Photograph: Riku Foto
Q: Tell us a little about your background. How did you make the decision to become an actuary?
A: I started out as a physics major in college with the intention of becoming an engineer—I knew very little of the actuarial profession then. It just so happened my college was the only school offering a bachelor’s degree in actuarial science at the time. My calculus professor recommended that I sit for exam 100 (the calculus exam). I passed it on my first try, and once I learned that I didn’t need to pay to go to graduate school and could find a job that paid me to continue my education, I switched to be an actuarial science major and haven’t looked back. Little did I know, however, that studying for actuarial exams would have a severe negative impact on my golf handicap.
Q: Why did the actuarial profession attract you? What sparked your interest in the health care industry?
A: I liked the versatility of the work. You could get into a number of areas within insurance and finance, and the work would always be diverse and allow for creative thinking. I was originally a pension actuary, but a big chunk of my work was Financial Accounting Standards Board (FASB) Statement 106 (the calculation of post-retirement health care benefit obligations), which was enacted a few years before I began working. Companies were interested in how they could change their retiree benefits to help reduce their liabilities. This generally led to discussions about active benefits, and next thing I knew I was spending 100 percent of my time doing health care benefits consulting.
Soon after, I was hired at a managed care consulting company where I specialized in provider contract analytics—especially converting bad capitation deals back into fee-for-service reimbursements. Additionally, I helped provider-owned health maintenance organizations (HMOs) develop products and price them, and I also worked in underwriting and sales. It was a tough environment for these entities, and I loved the challenge of trying to help them succeed.
Q: How did your current company get started?
A: My current company is a spin-off from a more traditional actuarial employer benefits consulting company. The benefits consulting company was constantly being asked to do more deep-dive analytics. This would have been a distraction from the core competency of benefits consulting, so the company created a new entity that specialized in analytics, specifically working to bring new ideas into the health care analytics space. We started with four people in 2013 and partnered with a company that had a machine learning platform to help with building the predictive models. Given our rapid success and growth, we quickly needed to add a more robust IT function to handle the massive amounts of data we were getting from clients, so we purchased another company. At the same time, we brought the data science work in-house and built our own machine learning platform. Now we have more than 50 employees, and we’re still growing.
The biggest challenge has been trying to meet client deliverables while exploring new product ideas with research and development (R&D) work. At a smaller company, striking a balance between the two with limited resources can pose all sorts of issues. Luckily, we have been able to keep our product development pipeline flowing while continuing to deliver high-value and interesting predictive models to our customers.
Q: What are the main functions of your job?
A: Contrary to what my colleagues would claim, it is not playing golf every day of the week. In all seriousness, I design our models, products and solutions. I work with customers to understand the problems they are trying to solve, and then provide input to our data science team as to which features make sense to include or exclude in the solution and which data should or should not be used.
Then, I confirm that the results of a trained model make sense for the problem we are trying to solve. I gather input from our clinical team to design the product, and from our IT team to make sure the necessary data infrastructure is in place. I also work to ensure the completeness and accuracy of the incoming data to support operationalizing the solution. On the customer-facing side, I meet with prospective clients and describe and demonstrate our products and solutions.
Q: You’ve worked for a number of health care companies during your career. What information have you taken with you that has been the most helpful?
A: There is no substitute for hands-on experience. Like many other health care actuaries, I’ve sat across the table from a group of unhappy providers when they needed an explanation about risk adjustment when they didn’t hit their shared saving target. I’ve sat in my CEO’s office with the head of sales and underwriting to try to work through balancing profitability and growth through rating actions on large groups. I’ve had to hunt down an answer from operations when my analytics were showing an increase in unit costs when I was expecting a decrease. I worked on many mergers and acquisitions (M&A) deals that had a variety of issues and problems to solve across the managed care spectrum. Through this work and exposure to many other topics, I learned what worked and what didn’t work, and what could have been done better with superior information, tools and models.
Q: How are you using predictive analytics in your job, especially as it applies to the health care industry?
A: There are a lot of opportunities in the population health management (PHM) space. Many of the existing tools and solutions used by managed care entities were developed years ago, and they only consider a handful of data elements and contain static algorithms and rules. To compound the issue of being ineffective, the processes that were built around these tools are broken. These entities have low patient engagement and clinical compliance rates, and they cannot demonstrate the value of their clinical programs. Finally, many prospective clients have a broad array of different kinds of data and want to put the data to good use. Given the nature of their current operational state, the incorporation of new data elements into their existing tools is difficult.
We combine our predictive modeling expertise with efficient process management to help improve these important metrics of a PHM program. This enables us to achieve great success in operationalizing predictive models, turning the theoretical into actionable. Predictive models allow for a more dynamic approach to PHM by continually updating rules as more data rolls in. In addition, the predictive models I design contain elements of traditional health care data, while incorporating a customer’s nontraditional data elements as well as any other data I can get my hands on from various domains. Because there is so much information out there, it can be challenging to weed through it all to find those data points that can improve the predictive power of the model.
Q: How do you see the role of predictive analytics in health care changing in the next five to 10 years? Where will actuaries fit into the equation?
A: Many individuals in predictive analytics are not actuaries. They get their undergraduate or graduate degrees in data science or business intelligence and can be considered for predictive analytics positions. Actuaries need to continue to differentiate themselves with superior predictive modeling skills and strong business acumen. With a robust curriculum from the Society of Actuaries (SOA), I think actuaries will continue to show the value we can add.
In terms of predictive analytics in health care in general, I think the integration of data across different disciplines will continue and help health plans and providers improve their outreach to patients. In addition to answering the question of which patients to contact, predictive analytics will improve choices in method of communication—what to say specifically, and how to say it. I think you will see entities putting more provider and payer fees at risk, based on the improvement in metrics that are first calculated using predictive analytics.
The iPhone was released 10 years ago, and the amount of additional data the smartphone has generated has been profound. It will be interesting to see how much additional data will be available in the next 10 years.
Q: What is the most challenging aspect of your work?
A: In my role as a product and model designer, there is innovation and an opportunity to think about solutions to problems differently. Innovation is extremely rewarding, but it can also be challenging. It is exciting to watch a customer adopt one of your ideas; implement a solution; and have a positive impact on financial results, operational efficiency and member satisfaction. But it takes a lot of trial and error, persistence and confidence to get to that point. In my mind, the hardest part of my job is letting go of an idea and going in a new direction, especially if I have spent a lot of time trying to make the original idea work. It is having the ability to pull out of the details, take a step back and reevaluate. Many times, the final product will contain elements of the original design, but being able to switch gears along the way is the key to success.
Q: What is most gratifying about a career as an actuary?
A: The day-to-day challenge of being presented a problem, wrestling with it, devising a solution and presenting results is the most gratifying part about being an actuary. Combine that with the diversity in application of domain knowledge and the use of problem-solving skills, and you have a universe of opportunities to be continually challenged. Specifically, in the predictive modeling space there are so many untapped areas to apply concepts. The universe will continue to expand and only make the actuary’s job more interesting and solving problems even more satisfying.
Q: What are the main skills actuaries need for work in predictive analytics? What skills do you think actuaries bring to analytics that other professionals may not bring to the role?
A: Predictive analytics is such a broad term and covers so many different areas, that It depends on what aspect of predictive analytics an actuary may be interested in. Actuaries are trained in such a way that they can fill a variety of different roles, so a variety of skill sets can work. For example, an actuary who likes to program, embraces math theory and enjoys building tools may like to serve in the role of a machine learning expert. The machine learning expert will write code to allow for the building of different types of algorithms. If the interest lies more in training, testing and validating models, then the actuary will be functioning as more of a data scientist. Actuaries can also assist in data science with domain expertise by providing insights on what data should be used and which solutions are feasible, and providing expert insight into the behavior of insurance-specific data that can lead to more effective and efficient models. Of course, in many cases, these functions can be combined in different ways, depending on the personnel involved and the particular goals of the predictive modeling function.
The predictive analytic process is just as important as the resulting model. From taking in new data all the way through to development of the final model, having a process that is repeatable will save time and money and result in a more consistent end product. In any of these roles, actuaries can bring a stronger business acumen to the predictive analytic process than a nonactuary. This allows for the predictive analytic processes to be better thought out from the beginning of a project and will result in better models, collaboration and implementation.
Q: Any advice for actuaries interested in the health care/predictive analytics arena?
A: If you are interested in how you can incorporate predictive analytics into your day-to-day work, or have an interest in focusing 100 percent in the predictive modeling area, take advantage of the webinars, publications and meetings the Society of Actuaries offers. It has done a great job positioning actuaries to be leaders in the predictive analytics space as it continues to be such a critical part of an analytics competency. In addition, the materials generally have a nice balance between business application and the technical underpinnings of predictive analytics.
There are also a number of online data science programs from Coursera, Caltech and Johns Hopkins that get into some of the nuts and bolts of the modeling and algorithm process if that is what interests you.
I also have found a lot of great little problems to generate ideas from Kaggle. There are a number of different data sets, discussion forums and contests on Kaggle for the actuary to see the types of problems that predictive analytics can solve.
Finally, there are applications of predictive analytics in every facet of the health care space, so try to get as much health care domain knowledge as you can, regardless of which aspect of predictive modeling you want to explore. If your main job function is in the Medicare bid process, for example, I can guarantee there are predictive modeling applications hidden in the space.
Q: What are some of your best professional memories/experiences as an actuary that may inspire others to explore different actuarial paths?
A: I’ve worked in all aspects of managed care, from product pricing and development, Medicare Advantage, Medicaid, Mergers and Acquisitions, underwriting, provider contracting and so on. I’ve worked behind the scenes building financial and actuarial models for various companies in these areas, as well as on the front lines helping sales win new business or retain existing business. Some of the models I built made huge impacts on efficiency, complexity and accuracy for their intended purpose, and I found that work to be very satisfying.
I don’t know if this will inspire or scare someone away, but my most memorable experience was soon after I finished my FSA when I worked at a Tricare contractor as the lead actuary that provided benefits for retired military and active-duty dependents. The government would issue changes to the original contract (“change orders”) regularly, which we would need to implement, and then calculate the economic impact of the change. All of these change orders resulted in more coverage and better benefits so they increased health care costs.
The problem was, the Department of Defense (DoD) needed to wait to get money through Congress before it could actually pay for the change orders. The government would hire economists to review our work-up of the cost impact, and there was always disagreement in assumptions, methodology and so on. Some of the disagreement was merely a stall tactic while the DoD waited to get money appropriated.
Finally, after budget negotiations with the president and Congress, the DoD received a large chunk of money to pay for the change orders, and then we went from settling at a snail’s pace to rapid fire. And all of the back and forth, which had previously been done by letters and emails, was now done face-to-face. After having gone more than a year without settling one change order, we settled more than 50 in a matter of days, worth tens of millions of dollars to our company. We had to be quick on our toes, because the government economists would fire questions about our methodology and assumptions, and if we could justify, it would make a huge difference in the financial results. The whole experience was pretty incredible, given the high stakes that were in play.