Medicare Advantage (MA) health plans recently finished their annual MA product design exercise and submitted their bids to the Centers for Medicare & Medicaid Services (CMS). The plan design process is a year-long exercise in which plans start to think about the following year’s benefit package after a short break. This interval also is a good time to reflect on the opportunities for health plan teams to improve the plan design process and approach.
“Quantifying benefit importance for each customer segment enables plans to customize their offerings.”
Data-driven insights help plans identify the benefits members want
A key dimension or element of product design is to understand member needs and the supplemental benefits that move the needle from an acquisition perspective. Generally, the feedback from agents, sales teams or syndicated market reports can be used to understand member preferences. While that feedback is valuable, it takes a lot of time and mature processes, including market research capabilities, to gather the right insights. Also, agent or sales team perception and bias are baked into most of these insights. Health plans therefore have an opportunity to supplement a sales team’s qualitative feedback with a data-driven approach to have a complete and unbiased picture of member preferences. The data-driven approach should look at the movement in enrollment and how it corresponds with prior plan updates to understand which benefits have significant impact.
AI- and ML-based solution approach
A data and AI- and ML-based solution approach uses product and enrollment data published by CMS and county-level social drivers of health (SDOH) data. It involves three key steps to understand the impact of benefit inclusion on enrollment:
- Identify county profiles
- Understand the impact of benefits for each county profile
- Validate the results
Identify county profiles
The U.S. has approximately 3,200 counties. These counties differ in terms of population density, MA penetration, access to care and population demographics, factors that impact benefit preferences of members. For example, transportation benefits may be more popular in a region where car ownership is low, and a meal benefit might be more in demand in a low-income area. Studying benefit preferences for similar counties provides a more accurate picture of member preferences than a national analysis or report.
We have created nine county profiles using SDOH data such as population, access to healthcare, demographics, education, employment, food insecurity, housing status, income, language, lifestyle and neighborhood safety. The profiles also account for access to transportation, percentage of veterans, health status and healthcare utilization.
FIGURE 1: County profiles group counties by social drivers of health to help guide healthcare interventions.
Understand the impact of benefits for each county profile
The next step is to create an AI and ML model to understand the impact of different MA benefits on enrollment for each county profile. The model for each profile helps create a relationship between the benefits offered by a plan and its enrollment share.
The model studies the impact of introducing a benefit on enrollment share for each plan, for the last five years, while controlling for factors such as size of the plan, brand recognition, degree of competition and current enrollment share. The benefits that appeared to make a statistically significant impact on enrollment share, for most plans for a segment of counties, are identified as important benefits or benefits with significant impact on enrollment. The model also can quantify the impact of a benefit as a percentage in enrollment left. This numerical relationship can then be used to simulate expected enrollment for a set of benefits or a version of a benefit package under consideration.
The benefit impact or significance also can be plotted against the prevalence of these benefits in a county to understand where the opportunities are, that is, benefits that are valued more but are less prevalent. The illustration below shows prevalence and benefit impact views for counties in profile 9, defined as counties in metropolitan regions with a high per capita income and a healthy population, against benefit significance for profile 2, defined as counties with lower income and education levels and a less healthy population.
FIGURE 2: The supplemental benefits most likely to drive MA enrollments in county profiles 2 and 9
Validate the results
To gain confidence in the benefit impact models, we ran multiple tests that ranged from a statistical analysis to understand model performance using percentage error to plotting enrollment differences between plans with and without significant benefits (see Figure 3). Through literature research, we also made sure the difference in important benefits across county profiles made business sense.
We defined significant benefits, flagged important by the model and lower in prevalence. We observed that the mean value of enrollments is significantly higher with an increase of 10% to 50% across all plan configurations with significant benefits.
FIGURE 3: The significant benefits that affect MA enrollments as seen in clustering profiles 2 and 9
What’s next with advanced analytics in MA plan design
The above process looks at only one part of the equation, the impact of adding a benefit on enrollment lift. But while designing the product, plans must evaluate the marginal value added against the marginal cost incurred at every step to find the optimal balance. This update can be done by using a plan’s internal cost and utilization data on plan benefits to arrive at the marginal value of adding a benefit.
Rather than using county profiles, we also can build benefit importance models for targeted member segments. Quantifying benefit importance for each customer segment enables plans to customize their offerings.
While bringing advanced analytics into MA product design approach can feel overwhelming, plans may benefit from thinking of the journey as one step at a time. Every step should provide incremental value and help plans design competitive products customized for their target population.
Add insights to your inbox
We’ll send you content you’ll want to read – and put to use.
