More than a vision: The data robot is watching you
When Jose Pedro Almeida, Head of Big Data Analytics and Machine Learning at Sao Jao Hospital Center, wants to illustrate what he will be talking about he uses an example: “At this moment, a doctor of ours might be walking to the ambulatory department to see a patient. He might be contacted by a data robot that informs him by text message that another patient of his is experiencing an increase in heart rate and a drop in blood pressure. Given that this is a dangerous condition, he might go back and see this patient first. This might be happening as we speak.”
Data robots are one possibility to turn big hospital data into action. At Sao Jao, a 1,100 bed hospital with nine floors and 5,400 employees, a big data platform was built over recently to make the kind of scenarios possible that Almeida talked about at HIMSS Europe 18 in Sitges. At the heart of everything is a big data warehouse that captures everything from lab data to medication data, vital signs, clinical notes and imaging. “We are talking about more than 560 billion data points that are waiting to be leveraged,” said Almeida.
Just throwing data into a data lake won’t do the job, though: “We need the input of doctors and nurses to built tools that can really change the course of a patient for the better.” One tool that was developed in close co-operation with the clinical specialists is the data robot mentioned in the beginning. It monitors all patients in real-time and gives alerts when certain parameters cross certain thresholds. The tool has been running for 2.5 years now according to Almeida: “We have shown that we are able to anticipate 30% of all ICU admission up to 7 days in advance.”
Another area of opportunity that is addressed by data robots in Porto is MRSA infections. The tool automatically plots MRSA cases on a hospital ward map, helping the hygiene staff to identify areas in which larger scale decontamination efforts might make sense. A third robot that is active already advises physicians on antibiotics choice based on individual patient and local hospital and resistance data.
By moving from conventional analytics to deep learning, Sao Jao Hospital Center is currently taking its data robots program to a new level. The goal is to improve risk prediction. The big data team can already show that the AI-powered data robots are getting better and better in identifying patients that are on their way to develop life-threatening conditions – based on a complex set of data that includes, among others, potassium and creatinine levels, CRP, heart rate, blood pressure, leucocytes and many other parameters.