- Few people like to hear the old adage that they must learn to walk before they can run. The promises of a mature, integrated data analytics information system are tantalizing, but the majority of providers are still in the first stages of building up their data collection and warehousing infrastructure, puzzling through wave after wave of information, and struggling to turn clinical and financial information into actionable insights served up quickly at the point of care.
David Delaney, MD, Chief Medical Officer of the US Healthcare Division of analytics giant SAP, believes that the industry can make those promises a reality – but only after providers make the effort to truly understand how their organization’s data can illuminate patterns in healthcare costs, quality, and patient safety.
What are some of the major issues facing healthcare organizations as they plan and build their analytics infrastructures?
The time it takes for data to double is compressing pretty rapidly. Some organizations are seeing a doubling of their data every two, three, or four years. And, of course, having genomics on the horizon threatens to accelerate it further. Then we have the fact that many organizations are struggling trying to shorten the time it takes for data to help them make a better decision in the organization. Given the ‘70s era disk space data storage methods that they’re using, there’s been an awful lot of effort and time involved in trying to make sure your data is available in a timely fashion.
The byproduct of that is that we often see a five-fold explosion for every data point that someone has in the organization. We’ll often talk to CIOs, and they might say that they have a petabyte of information. But when you drill down a little further, it turns out that in terms of unique information, they might only have about 200 terabytes. And that bloats up to a petabyte because they copy all that data into a staging area, and then they copy that into data warehouse, and then they create data marts on top of that. And then, to try to improve the time between creation and consumption, they add a tremendous number of indices, rollups and various other materializations with the data to try to help it run fast enough to be effective.
And it gets them, for the most part, where they need to go, but what it does is it explodes that data five times or more in terms of size. To handle this entire process, they need to have armies of analysts that focus on optimizing queries and grading queries to try to get the answer quick enough. The organizations end up becoming very internally focused. Most of them have a tremendous backlog of reports. They often have six months, nine months backlog of reports they’re trying to work on, and, of course, that leads to a lot of end user dissatisfaction.
The other side of the coin is with the end users: the clinicians and the administrators. Despite the CIOs drowning in data, they’re starving for contextual information to help them make better decisions doing their job. Even when people are using an EHR, you’ll often see them using Google to try to figure out the answer to a question or find a reference. It’s great that they’re using knowledge sources which are online and digital, but they’re not contextualized. There’s a lot of potential for errors. And when they do manage to get information from their analytics, most dashboards and reports are backward looking. They can tell you what happened last quarter or last month, but they really aren’t driving real-time decision making.
What are providers looking for when they invest in analytics tools?
On the provider side, what people are hungry for is what I would term ‘inline analytics.’ They want something that can help drive them towards making a better, smarter, more informed decision based on all available data. The hope with EHRs was that they would kind of change healthcare from a cottage industry to something that would be much more data driven.
Certainly we’ve seen the market move. I can’t remember the last organization I went into that doesn’t have an EHR, so we have seen a successful shift from paper-based to electronic data. The difference there is not small. Basically you’ve moved from completely written and unstructured data to digital data, although EHRs are probably still roughly 60 or 70 percent unstructured text.
So even though it’s digital, trying to get data out systematically still remains challenging. That’s where we really see the opportunity. It’s really taking to the next level and enabling analytics of all relevant information: the EHR information, the financial information, and other data stores to provide a complete view of what’s going on with a patient to help the provider fully understand care delivery.
What is the first major step towards really leveraging an organization’s data streams?
You have to understand the cost and quality of your care delivery. And that involves looking at data in a very rich format. EHRs are necessary, but they’re not sufficient for that. You have to be able to pull in financial information, including information from the payers in other systems, integrate it, and then be able to make sense of it and be able to look at how you’re delivering various services.
Take a knee replacement, for example. That can vary widely in terms of cost. You’ll see anything from $30,000 up to $80,000 in terms of the cost. And when you plot the cost versus quality, you’ll see a scattergram. For every organization delivering high quality at low cost, you’ll see another organization delivering a high cost and low quality. What you need to be able to do is pull that information all together and begin understanding your own care delivery service line by service line. Who are the outliers? What correlates with a bad outcome, and how can we create pathway that’ll help us be able to achieve good outcomes in a more systematic fashion?
When you look at most other sectors, like manufacturing, generally you can get higher quality, but it’s at a higher cost, or you can accept lower quality and get lower cost. It’s very difficult to actually increase quality while decreasing cost, and the only real lever we have to do that in healthcare is beginning to understand how we deliver care, what works, and what doesn’t.
But once that new care pathway is in place, we can begin to bring in other datasets as well. Maybe we want to look at how a patient’s genes impact their response to therapy. There’s been data looking at credit scores and other socioeconomic factors to predict compliance. So you start getting a richer and richer data set to understand how it all impacts a patient’s health.
How does meaningful use tie into the industry’s efforts to standardize and utilize analytics?
The first two stages of meaningful use are basically getting everyone on board the train. It’s slowly pulling out the station. And once everyone succeeds in getting on the train, we can start to figure out how to realize value. Not everyone is seeing the value that everyone had hoped for. People are getting very fatigued because we’re not seeing a major return on the investment.
But we have that foundation laid, and that’s important. Now we really need to begin building value on top of it, and we’re going to find it in the analytics piece. I’m very confident about that. We’re going to begin seeing meaningful use regulations become more meaningful. Right now it’s very rudimentary, but once everyone’s on board the train, it’s going to be accelerating, and the area it’s going to accelerate toward is the use of analytics.
Where do you see the healthcare analytics industry heading in the next three to five years?
Five years? I mean, who knows? Maybe we’ll all have our own personal hovercraft. That would be cool. But seriously, I think we’ll see the final stages of the shift in healthcare from a cottage industry to a value-based endeavor where reimbursement has largely shifted towards accountable care. You won’t get paid for volume. You’ll get paid for value, and there will be an increase of organizations taking on risk. Because of that, they will develop competency around delivering and understanding their own care delivery, and so everyone will become very focused on becoming learning organizations.
For those who do it well, they will succeed and gain market share. As a result of this, we will probably see a coalescence of the market. You’ll see an increase in the number of regional powerhouses emerging, and a lot of these smaller, less agile, less able-to-adjust organizations being swallowed up.
When it comes to analytics, I think you’re going to see healthcare becoming a very safety-focused endeavor. The number of people lost to accidents is just shocking. If you had an equal number of people going down in planes every year, no one would fly. I think you’re going to see an awakening to that, and you’re going to see a push toward visibility when hospitals make mistakes.
As part of this strong drive toward quality, there will be information systems that become more and more like what you see like at the FAA or in the aerospace industry, where there’s just a tremendous amount of safety built around to help us not do dumb things as humans. We won’t rely on doctors just being smart men and women who don’t make mistakes. We’ll realize that we’re fallible. That’s largely what the EHR has become, I think. It helps us practice safely and avoid mistakes.