Significance of benefits and other factors on enrollment gain
December 13, 2021 Whitepaper
Multiple factors come together to form a successful plan. While some factors strongly correlate to net enrollment gains, other factors serve as a value add to make plans more attractive to beneficiaries. Therefore, identifying these factors and correctly basing them as Significant or Insignificant factor is crucial in designing attractive plans and improving enrollment gains.
The various attributes have been segmented into three levels –
- Significant – Important factor in any given region which have a high correlation with enrollment gain.
- Insignificant – Factor that has low correlation with enrollment gains but can be used as a value add to improve plans.
- Table Stakes – Benefits present in nearly all the existing plans. Can be considered the bare minimum a plan must offer to be regarded as a competitive plan.
This study focuses on Non-SNP, MAPD plans for HMO & PPO Plan Types.
The results of this study were derived from a combination of black box and explainable AI (xAI) models. Random forest, neural networks, and multiple regression, along with regularization techniques like Lasso and Ridge regression, were some of the methodologies used to develop the significance bucketing AI/ML models. The results of these complex algorithms were then used as inputs for xAI methodologies such as LIME (Local Interpretable Model Agnostic Explanations). Finally, the results were verified using a sound business understanding foundation.
Fifty-four factors comprising of Original Medicare benefits, Optional benefits, and Cost & Other factors were used as inputs for the significance bucketing AI/ML models. While all 54 factors were segmented as Significant, Insignificant, or Table Stakes, this whitepaper focuses only on Optional benefits, and Cost & Other factors. With Original Medicare benefits being included in each plan irrespective of its significance towards net enrollment gain, those benefits were not considered in this analysis. However, evaluation of those benefits is equally important, and the influence is measured for the variables of the benefit, such as member cost, which was not considered in this whitepaper.
The results found below are the result of an in-depth, nationwide analysis of each factor, which have been rolled up to the nine census regions of the United States. Some factors are broken by plan type (HMO & PPO) to demonstrate how the significance, or influence of variables, can vary between the two segments, while other factors are shown at an overall level when the two had similar results.