Machine learning (ML), a type of AI in which algorithms are continually refined as additional data makes them more predictive, is taking hold in supply chain management.


Kishore Bala

First came the promises and hype about artificial intelligence (AI) in healthcare. Then came its practical application in select areas such as imaging and population health. Now, machine learning (ML), a type of AI in which algorithms are continually refined as additional data makes them more predictive, is taking hold in supply chain management.

The supply chain in hospitals hadn’t received much attention from healthcare executives until recently, when narrowing margins has made driving efficiency in all aspects of hospital operations essential. Because it’s highly complex, running both wide and deep, it can be hard for executives to grasp its importance and understand how to better manage it. Both operational and clinical leaders today need a better understanding of what supplies are stored in inventory, what’s consumed, how much is on hand and what was charged. These distribution, warehouse and clinical components are critical to making healthcare more efficient and less costly.  A study by Navigant has found that better supply chain management can cut supply costs by over 17% for the average hospital, and ML-enhanced supply chain management can accelerate those savings.

ML can help to make supply chain management more sophisticated while making it more manageable. By rapidly processing vast volumes of data to detect trends and reveal insights that are too complex for the human mind, ML has the potential to enable us to consistently deliver the right supplies at the right time, place and cost.

ML makes it possible to digest all the varied inputs that change dynamically, hour by hour and day by day, and determine their trends over time—the procedures, surgeons, techniques, and regulations all change. It enables algorithms to become continually smarter, better able to make sense of the mountains of data and the large amount of variability than humans can. Before we had ML, variability was too complex to be effectively managed. But because we now have the capability for the algorithm to become stronger, not only with more data but with more input, the potential holds great promise.


Six supply chain management areas where ML can make an impact

ML and analytics can make a significant impact in six areas of supply chain management:

  • Cost-per-case capture
  • Inventory level optimization
  • Supply standardization
  • Perioperative and procedural throughput
  • Labor efficiency
  • Expired and recalled supply management

Cost-per-case capture

In a fee-for-service world, hospitals could survive without knowing their true cost per case, since they could typically pass through volume or cost increases to health insurers and consumers. As value-based reimbursement models such as Bundled Payments for Care Initiative (BPCI) and BPCI-Advanced become more prevalent, however, capturing accurate cost-per-case data becomes critical to ensure that hospitals don’t underestimate their costs and jeopardize their margins.

But calculating total cost-per-case can be difficult, as operating room staff may neglect to scan one or more supplies or a physician may modify the procedure after it’s underway. Using an AI-enhanced program can reduce variation and capture true case costs by alerting staff when expected supplies aren’t on the consumption report. The program can also calculate consumption data in near real time to provide rapid cost feedback to surgeons and proceduralists.

 Inventory level optimization

The traditional method of forecasting inventory demand has been the time-series method, which predicts future demand by analyzing past patterns. However, this method enables a few data points to skew the prediction and fails to account for seasonal variations and other variables. That leaves hospitals vulnerable to avoidable waste and high costs.

Basic statistical and mathematical approaches aren’t proving adequate to optimize supply chain management. ML-enhanced algorithms can provide more accurate cost variance analysis as well as procedure and inventory demand intelligence. The software continually updates its model to incorporate additional data and create more refined projections.


We developed an ML-enhanced demand forecasting methodology that combines four advanced algorithms: Autoregressive Integrated Moving Average (ARIMA), simple exponential smoothing, Holt-Winter method and moving average method. Approaches such as these are more accurate because they show the range of usage as well as the mean usage from one time period (such as a month) to another, enabling users to more clearly see how their actual usage deviates from the forecast.

Supply standardization

The holy grail of supply chain management is standardization of supplies across the entire health system enterprise. That’s no small task when Provider Preference Items (PPIs)—e.g., items such as prosthetics, drug-eluting stents and orthopedic implants—account for a whopping 56% of the total supply spend. PPIs have proven notoriously difficult to impact, in large part because data-driven providers are reluctant to change their ordering habits without credible evidence that patient outcomes won’t be harmed by switching to a standard product.

ML-enhanced supply chain management gives providers the rapid feedback they need to trust the data and take action. It enables them to quickly see the relationship between supply variances and patient outcomes so that health systems can meaningfully reduce costs while preserving or enhancing quality. Since the price of costly supplies like knee implants can vary by a factor of four, each high-priced implant can cost the health system thousands of additional dollars. One health system found that three of its physicians alone accounted for $750,000 of additional supply costs that were not driving better care.

A hospital with $70 million in supplies can potentially save nearly $2 million by standardizing perioperative and procedural supplies and an additional $400,000 by standardizing warehouse supplies.

Perioperative and procedural throughput

Effective supply chain management can also play a critical role in making the lives of care teams less frenetic while increasing throughput in the OR and procedure areas. It lessens the number of ‘supply safaris’ in which nurses must leave the procedure room to fetch a missing supply or exchange the wrong equipment for the right one. That improves their working environment as well as efficiency. The health system can increase the number of cases it can schedule. Further, patients benefit when procedures aren’t delayed because a supply is missing and when the OR consistently runs on time. That can impact not only their satisfaction but potentially their safety.


Labor efficiency

Staff can save critical time when ML can efficiently process and assess vast amounts of supply data to help them analyze trends and accurately predict optimal inventory levels. Using ML-enhanced analytics, we’ve been able to help hospitals save nearly $400,000 a year due to improved supply chain management labor efficiency.

Expired and recalled supply management

Improving the identification and management of supplies that are past their expiration date or that have been recalled affects roughly 10% of total supplies. ML can quickly identify expensive supplies that have been recalled or expired before they end up in the OR where they could harm a patient or delay a procedure. For one 342-bed Mid-Atlantic hospital, applying ML in this way saved them an estimated $2.25 million annually.

Everyone benefits from ML-enhanced supply chain management. Clinicians can spend more time on patient care and less time hunting down the right supplies, patients get safer, more efficient care, and administrators and executives can reduce costs and improve the value they deliver.

Kishore Bala is chief technology officer at Syft. Corp. Connect with him on LinkedIn.