Guest post by Andrew M. Webster, MS, ASA, MAAA, Associate of the Society of Actuaries, actuary, Validate Health.
Predictive modelers working in the healthcare industry need to be one-part physician, epidemiologist, economist, and data scientist. Because healthcare is offered in a variety of settings, by a diverse set of highly-trained professionals, it requires health actuaries to model future healthcare cost and utilization with a high degree of precision.
It also requires a hands-on approach to data-mining.
During my decade-long career in healthcare, I’ve had the opportunity to work alongside clinicians while programming at an electronic health records (EHR) company and onsite in a hospital’s skilled nursing facility. Through those experiences, I gained firsthand knowledge of patient care delivery.
Because of that experience, I can recognize a sequence of patient events and care transitions when I see them in patient data. Observing how healthcare is delivered helps non-clinicians recognize which problems are most relevant to physicians.
Most clinicians want to know how data analysis will help improve patient outcomes instead of merely focusing on short-term cost reduction. Communicating the modeling results in a way that is meaningful to physicians and integrating results into their daily workflows is essential. While most physicians are not mathematicians, they are highly trained in the scientific method and ask insightful questions when reviewing modeling results from actuaries and data scientists.
As an example of the benefits of predictive modeling, my team helped a 20-physician independent practice determine which segment of its patient population was the most costly. By mining the practice data, we identified a specific Medicare Advantage Plan for patients with Chronic Obstructive Pulmonary Disorder (COPD). The medical practice then used the data to redesign its discharge protocol and develop a COPD care management program to help keep patients out of the hospital, improve quality care and lower costs. The solution was effective because of the hands on approach from everyone involved.
In healthcare, simply delivering a risk adjustment model is no longer sufficient. Seeing how physicians apply the data is critical. It also allows actuaries to tailor models to ensure they are adaptive to risk and deliver optimal outcomes to benefit clinicians.
Healthcare data is not static. It is a time series. It is a patient story. To understand a patient’s healthcare journey and predict that patient’s risk, an actuary has to analyze patient EHR and claims data. The physician is an indispensable resource for translating medical codes and terminology, explaining possible root causes for patterns of care, and framing model results within a context of change. In short, actuaries help to statistically analyze data while the physician can translate modeling results from paper to practice. Both clinicians and data scientists benefit from the other’s unique perspective and approach.
Physicians are used to relying on clinical research and randomized clinical trial data. They are trained to use data to diagnose and treat patients. The administrative and clinical data actuaries rely on is seldom used to treat individual patients – rather it is often used to analyze population characteristics.
The technology that is spawning in healthcare has led to unprecedented opportunities that will only continue to improve as data analytics grows in sophistication. Medical and information technology is revolutionizing patient care and substantially improving benefits. Once predictive-modelers gain firsthand knowledge by stepping out from behind their desks and standing next to physicians at patients’ bedsides, claims data will never look the same. The entire industry stands to benefit from an increased focus on predictive modeling and data analytics.