How predictive analytics work – and why they help

Predictive analytics isn’t just the latest healthcare buzzword. It’s being used today to provide better care and it has groundbreaking potential to enhance healthcare delivery. With all of the lofty hopes, sometimes the “real application” of predictive gets lost.

Here’s an example. Take 100 patients who are candidates for colorectal screening. We know that a group of them will try to avoid the procedure or postpone it for as long as possible. Some evidence of this would be regularly rescheduling appointments.

We also know that among this 100 is another group of people who need a colonoscopy more than others such as those with a family history of the disease.

Finding where these two groups overlap and focusing more time on outreach to those patients to get screened – and you have a simplified example of how predictive analytics can improve population health and cut patient healthcare expenditures.

That, in essence, is what it means to deliver value for all concerned.

What Are Predictive Analytics?

Basically, it’s the ability to look at data sets and learn where to expect problems within given patient populations we manage. Selected data can help guide us on where we can act and positively change outcomes. When done right, predictive analytics can keep patients healthier longer, help them avoid expensive medical care and identify cost drivers and help the medical group better control costs.

Before predictive analytics, we tended to react to a patient health issue. Here’s an example: A patient with hypertension and diabetes. We know that, over time, this patient will more than likely have heart disease. You could write prescriptions and watch for symptoms, but otherwise wait until the onset of heart disease before intervening.

You could do this within an information vacuum. Do you know if this patient is refilling prescriptions regularly? Taking their medicine at the right times, in the right dosages?

How much do you know about this patient’s lifestyle? Familial support? Financial situation? How might all this then impact the patient’s self-care to prevent heart disease?

Determining What’s Appropriate

Predictive analytics can ascertain for us how often such circumstances can be expected and in what sub-populations within this group. They can point to other areas of concern. They can guide us to ask appropriate questions, reach appropriate conclusions and provide appropriate care at the appropriate time at the appropriate facility.

The potential is great. We can make patient population management more effective with the least amount of adverse effects going forward.

These types of practices can lead the way in bending the cost of healthcare downward. For Austin Regional Clinic and others, it also creates an opportunity to separate ourselves from competitors.

Already, we have some evidence of success. Using the colonoscopy example, among Austin Regional Clinic’s Medicare population, we have increased the percentage of those up to date on their colorectal screenings to 73-percent, up from 65-percent just a few years ago.

This was accomplished without having to add more staff to our outreach teams. We used better analytics to uncover which patients required extra attention to get them to act, offered more options for patients to “get screened” more conveniently such as with the Cologuard at-home screening tool, and integrated these practices with existing resources.

Improvements in predictive analysis will be coming – and at a faster pace. Already, there are approaches available that utilize artificial intelligence and machine learning. In many ways, all this is mindboggling.

I’d like to say we use sophisticated data mining, but Austin Regional Clinic is not there. Nevertheless, we are able to use data in ways that wasn’t even possible ten years ago. And the prospects ahead are exciting to contemplate. New and improved predictive analytics are in our future.

Physician-Patient Bond

Amidst all of this, we must remember predictive analytics are informed calculations to help us manage patient populations. Still more important or, rather the most important is the one-on-one relationship physicians have with each patient.

“Population heath” cannot occur without a relationship. Without the physician-patient bond, the data that drive predictive analytics cannot be accumulated and the benefits cannot be realized. Jay R. Zdunek, DO, MBA, is a Family Medicine physician and the Chief Medical Officer at Austin Regional Clinic.

Tags: ARC Leadership blog