The 5 Pillars of Health Data Analytics


Guest Blog by Yiding Jiang, FFA, MSc

Analytics skills are foundational to an actuary’s career. For health actuaries, especially in recent years, data analytics and predictive modeling skills have become critically important. Knowing how to effectively find actionable insights from health data is enormously empowering and companies seek out those with prowess in these skills.

For most junior actuaries and data analysts, they are picked up informally on the job, over a long period of time. But certainly, using a more structured learning approach provides a stronger knowledge base and expedites the process.

So, what comprises superior health data analytics skills?

Asking the right questions

The ability to ask the right questions is a key skill in the business world. With a thorough understanding of the background and the motivations that drive various stakeholders, your ability to frame the analytic mission and intelligently perform the analyses that lead to actionable insights will enable you to provide value to your employer or clients.

Knowledge of the data elements

Data is the lifeblood of health analytics. Health care systems generate complex, interconnected datasets, of varying type, detail and accuracy.

While most actuarial analysts are familiar with insurance claims-based ICDs, procedures, drug codes, etc., few truly understand the inherent structures beneath the surface that can generate great insights.

Many other sources and types of data are becoming more available and relevant today, e.g. clinical notes and laboratory codes. Keeping your knowledge of data up-to-date is a continuous task that will reap generous rewards.

Work with data

There is no such thing as “perfect data” in the real world. Despite rapid developments in data generation and management technologies, the old programming adage: “garbage in, garbage out” is as relevant today as ever.

Your ability to identify issues quickly and then pragmatically correct errors and efficiently manage the data process will enable you to more effectively and efficiently meet the many demands of your profession.

Create features from raw data

To get actionable insights out of analyses or predictive models, you might want to spend the bulk of your time, maybe as much as 90%+, on feature engineering.  Whether statistical or expert-based, this process is where you turn raw data into new variables that are better predictors of outcomes.

Usually, features that mirror the real world tend to be better understood and do a better job at leading to actionable insights. The bottom line here is that features in data are where the answers lie.

Build and test predictive models

There are no perfect models. At times, analytics and predictive model building can feel like an art rather than a science. How you gauge whether a model is working depends on a lot of technical as well as practical considerations.

In my upcoming ACTEX webinar Making Your Health Data Confess, I will elaborate on these skills and demonstrate how to apply some of these methods using an actual case study, where we will get predictive about patients most at risk of opioid drug abuse. Please join us on November 15th at 1:00 PM Eastern Time.