Hospital Business Decision Support System for Improved Operations

Below is the continuation of the transcript of a webinar hosted by InetSoft on the topic of Using Analytics to Increase Staffing Productivity and Improve Hospital Operations. The presenter is Abhishek Gupta, Chief Data Scientist at InetSoft.

For the hospital business decision support system, the data came to us from Oracle. Originally we had a PM project product, and we purchased that product to support decision support. However the users throughout the systems, all the financial analysts and analysts and management wanted to have data that they can use InetSoft on top of. Management had decided to put the data into Vertica and put InetSoft on top.

My team has integrated data from 38 large tables combining 800 columns into one huge integrated table with millions and millions of rows and our users today just put InetSoft right on top of it. The performance is very well received. I am talking about seconds not minutes. If we have some complicated report that generates for 19 seconds, it will break our heart, because it against everything we believe.

To use this technology for this particular business intelligence project is really complicated. At first we used SQL and Microsoft SSAS, but it didn't work. We used InetSoft, and it worked. We had a question: how do we integrate large tables together not summarized but in detail format in a table with 800 columns and hundreds of millions rows and InetSoft on top. My team worked together, and we utilized technology that Vertica has.

Originally we ran out of memory all the time. The server hung up because the memory was not enough. The capacity was not there. We learned quickly that we needed to utilize a parallel process than the Vertica has. We break up the data, and it runs 16 or 17 times at the same time. We break the queries into quarterly or monthly and push all the queries into Vertica and run all at once.

Originally this particular task would take two weeks. After my team took it over, it went out to five days. Today it is a one-day process and two-day data validations and one-day for production. And thus we processed three years of data for 16 hospitals. Once again our goal will be just three days, if we have the right technologies.

The storage is amazing, the data is huge, but the data compression on Vertica is amazing. Right now the ratio is 1:7, so the seven times the size of the data. Once it finished processing, it will be the size of one. So the ratio we have is 1:7. The Sepsis project is using AI, the IDOL project that we purchased from Micro Focus, large data from Vertica and InetSoft on top. As you know that IDOL is an AI product, which is great topic that we are interested in.

View a 2-minute demonstration of InetSoft's easy, agile, and robust BI software.

Our next project for this IDOL project would be Stroke or Denial. Using the functions that AI provides we would be able to read a lots of text data, listen to notes, in addition to a structured data we were able to map. Every single hour in a 24 hours day, every hour and every 15 minutes represent a single rectangular box. The blue colors mean the patient doesn't have the sepsis symptoms. The white color means we did not have data or maybe the clinician did not document it in the EMR. The red colors means the patient has a sepsis condition.

The challenge that we had with this project would be that the AI tool is complicated and is new to us. And the request to do this project from the first proof-of-concept is really a challenge for our team. We have the support from the management team and physicians and residents from the ED. And they help us with the logic, not just the logic, the beginning of the logic is simple, but somehow you need to read and help AI to think and make that kind of decision and help us with information.

The screen you see on top, the patient on the left which we covered, the information of each patient for 15 minutes for 24 hours a day, the red, green and yellow towards to the left-hand side is how many minutes you have left in your three hour bundles. The physicians help us to teach IDOL. We display this information on their computer. Usually they have a quick screen and determine this patient yes, has sepsis, or no, he doesn't. Every time the physician indicates that this is not a sepsis patient, information is fed back to IDOL, and IDOL will learn and improve for the next time.

View live interactive examples in InetSoft's dashboard and visualization gallery.

And now I would like to send it back to Abhishek, to share with you more about InetSoft product. Abhishek?

Thank you so much for that great presentation, it's really exciting all the work you are doing and the Sepsis Project in particular. I know sepsis is a really challenging disease to find out when patients come in so being able to use AI and Machine Learning and analytics to catch those cases that is just incredible. For those on the call who are not familiar with InetSoft, I did just want to provide a brief overview before handing it back over to everyone for Q&A.

InetSoft is an advanced analytics platform, and it lives wherever your data resides so you can run high performance analytics wherever and whenever you need it. And although InetSoft is a massively parallel processing data mashup engine at its core, we integrate with all of the leading vertical data sources like Vertica, popular programming languages like R and Python, Open Source Solutions including Hadoop and Spark. And it's available through a broad range of deployment and pricing models.