- To many healthcare organizations, big data looks like an all-you-can-eat buffet. On the first go, they try to pile as much as they can on their plates, only picking through the bounty when they get back to the table to figure out what they like to eat and what was a mistake. The second time they address the banquet, they might be a little more discerning. But unlike a favorite restaurant, data analytics doesn’t have a one-time admission fee, and mistakes can be costly in more ways than one.
In order to set themselves up for future success, healthcare organizations need to figure out a better strategy than hoping they’ve collected enough to satisfy themselves on the first try. Shane Pilcher, Vice President at Stoltenberg Consulting, spoke with HealthITAnalytics to give healthcare providers some tips about how to avoid stomachaches by going smart instead of just going big
What’s the difference between big data and smart data?
Big data represents data sets that are so large and complex that they exceed the processing capacity of the conventional systems that we have. It also refers to the tools, processes and procedures that must evolve to allow us to analyze and process those extremely complex data sets that are involved in big data.
Today, only about 5% of organizations are leading the way in healthcare data analytics, and the other 95% are very early in their analytics process, or haven’t even started. For them, data collection is the focus. Data collection is the initial phase of data analytics.
Smart data is the intermediate phase. It focuses on the type of data that they have, the volume of data, and also the validity of that data. Some organizations that are collecting data aren’t really sure how they’re going to be using it, and it could be years down the road before they actually turn to their data warehouse and start to analyze it. Maybe at that point they’re looking for five years’ worth of trending data, but they dig into that warehouse and find that they don’t have enough data, it’s of a type that they can’t use, or it’s not accurate. So, you’ve got to have the right type of data, the right amount of it, and you’ve got to be able to trust the validity of that data – or you can’t trust the results from analyzing it.
Are hospitals collecting the right data in sufficient amounts?
I would think that the majority of hospitals are falling short of the data that they need right now. Hospitals see big data as something that’s off in the very distant future. For most of them, anything outside of five years is almost nonexistent. Due to the overwhelming demand that they have from all the projects on their plate, as well as the limited financial and employee resources available to them, most organizations have to prioritize how they spend their time. Anything perceived as being in the distant future receives very little time or resource bandwidth. But it’s a false sense of security that they have time. Big data in healthcare can be compared to looking in your side view mirror. The little warning that objects may be closer than they appear is quite fitting – big data is much closer than most organizations think.
As they start moving down their roadmap, organizations can’t simply be implementing for today or what they think they might need in the next year. Most organizations are looking at the very near future and only focusing on what they need to do today to reach that next milestone or get over the hump of that next project. Because of that, when they go to look at their data when they actually need it, they realize they’re woefully short on what they’re collecting.
What organizations are finding is that what they thought they were collecting isn’t what they were actually getting. For example, two to three years ago, when EHRs were really starting to be implemented, there was a big push to avoid controlling how clinicians entered their information. You could create drop-down boxes with options in the box, but because we didn’t want to spoon feed them or put words in their mouth, the typical approach was to leave an option for free text as well. If they didn’t like what they were seeing as options, they could write their own answer.
You think of that as a great option, you build your system and start collecting the data – but when you go to look at what you’ve got, you have a lot of unstructured data that is almost unusable in its current form. Yet, 80% of hospitals expect to use unstructured data in their data warehouse. When they realize how much data they have and that it doesn’t lend itself to being analyzed well, they begin asking for tools to turn it into structured information.
How will meaningful use aid hospitals as they start developing analytics?
Meaningful use is definitely putting us on the right track. Will we be where we need to be for analytics at the end of Stage 2? Probably not. But we’ll at least be moving down the right trajectory. One of the requirements that caught people by surprise for Stage 2 is that everyone has to track quality measures a different way. Quality measures require certain lab work, orders, and documentation to be done. The CQMs for 2014 and Stage 2 will be looking for specific verbiage and specific elements within that documentation. So, we’re going to have to move away from free text and to more structured data with elements to allow providers to qualify for those measures. A single typo could cause them to fail the measure.
What should hospitals be focusing on right now?
The first thing they’ve got to do is understand the business goals of the organization and try to align data analytics to fall in line with those business goals. The next step, especially for those developing their data analytics process, is to find some low hanging fruit that they can pursue. They can use historical data to find areas where they can make a change and, following that change, they can reanalyze and assess to determine what effects it’s had.
An example of that is identifying a department that has a high operating cost. They can look at the quality of patient care and the results provided by that department, and the reimbursements that they get back on it. In many cases, what they might find is that the treatments could be changed slightly to be less costly and keep patient outcomes the same or higher. Overall, it would drop the operating costs and produce more profit while having as good or better patient outcomes. With that, you’ve directly affected the bottom line and also grabbed the attention of the executive leaders, to get them on board with analytics. The savings realized could potentially be applied toward the budget for data analytics to allow the organization to increase their effectiveness. It could help you grow your analytics tools and grow your resources so you can start moving down the maturity level of analytics to start doing bigger and greater things.
What would you say to physicians to help convince them to get on board with analytics?
There are three levels of data maturity. You’ve got the descriptive level, followed by the predictive level, and eventually, you get to the prescriptive level. For clinical practice, having the right data and right data sources – lab work, genomics, diagnosis codes, and all sorts of historical information – can help us identify so much.
Take a patient with breast cancer. By using all the data available, including genomic data, a physician can identify a specific treatment that will have the least toxic effect and better outcome for that patient. They can customize their care while providing fewer side effects. That may or may not have a direct impact on the cost or the bottom line for an organization – but it does have an impact on physicians and their patients. Being able to show the benefits of analytics can keep the doors open is important for the organization, but it is also good to show caregivers how analytics can directly improve the care they can provide to their patients.