Moved by Metrics EHRs,Predictive Models Mainstreaming Analytics in Healthcare: What’s Behind the Slow Adoption?

Mainstreaming Analytics in Healthcare: What’s Behind the Slow Adoption?

“The prospect of integrating disparate sources of information into a multifaceted canvas of patient experiences is a tantalizing one, yet basic concerns with the usability of electronic health records, the availability of health information exchange, and chronic lack of time, knowhow, and funding have all contributed to keep big data on the back bench.” Source: Top 4 Emerging Tech Trends in Healthcare Big Data Analytics by Jennifer Bresnick, Health IT Analytics Journal June 27, 2016

Predictive Analytics World  moderator, Jeff Deal recited Ms. Bresnick’s statement to the panelists and asked “Is that statement fair? Is analytics still on the backburner in healthcare? How are we going to get adoption, get beyond ad hoc activities and weave it into the culture of our organizations?”  Guest expert panelists:  Martin Kohn, MD, Chief Medical Scientist at Sentrian, Wasim Malik, Ph.D. from the Harvard/MIT Laboratory for Neuromotor Signal Processing, and Nephi Walton, MD, from the Washington University School of Medicine and co-founder of Brain Spin, all shared their personal views, experiences and challenges with an audience of data patriots. Here are some of the session’s takes-aways for data scientists, statisticians or persons interested in getting a green light for setting up analytics in the everyday practice of healthcare.

Seeing Top Value: The Clinician vs. The Administrator
“My viewpoint is that analytics is not on the back burner and is not on the front burner either. There are a lot of important issues that need to be understood. And if we, as data science people, can understand them and work with clinicians and hospital administration then there are not a lot of road blocks,” expressed Dr. Malik.  From his perspective, healthcare’s clients are in two distinct camps.  One group is the hospital administrators and CFOs who are primarily concerned with dollar savings and economics.  Their focus is usually not on innovation but rather on cost savings. “When show dollar value improvement using predictive analytics to an individual hospital usually people will listen to you,” he told the audience.   Clinicians, the other group, look to see what is demonstrated from analytics.  Making a change or improvement in their procedures, work flow or clinical practice is key.  “Data should show some promise of clinical value. Once you do that, it is not that generally challenging,” he said with respect to engaging clinicians.

Pinpointing the Motivation for Building Models
Dr. Martin Kohn believes that we haven’t done a good job in addressing real world issues. Also, healthcare is an inherently changing system because there are so many mixed motivations. He reflected back to the days when he was an emergency physician. The goal was reducing emergency room visits. “In a fact,” he stated, “ER medicine opposes a lot of these initiatives where you have to get permission from your doctor to go to an emergency room. So when you are doing things to threaten some else’s self-interest, there is going to be a lot of push back”.  The challenge, as he sees it, is to demonstrate that which produces value for the end user.  It is one of the reasons why his company, Sentrian, does not work with individual physicians who are in fee for service models since they have less interest in reducing encounters, but ACO-like organizations do. Dr. Kohn highly recommended that you need to direct what you are developing in order to show real value for the target audience.

Barriers for Data and Implementation 
Barriers for analytics projects start with trying to get all the data. In Dr. Malik’s experience critical things like clinical data, particularly in the form of EHR data, is noisy.  “Typically when we work with EHR data, people have to sit down and manually curate and clean up the data.  We spend 80% of our time doing janitorial work.  It is not fun but has to be done to make sure. Otherwise, it is garbage in and garbage out,” he emphasized.

According to Dr. Walton, Washington University has a lot of data. Securing resources to build models can be worrisome. But perhaps the greatest challenge, once models are built, is implementing them into the infrastructure. “The universal answer – No.  You have to fight through layers of bureaucracy to actually get something into the system,” volunteered Dr. Walton.   He mentioned that another handicap is a thinly staffed IT group, and this situation limits the available resources for model implementation.  Since it is a cost center, the people who run may not see the benefit of what you are doing in order to make a priority.  “You have to translate what you are doing again into dollars or something that the other stake holders or other people understand in order to implement it. That is a major challenge,” he stated.

Establishing Trust in the System: Starts with Collaboration
When Dr. Walton first proposed changes to the standard genetic testing algorithms, he got a lot of resistance from physicians.  Having given the situation some thought, he realized they were on the peripheral of the project and not at the center making self-invested decisions. They had never even seen the data.  From his own experience, he imparted “If you show them a better way to do something, a way that they can believe in then make them apart of the solution. I think you can gain some acceptance. Tell them this is the data and ask how you [the clinician] want to fix it. You want them to be a part of the conversation. Making sure their concerns are addressed brings them into the conversation. That can improve your models.”

The Predictive Analytics World conference took place at the Jacob K. Javits Center in New York City on October 23-27, 2016.  Find out more about the Predictive Analytics World.

 

Related Post