Artificial Intelligence at the Top Medical Device Companies

Niccolo Mejia

Niccolo is a content writer and Junior Analyst at Emerj, developing both web content and helping with quantitative research. He holds a bachelor's degree in Writing, Literature, and Publishing from Emerson College.

Artificial Intelligence at the Top Medical Device Companies

Predictive and prescriptive analytics seem to be the most common types of AI applications at the top medical device companies. These companies, particularly Becton Dickinson, Medtronic, and Fresenius Medical Care, tend to partner with AI firms and vendors to create software that they intend to sell to hospitals most often.

In this article, we present the current AI initiatives at Becton Dickinson, Medtronic, and Fresenius, three of the top medical device companies in the world, as well as the AI firms they’ve partnered with to make their AI offerings possible. These projects include:

  • Becton Dickinson, their partnership with BERG Health, and how machine learning could help monitor patient condition and care practices.
  • Medtronic, their collaboration with IBM Watson Health, and how predictive and prescriptive analytics could help patients manage diabetes with a mobile app connected to a glucose sensor.
  • Fresenius Medical Care, their clinical trial for predictive analytics software intended to manage symptoms of chronic kidney disease (CKD), and demonstrations of their research and development.

The first medical devices company we discuss is Becton Dickinson, who has recently integrated an AI application into their MedMined “surveillance” offering.

Becton Dickinson: Predictive Analytics for Patient Monitoring

Becton Dickinson (BD), one of America’s largest and oldest medical device companies, offers patient monitoring software under their MedMined brand. Solutions include what the company calls “surveillance,” although it is unclear what they mean by this because it’s unlikely that hospitals will set up cameras in patient rooms. The company’s software seems to help caregivers identify signs of healthcare-associated infections (HAIs) and conditions, complete with alerts so personnel may react quickly.

In 2017, Becton Dickinson signed a multi-year agreement with AI firm BERG Health to collaborate using BERG’s predictive analytics solutions. The collaboration produced an algorithm made to assist healthcare workers with recognizing patient populations with the highest risk of not following their prescription medication regimens once discharged from the hospital.

BERG Analytics, a branch of BERG Health, claims this solution will help clinicians prioritize patient interventions and find the appropriate outside staff to improve outcomes post-discharge. BERG Analytics’ AI-based platform is purportedly used to gather what the company refers to as “patient intelligence,” which is likely patient information regarding their biometrics and medical history.

This data would then be used to create risk models and visualizations for showing statistics on how adherent patients of chronic and acute conditions are to their medications.

Regarding the collaboration and its purpose, Slava Akmaev, Chief Analytics Officer at BERG said, “we look forward to advancing clinical decision making through predictive analytics. This pro-active surveillance solution will drive targeted medication adherence programs and ultimately, improve patient outcomes.”

BD’s solution for mining healthcare information may allow clients to detect the pathogens that pose the highest risk to their patient population and prevent outbreaks as well. The company also claims their solution can recognize trends of what they call “organism occurrence” within hospitals, which we can infer as the development of harmful bacteria or pathogens.

Similar Use Case: Drug Discovery

The AI capability of analyzing and making predictions about data from clinical samples is also used in drug discovery. Pharmaceutical companies may also need to identify pathogens relevant to a clinical trial, along with the best patients to recruit for a clinical trial. Relevant pathogens include those that are present in the daily lives of patients that cause or affect the disease and treatments being studied.

Below is a short video featuring Leila Pirhaji, CEO and founder of ReviveMed. She explains the possibilities for artificial intelligence and predictive analytics in drug discovery. ReviveMed is a 2016 startup developing AI drug discovery software using metabolomics, or data recorded from metabolic processes. In the video, Pirhaji details her claims that the company’s AI platform uses biological attributes that were not usable by earlier AI software:

Another use case Pirhaji discusses is drug alternative use discovery, or finding treatment for certain diseases in drugs that already exist. This is very similar to BD’s collaboration with BERG in that they each study existing parameters to lead healthcare leaders in the direction of the best next step, as opposed to how to react to new drug compounds and treatments.

Medtronic: Prescriptive Analytics and Mobile Apps for Managing Diabetes

Medtronic also has a dedicated diabetes branch that covers insulin pumps and a range of treatments for both type one and type two diabetes.

The company partnered with IBM to use their Watson Health platform for more detailed analytics on their diabetes patients.

Medtronic’s Turning Point program was the first of their products to run on IBM Watson Health, which is a patient proactive diabetes management program with smartphone tracking.

IBM Watson’s machine learning solutions purportedly help Medtronic analyze data from blood tests and insulin pump activity to determine risk factors for blood sugar spikes. The system also uses that data to improve future predictions as well.

Patients can log into their mobile app to receive recommendations for the best next step to take in managing their diabetes, such as changing eating habits or more scheduled exercise.

The software likely uses prescriptive analytics to analyze this data and then provide feedback to the patient about when their blood sugar spikes or dips or when they need to inject insulin.

Medtronic’s Guardian Connect CGM, a continuous glucose management (CGM) system, is likely integrated as well. The system purportedly alerts diabetics on when to inject their insulin via two smartphone apps and a bluetooth transmitter worn directly on the skin and held with an adhesive.

The Guardian Connect App displays blood glucose data from the transmitter, details insulin level trends over time, and gives patients alerts through sound or push notification.

A second app called Sugar.IQ-diabetes assistant purportedly offers patients a more detailed look at their glucose patterns across the day and which factors affect their glucose levels the most. Because they claim the app offers a “fuller picture of [one]s] diabetes,” we infer the data from the glucose transmitter is also used to reach these statistics.

Below is a video from IBM Watson Health demonstrating how their machine learning capabilities could help manage diabetes through a smartphone platform. Any smartphone app imaged in this video may have a different display and menu if accessed now:

The improvements to Medtronic’s diabetes management systems from the IBM Watson Health integration likely come from recognizing finer details and less common trends in patient biology.

This is where machine learning could improve diabetes analytics because it could recognize minor changes in blood sugar in response to easily missed activities such as a small drink of water.

For example, if a patient had made a habit of exercising at unusual times or minutely overhydrating, a machine learning solution could recognize this trend and alert a patient to it before they sustain harm.

In cases where new patient behavior results in a spike or a severe dip in blood sugar, a machine learning-based alert system may be able to alert a patient before their condition drops too far for them to be able to seek help.

As with all AI-and ML-based software, continuous intake of new data will help it “learn” as it works, and in this case come to better “understand” the factors that contribute to blood sugar spikes. Medtronic claims this collection of data also helps them discover new treatments for diabetes, as well as improvements to current treatment methods.

This could include ways to produce and dispose of diabetes-related medical goods more sustainably and thus improve best practices for insulin injection.

Fresenius Medical Care: Predictive Analytics for Managing Chronic Kidney Disease

Fresenius Medical Care is a medical device company specializing in kidney disease and care, dialysis machines, and the management of anemia. The company is home to a network of dialysis facilities, outpatient labs for cardiovascular conditions, and urgent care centers.

They claim to offer specialty pharmacy and laboratory services to hundreds of thousands of patients in North America, along with manufacturing and distributing dialysis equipment, disposables, and renal pharmaceutical products.

A clinical trial from Fresenius was published to the website of the U.S. National Library of Medicine in 2017 where the company is researching an AI application to help manage anemia in chronic kidney disease (CKD) patients.

The trial documents state it will most likely run between December 2017 and September 2019 and include 240 CKD patients. It also states that the software is made in conjunction with an algorithm trained on large amounts of patient data and can recommend medication dosages based on patient dosage history and that of people of similar body types and constitutions. This would likely be a prescriptive analytics software that takes in kidney-related medical history.

The description of the clinical trial states that the neural network used to build the machine learning algorithm “complies with the European requirements for medical devices.” Proof of concept trials were held in three Fresenius dialysis clinics across Europe.

The purpose of this clinical trial is to assess how well the software works for anemia management within a typical clinical practice setting. If the software is successful, it could prove that AI applications have a valid use case in providing care to those with CKD.

Len Usvyat, Vice President of Integrated Care Analytics at Fresenius North America, spoke about the importance of artificial intelligence to the company and their patients moving forward in a 2018 interview. When asked about how artificial intelligence can help facilitate CKD management for the provider and the patient, Usvyat said,

Peritonitis is one of the top factors for why patients are unable to continue peritoneal dialysis at home. Can we use the wealth of data at our disposal to better predict when patients will have an infection? We are starting to see some early positive results on a predictive model that looks out about a month in advance. The next phase will be seeing whether we can use those insights to actually prevent peritonitis episodes. Possible interventions may include proactively visiting patients at the highest risk for peritonitis in the next month or increased testing to identify and treat infections early.

While this quote does not refer directly to the AI application being tested in the previously-mentioned clinical trial, it is clear that Fresenius Medical Care is attempting to adopt AI applications across the CKD management workflow. Fresenius would go on to demonstrate these capabilities and other advances in kidney care at the 2018 American Society of Nephrology’s (ASN) Kidney Week Symposium.

Fresenius presented 71 abstracts for different applications of AI technology and connected home therapy methods. The company’s experts purportedly demonstrated how their AI applications could improve patient outcomes. They demonstrated the following capabilities with AI:

  • Using artificial intelligence to help predict imminent hospitalizations in patients with end-stage renal disease (ESRD). This endeavor explores the quality improving capability This endeavor explores how a predictive model trained on more than 1,500 variables may more effectively discern a patient’s likelihood of hospitalization after seven days of hemodialysis.
  • Using machine learning to help predict elevated serum phosphate levels in patients with ESRD. This project uses data analytics to predict high phosphate levels in hemodialysis patients within one month.
  • A machine learning model to predict patient risk of peritonitis episodes. This quality improvement project is studying the ability of machine learning models to predict how much peritoneal dialysis patients are at risk of a peritonitis diagnosis within the next month.

While the company is still studying the capabilities of AI in the healthcare sector and only using preliminary technology internally, Fresenius Medical Care may be able to use AI for helping patients with kidney disease.

Possible Challenges with Adopting AI at Medical Device Companies

It is important to note that incorporating this type of solution without proper preparation could disrupt a healthcare company’s workflow and lead to more difficulty before success.

Business leaders in healthcare would have to examine how this would change the jobs of their employees as well as the care they provide their patients. If too many healthcare providers are not familiar with the new systems, it could lead to slower or less attentive care.

We spoke with Shelley Zhuang, founder of Eleven-Two Capital about how healthcare companies use AI to put their patients first. When asked about what is most important about integrating AI systems into healthcare operations, Zhuang said:

I think it’s absolutely critical that companies need to clearly understand where their solution fits into existing clinical workflow…and that can only be done through a lot of interviews, a lot of meeting with key opinion leaders, and it would be great if you could get several KOLs (key opinion leaders) to get behind your value proposition; another way we’ve seen companies get product market fit is partner with an existing larger company…and to get their new technology into the market.

Coming up with a plan to integrate a solution similar to this collaboration will be key in making the process as smooth as possible. It would require discussing how to minimize the negative effects on patient care that the transition will have, while still familiarizing staff with any new systems they will need to interact with.


Header Image Credit: Wired

Stay Ahead of the AI Curve

Discover the critical AI trends and applications that separate winners from losers in the future of business.

Sign up for the 'AI Advantage' newsletter: