AI Protects Patients From Dangerous Drug Interactions

By John Jeffay, NoCamels December 05, 2022 5 minutes

Courtesy Polina Tankilevitch at Pexels

Algorithm tackles problem that hospitalizes 12M people a year in US

Millions of patients are taking combinations of drugs prescribed by their doctor that could be doing them more harm than good.

A startup in Israel is tackling the problem, and saving lives, with a sophisticated algorithm that can analyze the almost infinite number of possible conflicts.

Ask any doctor or pharmacist and they’ll tell you about the minefield called “drug interaction”. That’s when a medicine designed to resolve one problem causes an adverse reaction with another medicine designed to resolve another problem. Or with anything else the patient puts in their body.

Doctors constantly have to check that prescribed medications won’t interact to harm their patient. Courtesy RODNAE Productions at Pexels

MDI Health has created what it calls an ‘artificial pharmacology intelligence platform’, which understands the vast and complex network of relationship between drugs and an array of lifestyle factors, and is able to flag concerns in a matter of minutes.

The company was founded in Tel Aviv in 2019 to combat the increasing number of deaths and the costs – both personal and financial – associated with adverse drug interactions.

Every year they collectively lead to 12 million hospitalizations in the US, and cost $528 billion.

“MDI Health has made it possible to improve the quality of medication reviews and intervention via its artificial pharmacology intelligence, which takes into account individual data such as lab results, lifestyle factors, and polypharmacy information (the use of multiple prescription multiple medications), and does so at scale, saving and improving the lives of millions,” says Brad Margolis, Chief Operating Officer at MDI.

Many patients take several medicines for a range of conditions. Courtesy Polina Tankilevitch at Pexels

The problem of drug interactions is more common that you might imagine. Anyone over 65 in the US is probably taking at least five prescription medicines.

A typical example is a patient with diabetes, high blood pressure and epilepsy, for example, who could be taking several medications for each condition.

Their doctor may be aware of some possible drug interactions, and will check online directories for others, but to fully investigate the drug regimen of each patient, especially those with multiple conditions, it could take hours.

And there’s an array of other factors that could all affect the way a patient reacts to any mix of medicines. Their lifestyle, diet and alcohol consumption could all make a difference.

Navina — A comprehensive review of a patient’s medical regimen could take up to four hours. Deposit Photos

A surprising number of patients are non-compliant – they say they’re taking their meds as they should, but they aren’t. Add into the mix the fact that every medicine comes with a long list of possible side effects, from relatively minor (nausea, dizziness, skin rash) to very major (abnormal heart rhythms, internal bleeding, suicidal thoughts).

Doctors simply don’t have time, or possibly the full set of skills, to comprehensively assess each patient’s medication.

“These days a physician has 10 or 15 minutes with a patient every month or so,” Avishai Ben-Tovim, MDI’s CEO tells NoCamels.

Sign up for our free weekly newsletter

He says doctors do their best to gather the relevant information, but to do a proper, in-depth analysis could be a three or four hour job. They rely instead on online alerts, designed to flag possible interactions. But he says they throw up 80 per cent false alerts, flagging “problems” that aren’t problems, which makes them neither effective nor practical.

There are online directories to check possible interactions, but they throw up many false alerts.Courtesy Ivan Samkov at Pexels

He says 275,000 people in the USA are currently taking a combination of drugs designed to help them, but which, he says, “it is probably going to kill them”.

He says MDI’s technology offers a unique solution. “Other systems look at one-to-one drug interactions which account for only 20 per cent of the problem,” he says.

“We take account of all possible permutations of both drugs and personal information including genetics. We finally solve the problem of many different drugs interacting with different types of patients.”

The algorithm that powers MDI incorporates a mass of anonymized patient data, together with expert input on the 400 drugs most commonly-used outside hospitals.

Millions of patients take at least five medicines a day. Courtesy Pixabay

“The beauty of our algorithm is that it’s evidence-based, unlike other AI systems, which are ‘black box’ – they say trust us, but don’t explain why. With our algorithm we provide the literature behind the conclusion.”

It looks at all the medication that a given patient is taking, analyzes it on a pharmacological basis and decides whether whether changes need to be made.

“We first provide something which is laser focused, and actionable, or details exactly what needs to be done,” says Ben-Tovim.

That translates to very practical outcomes. The algorithm will, for example, say a patient should stop taking a certain drug, or alter the dose, or switch to another drug.

The algorithm finds that around one in 20 of the general population, aged 18 to 65, are taking a combination of drugs that could cause them harm. That figure rises to one in five among the over-65s and those with serious illnesses.

The MDI Health team that developed AI to prevent drug interactions. Courtesy

It took a huge amount of work to develop the algorithm, but it is now up and running.

“We are currently live with our first clients both in Israel in the US, and working with one of the top three health insurance companies in New York,” says Ben-Tovim, but has plans to expand further.

“Our system will help physicians and clinicians to take the most personalized and optimized decision regarding medication treatment.”