When Richard Eden arrived at the online doctor service DrEd: “their data system was pretty much in its infancy,” he remembered. “They basically had to write SQL on the production database to download data.” Unfortunately, every time they did so, it slowed the website down to the point that the doctors couldn’t even look at patient records and were nearly unable to do their jobs.
To remedy this situation Eden, with the help of another person, built a data warehouse using Redshift. Big data analytics platform Looker sits on top of that. “Three quarters of our staff have access to Looker at various security levels, so not everyone can see patient data. It’s all in AWS and all the machine learning stuff is done using Python,” he said.
Eden explained that the introduction of Looker has been revolutionary. “Looker has transformed the business, nothing short of that. It’s suddenly given everyone the ability not just to get reports instantly but create more interesting things and create analysis I haven’t even had the time to think of,” he said.
All that is left for Eden to do is to make sure that sensible data is organised in a sensible way. It is then left up to the other people in the business like the marketers and doctors to do the “cool stuff.”
The ability of the 100 or so people working at DrEd to access and analyse the data themselves is crucial to Eden as he is part of a very small team of two. Three months ago, he recruited a data scientist, hence his rise to the “heady heights of head of data.”
“We really do take security very, very seriously.”
The fact that they are dealing with medical records means the business takes security “very, very seriously” and so there are no names, addresses or email addresses in Looker.
When he first started, Eden said the focus was on business metrics and KPIs because “if you don’t have that, you don’t have anything.” As time went on, with online doctors being a new area of medicine closely watched by the regulators, the focus shifted to clinical safety.
“We can see all the information coming in from the patients, all their bio-metrics, how they’ve answered questions, see the decisions that people are making and compare different doctor’s practices to each other,” said Eden. That way, they can ensure that doctors are prescribing safely and learning from each other. But also to question a doctor who might be prescribing the contraceptive pill at a higher rate than others , for example.
“There is a lot about checking that doctors are making safe decisions.”
“There is a lot about checking that doctors are making safe decisions but also setting up alerts directly in Looker. If a doctor were to break an obvious rule, like a maximum age for a treatment, instantly various senior people in the company get an email alert,” he said.
They also have aa preventive system in place to avoid certain patient behaviours, as some shop around for treatment by creating duplicate accounts. Eden said this is a problem with emergency contraception with some patients attempting to get prescriptions for it weekly.
“We were inspected by the CQC in 2017. They said it was the first automated alerting system they had seen."
DrEd has received praise in the medical industry for this alert system. “We were inspected by the CQC in 2017. They said it was the first automated alerting system they had seen. It just blew them away. I heard they were basically saying they want other people to set this up because it should obviously exist. That was good,” said Eden.
One of the effects of the data team’s data analysis has been to help the marketing team understand how realistic their targets are. “We built various machine learning models that take into account the day of the week, closeness to payday and bank holidays and predict sales each day.”
Eden also used Looker to create some charts that quantified the frequency of late deliveries and showed it to the head of operations. Within a month or two, they ditched and switched to a new delivery partner, thanks to the agility of the company.
If Eden had a magic wand, he would like to bring everything they know about each patient together into one easily digestible place for doctors and be able to assist their decision-making in a smarter way. He would like Dr Ed to be even better at flagging up anything suspicious. When a patient gives their height, Eden would like to plot that against the normal distribution of heights in the general population. “Rather than self-driving cars, it’s more like driver assist. I want doctor assist.”
Richard Eden was speaking to DataIQ at Looker’s Join The Tour conference.