Reducing device downtime using actionable intelligence on AWS
Blog guest authored by Michael Petrillo of Becton, Dickinson, and Company (BD)
Overview of Solution
There’s never a good time to take an in-service device offline, especially when healthcare practitioners depend on the device to deliver care to patients. Yet, to keep customer devices operating at optimized throughput, performing maintenance is essential and often requires temporarily taking devices offline. Inherently, the conflicting needs for continuous device access and the need to perform maintenance result in a lack of alignment.
At Becton Dickinson (BD) Medication Management Solutions (MMS), our data science operations (DataOps) team set out to solve this challenge and ensure that our customers have continuous access to the tools and services they need helping to ensure that patient care remains uninterrupted.
The answer, we hypothesized, was a more proactive, data-driven approach to maintenance that enables our technicians to spend less time diagnosing the issue and more time helping our customers. Upon discussing our hypothesis with our technical team, the response was overwhelmingly positive–labelled a game changer by executive leadership.
“When our field service technicians demand to know when they can start using the tool, we’ll know we’ve got it right.”―Michael Petrillo, Lead Data Scientist at Becton Dickinson
The first criteria we outlined was the need to perform this new method of maintenance on devices already in the field. We turned to the guidelines of a connected device strategy (Porter, 2015) to turning device performance metrics into actionable insights. In order to deliver value to our customers with existing devices, we sought to leverage technology already integrated into our on-market devices
With a clear direction in-mind, we saw how we could use Amazon Web Services (AWS) to help us modernize our approach to medical device maintenance and customer care. Our first point of engagement with AWS was the utilization of Amazon SageMaker for data exploration. Understanding patterns in our underlying device performance metrics help turn the data models into persuasive insights to help impact field operations.
With a data driven direction formed, our DataOps team began deeply engaging with our field technicians and customers to fully understand both the requirements for uptime and the need for performance maintenance, which informed our solution detailed below.
Developing a data-driven solution to reducing device downtime
We selected AWS for their service of stackable offerings, which we felt could complement our approach. “Leveraging the services and tools AWS offers, we only need to code a fraction of what we would have needed to otherwise, saving us so much time.” Warren Hall, Software Engineering Manager.
Working backwards from our customers and field representatives, the team developed Machine Performance Assessment Report (MPAR), a solution on AWS that provides insights into potential fleet disruptions prior to an on-site customer visit.
Our workflow is outlined below:
- Integrating data: Using Amazon AppFlow, we integrated our SalesForce.com workflow event data, which provides visibility into when a new work order is assigned to a field technician.
- Generating insights: Once data has been aggregated from SalesForce, Amazon AppFlow triggers an AWS Lambda function, which then gathers insights by utilizing machine learning outputs. As soon as results have been generated, AWS Lambda aggregates the results and translates them into plain-language reports that can be understood by our field services team.
- Proactive opportunities: A second AWS Lambda organizes thoughts and actions as experiments for improvement for customer support. This model is used by our teams to help prioritize opportunities for proactive device maintenance that can then be performed while the technician is on-site at the customer location. These insights are included in an email to the assigned field service team member.
- Identifying usability issues: Our team leverages Amazon Athena to run ad hoc queries against device telemetry data. The insights are then populated into an Amazon QuickSight dashboard for the DataOps and Global Support teams to evaluate. Data generated from this step is used to better understand fleet performance and identify ways to further enhance products by identifying potentially unreported usability issues.
- Usage: MPAR deploys hundreds of emails daily to field service teams with actionable insights on how to help our customers. During a visit at one facility our field team performed at a rate which had never been done before. Customers and BD are pleased by the impacts to device downtime and field service teams.
“Losing time away from patients, when a nurse has to trouble shoot a device, creates debates about resources. This stops that.”―Michael Petrillo, Lead Data Scientist at Becton Dickinson
Leveraging AWS’ portfolio of stackable offerings helped us to bring this project from concept to launch more quickly than normal. “We only code the parts that aren’t already done for us (by AWS) and that saves us so much time.” Warren Hall, Software Engineering Manager. “That’s the real power of AWS—the offloading of all the small, mundane, yet necessary tasks. It’s like Legos and you can build great things for pennies on the dollar—that’s the beauty—do it quickly, at a low cost, so we can focus on delivering a better customer experience.”
Now 12 months after releasing dashboards, and months of actionable intelligence email reports, we have seen substantial adoption by our field-services teams.
Implementing our new dual approach of data-driven reactive diagnoses and proactive improvement identification has delivered faster maintenance for increased device up-time and material savings for BD and our customers.
We are looking to the future with AWS because with AWS, we have been able to help our teams and customers.
To know what AWS can do for you contact your AWS Representative.
- Predictive Maintenance Using Machine Learning
- AWS for Healthcare at HIMSS22: Healthcare Analytics & AI/ML
- Philips Uses AI and ML to Improve Medical Imaging Diagnostics for Philips HealthSuite Built on AWS
Porter, Michael E., and James E. Heppelmann. “How Smart, Connected Products Are Transforming Companies.” Harvard Business Review 93, no. 10 (October 2015): 97–114.
BD’s operations consist of three worldwide business segments: BD Medical, BD Diagnostics and BD Biosciences. BD Medical produces a broad array of medical devices that are used in a wide range of healthcare settings. BD Medical’s principal product lines include needles, syringes and intravenous catheters for medication delivery (including safety-engineered and auto-disable devices); prefilled IV flush syringes; syringes and pen needles for the self-injection of insulin and other drugs used in the treatment of diabetes; pre-fillable drug delivery systems provided to pharmaceutical companies and sold to end-users as drug/device combinations; regional anesthesia needles and trays; sharps disposal containers; and closed-system transfer devices. The primary customers served by BD Medical are hospitals and clinics; physicians’ office practices; consumers and retail pharmacies; governmental and nonprofit public health agencies; pharmaceutical companies; and healthcare workers.
At Becton Dickinson, Michael Petrillo works as a Data Scientist in a DevOps group with a passionate about improving healthcare with automation. Michael’s focus is in utilizing existing data to improve device capabilities. In his spare time, he loves to collect Legos, trying to be a frugal Foodie and enjoys to travel with his wife and young children.