Getting Started With Predictive Analytics

Lauren Drell and Julie Davis
Marketing Health Services
Current average rating    
Key Takeaways
  • The health care industry is increasingly leveraging data to forecast trends and behavior patterns, a practice known as predictive analytics.
  • Marketers can use data from electronic medical records, billing systems, prescription histories and social media to derive actionable insights, which could improve quality of care and boost a health care organization's bottom line.
  • Among top performing companies, 65% use analytics for sales and marketing, compared with only 40% for lower performing health care organizations. ​

Imagine that you could pull out a crystal ball and predict the future of your health care organization. What types of insights would you ask to see?

For medical professionals, seeing the future could mean predicting how busy the emergency room might be in two weeks and adjusting the schedule in advance to make sure it will be fully staffed. For health care marketers, it could mean developing a safe jogging program that launches before exercise-related knee injuries are expected to spike.

The health care industry is gaining this valuable, hyper-focused foresight by using predictive analytics, a technique that leverages data to forecast trends and behavior patterns. The insight gleaned from the resulting data could help organizations take a proactive approach in making accurate and timely health care decisions and allow marketers to target their health care initiatives accordingly.

Predictive Analytics Primer

As the digital health revolution continues to unfold, electronic data flows from nearly every direction of a patient’s health journey, from electronic medical records to billing systems, prescriptions, social media and even wearable devices and smartphone apps. Marketers can use this data to get actionable insights about their patients, improve quality of care and patient outcomes, reduce costs, and boost efficiency—all adding to an organization’s bottom line.

Eric Siegel, the author of Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die and founder of the Predictive Analytics World conference, says that predictive analytics help marketers decide how to target an offer or which specific offer to give a patient population. “There’s an interesting parallel between marketing applications of predictive analytics and health care applications because in both cases, the whole point is to improve treatment decisions,” he says.

There’s an important distinction to be made between what will happen and what could happen, adds Mike Gualtieri, principal analyst at Cambridge, Mass.-based Forrester Research. “We’re always talking about probability,” he says. “So if we say we have a predictive model, it’s never about absolutes. It’s never 100%. There’s always some probability.”

According to Gualtieri, two categories of patient data are critical for predictive analytics: 1) demographic information about patients, including their geographic location, hobbies and lifestyle; and, 2) the patient’s behavior, such as Web-browsing habits and hospital admission rates. Marketers and health care professionals, therefore, can use this information to predict certain actions, such as which patients will likely be readmitted to the hospital, which patients are least likely to follow through on their prescribed medication regiment, which patients are most likely to manage their chronic disease, and more.

Health care organizations also use reports and trend analysis, patient results and recent studies to determine what could become a problem, says Karen Butner, an associate partner at IBM’s Institute for Business. “[It’s important to] be able to digest some of that information so that they have leading indicators as to what’s around the corner,” she says. “We can look at a lot of things, but the customer analytics … is what it’s all about.”

Real-World Applications

In The New York Times article “How Companies’ Learn Your Secrets,” author Charles Duhigg reported on how retail giant Target was using predictive analytics to determine which of their customers were pregnant. According to Duhigg, Target’s statistician, Andrew Pole, analyzed customer data and noticed that although lotion is a popular item for all consumers to buy at the big-box retailer, the women on Target’s baby registry were buying larger quantities of unscented lotion around the beginning of their second trimester. Additionally, he found that pregnant women were purchasing supplements like calcium, magnesium and zinc within their first 20 weeks of pregnancy. According to Duhigg, Target used this data to create a predictive model that could deliver insights about customers who weren’t in the registry but exhibited similar characteristics, allowing the retailer to market to a larger swath of pregnant women. According to Duhigg, Target’s methods created a controversy after a Minnesota father reportedly found out his teenage daughter was pregnant after receiving pregnancy-related marketing materials from Target.

Gualtieri insists that Target’s foray into predictive analytics is not a cautionary tale for marketers. Instead, he says it should be a reminder for companies to think through how they collect and use information. He recommends marketers first approach prospects indirectly and confirm that they may be interested, and then make additional offers.

According to Gualtieri, a majority of consumers don’t mind if companies use their personal data to improve their targeted marketing efforts. “The risk of another competitor providing a more individualized service is higher than a customer being creeped out by a more personalized service,” he explains.

Bringing Predictive Analytics to Health Care

Beaufort Memorial Hospital started incorporated predictive analytics into their marketing strategy in 2011 when the hospital partnered with Birmingham, Ala.-based health care data analytics company Medseek to determine the right audiences for their preventative care campaigns.

“We don’t have the marketing dollars to compete with the major medical centers that are also targeting in our market,” says Courtney McDermott, Beaufort’s director of marketing and communications. “We wanted to make sure that we had the tools available to reach the right patients with the right message at the right time.”

Medseek and Beaufort have executed a variety of campaigns, some as simple as reminders to come in for an annual check-up, and some, such as campaigns promoting new services, which tap into Medseek’s predictive tools. The team is currently working on a campaign promoting Beaufort’s new tomosynthesis or 3-D mammography equipment. Using its predictive data, Beaufort’s team is developing a direct mail piece that will target women who may be considering a mammogram.

Some of Beaufort’s past campaigns have been so successful in drawing new customers that the hospital has actually had to put one campaign on hold until the organization had time to hire additional staff. Over a period of 13 months, Beaufort acquired 1,966 new patients and generated a total net profit of $696,331, according to Medseek.

Health insurance companies have been using predictive analytics to estimate consumer health and medical expenses, but its move into health care marketing has been more recent. Blue Cross of Northeastern Pennsylvania (BCNEPA) began using Deloitte’s predictive analytics models to target a direct mail campaign in late 2013 and early 2014. Paul Holdren, senior vice president of sales and marketing at BCNEPA, says there were 120,000 uninsured people in northeastern Pennsylvania, and by using predictive analytics, they targeted their outreach to a little more than 10,000 contacts and determined that 10-12% had a high probability of enrolling, according to BCNEPA. Holdren’s team implemented a direct mail campaign promoting BCNEPA’s individual insurance plans aimed toward patients eligible for the plan and who had a strong likelihood of being interested in the company.

Holdren’s previous direct mail campaigns generated 1-2% enrollment, on average, but this campaign netted 20% enrollment. “I only wish I had it 20 years ago,” he says. “It would have made a big difference in how I spend my advertising and direct marketing dollars in order to secure people.”

New York-based EmblemHealth’s use of predictive analytics has prompted the company to begin building an in-house analytics team. The team has supplemented the ongoing relationship between EmblemHealth and Columbia, Md.-based Merkle Analytics for the past year and a half. “An agency is primarily tasked with analytics around the campaigns, but we use analytics for lots of strategic purposes beyond running our direct marketing campaigns,” says Shirish Dant, EmblemHealth’s head of consumer marketing and market insights. “So our in-house analytics capabilities are really critical in doing those strategic analytics.”

EmblemHealth saw strong results using predictive analytics for their Medicare open enrollment campaign in 2013. According to Dant, during the 10-week period of the direct mail campaign targeting Medicare members interested in switching health care plans, their costs per lead went down 50% compared with the previous year. The team was able to generate more leads while sending 25% fewer pieces of mail.

As a business serving a large local market—New York City—EmblemHealth’s marketing strategy has always revolved around neighborhoods. “We are very specific and very targeted when it comes to our marketing strategies and our campaigns,” says David Mahder, EmblemHealth’s vice president of marketing and communications. “Predictive analytics really fits right in because it allows us to be even more effective as we target these different populations.”

The team is now working on incorporating more data—tracking not only responses, but also leads and conversions—to better refine their predictive models.

The Analytics Edge

Making good use of data and analytics is a trait shared by the industry’s leading organizations, according to 2012 research from IBM. Among top performing companies, 65% use analytics for sales and marketing, compared with only 40% for lower performing health care organizations. Graham Hughes, chief medical officer at the Cary, N.C.-based SAS Institute Center for Health Analytics and Insights, says that seemingly small improvements, such as understanding high-risk patients more accurately or increasing retention or engagement by 5%, can be a huge advantage. “That translates into a phenomenal amount of differentiation if your competitor isn’t doing it,” he says. It’s not a perfect science, but it’s one of those things that one or two percentage points … can have a huge influence on understanding your business.”

Many organizations are limited because they don’t understand the kinds of questions data can answer, Gualtieri says. He recommends companies look at their customer journey map and ask detailed questions. “Don’t constrain yourself,” he says. “Ask yourself the question, ‘What could I predict about this customer that would allow me to perfectly provide them some product information that they need to know?’ ”

The availability of data will continuing to grow rapidly, but many health care companies will need to improve the way they make use of that data in order to reap the rewards. “We really wouldn’t be where we are today without predictive analytics, so it’s a really important tool,” Mahder says.

Best Practices for Predicting

When putting a predictive analytics strategy together, Karen Butner of IBM’s Institute for Business says health care organizations should think of it no differently than undertaking a marketing campaign. “It’s not unlike Procter & Gamble selling Tide,” she says. “It’s a marketing campaign in that it communicates awareness—all those things—so that they’re increasing their [brand] loyalty, increasing their customers, and increasing their products and services.” In order to successfully incorporate predictive analytics, marketers should understand a couple best practices:

1. Outline the business goals.

Butner suggests to clients that marketers start with what IBM calls “a Big Data strategy,” which helps determine an organization’s vision and goal for using predictive analytics. Once you’ve outlined your vision, she says, make sure to communicate it with everyone involved. “We need to first decide what the business need is, what we’re trying to accomplish and, of course, work with our IT and researchers … to create the blueprint and say, ‘Here’s what we want to try to achieve, but to achieve that, this is the information we need,’ and get everybody on board with the overall plan,” Butner says.

2. Leverage existing data.

The most logical, cost-effective place to look for new insights are within an organization’s existing data, says Butner, so start there. “Don’t try to boil the ocean. … [Wait] until you first tackle your existing data and bite off what you can chew and digest,” she says.

3. Build on your analytics capabilities.

Butner says that building on analytics capabilities should all come from the organization’s original roadmap—its current needs and goals. By fitting a predictive analytics strategy within your already-existing parameters and business requirements, organizations can build out their analytics efforts appropriately. 

4. Support the business case with proof.

After you’ve taken on a project or two using predictive analytics, communicate your discoveries and insights with the business team. By doing this, Butner says, you’re ensuring active involvement and sponsorship from top-level staff members, which could pave the way for larger projects that could show more definite ROI.

5. Think about enterprise.

Keep in mind that you may need to scale your analytics strategy over time, says Graham Hughes of the SAS Institute Center for Health Analytics and Insights. Once you get a five-year strategy in place, invest in a platform that gives you the ability to solve hundreds of different predictive cases. Don’t piecemeal yourself one at a time with different solutions, he says. “I think the most sophisticated organizations are looking at marketing automation and marketing analytics and marketing operations [using] customer intelligence platforms that will be able to cover a broad spectrum of predictive market analytic needs,” he says.

6. Consider recruiting help.

Most likely, you’re going to need someone to mine the analytics, or a tool to help define your predictive models, or both, says Forrester’s Mike Gualtieri. He suggests looking into software—such as programs by the SAS Institute or Alpine Data Labs—and consider hiring a data scientist or analyst if you really want to invest in your predictive analytics capabilities.

7. Learn from the pros.

Network with a marketing leader from another industry, such as a chief marketing officer or a digital campaign manager, to learn about their journey with predictive analytics, and then apply those learnings to your own environment, Hughes says.


​This ​was originally published in the Fall 2014 issue of Marketing Health Services​.​​​​​​​​


Author Bio:

Lauren Drell and Julie Davis
Lauren Drell and Julie Davis are staff writers for the AMA's magazines and e-newsletters. Send them an e-mail at
Add A Comment :

Become a Member
Access our innovative members-only resources and tools to further your marketing practice.