Prospective Analytics: The Next Thing in Healthcare Analytics

My Folder

prospective analyticsWhen the experts and pundits talk about what’s new and hot in healthcare, a new term that is quickly gaining traction is prospective analytics. And for good reason. This new type of analytics offers an unprecedented opportunity to use data to affect decisions, actions, and outcomes at the point of care. It takes advantage of retrospective and predictive analytics to support clinical decision making in a more expansive way, by looking at the potential results of all proposed options. For anyone interested in graduating to the next level of outcomes improvement, this new methodology warrants closer examination.

But it’s important to understand that for an organization without a proven track record in retrospective and predictive analytics, jumping right into prospective analytics is like walking onstage to perform a Beethoven symphony with a world-class orchestra at Carnegie Hall with no prior musical experience.

To get to this new level, organizations must first establish a foundation by gaining experience with other types of analytics. Just as becoming an accomplished musician – for all but the very rare savant – requires learning to read music, working with an expert instructor, and practicing for many, many hours.

We must view the process of prospective analytics as a journey. Currently, there are three types of analytics that organizations should use with the petabytes of clinical, financial, and operational data they currently generate to elevate quality, improve outcomes, and lower costs. It’s important to not only become proficient in all three, but to do it in the proper order. Let’s look at each of them in depth.

Retrospective Analytics

Healthcare organizations need to perform this as the foundational layer for all other analytics. This must be mastered first. In our orchestra example, it’s the equivalent of learning to read music.

As the name implies, retrospective analytics provides a look at what has already happened, helping a healthcare organization understand why those events happened. It is also the type most familiar to clinicians. They are comfortable with it because it draws empirical conclusions.

Clinicians can use retrospective data to view past actions – such as whether a panel of patients with sepsis received medication A or medication B – and what the outcomes were with each. They can also confirm that the sample size used in the analysis was statistically valid.

Retrospective analysis can be extremely effective at helping the organization standardize care and remove variations. It is also very effective in other areas such as staffing, inventory control, and billing.

Its two main limitations are 1) it leads to conclusions that are restricted to the choices the organization has already made, and 2) it takes a long time for those conclusions to become policy at the point of care, usually after many rounds of discussion. There is no “what-if” element. That’s where the second type of analytics comes into play.

Predictive Analytics

Predictive analytics takes a higher level, forward-looking view. It takes the conclusions from retrospective analytics and gives the organization the ability to speculate on options. Let’s take a heart failure readmission scenario, for example. Predictive analytics help identify heart failure patients with a high risk for returning to the hospital. Once we identify those patients, we alert a case manager and interventions can focus on decreasing the likelihood those patients would return as readmissions. In our music analogy, predictive analytics is the equivalent to practicing scales and techniques that improve the musician’s skills.

Predictive analytics also shows what is likely to occur if the healthcare organization continues down the course it is already on, i.e., the cost of inaction. It’s like the Great Horse Manure Crisis of 1894, which led the Times of London to speculate that by 1944 “every street in London will be buried under nine feet of manure.”

With predictive analytics, the outcomes aren’t known or guaranteed. The organization is simply looking at the likelihood of an event to occur if it follows a particular course. It then relies on other processes to determine what action to take.

Here’s a good example from my personal experience. Remember all the angst around the turn of the millennium and what would happen to all the computers and medical devices because of Y2K? I was working in an emergency department at that time. The hospital wanted to be prepared for whatever might happen, and since there was no retrospective data—given that the last time the millennium changed, “medieval” described the era rather than a theme restaurant—all anyone could do was take their best guess.

My organization took the data it had, made predictions based on the worst-care scenario, and prepared accordingly. Of course, when January 1, 2000 hit, it turned out to be just another day in the ED. Had something gone wrong, however, at least we had a plan in place.

While predictive analytics can generate new possibilities, it is still not a decision-making tool..

Next Level Analytics

Prospective analytics takes the knowledge gained through retrospective and predictive analytics, then drills down to show bedside clinicians (or administrators) all available options for changing the current state, as well as the associated consequences. This is why organizations are so excited about this new methodology.

Here’s a good clinical example. There are two basic types of appendectomies: simple and complex. A simple appendectomy is a surgery that removes an inflamed appendix, versus a complex appendectomy, which removes an appendix that has ruptured. Because both kinds present with symptoms of lower abdominal pain, fever and nausea/vomiting, it’s impossible for a clinical team to know which type they’re addressing until they begin to operate. As a result, the default is to code all appendectomies as simple upon admission, which makes sense.

The challenge is that the surgical procedure and the post-op care for a simple appendectomy are very different than for a complex one. For example, different medications are required and the length of stay is longer for complex appendectomies. If the procedure isn’t re-coded, the quality measures will be based on the wrong set of standards and the organization will likely be penalized for not meeting its requirements when, in fact, it did everything right, except use the correct code.

Prospective analytics collates data as it is entered into the electronic health record (EHR) and alerts the clinician to the possibility that this procedure initially coded as a simple appendectomy may actually be a complex appendectomy. Also provided are the best practice care options for the complex appendectomy patient vs. the simple appendectomy. It’s a form of decision support.

There are also has many applications on the operational side. One ED I worked in would tend to get an influx of patients, especially orthopedic patients, whenever amateur rodeo night rolled around. Retrospective analytics told us how many patients we’d treated on that Saturday night, and how it compared to the previous 20 Saturdays. Predictive analytics told us the likelihood that we would need to increase certain areas of service, such as having more orthopedists available or on-call, and limiting the number of elective orthopedic cases in the OR. It was still an educated guess that allowed us to plan for future events

Prospective analytics would have been able to show us how to adjust resources if we were suddenly overwhelmed. For example, if the x-ray suites were inundated, this type of analytics could determine which radiology exams to perform at the bedside (portable x-rays) based on available materials and capabilities of the clinicians on staff at that time.

In short, while retrospective analytics are good at identifying problems, and predictive analytics are good at anticipating problems, the prospective approach delivers its value by validating the gut instinct of clinicians and healthcare administrators, with real-time, evidence-based solutions to problems, i.e., empirical data.

Sharing Knowledge

One advantage of prospective analytics is that it can integrate the multiple variables associated with each patient and disease process, and identify likely outcomes based on existing data and past analysis. Considering the heart failure example above, the system could secure a follow-up appointment in a timeframe most suited for the patient.

Blending Prospective Analytics into Workflows

Of course, healthcare is filled with great initiatives that never come to fruition because they don’t fit within the workflows of healthcare professionals. That’s why it’s important to proactively bring the results of prospective analytics to clinicians and others as part of their normal course of work, rather than forcing them to seek it out.

As described earlier, a simple alert in the EHR or other clinical system that indicates when a procedure is improperly coded, and then offers a one-click option to correct, is likely to drive more positive outcomes than holding that information or capability in another system. After all, technologies that create extra work are among the biggest complaints of clinicians everywhere.

Consider that the typical primary care physician has roughly 12 to 15 minutes to spend with a patient. He/she doesn’t want to spend that time looking up information. He/she needs the most updated information presented at the point of care, including all possible outcomes. This is what drives clinical quality improvement and ensures consistency at the lowest reasonable cost.

Driving Adoption

Prospective analytics offers tremendous possibilities. But keep in mind that not everyone in the organization is forward-thinking. Some will need convincing. This is one more reason to approach the move into this new process as a journey.

If the organization starts with retrospective analytics, and uses that to support the conclusions drawn by prospective analytics, it will be much easier to sell to the skeptics. That’s why it’s important to take the full journey instead of jumping right in. Use the evidence to get everyone onboard, and soon you’ll have them making beautiful, and cost-effective, music together.

PowerPoint Slides

Would you like to use or share these concepts? Download this quality improvement presentation highlighting the key main points.

Click Here to Download the Slides

Loading next article...