systems with missing, incomplete and rather cryptic bits and pieces of information, as in ‘CFN56A, TUBE SUCT CONN 72X3/16IN,’” indicated Vasco Kollokian, director, Innovation and AI, Tecsys. “Smart search engines powered by purpose-built medical/surgical knowledge bases/ graphs can handle such incomplete keywords searches, identifying not only the most likely product candidates represented by the cryptic keywords, but their equivalencies in case of shortage replacement considerations.”

Yet it extends beyond a traditional translation software program or value-added network linking product identifiers and data standards, according to Kollokian. “I would say it is more of a search-centric intent detection rather than translation of nomenclature,” he added.

Supply chain also could move into modeling for operations and workflow.

“Autonomous AI agents can create and maintain digital models that mirror physical supply chain entities, such as levels of inventory, patient demand, supplier information, etc.,” Kollokian said. “By ingesting and continuously updating the models from various data sources, such digital supply chain twins provide a real-time digital reflection of the supply chain, capturing every nuance for precise optimization, identifying flows, interactions and bottlenecks, highlighting mechanisms to overcome them.”

Kollokian further defines the concept of autonomous AI agents. “By ‘autonomous’ in this case I mean agents that do not require constant human intervention but are under the control of programmed algorithms driven by rule sets to specifically carry out their tasks,” he indicated. “Like real-time data collection and updating of various aspects of the supply chain digital twins. It does not mean they are free-roaming and able to conduct arbitrary tasks that can be manipulated. An analogy would be a smart thermostat that takes into consideration the cost of the energy at various times and how to mitigate heating/cooling to meet the temperature expectations.

Supply chain could use AI to navigate through crises or daily demand spikes.

“AI-powered process modelling can simulate what-if scenario analyses to anticipate supply chain disruptions and craft potential strategies to ensure a resilient and efficient supply chain,” he continued. “For example, answering what-if questions like ‘Among all the scheduled surgical cases for the next two weeks, highlight all the affected cases if item #12345 is on back order.’ Then, med/surg-specific recommendation engines, powered by smart searches, can go to work to identify potential replacements to avoid shortfall and identify their sources of procurement.”

This also might entail upstream supply chain views beyond the manufacturer, according to VUEMED’s Chazal.

“Automate the purchasing and replenishment of stock with greater visibility in the upstream supply chain all the way to manufacturers and even their own sourcing channels and include the logistics of the delivery of goods,” he noted. “It could figure out alternate channels and suppliers based on specific requirements including availability, price of goods, delivery mechanism, and cost to source the necessary products with optimal time/quality/cost benefit.”

What if “SCarlet” could make sense of operational variability and decision turnarounds or updates and oversee aspects of ordinary workflow? Josh Wolfe, senior vice president, Inventory Management, Medline, believes it’s possible.

“With increasing volumes of product substitutions, dual sourcing initiatives and supplier conversions, product data governance and consistency are more important than ever,” Wolfe noted. “AI can assist healthcare supply chain professionals in identifying data errors and ensuring accuracy during product conversions or channel shifts.

“An often overlooked and relatively straightforward application of AI is to not necessarily automate entire processes, but instead simply parse out daily emails or tasks into specific work groups,” he continued. “Just the act of automatically funneling that inbound workflow can make our teams that much more efficient in processing the work itself.”

Counterfeit detection, managing risk

Ferreting out counterfeit and knockoff products anywhere along or within the supply chain can be directed by AI, according to JumpTech’s Freund.

“Something that has been around for a while but really came to be during COVID was the problem with counterfeit products,” he noted. “Counterfeit medical products seriously threaten patient safety. AI can detect counterfeit products by analyzing product packaging images and labeling. AI algorithms can identify subtle differences between genuine and counterfeit products, such as font size, color or logo placement variations. This can help to prevent counterfeit products from entering the healthcare supply chain.”

This likely would free up staff from scanning bar codes, radiofrequency identification (RFID) chips and tags or other relevant tracking data due to the visual element deployed by AI. But security remains a concern.

“Product images and imaging, in general, will play a big role in AI in the future,” Freund said. “Yes, the AI will need to be trained. This is a great use of generative AI, and all of healthcare needs to share data and work together on this problem. This opens up the huge problem in healthcare of data standards and governance. As for the security issues, that is something that is always of concern, especially in these early days. Generative models will only be valuable to a hospital if they are large enough. This means an entity that gathers and consolidates data from many sources and trains the models. This will not be a single hospital or health system but rather a third party. How data is exchanged in that environment is something hospitals will need to pay attention to when it comes to security, patient knowledge of generative models, data ownership and the like.

“Hospitals and health systems will need to apply the same level of security on their data as they do today, but then add in all the issues associated with the large data sets associated with generative AI,” he added.

AI-powered risk management represents another not-so-obvious possibility, according to Freund, where AI can be used for automated supplier screening, real-time supplier monitoring, and predictive risk modeling. “These three applications are well-defined and can be broken down into smaller, more manageable tasks,” he added. “There is a growing body of data available to train AI models for these tasks, and there are a number of AI tools available that can be used to implement these applications, making them relatively easy to implement.”

AI could aid in regulatory compliance using Natural Language Processing, says James Kollai, senior director, Client Solutions, PartsSource. “AI can scan regulatory documents and compare them with internal practices to ensure compliance,” he noted. “This is usually a manual and time-consuming process and using technology tools in this way can increase efficiency and accuracy.”

Henry Schein’s Lieberthal notes that AI could tap Natural Language Processing (NLP) and Optical Character Recognition (OCR) to fortify Robotic Process Automation (RPA) to streamline ordering, repetitive processes, claims management and other related tasks.

Tracking equipment performance

Imagine if “SCarlet” could oversee equipment operations to monitor uptime, regulate downtime and ensure consistent running?

“AI [could handle] equipment monitoring to inform the preventive maintenance process and inform decisions on what spare parts are needed,” said Tom Redding, senior managing director, Healthcare, St. Onge Co.

Henry Schein’s Lieberthal concurs. “AI can be coupled with Internet-of-Things (IoT) sensors and software to closely track the status of critical equipment and to predict when to replace parts. It can also help detect very early signs of issues to prevent failures,” he added.

Predictive maintenance may not be directly related to supply chain, but it’s connected, according to PartsSource’s Kollai. “Understanding when equipment might need repair or maintenance ensures the health of that equipment. By using AI to predict when equipment needs maintenance, unexpected downtime can be more preventable, affecting supply chain efficiency,” he indicated.

Fine-tunning supply chain operations

As supply chain expands beyond the hospital walls and acute-care boundaries, St. Onge’s Redding feels that AI may facilitate the transition.

“As health systems support a decentralized model of care with many more non-acute care locations, including deliveries to patient homes, the ability to monitor traffic patterns and inform their daily courier and transportation activities will be critical to optimize resources and ensure timeliness of deliveries and pick-ups,” he said.

PartsSource’s Kollai calls for “cognitive procurement” as another way AI can be used. “AI can enhance decision-making in procurement by providing insights from unstructured data, such as new reports, market trends, current events and social media,” he noted.

Redding recognizes the possibilities, too. “Capturing process data through systems and technology can provide insights to identify process variation and process reliability issues,” he indicated. “Supply chain leaders are under constant pressure to provide reliable and efficient services throughout their health systems. Supply chain leaders can’t simply hang their hat on saving the organization money on the price of the products and services, they need to continually deliver their services to enable patient care. The first time the supplies and equipment are not available, trust is lost and is hard to regain, and sometimes never regained.”

Carl Natenstedt, CEO, Z5 Inventory, questions whether AI can eliminate PAR assessments altogether. “Simply have AI develop orders for specific inventory areas based on caseloads, census figures, and annual, bi-annual or quarterly inventory assessments and other quantitative inputs it can learn from to develop order patterns and frequencies. If AI can write your college term paper, it can certainly figure out what products and how many to buy for the average department's inventory shelf,” he quipped. “AI can be trained to conduct sourcing [requests for proposals] and to select and negotiate the best vendor contracts. AI is currently being used to write legal briefs and other complex legal and business documents. Using it to increase the speed and simplify the sourcing RFP process is a likely future application.”

Archie Mayani, Chief Product Officer, GHX, expands AI’s boundaries to include three areas: Resource tracking, environmental sustainability and streamlined management of specialized materials.

Expiration Date Prediction in Pharmaceutical Supply Chain Management: “AI can be leveraged to predict the remaining shelf life of pharmaceuticals, aiding in the efficient management of inventory and minimizing wastage due to expired products,” she said. “By analyzing factors such as storage conditions, product characteristics and historical usage patterns, AI algorithms can predict expiration dates more accurately than traditional static shelf-life assignments. This can help reduce waste by eliminating the discarding of still-usable items.”
Carbon Footprint Optimization: “AI can help optimize the carbon footprint of the healthcare supply chain by providing insights into the environmental impact of transportation, packaging and procurement decisions, Mayani continued. “AI algorithms can analyze data related to transportation routes, packaging materials and supplier locations to suggest environmentally friendly alternatives including the last mile. This not only aligns with sustainability goals but can lead to cost savings by identifying more efficient and eco-friendly supply chain practices.”
Clinical Trial Material Forecasting: “AI can be applied to forecasting demand for materials used in clinical trials, optimizing the supply chain specifically for trial-related products,” she indicated. “Clinical trials often involve unique and specialized materials, and predicting the demand for these items can be challenging. AI algorithms that leverage historical trial data, patient enrollment patterns and external factors like patient demographics, disease prevalence and incidence rates, economic conditions, and geopolitical events, can provide more accurate forecasts. This helps ensure the necessary materials are available when needed, preventing delays in trial timelines and improving overall trial efficiency.”

Provider practicality

Healthcare supply chain executives within integrated delivery networks (IDNs) and hospital systems know how to balance pie-in-the-sky possibilities with practical workflow improvements.

Joe Dudas, division chair, Supply Chain Innovation and Strategy, Mayo Clinic, envisions AI assisting in four key areas.

First is visibility as he believes AI can help him process and synthesize information across the supply chain as opposed to a single product and location. The second involves contracting: Imagine being able to “redline a contract and/or identify contract risk,” he indicated. Third is customer experience where you can “use voice and/or [virtual reality] requisitioning and eliminate today’s shopping at work user experience,” he noted. Finally, there’s consulting where the AI facilitates asking a supply chain management expert on a whole host of topics and benchmarks of best practices, he adds.

Jack Koczela, director, Sourcing & Transformation, Supply Chain, Froedtert Health, highlights three worth noting.

“Optimizing clinical efficiency through preference card review could lead to significant savings and operational efficiencies,” he told HPN. “Unfortunately, the data accuracy on preference cards and the lack of robust, barcode-based documentation makes this very difficult to achieve.”

Koczela also favors AI “matching clinical need to implant or supply selection,” and “creating AI- or bot-based customer service systems for our internal supply chain customers.”

George S. Godfrey, Chief Supply Chain Officer and corporate vice president, Financial Shared Services, Baptist Health South Florida, suggests three potential process improvements:

AI can be used to manage the resolution process associated with invoice match exceptions.
AI can be used to assist the contract activation process after you start the contract renewal proceedings.
AI can be used to monitor each nurse station supply cabinet and adjust min/max inventory levels based on specific demand patterns for a particular nurse station, he adds.

How much clear and present danger lurks amid AI fog?

Healthcare supply chain must deepen understanding before use

By Rick Dana Barlow

Consider an enterprising but secretly nefarious provider executive sitting in the audience of a seminar at a trade show, listening to a supplier executive give a presentation and recording that executive’s voice only to return to the office to use artificial intelligence to manufacture a seemingly authentic voice mail where that supplier executive promises the provider executive’s organization a 25% discount above the negotiated contract price to satisfy the CFO’s cost-cutting demand of supply chain.

Rest assured, Vasco Kollokian, director, Innovation and AI, Tecsys, doubts something so dastardly could happen, foiled by clear indications.

“To generate deep fakes, specialized generative neural networks (called GANs) require quality data of substantial size to truly impersonate an individual,” he assured. “Capturing someone’s voice over a mobile device, in a crowded conference, will undoubtedly result in poor quality fake, if ever.

“Again, the quality of the recorded message should raise concerns,” Kollokian continued. “To [make] real fakes, sizable data of high quality is required and unlikely to come from short cellphone recorded snippet of voice in crowded conferences. [With] actors’ examples, there is a preponderance of quality data (video and audio) widely available.”

Further, he questions access to provider information. “The so-called vendor impersonator having access to contract pricing to offer competitive counterbids, these contract details are likely inaccessible to external parties. The low probability that such an engagement could come in voicemail form, from a legitimate vendor. This is akin to banks saying things like ‘the bank never asks you to authenticate yourself through an email…’. The challenge is vendors may not be suffering from wrong impersonations. At least not yet.

“If I go out on a limb and do some conjectures, I’d say deep-fake detection will likely be a market for potential AI-based counter-offensives,” Kollokian noted. “Like modern smartphones currently detect inbound spam calls and filter them. These would come in the form of nuanced fake recordings detection capabilities (easy for AI to make, easy for AI to detect) embedded into communication devices (email, voice, others). However, as time passes and crooks evolve, the countermeasures will not just need to keep up, but surpass too.”

But John Freund, president and CEO, Jump Technologies, wonders about other applications.

“I have already used a third-party tool to take my corporate headshot found on our website, wrote a script I wanted that image to say, and generated a video of me reciting that script,” he indicated. “With my voice being out on the web with the various webinars and panel discussions I have done, one could use that voice to make it sound just like me, and in the end, I wouldn't be able to stop it until the damage was already done. Google, Microsoft, Amazon, Meta and Apple all have voice recognition, facial recognition and encryption software designed to help with this, but they are not perfect. Elon Musk and others have been vocal in saying this is an area of AI that needs some work.”

In a world increasingly susceptible to the pandemic of cybersecurity threats and “hackfluenza,” the pride and prejudice of AI seems to stain the promise and productivity it promotes.

To wit, a recent report shows that the blazingly popular generative AI “chatbot” ChatGPT has cultivated nearly 15 billion visits globally for the year ended August 2023.

Obvious dangers

Against the backdrop of data and information privacy and security as well as transactional integrity, some express concern about the safety and responsible and ethical use of AI in the workplace even as organizations across multiple industries establish strict user guidelines for the technology in advance of any potential regulatory requirements.

To appropriate a popular line of dialogue from the 2000 film, “Meet the Parents,” it’s all about establishing a “circle of trust.” Imagine the two-fingers-to-eye motion with rotating wrists on your own.

“In my opinion, this really comes down to building stronger relationships with the people who are using AI,” Jack Koczela, director, Sourcing & Transformation, Supply Chain, Froedtert Health, told Healthcare Purchasing News. “AI should never replace human relationships or human interactions; it should truly augment them. Of course, we are living in an age where this is a new and emerging technology. We are still trying to figure out what the rules of the road will be. We should see this as an opportunity, not as a threat. With that said it is important for us to train our associates to understand the benefits and the potential pitfalls of artificial intelligence. By now, we are all used to cybersecurity training, I imagine we will start to train our teams on AI, too.”

Aside from human behavior modification, healthcare organizations should pursue developing technology control measures, according to Archie Mayani, Chief Product Officer, GHX.

“Data is a cornerstone of AI development,” Mayani asserted. “To harness its full potential, healthcare organizations must invest in a robust data infrastructure that not only encompasses data collection, but also accounts for storage, processing and security measures. A robust data infrastructure is critical to safeguard patient data, which is increasingly vulnerable to cyberattacks. To improve data integrity, it is imperative to obtain diverse data from sources ranging from electronic health records (EHR) to medical imaging and ensure their cleanliness and accessibility. Organizations must also establish and routinely revisit data governance protocols to comply with healthcare regulations and maintain privacy standards.”

Arnold Chazal, CEO, VUEMED Inc., cautions against unbridled AI access and use, citing “an over-reliance on AI and machine learning without human safeguards and oversight” as an inherent and obvious danger sign.

“We need to be able to understand how the information has been generated and if the logic and algorithms are aligned with the ultimate objectives that the information is used for,” he indicated. “We also need to be able to guarantee that AI is not using data that, in the first place, is questionable, which will hinder its learning ability and provide mediocre results or false information. And we need to ensure that AI is used in a framework that is serving the purpose of human decision-making and not replacing it, at least not initially and only within limits.”

Even Joe Dudas, division chair, Supply Chain Innovation and Strategy, Mayo Clinic, recognizes the obvious and inevitable: “[AI] is not perfect and only as good as the information it consumes.”

Healthcare organizations should recognize the concept of “caveat emptor” and trust their gut, according to Josh Wolfe, senior vice president, Inventory Management, Medline.

“When it comes to accuracy and validity of the output of AI tools, many of the obvious warning signs or red flags to monitor are the same things we’d look for in non-AI driven processes,” he said. “Outcomes or proposals that seem too good to be true, are misaligned with organizational goals, or clearly conflict with expected results should be scrutinized heavily.”

Froedtert Health’s Koczela concurs. “It's probably important to follow the old adage if it sounds too good to be true, it probably is. This applies for seeing through AI dangers, such as deep fake videos. It also applies to startup companies promising AI solutions that are just vaporware. Finally, we should always be skeptical of companies promising AI models that do not need any human intervention. Some of these models are very accurate. But I have still seen them fabricate results or draw conclusions that no reasonable human would,” he said.

George S. Godfrey, Chief Supply Chain Officer and corporate vice president, Financial Shared Services, Baptist Health South Florida, echoes warnings of “too good to be true” offers. Further, he advises to “be cautious with new suppliers offering a good or service that meets an immediate need. How did this supplier know we had a specific need? How did they obtain this information? Are they seeking payment prior to the delivery of goods and services?”

A lack of transparency and accountability within AI projects represents an obvious danger, alerts JumpTech’s Freund. “It is important for organizations to have a clear understanding of how AI algorithms are being used and the data that is being collected, processed and stored. Algorithms should be transparent and auditable, and organizations should be able to hold those responsible for any data breaches accountable.”

Such efforts will require testing and training, Freund continues. “Hospitals must make sure that there is a proper amount of algorithm training and testing being done. AI models are only as good as the data that they are trained on. Organizations need to carefully select and prepare training data to ensure that AI models are accurate. Models should also be thoroughly tested before deployment to identify and address any potential flaws or vulnerabilities. We are all very aware of the data quality issues in healthcare. Hospitals much take this into consideration when training their AI models.”

Finally, AI should not be used in isolation, he urges. “There should always be a human in the loop to supervise and interpret the outputs of AI models. Humans should be able to override AI decisions when necessary and ensure that AI is not being used in a way that violates patient privacy or ethical standards,” he added.

At the same time, not every human but selected humans should be involved, according to James Kollai, senior director, Client Solutions, PartsSource.

“One of the obvious warning signs to watch out for is unauthorized access to data,” he advised. “Some ways to avoid this is by never uploading sensitive or confidential data or information to the web. Additionally, avoid using publicly available platforms like ChatGPT to address business specific problems, unless you have established very clear guidelines on what can and/or cannot be done.”

Kollai also wars about a lack of transparency. “Without a clear understanding or reasoning behind insights generated from AI, mistrust toward the information can develop from stakeholders, healthcare professionals, etc.,” he indicated. “Also, insufficient user training and security practices can jeopardize the safety of AI in an organization’s healthcare supply chain. By not properly educating staff and/or not encouraging the implementation of cybersecurity practices, cyberattacks through phishing and other methods can take place.”

GHX’s Mayani acknowledges that various technical and ethical challenges accompany implementing AI and generative AI in healthcare supply chain management. Three come to mind to her.

Data and privacy security concerns: “Healthcare data, including supply chain information and data, is highly sensitive. Ensuring the privacy and security of this data is crucial. AI systems must comply with strict regulations such as HIPAA, GDPR and other regional laws that govern data privacy and security.”
Interoperability and data standardization: “Healthcare data often resides in diverse systems and silos that may not easily communicate with each other. Integrating AI solutions with existing infrastructure and ensuring data standardization across platforms can be particularly challenging, leading to interoperability issues.”
Data quality and availability: “AI algorithms rely on high-quality, comprehensive datasets. In healthcare supply chain management, obtaining accurate and up-to-date data can be challenging due to variations in data sources, data entry errors and differing data formats,” she indicated.

New role brewing, emerging?

Should healthcare provider organizations employ dedicated experts to manage, regulate and train this technology, such as an AI manager as part of the information technology or supply chain staff?

“Yes, I believe this may follow the same process of development that we have seen with analytics,” Froedtert Health’s Koczela noted. “I think it will start with highly specialized and centralized teams. As the technology develops, AI will become part of the fabric of the way we do business. At that point, these centralized teams can become centers of excellence for the rest of the organization to use as they adopt AI enabled solutions.”

JumpTech’s Freund agrees. “Large health systems are hiring data scientists today to champion their AI strategies,” he said. “They will be part of the tech team but will work with areas of the hospital, like supply chain, to come up with their AI strategies. This role will help with the ‘make versus buy’ decisions associated with their AI strategy. You will also have IT staff responsible for the security and governance of AI data, whether the health system owns data or uses generative models they license from third parties.”

Tecsys’ Kollokian considers the prospect interesting. “My take on it is that the effectiveness of AI in healthcare supply chain – or any other vertical for that matter – is to have AI expertise among staff, who are in position to scrutinize the nature of vendor-pitched AI-based solutions. Specifically avoid ‘AI-washing’ initially, which would lead to higher probability of a successful business outcome in the context of the particularities of the supply chain problem they are intended to solve.

Not-so-obvious dangers

Perhaps the most obvious of any “not-so-obvious” dangers ironically starts at the beginning.

“One warning sign would be companies that do not have an AI strategy,” Froedtert’s Koczela noted. “This technology is rapidly advancing, and we need to be ready for the improvements that it can bring to our supply chain.”

Healthcare organizations should see inherent value in understanding how AI works, according to GHX’s Mayani. “Due to the lack of domain expertise, it can be difficult to explain how the AI algorithm reaches a certain conclusion,” she said. “Additionally, it’s important to consider implicit bias in the way AI models are set up or trained. Therefore, AI should not be the ultimate decision maker in patient care – care decisions still require a clinician’s expertise and needs to adopt a human-in-the-loop approach, especially with decisions where clinical outcomes can be impacted.”

Understanding how AI works and can be manipulated can motivate you to develop proactive protectionary measures that also can help you react to troubling situations, according to Tecsys’ Kollokian.

“Falling under the ever-expanding umbrella of cybersecurity, organizations must remain vigilant as ever to root out threats and misuse,” he urged. “They should have tools, training and processes in place to proactively snuff out suspicious activity. The current problem is that many aren’t trained to spot those security threats because it is all so new. The key lies in adapting their understanding and reflexes to this evolved digital landscape. It’s about recognizing that the digital threats faced today are simply the latest versions of the fundamental security challenges that have always existed. By embracing continuous education, investing in updated security technologies, and fostering a culture of awareness, organizations can better prepare themselves to identify and combat these threats effectively.”

In fact, Kollokian encourages healthcare organizations to use AI to battle hackers because “as hackers extend their penetrative techniques to include AI, so will the countermeasures.”

PartsSource’s Kollai concurs. “Something to watch out for are hackers when using AI,” he warned. “Hackers are exploiting vulnerabilities in [generative AI] to develop workarounds to cybersecurity measures. So, healthcare provider IT teams and cyber experts must be continuously vigilant to new attack methods that can be creatively derived and implemented frequently,” he said.

Kollai also emphasizes deep-fake utilization of AI. “This is when AI is used to manipulate or generate images, audio and other media that look real,” he explained. “Deep fakes could potentially be used to bypass facial recognition or other authentication systems. Some ways to protect against this is through the education and training of employees, using multi-factor authentication for added layers of security and being vigilant in updating security software.

“Lastly, something to consistently watch for is unrecognized bias,” he continued. “Without proper collaboration with various stakeholders including professionals in healthcare supply chain, there could be an inherent bias and/or a decrease in acceptance of the use of AI. To avoid this, encourage and practice collaboration and engagement among all stakeholders and implement regular audits and diverse testing scenarios.”

JumpTech’s Freund raises the bias issue as well. “Many of us have heard of the concerns with data bias and discrimination in AI models. This does not always mean discriminating based on race, gender or political belief. AI algorithms are trained on large datasets, and if this data is biased, the algorithms will also be biased. This can lead to discriminatory outcomes involving vendor selection and product use. Ensuring that there is no bias in your models is very important.

Data need protection, too, specifically the quantity that AI accesses, according to Freund. “AI systems often handle large amounts of sensitive patient data, making them attractive targets for cyberattacks,” he said. “It is also important to remember that data used in generative AI models comes from very large public data sets and that the results of what you generate there become part of the public data set. Organizations need to have robust data security measures in place to protect this data from unauthorized access, disclosure, alteration or destruction and be mindful of ownership of the data being used.

“Finally, it is important for organizations to be transparent about how AI is being used in their healthcare operations,” he concluded. “This includes providing patients with clear explanations of how their data is being used and how AI algorithms are making decisions about their care. It is also important for organizations to have mechanisms in place to hold AI developers and deployers accountable for any harm that their systems may cause.”

VUEMED’s Chazal recommends immersion in AI before active use and avoid “not auditing the AI algorithms or fully understanding how they work and what they are intended to do in terms of providing humans with information. We need to understand to be able to trust AI and be able to check on the process, not just the results.” Taking the human out of the decision-making process and blindly relying on information and recommendations made by AI is a red flag, he adds.

“We need to proceed carefully with full automation guided by AI by putting checks and balances in place to ensure that the information or recommendation is fully aligned with the values and ethics of the organization that uses them,” he said.

“We see some of the not-so-obvious risks with AI as being less about overt risk of fraud or misuse, and more about the unintended consequences for the teams and processes using AI,” noted Medline’s Wolfe. “If there’s any selection bias in the data feeding the AI, even if unintended, it can result in biased outputs leading to poor results and organizational risk. And even when the data is reliable and the AI is improving efficiency, there’s a real risk of becoming over reliant on it, and users losing the business acumen and subject matter expertise that makes them unique.”

AI’s meaningful contributions to key supply chain issues

Healthcare Purchasing News concentrated on several key areas – departments and functions – where artificial intelligence can play a pivotal role in reshaping and improving practices and processes and asked provider- and supplier-based corporate supply chain executives what they reasonably might expect. Here’s what they shared.

1. To cleanse their data routinely and/or seek out and/or use only authenticated, trustworthy, validated data and potentially obviate the need for or replace data-cleansing services and software

“AI algorithms can identify and remove duplicate records, incomplete data and erroneous entries from hospital databases. This can help to improve the quality and accuracy of data, making it easier for clinicians and researchers to access and analyze reliable information. AI can be used to identify unusual patterns or outliers in data, which may indicate potential errors or fraud. This can help to proactively address data quality issues before they cause problems. AI can be used to monitor data streams in real time, identifying and flagging potential issues as they occur. This can help to prevent data quality problems from escalating and disrupting workflows.” On replacing data cleansing services and software: “Yes.”
--John Freund, president and CEO, Jump Technologies

“Having inconsistent item master data with missing or incorrect Universal Device Identifiers (UDI) like GTINs will cause all sorts of supply chain hurdles. No GTIN means that items cannot be scanned, ordered in the right quantities, received or billed. Using a variety of techniques such as fuzzy searches and semantic associations, abbreviation disambiguation, AI can leverage trusted data sources to continuously resolve and cleanse UDI’s in item masters, rectifying their package multipliers so proper items and quantities can be ordered.” On replacing data cleansing services and software: “Progressively yes, although they themselves might leverage AI techniques under the covers.”
--Vasco Kollokian, director, Innovation and AI, Tecsys

“While some providers may use artificial intelligence to cleanse and/or update their data, many providers will probably look to third parties to provide this service for them. It would be my expectation that any third party providing this service would have an AI-enabled solution. This should allow them to scale their solutions to the size of modern item masters and SC data needs. Once, a provider’s item master only contained 10-20,000 items. Now these files have grown to 500,000+ lines long. This amount of information is impossible for a human to cleanse. But using AI rules it should be relatively straightforward, efficient and scalable.” On replacing data cleansing services and software: “No, it will not. AI will enable data cleansing at scale. The amount of time we spend cleaning data is a huge expense. If we can get AI to clean huge amounts of data, it will make our downstream systems and reports far more efficient.”
--Jack Koczela, director, Sourcing & Transformation, Supply Chain, Froedtert Health

“[AI] can use more information that may not be considered today including unstructured information such as sound and image.”
--Joe Dudas, division chair, Supply Chain Innovation and Strategy, Mayo Clinic

“By surveying many sources of data available and ensuring the comprehensiveness and accuracy of product information, including UDI data.”
--Arnold Chazal, CEO, VUEMED Inc.

“Something AI could potentially be used for is cleaning and normalizing asset lists to understand the make and model of equipment in use by the provider. This could lead to improvement in inventory tracking and management of supplies and resources and a more streamlined procurement process by having more information to identify the best suppliers and parts.

“Also, one method our teams at PartsSource are working on integrating AI is through the comparison of vendor agreement terms and conditions that can also be normalized in the pricing process. By doing so, PartsSource members that have access to our Repair and Service Agreements solution will be able to increase efficiency by using an automated process to compare vendors and consolidate them into one contract at one price. This in turn, increases not only cost savings and efficiency, but also capacity for healthcare technology management and clinical engineering teams by freeing up time and resources that would normally be used to compare and manage vendor contracts.”
--James Kollai, senior director, Client Solutions, PartsSource

“AI can assist in providing an auditing function for many of the underlying datasets that exist within the healthcare supply chain. For example, if the product unit price was added incorrectly and/or the unit of measure doesn’t match similar items then there are opportunities to proactively address data integrity concerns to eliminate the potential of downstream rework with order processing and accounts payable. Additionally, there may be opportunities to identify duplicate data and anomalies, product naming nomenclature standardization and audit data files for incomplete and/or inconsistent information, to name a few.”
--Tom Redding, senior managing director, Healthcare, St. Onge Co.

“AI could review contract activation data at the item level ,– review price and unit of measurement information versus benchmark data and generate an error message if the data is out of line.”
--George S. Godfrey, Chief Supply Chain Officer and corporate vice president, Financial Shared Services, Baptist Health South Florida

“Data integrity and data completeness and usefulness is one of the primary areas for AI to benefit the healthcare supply chain. We use AI to authenticate unit of measure data from customer files, find and remove out-of-tolerance prices, sort out mismatches between purchase unit of measure requests and price. For example, did the customer really buy a case or box of a certain product? Also, AI can be used to ensure data is properly synced between disparate operating systems like the ERP and the EHR or other supply chain tools like the inventory systems and the ERP.”
--Carl Natenstedt, CEO, Z5 Inventory

2. To adopt and implement supply data standards and escalate UDI application

“AI can automate the exchange of supply data between different systems and stakeholders. This can streamline the UDI compliance process and ensure that all parties have access to the necessary information.”
--John Freund, Jump Technologies

[AI] can find the not so obvious patterns and can also provide information in context (as it is needed).”
--Joe Dudas, Mayo Clinic

“By expanding the use of standards such as UDI, filling the data gaps and ensuring the dissemination of accurate data across hospitals IT systems cohesively and in real time.”
--Arnold Chazal, VUEMED

“FDA mandates that medical/surgical equipment obtain proper clearances before being utilized in healthcare practices. As such, manufacturers need to abide by UDI standardizations. However, in the real world, different standards, like GS1, HIBCC or UNSPEC, are adopted by various manufacturers. AI can help by lowering the impedance mismatch among various standards. AI-built and managed equivalency knowledge representations will help with proper device identification, tracking, inventory control and recall management.

“Having the proper device identity is critical to any tracking and recall management, especially if the item is serial or lot controlled. This is where universality and uniqueness come into the picture. Where AI comes in is through a decision support system, where it can facilitate the identification of all the relevant, related and affected items and their distribution across the health care provider network.”
--Vasco Kollokian, Tecsys

“AI can help through the development of compliance screenings to determine where gaps exist in UDI implementation. By doing so, it can increase supply chain visibility by identifying inaccurate or incomplete data. It can also assist in driving strategic decision-making by providing key information that could benefit sourcing strategies and other data-driven decisions. AI can also be used to streamline procurement. For instance, AI can determine where items are the same and establish standardized vendor procurement to maximize cost based on purchasing power.”
--James Kollai, PartsSource

“Adoption of data standards and solutions, like UDI, HL7, and FHIR make data more streamlined and enable for the safer sharing of sensitive data. UDI enables the tracing of data back to its provenance to ensure it is pure and has not been tampered with. AI can be of help here, for example, by:

Helping to convert non-standard data to data compliant with data standards.
Working with UDI solutions to identify, track, and trace medical devices and data.
Dynamically analyzing standardized data to identify new and emerging threats.”

--Bruce Lieberthal, vice president and Chief Innovation Officer, Henry Schein

“Using AI to implement data standards could be achieved through use of Natural Language Processing and a rule-based approach to create the necessary data structure. Without using AI, the process of implementing data standards can be time consuming and challenging to maintain. When considering how many different people are making changes daily to the ERP system and/or other subsystems, there is a need to have a reliable method to monitor and adjust data. Health systems can’t continue the current process of cleansing their data is a quarterly and/or semi-annual basis and keep up with all the data requirements.”
--Tom Redding, St. Onge Co.

“AI may assist in escalating UDI application or adoption, but not only through gathering standardized data. Rather, I think providers will see the value of broad UDI adoption as a tool to provide the better data quality needed for large AI models.”
--Jack Koczela, Froedtert Health

3. To integrate computer systems (e.g., with accounting and billing, EHR/EMR, manufacturer/distributor ordering systems and/or device and equipment technologies internally and with suppliers for service

“Among other things, AI can be used to standardize data from disparate sources, map and match data elements from different systems, cleanse and validate data, and integrate data from different systems into a single repository or federated database.”
--John Freund, Jump Technologies

“Modern interfaces (API) are built on standards that are a lot easier to implement. APIs allow real-time transactions that can speed up the supply chain as well as make it much smarter.”
--Joe Dudas, Mayo Clinic

“By being able to set and maintain data standards across internal and external systems to ensure better communication and facilitate process integration upstream in the supply chain and downstream in the clinical areas with clinical data.”
--Arnold Chazal, VUEMED

“Integrating multiple computer systems that deal with various aspects of supply chain in healthcare, ranging from accounting to electronic medical records to supplier portals, is challenging not because of the technology’s complexity. The real issue lies in the human reluctance to embrace such integration. While AI can fill in some gaps and facilitate data standardization and exchange, the underlying problem can often be attributed to competitiveness among organizations, which suffer from the ‘’Not-Invented-Here Syndrome” which resist using ideas from outside their own walls.

“If AI has to access historical ordered items and their quantities, and has real-time visibility into levels of inventory and their shelf life, is able to calculate consumption rates at various point-of-use locations, then in principle it can efficiently combine these multi-dimensional sources of information to automatically procure what is needed, when and in what quantity, controlling levels of replenishments where they are needed at the right time and cost, avoiding unnecessary expiration-driven waste.

“The adoption of AI [requires] trust. While these systems may be a certain level of maturity (or lack there-of), humans typically would like to be in control, let AI suggest and humans act. Which characterizes AI in this case as ‘Augmented Intelligence’, akin to GitHub copilot for developers from Microsoft. It also depends on whether the AI is simply ordering an incorrect amount of inventory versus delivering an ionizing dose of radiation to a patient – different consequences.

“I would characterize the role of AI – especially in this case- as ‘glue’ that comes on as a set of components, disjointedly under the control of potentially competing vendors. The overarching end-of-end value promised by AI in a domain such a supply chain, partially controlled by more than one vendor, seems to remain – at least in the near term – relatively fractured. However, as seamless access to data directly (through API) or otherwise (through extrapolation and third-party sources) becomes more prevalent, the fractured nature of AI—based value will undoubtedly coalesce, albeit at a slow pace.
--Vasco Kollokian, Tecsys

“The normalization of the data across platforms is one way the integration of AI with computer systems can be beneficial. AI can help in identifying duplicate patients, potential insurance fraud, revenue opportunities, etc. as the data becomes cleaner and more integrated across platforms. This enables pattern recognition algorithms for Deep Learning to be utilized for issues and opportunities. Also, with system integration, there will be a corresponding productivity benefit for the patients and providers as scheduling improves, spend is more visible, and staff can be better utilized.”
--James Kollai, PartsSource

“AI can automate data entry, extraction, and use in health care information systems. AI is used to harmonize data across various data formats and systems and can help healthcare systems safely share data. AI can identify and fix data errors. AI can assist in data localization from one language/geography to another. AI may be used for clinical decision support solutions that assist providers in making faster, more accurate diagnoses and therapeutic decisions. AI automates patient engagement and safe access for patients to their health care data. AI may also help patients make better decisions as to what the data means, how to act, and who to speak with.”
Bruce Lieberthal, Henry Schein

“AI is critical to ‘connect the dots’ between all the various systems and proactively identify data gaps and resolve independently to ensure a healthy supply chain. The current approach of waiting until an error occurs then addressing the error doesn’t resolve the underlying reason for the error in the first place. AI will become the ‘quarterback’ for all the various internal and external systems to be the play caller on where best to apply the necessary resources to streamline the operation.”
--Tom Redding, St. Onge Co.

“I think this type of application will more often come from a third party such as ChatGPT or other large language model (LLM) solutions. If providers adopt an LLM solution, I believe that they will be able to gain incredible efficiencies through allowing artificial intelligence to sift through the mountains of data that we already have captured in our many, but disconnected, systems. I think that AI deployed through [robotic process automation] can help link systems together. It is more efficient for the [software-as-a-service] vendors to develop a prebuilt/linked system at scale, rather than each of us developing one off solutions.”
--Jack Koczela, Froedtert Health

“AI could link to our recall system, generate messages to internal clinical departments and review contractual commitments that are missed by manufacturers.”
--George S. Godfrey, Baptist Health South Florida

“AI is best used to analyze and correct cross platform data inconsistencies. AI can identify the mismatched transactions and suggest possible corrections for human review. It can also be used to prevent future mistakes by identifying corrections to previous mismatched transactions and fixing the source data causing repeat mistakes. In this way, AI can help humans to improve their systems and generate increased efficiency in their supply chain teams.”
--Carl Natenstedt, Z5 Inventory

GHX offers quick tutorial on AI-supply chain links

Archie Mayani, Chief Product Officer, GHX, appreciates how artificial intelligence can assist healthcare supply chain performance improvement as well as how supply chain executives and professionals show growing interest in the technology. But she recommends that the industry takes this technology extremely seriously. Here’s why.

Building Generative Artificial Intelligence (GenAI) and [Machine Learning]/AI healthcare applications involves handling sensitive data, and implementing robust data governance practices is crucial. Here are some best practices for data and governance in the development of GenAI and AI healthcare applications:

Data Privacy and Security: Implement encryption and other security measures to protect genomic and healthcare data. Adhere to industry standards and regulatory requirements (e.g., HIPAA, GDPR, UDI) for data privacy.
Data De-identification Techniques: Implement robust data de-identification techniques, such as anonymization or pseudonymization, to protect patient privacy while still allowing for meaningful analysis and application functionality.
Informed Consent, Data rights and Ethical Considerations: Clearly communicate data usage policies to patients and obtain informed consent for customer data collection. Adhering to ethical guidelines to ensure responsible and transparent use of data with appropriate data rights is critical.
Interoperability Data Quality and Standardization: Implement data validation processes for high-quality healthcare data and standard data formats and coding systems for consistency and interoperability. Design applications to be interoperable with existing healthcare systems and databases. Utilize standards like EDI, GTIN, FHIR (Fast Healthcare Interoperability Resources) for seamless data exchange.
Data Governance Framework: Establish a comprehensive data governance framework that includes policies, procedures, and guidelines for data management, usage and compliance. Regularly review and update these policies to adapt to changing regulations.
Transparent Explainability: This is most critical! Ensure transparency and explainability of AI algorithms. Provide clear documentation on how algorithms operate and make decisions, especially in applications involving patient care.
Continuous Monitoring and Compliance: Implement continuous monitoring mechanisms to help ensure ongoing compliance with data protection regulations. Regularly conduct risk assessments and audits to identify and address potential vulnerabilities.
Collaboration and Stakeholder Involvement: Collaborate with relevant stakeholders, including healthcare providers, researchers and regulatory bodies. Involving stakeholders helps ensure that the application aligns with industry standards and addresses specific needs.

By integrating these best practices into the development and deployment of GenAI and AI healthcare applications, developers can ensure the responsible and ethical use of genomic and healthcare data, fostering trust among patients, healthcare professionals and regulatory bodies.

Healthcare is still early in its journey of leveraging AI to transform the supply chain. AI will help providers in a multitude of ways moving forward; however, it is still important for organizations to tread carefully in these new waters. While providers will consider how and why they should implement AI, it’s important they identify a trusted resource as they embark on this innovation journey. Technology partners have an opportunity to act as an enabler for providers implementing AI, leveraging the right technologies to solve customer problems, whether that is to reduce clinical burnout or driving overall value-based care.

AI and healthcare: A courtship of convenience, convergence

For better or for worse, artificial intelligence and all its ilk, including generative and machine-learning versions, have captivated the hearts, interests and minds of consumers and industries everywhere – healthcare being no exception.

From a content perspective, educators fear (but students love) its ability to craft essays and research papers; Hollywood screenwriters despise (but studios secretly favor) its ability to pen scripts; media outlets quiver about promoting AI-created photorealistic fake images and videos and issue codes of ethics as guidelines for use; corporate executives appreciate its ability to collect and crunch data for analysis even as they establish usage guardrails for employees; and healthcare organizations espy a litany of clinical, financial and operational benefits – from providers to payers to suppliers – even as they worry about burgeoning cybersecurity threats.

During the last three years, the global AI industry reached $240 billion in value and 250 million users worldwide, according to Altindex.com, which also projects the global AI market to exceed $500 billion in value by 2027.

Based on the quantity of media releases that Healthcare Purchasing News has received during the last year alone, healthcare seems quite enamored with the promise that AI brings. A curious headline from the public relations segment of the media industry speaks volumes: “Made by Humans: The New Gold Standard for Marketing and PR Output?”

What follows are a sampling of curated headlines and media release promotions with live links trumpeting noteworthy applications, some delicately seasoned with a pinch of hyperbole.

“Getting technical with US consumers on healthcare,” https://7796197.fs1.hubspotusercontent-na1.net/hubfs/7796197/GRAPHICS/23_CAR011_Consumer_Survey_Infographic_F%20%281%29.pdf

“AI is revolutionizing the visualization of heart conditions, enabling doctors to predict the likelihood of a heart attack or the development of cardiovascular disease like never before.”

“AI technology is playing an increasingly prominent role in clinical trials in the U.S., with large biotech and pharma companies training AI to find suitable trial patients.

“To increase patient participation diversity, especially with African American participation in clinical trials as low as 4.8%, drug giants are looking toward AI for success.”

“To streamline cardiac monitoring and virtual telemetry, InfoBionic, a digital health company providing AI-powered diagnostic remote patient monitoring solutions, recently announced its collaboration with Mayo Clinic to incorporate its extensive know-how in cardiac patient monitoring, AI-ECG, and virtual telemetry to optimize and build upon its existing monitoring platform.”

“IBM Study: Businesses Work on Adapting to Generative AI, Hybrid Cloud,” https://www.techrepublic.com/article/ibm-hybrid-cloud-user-survey/?utm_source=Sailthru&utm_medium=email&utm_campaign=e312_11.21.23_DT_Update_Paylocity_Spotlight&utm_term=daily-tech-insider-active

“Premier Weighs in with FDA on Use of AI in Drug Manufacturing,” https://premierinc.com/newsroom/policy/premier-weighs-in-with-fda-on-ai-in-drug-manufacturing

“AI is helping translate disparate medical records.”

“Avoiding the GenAI hype machine,” https://www.oliverwyman.com/our-expertise/perspectives/health/2023/november/avoiding-the-gen-ai-hype-machine.html?utm_source=dynamics-marketing&utm_medium=email&utm_campaign=owhealth-newsletter&utm_content=17-nov-2023&utm_id=cmp-15851-j9l8k4#msdynttrid=OQshi39mSGBlQBf2_6KRrA9CwrOtUdAXoL8qQ7KUF8g

“Free AI health assistant app that allows you to scan your face and eyes and measure vitals with your phone camera.”

“The potential power of AI in bridging the gap between our lifespan and healthspan.”

“FDA Launches Pilot Program to Help Further Accelerate Development of Rare Disease Therapies,” https://www.fda.gov/news-events/press-announcements/fda-launches-pilot-program-help-further-accelerate-development-rare-disease-therapies