Three Examples of How Predictive Analytics is Being Used for Better Healthcare

Healthcare analytics data is useful in many ways: 

1. It helps hospitals offer better patient care 
2. It helps patients get diagnosed faster and more accurately 
3. It helps healthcare payers (like employers and the federal government) budget for healthcare expenses 
   

Benefits professionals are also turning to healthcare data as a means of predicting the future. They’re using medical claims data, prescription claims data, and wellness program data to track, measure, and predict the health and wellness of their employee populations. While we’ve all heard how useful “big data” can be, even “small data” generated by an employee population can be crucial for this new trend of predictive analytics. 

Let’s look at three examples of how predictive analytics is being used to improve healthcare for employees and their families.

Predicting Future Costs for Payers

At Artemis Health, we talk to a lot of benefits professionals, including consultants, brokers, and HR/Benefits directors. One thing they all have in common? They’ve been asked by C-suite leaders to explain why healthcare and benefits costs were higher than expected in a fiscal year. This year, due to the outbreak of COVID-19, it might be a different story. Unexpected costs will be a big challenge for benefit leaders, and Coronavirus will undoubtedly disrupt their ability to predict 2020 healthcare costs.

It’s not easy for healthcare payers like self-insured employers or even carriers to predict exactly what they’ll spend in a given year on healthcare benefits. After all, there’s a certain amount of variability and randomness. If we could predict who would get cancer or how many people would be in a serious car accident each year, we would be able to take corrective action to (hopefully) prevent it from happening in the first place. 

But, sadly for health and benefits professionals, this is not Minority Report. We cannot stop medical claims before they start. So the best we can do is use reliable data to predict and forecast these costs. 

Let’s look at a real-world example. Artemis has developed a tool called Trend Explorer within the Artemis Platform for questions just like this. It helps self-insured employers and benefits advisors look at trending costs, common conditions, and how they change over time. 

While a lot of this information is available to benefits analysts through manual number crunching, Trend Explorer makes it easy to calculate and draw conclusions all in one place. In this sample data, you can see that the cost of vision exams is trending down by 4.9%, while the cost of outpatient surgeries is trending up by 5.8%. Not only does this data give you the insights you need to impact future costs, but it provides a solid foundation for predicting future healthcare and benefits costs. 

Predicting New Disease Diagnoses 

Predictive analytics works best when you can compare it to larger trends and find out how your analysis compares to a larger picture. That’s where benchmarking comes in. Many of our clients use wider population health data to find out how their members stack up against the norm. 

A risk score is a way to determine an individual member’s overall health. Each person’s risk score is based on their demographics, health status, and potential healthcare utilization. For example, someone with a high risk score may have a new diabetes diagnosis, while someone with a low risk score may have seen the doctor for the occasional seasonal cold. When analyzed on a population level, risk scores can help employers assess the potential future health of their population, and even prepare for future disease diagnoses.

This sample population is just over 21,000 people, a relatively small data set. You can see for these members, they have an overall low risk score, which means a higher proportion of the population is classified as healthy based on their diagnoses and utilization. 

Now let’s look at medical claims and diagnoses for just high risk members.

39 members who were categorized as high risk were treated in the ICU during our analysis period, which resulted in over $100,000 in costs per member. This sample data points to “diseases of the heart” as the largest portion of ICU costs. A benefits advisor or self-insured employer may take this information into consideration when adding or measuring the success of disease management programs.

Predictive analytics is not just for benefits professionals, either. Hospital systems and providers can use similar data points as indicators of future disease states. Think about the strides made in Type II Diabetes management over the last 30 years. It wasn’t until 1979 that “pre-diabetes” was established as an umbrella term for elevated blood glucose levels. Some experts have calculated that as many as 60% of people diagnosed with pre-diabetes will develop Type II Diabetes without intervention. The data points and criteria used to diagnose pre-diabetes are crucial for providers who can intervene early and hopefully prevent a Type II Diabetes diagnosis. 

Predicting Provider Quality and Outcomes 

Measuring and predicting the quality of doctors and hospitals is tricky. It’s beyond tricky for a few key reasons: 

1. A lot of provider quality data is anecdotal or based on patient perceptions 
2. A high quality physician could practice as a less great hospital, or vice versa
3. A doctor or nursing staff could be really great at some procedures and less great at others 

There are simply a mountain of data points to consider, weigh, and measure in order to come up with a solution on provider quality. Luckily for benefits advisors and employers, some reliable metrics and data sets can help measure the performance of providers and help predict their future outcomes based on past performance. 

Artemis Health is working with Quantros to provide our clients with reliable, useful data on provider quality. Quantros’ Check+ rating system puts physicians and facilities into these quality categories: 

The data is based on a number of metrics, such as complications instances, mortality rates, patient safety ratings, readmission rates, and more. Benefits professionals can use this data immediately to look at their networks, and by tracking changes over time, can predict the future state of provider quality within their networks.  

While predictive analytics is an inexact science, there are data-driven actions HR and Benefits teams can take to improve the health of their populations. Through cost projections, condition tracking, and provider quality measures, employers and advisors are in a great position to impact the future of healthcare for their employees and their families.