A former NSA Intelligence Officer wants bots in every hospital to tackle the opioid epidemic
Almost three million Americans are addicted to opioids, an epidemic which continues to damage communities and claim lives across the nation.
Provisional data for 2017 suggests that the number of opioid-related overdose deaths rose from 42,000 in 2016, 33,091 in 2015, and 28,647 in 2014.
In October 2017, President Donald Trump declared a 90 day public health emergency to mobilize the federal government to tackle the opioid epidemic.
The declaration ran out on January 23rd, 2018, and opioid addiction is still widely considered to be an epidemic.
Sean Lane, a former US Air Force Intelligence Officer, who served five combat tours in Iraq and Afghanistan, believes that AI could be a game changer in combating the opioid epidemic.
His startup, CrossChx, has developed an AI bot called Olive, which can be deployed in a hospital’s electronic medical record (EMR) system and put to work as a hospital employee.
The bot can use almost any software by using computer vision technology to look at and understand the user interface (UI), and so can do many of the tasks that humans currently do, like checking insurance eligibility, prior authorizations, and scheduling.
Olive has two key advantages over a human, beyond the fact that it can work faster and more consistently.
Firstly, it learns globally. In the same way that when one self-driving car crashes, all self-driving cars around the world learn from the mistake, so too do all Olives in every hospital around the country learn together.
Olive’s other great strength is that it is armed with a comprehensive awareness and understanding of who a patient is globally based on enormous amounts of data on eight million patients that CrossChx gathered from 250 health systems in the US.
If a patient appears in two different hospital systems, then Olive can cross-check the names, social security number, date of birth, address, previous address, and even hair or eye colour to discover whether it is the same person.
By matching up these details, Olive could proactively identify potential problems like ‘doctor shopping’ and alert hospital staff.
‘Doctor shopping’, when addicts visit several ERs until eventually they get their prescription, precedes 21.4% of unintentional prescription-drug overdose fatalities, according to a study by the Centers for Disease Control and Prevention (CDC).
“The problem is we can only detect when someone has been prescribed the drug,” Lane said, “You don’t know that they’ve attempted to get the prescription 14 times before they were actually prescribed”.
This issue is personal for Lane, who is from a small, rural town in Gallia county, Ohio.
In the state of Ohio, there are more than 10 opioid overdose deaths per day, and Gallia county is one of Ohio’s poorest regions.
In terms of the national average it is 45th in college affordability and 22nd in education, but in the top five for opioid deaths.
Sean said, “I had friends and family who were affected by it. I had friends I went to high school with who overdosed and died.”
The Gallia county local government is one of many in the US who have filed nearly 200 lawsuits against drug companies, which are blamed for fuelling the nationwide opioid epidemic by flooding local communities with prescription drugs.
Sean, who has both an award-winning tech background (he was named Ernst & Young’s Entrepreneur of the Year in 2011) and a career in the National Security Agency (NSA), felt he had a responsibility to investigate the issues plaguing Gallia county.
“I wanted to know why. Why can’t we track an individual and where they’ve gone to the doctor? That was my number one, first question: how come we can’t know that?”
He knew the CEO of the local health system in his hometown, and asked for permission to peek into their setup.
It didn’t take long to see that the gap in the healthcare system was the internet. There was no internet for healthcare.
“Siloed systems, enterprise systems that aren’t connected…There’s no entity extractions, they’re not doing identity of patients, people, places and things. These are really big problems, but they’re symptoms of the bigger problem: you can’t really know any information about a patient beyond what’s in one specific hospital.
“It is really emblematic of where healthcare is today”, said Sean, “which is nothing like where it should be. It’s not modern in any way from that perspective, it should be very connected. And I started to think: how do we connect this? How do we make the internet for healthcare?”
That was when Sean and CrossChx built their database, healthcare’s equivalent of a routing table – a set of rules used in all internet enabled devices that decides where data packets will travel to.
Once patients were in the system, it became clear whether they had records at several different hospitals, and problems like ‘doctor shopping’ could be identified.
But even though the problems in the healthcare system were laid bare, there was still no capability to take any preventive action to improve things.
“We knew where people were, we knew how to identify people. We had the information and the knowledge, but we had no way to act upon it,” Sean said.
Sean’s team realised that the best way to enable the system to take actions was through an AI bot.
“The reason is, you’re not just going to replace the electronic medical record (EMR) systems that exist in healthcare. It’s just like any other big enterprise system – they’ve spent an incredible amount of money on them and they do a lot of tasks.
“This idea that we’re going to go in with some sexy user interface (UI) and just completely disrupt the Epics, Cerners, and Meditechs of the world is probably silly.
“What we decided was, let’s just use that software. Let’s have an AI bot log in like a human would.”
And so, Olive was created to fulfil the role of the AI bot which could monitor and optimize patient verification.
The system was in place for the whole of 2013, and the opioid doses in Sean’s home town dropped by 16.2%
The litmus test for CrossChx’s AI system was whether it could help to impact the opioid crisis in Sean’s hometown in Gallia county.
The team installed their identity solution, along with a biometric fingerprint reader, into every hospital in the local area.
People had to use their right index finger to verify themselves into a provider for a visit, and Sean says there are still thousands of these fingerprint readers out in the US.
The system was in place for the whole of 2013, and the opioid doses in Sean’s home town dropped by 16.2%. This was the biggest reduction in the state that year, remarkable considering Sean’s town was typically increasing opioid doses, not decreasing.
Sean said, “It was jaw dropping for the folks at the Ohio Department of Alcohol and Drug Addiction Services (ODADAS), and they called me and asked, ‘Did you see the numbers? I gotta show you this’.
“I’m not saying that there’s a direct causation necessarily between what we did and the drop of opioid doses, but I do think it had an impact, I do think it was part of the solution.”
Gallia county’s opioid woes have continued to worsen as powerful drugs like fentanyl and carfentanyl become increasingly prevalent.
Sean believes answers could be found in wider distribution and development of bots like Olive.
“Artificial intelligence needs its ‘personal computer moment’. I would like to see a million people have access to AI and a development platform, so they can build bots that we can’t even imagine today.”